Robotic Wireless Sensor Networks Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Abstract In this chapter, we present a literature survey of an emerging, cutting-edge, and multi-disciplinary field of research at the intersection of Robotics and Wireless Sensor Networks (WSN) which we refer to as Robotic Wireless Sensor Networks (RWSN) . We define a RWSN as an autonomous networked multi-robot system that aims to achieve certain sensing goals while meeting and maintaining certain communication performance requirements , through cooperative control, learning and adaptation. While both of the com- ponent areas, i.e., Robotics and WSN, are very well-known and well-explored, there exist a whole set of new opportunities and research directions at the intersection of these two fields which are relatively or even completely unex- plored. One such example would be the use of a set of robotic routers to set up a temporary communication path between a sender and a receiver that uses the controlled mobility to the advantage of packet routing. We find that there exist only a limited number of articles to be directly categorized as RWSN related works whereas there exist a range of articles in the robotics and the WSN literature that are also relevant to this new field of research. To connect the dots, we first identify the core problems and research trends related to RWSN such as connectivity, localization, routing, and robust flow of information. Next, we classify the existing research on RWSN as well as the relevant state-of-the-arts from robotics and WSN community according Pradipta Ghosh (*Corresponding author) University of Southern California, Los Angeles, CA-90089 e-mail: pradiptg@usc.edu Andrea Gasparri Universit degli studi ”Roma Tre”, Via della Vasca Navale, 79 00146 Roma (Italy) e-mail: gasparri@dia.uniroma3.it Jiong Jin Swinburne University of Technology, Melbourne, VIC 3122, Australia e-mail: jiongjin@ swin.edu.au Bhaskar Krishnamachari University of Southern California, Los Angeles, CA-90089 e-mail: bkrishna@usc.edu 1 arXiv:1705.05415v3 [cs.RO] 2 Sep 2018 2 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari to the problems and trends identified in the first step. Lastly, we analyze what is missing in the existing literature, and identify topics that require more research attention in the future. 1 Introduction Robotics has been a very important and active field of research over last cou- ple of decades with the main focus on seamless integration of robots in human lives to assist and to help human in difficult, cumbersome jobs such as search and rescue in disastrous environments and exploration of unknown environ- ments [1, 2]. The rapid technological advancements over last two decades in terms of cheap and scalable hardware with necessary software stacks, have provided a huge momentum to this field of research. As part of this increas- ing stream of investigations into robotics, researchers have been motivated to look into the collaborative aspects where a group of robots can work in synergy to perform a set of diverse tasks [3, 4]. Nonetheless, most of the research works on collaborative robotics, such as swarming, have remained mostly either theoretical concepts or incomplete practical systems which lack some very important pieces of the puzzle such as realistic communication channel modeling and efficient network protocols for interaction among the robots. Note that, we use the term “realistic communication channel model” to refer to a wireless channel model that accounts for most of the well-known dynamics of a standard wireless channel such as path loss, fading, and shad- owing [5]. On the other hand, the field of Wireless Networks (more specifically, Wireless Sensor Networks (WSN) and Wireless AdHoc Networks) has been explored extensively by communication and network researchers where the nodes are considered static (Sensor nodes) or mobile without control (Mobile AdHoc Network). With the availability of cheap easily programmable robots, researchers have started to explore the advantages and opportunities granted by the controlled mobility in the context of Wireless Networks. Nonethe- less, the mobility models used by the network researchers remained simple and impractical, and not very pertinent to robotic motion control until last decade. Over last decade, a handful of researchers noticed the significant disconnec- tion between the robotics and the wireless network research communities and its bottleneck effects in the full fledged development of a network of collab- orative robots. Consequently, researchers have tried to incorporate wireless network technologies in robotics and vice verse, which opened up a whole new field of research at the intersection of robotics and wireless networks. This new research domain is called by many different names such as “Wire- less Robotics Networks” , “Wireless Automated Networks” and “Networked Robots” . In this chapter, keeping in mind that the primary task of teams of robots in many application context might be pure sensing, we will refer to Robotic Wireless Sensor Networks 3 this field as “Robotic Wireless Sensor Networks (RWSN)” . According to the IEEE Society of Robotics and Automation’s Technical Committee: “A ‘net- worked robot’ is a robotic device connected to a communications network such as the Internet or LAN. The network could be wired or wireless, and based on any of a variety of protocols such as TCP, UDP, or 802.11. Many new applications are now being developed ranging from automation to exploration. There are two subclasses of Networked Robots: (1) Tele-operated, where hu- man supervisors send commands and receive feedback via the network. Such systems support research, education, and public awareness by making valuable resources accessible to broad audiences; (2) Autonomous, where robots and sensors exchange data via the network. In such systems, the sensor network extends the effective sensing range of the robots, allowing them to commu- nicate with each other over long distances to coordinate their activity. The robots in turn can deploy, repair, and maintain the sensor network to in- crease its longevity, and utility. A broad challenge is to develop a science base that couples communication to control to enable such new capabilities.” We define a RWSN as an autonomous networked multi-robot system that aims to achieve certain sensing goals while meeting and maintaining certain communication performance requirements via cooperative control, learning, and adaptation. Another important definition related to this field is “coop- erative behavior” which is defined as follows: “given some task specified by a designer, a multiple-robot system displays cooperative behavior if, due to some underlying mechanism (i.e., the “mechanism of cooperation”), there is an increase in the total utility of the system.” A group of cooperative robots is of more interest than single robot because of some fundamental practical reasons such as easier completion and performance benefits of using multiple simple, cheap, and flexible robots for complex tasks. Over the years, robotics and wireless network researchers have developed algorithms as well as hardware solutions that directly or indirectly fall under the umbrella of RWSN. Network of robots are already experimentally applied and tested in a range of applications such as Urban Search And Rescue mis- sions (USAR), fire fighting, underground mining, and exploration of unknown environments. The very first practical USAR missions was launched during the rescue operations at the World Trade Center on 11 September 2001 [2] us- ing a team of four robots. In the context of fire fighting, a group of Unmanned Aerial Vehicles (UAV) was used for assistance at the Gestosa (Portugal) for- est fire in May 2003 by Ollero et al. [6]. Communication links between the robots in such contexts can be very dynamic and unreliable, thereby, require special attention for an efficient operation. This requires careful movement control by maintaining good link qualities among the robots. Among other application contexts, underground mining is very improtant. Due to many difficulties like lack of accurate maps, lack of structural soundness, harshness of the environment (e.g. low oxygen level), and the danger of explosion of methane, accidents are almost inevitable in underground mining resulting in the deaths of many mine workers. Thrun et al. [7] developed a robotic sys- 4 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari tem that can autonomously explore and acquire three-dimensional maps of abandoned mines. Later, Murphy et al. [8] and Weiss et al. [9], also presented models and techniques for using a group of robots in underground mining. The field of cooperative autonomous driving, one of the major research focus in the automobile industry, also falls under the broad umbrella of RWSN. Baber et al. [10], Nagel et al. [11], Milanes et al. [12], and Xiong et al. [13] worked on solving a range of problems in practical implementations of au- tonomous driving systems and flocking of multiple unmanned vehicles like Personal Air Vehicle (PAV). Robot swarms [14, 15], which deal with large numbers of autonomous and homogeneous robots, are also special cases of RWSN. Swarms have limited memory and have very limited self-control ca- pabilities. Example use cases of swarms are in searching and collecting tasks (food harvesting [16], in collecting rock samples on distant planets [17]), or in collective transport of palletized loads [18]. Penders et al. [1], developed a robot swarm to support human in search missions by surrounding them and continuously sensing and scanning the surroundings to inform them about potential hazards. A swarm of robots can also be used in future health-care systems e.g., swarm of micro robots can be used to identify and destroy tumor/cancer cells. Military application is another obvious field of applica- tion. Many researchers have been working on developing military teams of autonomous UAVs, tanks and Robots, e.g., use of Unmanned Ground Vehi- cles (UGV) during RSTA missions. One example of such project is “Mobile Autonomous Robot Systems (MARS),” sponsored by DARPA. The works of Nguyen et al. [19] and Hsieh et al. [20] are mentionable on military application of Networked Robots. In the field of Exploration, the most famous example is the twin Mars Exploration Rover (MER) vehicles. They landed on Mars in the course of January 2004 [21]. Among other applications, hazardous waste management, robot sports [22], mobile health-care [23], smart home [24–26], smart antenna, deployment of communication network and improvement of current communication infrastructure [27, 28], and cloud networked robotics [29–31] are also important. In summary, there exists a huge range of appli- cations of a RWSN. In Table 1, we list the different types of applications of a RWSN. In this chapter, we present a literature survey of the existing state-of- the-arts on RWSN. We discover that while there exist significant amount of works [32] in both the ancestral fields (Robotics and WSN) that are also rele- vant to RWSN, most of these state-of-the-arts cannot be directly classified as RWSN related works, yet, should not be omitted. Nonetheless, there also ex- ist a range of works that properly lie at the intersection of robotics and WSN and, therefore, directly fall under the purview of RWSN. To draw a complete picture of the RWSN related state-of-the-arts, we first identify and point out current research problems, trends, and challenges in the field of RWSN such as connectivity, localization, routing, and robust flow of information. While some of these problems are inherited from the fields of robotics (such as path control and coordination) and WSN (such as routing and localization), Robotic Wireless Sensor Networks 5 Table 1: Application Summary of RWSN Applications Reference Search and Rescue [1, 2, 6, 10] Mining [7–9] Autonomous Driving [10–13] Robot Swarm [14–18] Military Applications [19–21] Robot sports [22] Mobile health-care [23] Smart Home [24–26] Deployment of Communication Network [27, 28] Cloud Networked Robotics [29–31] a new class of independent problems have also emerged such as link quality maintenance, radio signal strength information (RSSI) estimations in present and future location of a robot, and guaranteed proximity between neighbor- ing nodes. Secondly, we categorize the existing research in accordance to the problems they address. In doing so, we also include solutions that are not directly applicable but provide with a solution base to build upon. For example, in the contexts of connectivity maintenance, the traditional solu- tions involve representing the network as a graph by employing simple unit disk wireless connectivity model [33]. However, in practical employments, the wireless communication links do not follow unit-disk model and rather follow a randomized fading and shadowing model [5]. Thus, such existing solutions do not directly fall under RWSN yet can be modified to include more realistic communication model and, thus, should not be omitted. Lastly, we discuss what is missing in the existing literature, if any, as well as some potential directions of future research in the field of RWSN. In this chapter, we fur- ther discuss how and where do the existing works fit in the context of the layered architecture (Internet model) of a network stack in order to identify key networking goals and problems in a RWSN. 2 What is a RWSN? In this section, we illustrate the field of RWSN in details. Here, we address some core and important questions related to RWSN: What is an RWSN? What kind of research works are classified as RWSN related works? We define an RWSN as a wireless network that includes a set of robotic nodes with controlled mobility and a set of nodes equipped with sensors; whereas all nodes have wireless communication capabilities. Ideally, each node of a robotic sensor network should have controlled mobility, a set of sensors, and wireless communication capabilities (as illustrated in Figure 1). 6 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Fig. 1: Illustration of a Robotic Wireless Sensors We refer to such nodes (devices) as “Robotic Wireless Sensors (RWS)” . Nonetheless, an RWSN can also have some nodes with just sensing and wire- less communication capabilities but without controlled mobilities. We refer to such nodes (by following traditional terminology) as “Wireless Sensors” . Note that, every node of an RWSN must have wireless communication capa- bilities according to our definition. Moreover, an RWSN is typically expected to be able to fulfill or guarantee certain communication performance require- ments enforced by the application contexts such as minimum achievable bit error rate (BER) in every link of the network. To answer the second question, the existing research works related to RWSN can be subdivided into two broader genres. The first genre focuses on generic multi-robot sensing systems with realistic communication chan- nels (i.e., include the effects of fading, shadowing etc.) between the robots. To clarify, these are mostly the existing works in the robotics literature on multi-robot systems but with practical wireless communication and network- ing models. One application context of such a RWSN is in robot assisted fire-fighting where the robots are tasked to sense the unknown environments inside rubble to help and guide the firefighters. Now, if the robots are not able to maintain a good connectivity among themselves or to a mission control station, the whole mission is voided. Refer to Figure 2 for an illustration of such contexts where a group of robots are sensing an unknown environment to guide the human movements. In Figure 2, the network consisting of five robots and two firefighters needs to be connected all the time and also needs to have properties such as reliability and lower packet delays. Thus, we need a class of multi-objective motion control that will optimize the sensing and Robotic Wireless Sensor Networks 7 exploration task performance, and will also ensure the connectivity and the performance of the network. Some of the main identifiable challenges in this genre of works are: connectivity maintenance, efficient routing to reduce end to end delay of packets, and multi-objective motion control and optimization. Fig. 2: Illustration of Robotic Sensor where a group of five robots are sensing the environment around the firefighters to guide them in firefighting while also providing connectivity The second genre of works focus on the application of RWS to create and support a temporary communication backbone between a set of communicat- ing entities. In these contexts, we sometime use the terms “robotic router” and “robotic wireless sensor” synonymously, to put emphasis on the com- munication and routing goals. The main theme of these works is to exploit the controlled mobility of the robotic routers to perform sensing and communi- cation tasks. Note that, there exists a vast literature on multi-agent systems in robotics and control community that apply simple disk models for com- munication modeling and, subsequently, apply graph theory to solve different known problems such as connectivity and relay/repeater node placements. In order to be directly included in the RWSN literature, these existing works need to include the effects of fading and shadowing in the communication models which is likely to significantly increase the complexity of the problems as well as the solutions. An example of the second genre of works is in the application of RWS in setting up a temporary communication backbone. While sensing is 8 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari still involved for the robotic router placement optimization and adaptation, the main purpose of the system is to support communication, not sensing. In Figure 3, we present an example illustration where a set of two robotic routers form a communication relay path between two humans (e.g., two fire-fighters) who are unable to communicate directly. Some of the main challenges in this genre of RWSN research are: link performance guarantee (in terms of Signal to Interference plus Noise Ratio, SINR, or Bit Error Rate, BER), optimized robotic router placements and movements in a dynamic network, non-linear control dynamics due to inclusion of network performance metrics into con- trol loop, and localization. A special case of this would be robot assisted static relay deployments, where the robots act as carriers of static relay nodes and smartly place/deploy them to form a communication path/backbone. Fig. 3: Illustration of robotic routers where the two humans are not able to communicate directly due presence of a wall or some other blocking object Next, we identify a set of system components and algorithms required in an RWSN as follows. All these pieces are individual research problems themselves and thus require separate attentions. • RSSI Models, Measurements, and RF mapping: In an RWSN, it is important to estimate and monitor the quality of the communication links between the nodes (in terms of Bit Error Rate (BER), Signal to Noise plus Interference Ratio (SINR) etc.) in order to satisfy the communication related requirements (Note that, RF based communication is standard mode of communication in RWSN for obvious reasons). For practicality, these estimations must be either partly or fully based on online RF sensing such as temporal RSSI measurements in a deployment. Moreover, in some Robotic Wireless Sensor Networks 9 application contexts of RWSN, the sole goal of a robotic sensor network can be to sense and formulate an RF map of an environment to be processed or exploited later on. We present a survey of such RF mapping and modeling related works in Section 3.1. • Routing Protocols: Similar to any wireless sensor networks, routing and data collection is crucial in an RWSN. The concept of RWSN have opened up the door to a new class of routing protocols that incorporates the controlled mobility of the nodes in the routing decisions for more effec- tive communication. Moreover, end to end delay reduction and reliability improvement have become of prime interests. A brief survey of existing RWSN related works on routing protocols is presented in Section 3.2. • Connectivity Maintenance: In any collaborative network of robots, it is important to maintain a steady communication path (direct or multi-hop) between any pair of nodes in the network for an effective operation. This problem, traditionally referred to as connectivity maintenance problem, in very well studied by the robotics research community. A survey of such state-of-the-arts is presented in Section 3.3. • Communication Aware Robot Positioning and Movement Con- trol: As mention earlier, one of the application contexts of RWSN is in supporting temporary communication backbones. The most important re- search question in such contexts is to devise a control system that adapts the positions of the robotic routers throughout the period of deployment to optimize the network performance while optimizing the movements as well. Therefore, the main goal of these class of work is continuous joint optimization of the robotic movements and the wireless network’s perfor- mance. Moreover, the router placement controller should also be able to support network dynamics such as node failures and change in the set of communication endpoints. Another important application context of RWSN systems is distributed coordinated sensing using multiple robots. In such sensing contexts, the robotic sensing agents should be able to opti- mally sense the region of interest and route the sensed data to other nodes or a command center. This also require careful communication-channel- dynamics-aware positioning of the robotic sensors. We present a summary of such communication aware robotic router/sensor positioning works in Section 3.4. • Localization: Localization is a well known problem in the field of WSN as well as robotics. Thus, it is quite intuitive to be included in the field of RWSN. Moreover, the field of RWSN sometimes requires techniques for robots to follow each other or maintain proximity to each other. For that high accuracy relative localization is more important than absolute localization. We present a brief summary of such localization related works in Section 3.5. 10 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari 3 RWSN Systems Components: In this section, we present a categorical survey of all state-of-the-arts in the field of RWSN. The works are classified according to the problem addressed. 3.1 RSS Models, Measurements, and RF Mapping Radio Signal Strength (RSS) variation over a spacial domain greatly impacts the wireless communication properties, such as power decay and packet loss, between two nodes. Different properties of a physical environment, such as obstacles and propagation medium, affect radio signal propagation in dif- ferent way, thereby, causing fading, shadowing, interference, and path loss effects [5]. All these should be taken into consideration (via proper com- munication channel models) for proper selection of radio transmission and reception parameters to improve the communication quality in a RWSN. While there exists a range of standard communication channel models in the literature, such as simple path loss model and log normal fading model [5], the applicability as well the model parameters’ values depend on the actual deployment environments. For example, according to a log normal fading model [5], the received power is calculated as follows. P r,dBm = P t,dBm + G dB − L ref − 10 η log 10 d ( t ) d ref + ψ P ref r,dBm = P t,dBm + G dB − L ref + ψ (1) where P r,dBm is the received power in dBm, P t,dBm is the transmitter power in dBm, G dB is the gain in dB, L ref is the loss at the reference distance d ref , η is the path loss exponent, d ( t ) is the distance between the transmitter and receiver, ψ ∼ N (0 , σ 2 ) is the random shadowing and multipath fading noise which is log normal with variance σ 2 , and P ref r,dBm is the received power at reference distance ( d ref ). While some of the variables such as P t,dBm , d ref , and G dB are known or can be measured, the values of other variables such as η , L ref , and σ are dependent upon the deployment environment. Thus, it is important to identify or estimate the proper values of such parameters in the deployment environment in order to estimate the P r,dBm . Moreover, RSS models and maps are very important and useful database that can be used for a range of purposes such as localization of nodes based on received signal strength [34–37], mapping of an unknown terrain [38], and identifying obstacles [39]. The communication model based estimations of the RF signal strengths in the future and unvisited locations are also very important for maintaining connectivity among mobile nodes and for optimizing the network Robotic Wireless Sensor Networks 11 performance. In this section, we provide a brief overview of the state-of-the- art RSS modeling and mapping techniques that are applicable in a RWSN. What is already out there? In the context of mapping and modeling of the RF channels, the most common and practical class of approaches is to deploy the network of robots with a initial model of the communication channel. Then, the robots continually or periodically collect RF samples to update the communication model parameters in an online fashion. One can also opt for an offline modeling where the robots collect a set of RF samples over the region of interest followed by post-processing of all the samples to estimate the communication channel properties. Nonetheless, the later class of methods is unrealistic and of little interest to us as it can not cope with temporal changes in the communication channel properties, which is a quite common phenomenon. One of the key challenges of online RF mapping is in the sparsity of RF samples. Mostofi et al. [40] have done significant research in exploiting this sparsity to their advantage for RF channel modeling in networked robot sys- tems. In [41], Mostofi and Sen presented a technique of RSS mapping by exploiting sparse representation of the communication channel in frequency domain. They demonstrated how one can reconstruct the original signal us- ing only a small number of sensing measurements by employing the theory of compressive sampling [42]. Later, they utilized the compressive sampling based reconstruction of the signal in the domain of cooperative mapping [38] to build a map of the region of interest. Mostofi et al. [43] further pre- sented an overview of the characterization and modeling of wireless channels for a RWSN. In their works, all three major dynamics in a physical wire- less communication channel (small-scale fading, large-scale fading, and path loss) are considered [44–46]. In [46], Mostofi, Malmirchegini, and Ghaffarkhah also presented techniques for channel predictions in future locations of robots based on the compressed sensing based channel models. Since all these are completely based on wireless measurements, multi-path fading has a great in- fluence over the results. Later on, Yan and Mostofi [47] presented a combined framework for the optimization for motion planning and communication plan- ning. The concept of compressive sensing based signal reconstruction and ma- terial dependent RF propagation properties are also employed for RF based imaging and mapping of cluttered objects/regions [39, 48, 49]. In these works, RF signal propagation properties (mostly attenuations) of the communica- tion path between pairs of moving robots are used to estimate a structural map of an obstacle or a cluttered region. Hsieh, Kumar, and Camillo [50] have also demonstrated an RSS mapping technique using a group of robots in an urban environment. The goal was to learn the environment’s communi- cation characteristics to generate a connectivity graph. Later on, they used this model for connectivity maintenance problem in a network of robots [51]. In [52], Fink and Kumar presented another mapping technique using multi- ple robots. Their goal was to use the mapping for localization of an unknown source using a Gaussian process based maximum likelihood detection. They 12 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari further extended the Gaussian process based channel model to guarantee a minimum stochastic end-to-end data rate in a network of robots [53]. Signal attenuation factors due to presence of obstacles are taken into account in the work of Wang, Krishnamachari, and Ayanian [54]. Ghosh and Krishna- machari [55] have proposed a log normal model based stochastic bound on interference power and SINR for any RWSN. The works presented in [56–58] are also related to this context. What would be the potential future research directions? To our knowledge, a generic model for interference and SINR estimation is missing in the current literature. Interference modeling is a key to develop a more realis- tic model of wireless channels. Interference should also be taken into account for robotic router placements contexts, where the main goal of the network is to guarantee certain communication performance qualities. In our opin- ion, this should be a major focus of future research in this topic. Moreover, the effects of channel access protocols, such as Carrier Sense Multiple Access (CSMA) [5], needs to be taken into account in the interference and SINR models. The existing signal strength models could be extended to achieve this goal. Nonetheless, to our knowledge, this has remained an unexplored problem in the context of a RWSN. Another future research direction would be to perform an extensive set of measurement experiments in real (rather than simulated) mines, undergrounds, or fire fighting environments. This data can be used to identify the RF properties in such environments that can be exploited to the advantage of RWSN system design. Table 2: Summary of RSS Models, Measurements and RF Mappings Related Works Algorithm Classes • Online: Modeling during deployment [38, 40, 41, 43, 46, 47] • Offline: Modeling before deployment [50, 51, 55] Challenges • Sparse Sampling [38, 40] • Future Location’s Signal Prediction [46, 47] • Temporal Dynamics [46] Available Theoretical Tools • Compressive Sampling [42] • Gaussian Process [52, 55] • Fading Models [5] • Path Loss Models [5] Potential Future Directions • Interference and SINR Models • Account for Channel Access Methods like CSMA • Real World Data Collection and Analysis Robotic Wireless Sensor Networks 13 3.2 Routing Routing in an RWSN can be considered same as in Mobile Adhoc Networks (MANET) but with an extra advantage of controllability. In MANET, there are mainly two types of popular routing algorithms called reactive and proac- tive techniques. Among the reactive techniques, Ad-hoc On demand Distance Vector (AODV) [59, 60] and Dynamic Source Routing (DSR) [61, 62] are the most popular ones. On the other hand, in the class of proactive techniques, Optimized Link State Routing (OLSR) [63] and B.A.T.M.A.N. [64] are the popular ones. While any of these algorithms can be used for an RWSN, they do not take advantage of the extra feature in RWSN: controlled mobility. Ideally, the controlled mobility aspect should also be taken into account for optimized routing decisions. Nonetheless, it remained to be one of the less explored areas in RWSN. Moreover, a lower end-to-end delay and higher re- liability in packet routing (mostly control packets) are two important and required aspects in an RWSN. Delayed or missing packets can result in an improper collaborative movement control and task completion in an RWSN. To this extent, some researchers have modified existing routing solutions to adapt in a robotic network and proposed completely new routing solutions as well. In this section, we present a brief survey of the state-of-art rout- ing techniques in RWSN that are developed or modified with sole focus on robotics. MRSR, MRDV and MRMM: In [65], Das et al. presented three rout- ing protocols based on traditional mobile ad-hoc protocols, such as DSR [61] and AODV [60], for routing in a network of mobile robots. A brief description of each of these algorithms are as follows. • Mobile Robot Source Routing (MRSR): It is a unicast routing algorithm based on Dynamic Source Routing (DSR) [66]. MRSR incorporates three mechanisms: route discovery , route construction , and route maintenance . In the runtime of route discovery phase, each robot along the pathway of route reply message encodes its mobility information into the route reply packet. During route construction , MRSR exploits graph cache that con- tains the topological information of the network. The route maintenance phase is similar to the maintenance method applied in DSR. • Mobile Robot Distance Vector (MRDV): This is also a unicast routing algorithm based on the well known AODV [67] routing protocol. MRDV protocol adopts AODV features such as the on-demand behavior and hop- by-hop destination sequence number. Nevertheless, unlike MRSR, MRDV explores only one route that may not have the longest lifetime among all possible routes, thereby, resulting in high probability of route errors. • Mobile Robot Mesh Multicast (MRMM): This is a multicast protocol for mobile robot networks based on ODMRP (On Demand Multicast Routing Protocol) [68] for MANETs. 14 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Adaptive Energy Efficient Routing Protocol (AER): This protocol was proposed by Abishek et al. [69] to achieve optimal control strategy for performing surveillance using a network of flying robots. AER protocol is also subdivided into three phases (similar to DSR [66]): route discovery , route maintenance , and route failure handling . The residual energy levels and signal strengths at the neighboring nodes are the main route determining factors in this protocol. To model them, the authors defined two decision parameters: T (attribute value of the neighbor) and C (cost function). The best value of T decides the forwarding node. The nodes that are neither selected for message forwarding nor have sufficient energy, are switched to the sleep state to minimize energy consumptions. ACTor based Robots and Equipments Synthetic System (AC- TRESS): In [70], Matsumoto et al. proposed a robotic platform called AC- TRESS that consists of robotic elements referred to as robotors . They also proposed a routing protocol exclusively for that platform. The messages are classified into two different classes: messages to establish/relinquish a com- munication link, and messages for control and rest of the purposes. The first kind of messages use traditional communication protocols such as TCP/IP. The second type of messages use their proposed special protocol to estab- lish logical links, allocate tasks, and control cooperative motions. For this purpose, the authors have introduced four levels of messages: Physical level, Procedural level, Knowledge level, and Concept level. The common part of all four types of messages is referred to as the message protocol core , which is used for: negotiation, inquiry, offer, announcement and synchronization. WNet: Tiderko et al. [71] proposed a new multi-cast communication tech- nique called WNet that is based on the well-known Optimized Link State Routing(OSLR) protocol [63]. Similar to OLSR, WNet uses HELLO and Topology Control (TC) management frames to create and update the net- work topology graph stored in each robot node. However, the packet frames are integrated with some additional information of link attributes that is used for link quality estimation. Next, the Dijkstra algorithm [72] is applied to the topology graphs to determine the routing paths. Steward Assisted Routing (StAR): In [73], Weitzenfeld et al. pre- sented a new routing algorithm called StAR that deals with mobility and interference in an RWSN. The objective of this protocol is to nominate, for each connected partition, a “steward” for each destination. These stewards are noting but next hop robots toward destination that can store the data until a route to the destination is available. The message routing of StAR is based on a combination of global contact information and local route main- tenance. Also periodic broadcast messages with unique source identifiers are sent containing topological location of the active destination. At the begin- ning of this process, each node selects itself as the steward and then progres- sively changes the local steward based on advertisements from the neighbors. StAR uses a sequence number to maintain the freshness of information, sim- ilar to AODV protocol. Robotic Wireless Sensor Networks 15 Optimal Hop Count Routing (OHCR) and Minimum Power over Progress Routing (MPoPR): Hai, Amiya, and Ivan [74] are among the few researchers who leveraged controlled mobility of the robotic routers to assist in wireless data transmission among fixed nodes. This method is di- vided into two parts. The first, which they refer to as Optimal Hop Count Routing (OHCP), computes the optimal number of hops and optimal dis- tances of adjacent nodes on the route. Each node identifies its closest node by comparing the respective neighbors’ distances with the optimal distance. If a node cannot find any such neighboring node, it sends back a route failure message to the source. Otherwise, the second part of the routing, which the authors refer to as Minimum Power over Progress Routing (MPoPR), uses greedy routing on the results obtained from OHCP to minimize the total transmission power. Synchronized QoS routing: In [75], Sugiyama, Tsujioka, and Murata presented a QoS routing technique for a robotic network that is based on the Qos Routing in Ad-hoc network [76] and DSDV routing protocol [77]. In this method, they used the concept of Virtual Circuits to reserve a specified bandwidth. It is the job of the sender to reestablish the circuit in case of a broken connection due to topology changes. This method also includes a methodology for accelerating the transmissions of the control packets. Topology Broadcast based on Reverse-Path Forwarding (TBRPF): In the CENTIBOT project [78], Konolige et al. used a proactive MANET technique called Topology Broadcast based on Reverse-Path Forwarding (TBRPF) to deal with multi-hop routing in dynamic robotic network. This link-state routing protocol was originally proposed by Bellur and Ogiel [79, 80]. In this algorithm each node maintains a partial source tree and report part of this tree to its neighbor. To deal with mobility it uses a com- bination of periodic and differential updates. B.A.T.Mobile: Sliwa et al. [81] have also proposed a mobility aware routing protocol called B.A.T.Mobile which builds upon the well-known B.A.T.M.A.N routing [64] protocol for MANET. This algorithm relies on a future position estimation module for the next hops that uses the current and past position related information as well as the knowledge of the mobil- ity algorithms of the users. The estimation module is further used to rank the neighbors and estimate their lifetime. The neighbor rankings are used to change the route in a proactive manner for end-end to data transfer. Other Methods: There also exist some methods that have the poten- tial to be used in an RWSN after few modifications. Among such methods, the geographic routing algorithms and encounter based routing algorithms are mentionable. Greedy Perimeter Stateless Routing (GPSR) [82] is an ex- ample of geographic routing protocols that uses router positions and packet destinations for making forwarding decisions. If the locations of all nodes in the network are known, this algorithm can be used in RWSN. However, this approach faces many problems in mobile wireless networks. In [83] Son, Helmy and Krishnamachari identified two problems due to mobility in geo- 16 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari graphic routing, particularly in GPSR, called LLNK and LOOP. They also presented two solutions: neighbor location prediction (NLP) and destination location prediction (DLP); to solve those problems. Rao et al. [84] also iden- tified some issues with GPSR and proposed a lifetime timer based solution. In [85], Mauve et al. presented a generalized multicast version of GPSR like geographic routing. There are many other works on position based routing [86, 87]. For a more detailed and complete overview on position based routing algorithms, an interested reader is referred to [88]. Encounter based routing is another relevant group of routing, mainly used in delay tolerant networks (DTN). In general, DTN routing protocols are di- vided into two categories: forwarding-based or replication-based. Forwarding- based protocols use only one copy of the message in the entire network while the replication based technique uses multiple copies of the message. Replica based protocols are also subdivided to two categories: quota-based and flooding-based. Flooding is the most simple and inefficient technique. Balasubhamanian, Levine, and Venkataramani presented a flooding based technique of replication routing in DTN [89, 90], modeling it as a resource al- location problem. Another flooding based technique is presented in [91], called Maxprop. Spyropoulos, Psounis and Raghavendra presented two quota based replication routing techniques for DTN called Spray and Wait [92], and Spray and Focus [93]. There are many other papers on DTN routing [94–96]. Al- though, this group of techniques are not directly related, they can be modified to develop very efficient routing for RWSN. The research works related to data collection protocols in WSN commu- nity are also of interest. Among these protocols, a prominent and recent class of queue aware routing algorithms, called Backpressure routing algorithms [97, 98], has caught our interest. The Backpressure routing algorithms and a range of similar algorithms [99–101] are proved to be ‘throughput optimal’, in theory. One of the most recent Backpressure style routing algorithm is called the Heat Diffusion (HD) routing algorithm [102, 103] that has shown to offer a Pareto-optimal trade-off between routing cost and queue congestion (de- lay). The Backpressure routing algorithms, including HD algorithm, do not require any explicit path computations. Instead, the next-hop for each packet depends on queue-differential weights that are functions of the local queue occupancy information and link state information at each node. There have been several reductions of Backpressure routing to practice in the form of distributed protocols, pragmatically implemented and empirically evaluated for different types of wireless networks [98, 104–106]. Ghosh et al. have also developed a distributed practical version of the HD algorithm called Heat Diffusion collection protocol [107]. While these protocols perform effectively in a static WSN, their applicability in an RWSN are yet to be tested. Since these algorithms do not require any route calculation as well as routing tables, they will require less memory and computation in the resource constrained robotic nodes. Moreover, one extra advantage of such protocols is the adapt- ability in a dynamic network due to not relying on a single predetermined Robotic Wireless Sensor Networks 17 path. Besides these protocols, there exist a number of other prior works on routing and collection protocols for wireless sensor networks, including the Collection Tree Protocol (CTP) [108], Glossy [109], Dozer [110], Low-power Wireless Bus [111], and RPL [112]. These protocols can also be modified for application in RWSN. The work presented in Glossy [109] is of particular interest due to its simplicity and wide adaptability for high throughputs and low delays. What would be the potential future research directions? To our knowledge, there exist a significant amount of research on routing related to MANET and WSN that can be applied to an RWSN either directly or after some modifications. However, a significant focus of future routing algo- rithms need to be directed towards reducing delays, improving reliability, and incorporating the controlled mobility in the routing decisions. The emphasis should be on delay and reliability as on-time message delivery among different control system components is the key for a successful and efficient control sys- tem. One example of using the controlled mobility to our advantage is shown in the works of Wang, Gasparri, and Krishnamachari [113] where the robots ferry messages from a source to sink in a way similar to a postman. Another research direction would be to add node movements in routing decisions. For illustration, assume that there exists two possible routing paths and the rel- atively bad path can be improved considerably by slightly moving the node. Then, the routing decision should include movement into consideration. 3.3 Connectivity Maintenance Connectivity maintenance is a well studied and classic problem in the field of swarm robotics. In the connectivity maintenance problem, the main goal is to guarantee the existence of end-to-end paths between every pairs of nodes. The interaction between pairs of robots is usually encoded by means of a graph, and the existence of an edge connecting a pair of vertexes represents the fact that two robots can exchange information either through sensing or communication capabilities. Notably, the connectivity of the interaction graph represents a fundamental theoretical requirement for proving the convergence of distributed algorithms in a variety of tasks, ranging from distributed estima- tion [114–116] to distributed coordination and formation control [117–119]. Traditional Approach: In the context of robotic networks, where the connectivity of the interaction graph is strictly related to the motion of the robots, a fundamental challenge is the design of distributed control algorithms which can guarantee that the relative motions of the robots do not result in a network partitioning, by relying only on local information exchange. Two possible versions of the connectivity maintenance problem can be considered: 18 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Table 3: Summary of Routing Related Works Routing Algorithm Name/Class References Comments MRSR, MRDV and MRMM [65] Based on Dynamic Source Routing (DSR) [66], AODV [67], and On Demand Multicast Routing Proto- col [68], respectively Adaptive Energy Efficient Routing Protocol(AER) [69] Similar to DSR [66] ACTor based Robots and Equip- ments Synthetic System (AC- TRESS) [70] Main steps are negotiation, in- quiry, offer, announcement and synchronization. WNet [71] Based on Optimized Link State Routing(OSLR) protocol [63] Steward Assisted Routing (StAR) [73] It is a hierarchical routing proto- col where a group of robots act as ”stewards” that can store the data until a route to the destination is available. Optimal Hop Count Routing (OHCR) and Minimum Power over Progress Routing (MPoPR) [74] Uses the controlled mobility of the robotic routers to assist in wire- less data transmission among fixed nodes. Synchronized QoS routing [75] Based on the Qos Routing in Ad- hoc networks [76] and DSDV rout- ing protocol [77] Topology Broadcast based on Reverse-Path Forwarding (TBRPF) [78] It is a link state routing protocol. Each node maintains a partial tree. B.A.T.Mobile [81] Based on B.A.T.M.A.N rout- ing [64] Geographic Routing Algorithms [82–88] Employs locations of the nodes for efficient routing in WSN Encounter Based Routing Routing algorithms for delay tolerant networks • Flooding Based [89–91] • Quota-Based [92, 93] Data Collection Routing in WSN Data collection routing algorithms are used in WSNs for efficient routing of the sensed data to the data sinks. • Backpressure Routing [97, 98] • Heat Diffusion Routing [102, 103, 107] • Glossy [109] • Collection Tree Protocol [108] • Dozer [110] • Low-power Wireless Bus [111] • RPL [112] Potential Future Directions • Modify existing routing protocol to include con- trolled mobility • Focus more on delay reduction, data transfer guarantee, and energy efficiency local connectivity and global connectivity. The local version of the connec- tivity maintenance problem focuses on the preservation of the original set of Robotic Wireless Sensor Networks 19 links of the graph encoding the pairwise robot-to-robot interactions to en- sure its connectedness. The global version of the connectivity maintenance problem focuses on the preservation of the overall graph connectedness, i.e., links can be added or removed as long as this does not prevent the inter- action graph to remain connected over time. Historically speaking, the local version of the connectivity problem has been the first version of the problem to be investigated by the research community. However, it turned out that preserving each of the links of the interaction graph significantly constraint the robots mobility, while, in general, not each link is strictly required to ensure the connectedness of the interaction graph. For this reason, more re- cently the research community has focused mostly on the global version of the problem. Next, we present a brief survey of the available connectivity maintenance protocols that can be used in RWSN. As already mentioned, connectivity maintenance has been studied exten- sively in the contexts of distributed robotics and swarm robotics. Most of the state-of-the-art protocols for connectivity maintenance are based on a graph modeling of the robot-to-robot interactions. More specifically, let G = {V , E} be the interaction graph where V = { 1 , . . . , n } is the set of robots and E is the set of edges encoding the interactions between pairs of neighboring robots. In particular, the existence of an edge is often related to the spatial proximity between pairs of robots, i.e., an edge exists between two robots if the euclidean distance between them is less than a given threshold. By following this graph-based modeling of a robotics network, a natural metric to measure the network connectedness is the algebraic connectivity . More specifically, in the context of graph theory, the algebraic connectivity is defined as the second smallest Eigenvalue, λ 2 ( L ), of the graph Laplacian matrix, L , of the network. In [120, 121], it is shown that λ 2 ( L ) is a concave function of the Laplacian matrix and represents the network connectivity when L is a positive definite matrix. Thus, the connectivity optimization goal becomes simple maximization of the algebraic connectivity value, λ 2 ( L ). Another way to represent the connectivity is via a powered sum of adjacency matrix, A sum = ∑ K i =0 A i where A is the adjacency matrix of the network graph G . A sum basically represents the number of paths up to length K between every pair of nodes in the graph [122]. It follows that for a network to be connected, for all pairs of nodes and K = n − 1, A sum has to be positive definite (where n is number of nodes). Next, we briefly discuss some representative state-of-the-art local and global methods for connectivity maintenance. In [123], Dimarogonas and Kyriakopoulos presented one illustrative local connectivity maintenance ap- proach using two potential fields in the controller where the nodes try to maintain all the initial links throughout the time. Notarstefano et al. also presented a double integrator disk graph based local approach for connec- tivity maintenance [124, 125]. In [126], Spanos et al. introduced a concept of geometric connectivity robustness which is basically a function to model and optimize local connectivity. Another class of local connectivity related 20 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari works lies in the context of leader-follower robot architectures where the fol- lower robots try to maintain the connectivity to a designated leader or vice verse. The work of Yao and Gupta [127] is relevant in this context. They em- ployed a leader follower control architecture for connectivity by adaptively classifying the nodes into backbone and non-backbone nodes. Gaustavi et al. [128] have followed similar path by identifying sufficient conditions for connectivity in a leader-follower architecture of mobile nodes. On the other hand, there exists a range of global connectivity related works that are pro- posed over last decade. One of the earlier global decentralized connectivity maintenance techniques is the super-gradient and orthonormalization based approach by Gennaro and Jadbabaie [129]. Later on, Dimarogonas and Jo- hansson [130] proposed a control strategy using ‘bounded’ inputs. Another very effective approach for connectivity maintenance based on decentralized estimation is presented in [33]. There are many extensions to this framework such as the integration of additional (bounded) control terms [131] and the saturation of the connectivity control term itself [132]. Zavlanos et al. also presented a couple of important techniques on the distributed global con- nectivity control [133–135] along with a compact survey on graph theoretic approaches for connectivity maintenance [136]. In [137], a technique based on dynamics of consensus methods is presented. Schuresko et al. [138] also presented techniques for connectivity control based on information dissemi- nation algorithm, game theory, and the concept of spanning tree. A multi-hop information based global connectivity maintenance and swarming technique is introduced by Manfredi [139]. Gil, Feldman, and Rus [140] proposed a well-known k-center problem based connectivity maintenance algorithm for an application context where a group of robotic routers provides routing sup- port to a set of robotic clients. The concept of bounded velocity of the routers and the clients is employed in this work. In [141], the connectivity mainte- nance problem in multi-robot systems with unicycle kinematics is addressed. In particular, by exploiting techniques from non-smooth analysis, a global connectivity maintenance technique under non-holonomic kinematics is pro- posed, which only requires intermittent estimation of algebraic connectivity, and accommodates discontinuous spatial interactions among robots. Most of these global connectivity maintenance methods are built upon the concepts of graph theory and algebraic connectivity. Realistic Approach for RWSN: While all the previous mentioned methods for connectivity are relevant to the field of RWSN, most of them lack a communication channel model that includes the effect of fading and shadowing observed in a standard wireless channels. Rather, most of these methods employ the simple unit disk model for wireless communication links to model the network graph where every pair of nodes are assumed connected if and only if they are located within a communication radius, say R , of each other and disconnected otherwise. However, in reality, the communication links are very dynamic and unpredictable due to effects like fading and shad- owing [5]. Therefore, the unit disk model based connectivity maintenance Robotic Wireless Sensor Networks 21 are rather impractical and should be modified. Moreover, the connectivity maintenance algorithms should use an optimization function that takes into account communication link features such as signal strengths, data rates, realistic communication models, and line of sight maintenance to define con- nectedness. As an example, Mostofi [142] presented a realistic communication model based decentralized motion planning technique for connectivity main- tenance. A behavioral approach for connectivity that takes into account the locations, measured signal strength and a map based prediction of signal strengths is proposed by Powers and Balch [143]. Anderson et al. presented a line of sight connectivity maintenance technique via a network of relays and clusters of nodes in [144]. In [145], a spring-damper model based con- nectivity maintenance is described. In summary, there exists only a handful of connectivity protocols that incorporate the well-known characteristics of a RF channel such Fading and even fewer are practically implemented and evaluated. What would be the potential future research directions? One ob- vious future direction would be to extend the theory of traditional unit disk model based connectivity maintenance protocol to include the effects of fad- ing and shadowing. A modular or hierarchical approach would be ideal in this context where a graph identification module (by including fading and shad- owing effects) and a graph theory based connectivity maintenance module will work independently but with synergy. Another future direction would be to develop more protocols of second kind and evaluate them extensively to en- rich the literature. Lastly, but most importantly, there is a lack of real-world experiments with a physical RWSN testbed to validate most of the existing theoretical and algorithmic contributions. Thus, a future goal of connectiv- ity maintenance related research should be on the development of a cheap, scalable, and easily programmable physical system and demonstration of the feasibility of the well-known solutions. 3.4 Communication Aware Robot Positioning and Movement Control In this section, we present a summary of the state-of- the-art techniques on communication aware positioning and placement control of a group of robots in order to fulfill data routing and sensing requirements. 3.4.1 Multi-Robot Sensing One of the main focus of RWSN should be on multiple robots based sensing with realistic wireless communication constraints. Note that, every function/sub- problem in an RWSN, such as localization and movement control, involves 22 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Table 4: Summary of Connectivity Maintenance Related Works Importance • Exchange of Information • Proof of Convergence • Distributed Coordination and Formation Control • So on Versions • Local [123, 126, 127] • Global [130, 137, 138, 140, 141] Traditional Approach • Algebraic Connectivity: •• Second smallest Eigenvalue, λ 2 ( L ), of Graph Laplacian L . •• Represents Connectivity if L is positive definite. •• Maximize λ 2 ( L ) to improve connectivity • Powered Sum of Adjacency Matrix: •• A sum = ∑ K i =0 A i where A is the adjacency matrix •• A sum represents the number of paths up to length K between any pair of nodes in the graph • Issues: •• Relies on simple unit disk model for interactions. •• Lack realistic communication channel model; effects of fading and shad- owing [5] are ignored. •• No Focus on the communication link qualities. Realistic Approach • Features •• Accounts for location, signal strengths, interference, and data rates [142, 143] •• Realistic communication channel models [142]. •• Line of sight maintenance between neighbors [144]. • Examples •• [142–146] Potential Future Directions • Modify existing unit disk model based graph methods of connectivity • Develop more efficient algorithm for connectivity with the focus on the realistic link qualities • Develop hardware prototypes and test out the algorithm in real world scenarios some sort of sensing such as RSSI, SINR, or locations of nodes. However, in this section, we focus on the state-of-the-arts on multi-robot systems where the main purpose of deployment is to sense an environment. There exist many works in the field of sensor networks and distributed robotics that deal with distributed sensing and sensed data collection. However, most of these works have some idealistic assumptions about either the communication model or the robot control problem and, thus, not directly applicable in RWSN. In this section, we only focus on the existing works on multiple robot based sensing that involves realistic models for both communication and control. What is already out there? In the field of multi-agent sensing, dis- tributed coverage of the area of interest is a very well known topic of re- search in the contexts of both WSN [147] and multi-robot systems [148]. In the contexts of WSN, coverage control algorithms focus on the placements of static sensor nodes to optimally cover the area of interest. We do not present a survey of this well studied problem. An interested reader is referred to a Robotic Wireless Sensor Networks 23 survey such as [147]. However, most of these works do not use the controlled mobility of the robots to the advantage. On the other hand, there exists a class of coverage control related articles in the field of coordinated robotics that focus on the control and path planning of the robots. Most of these works employ graph theoretic tools such as Voronoi partitions to solve the cover- age problem [149–151]. However, these works do not address the problem of collecting and communicating the sensed data effectively. Only recently, a small group of researchers started to look into multi-robot sensing problem from both control and communication point of views. The work of Kantaros and Zavlanos is relevant in this context [152]. They looked into the cover- age problem of multiple robotic wireless sensors placement by formulating an optimization problem that combines placement optimization with realistic communication constraints and sensing efficiencies of the robotic nodes. In [153], Yan and Mostofi also looked into the problem of robotic path plan- ning and optimal communication strategies in the context of a single robot assisted sensing and data collections. Similar combined path planning and communication optimization in the contexts of multiple robot based system is presented in the works of Ghaffarkhah and Mostofi [154, 155]. The works of Mostofi et al. [156–158] on cooperative sensing and structure mapping are also related to this context. In these works, the authors leveraged multiple pairs of coordinated robots and their RF communication abilities to sense/map unknown structures. To this end, they employed the concepts of compress- ible sampling/sensing [42] and the well known propagation properties of RF signals such as path loss and fading [5]. The work by Le Ny, Ribeiro, and Pappas [159] also presents an optimization problem that couples motion plan- ning and communication objective for sensing. On similar note, Williams and Sukhatme proposed a multiple robot based plume detection method in [160]. In [161], a new formulation of the multi-robot coverage problem is proposed. The novelty of this work is the introduction of a sensor network, which coop- erates with the team of robots in order to provide coordination. The sensor network, taking advantage of its distributed nature, is responsible for both the construction of the path and for guiding the robots. The coverage of the environment is achieved by guaranteeing the reachability of the sensor nodes by the robots. What would be the potential future research directions? To our knowledge, there exist a very few works on robot assisted sensing that in- volved timely, reliable, and efficient delivery of the sensed data. A major focus of the future should be on such joint optimization of data collection and sensing tasks. Moreover, there might exist many dissimilar robots (each consisting of different sets of sensors) in an RWSN. This requires a simple, unified abstraction in terms of control as well as the sensed data collection. The popular publish-subscribe based frameworks (such as MQTT [162]) may be used in such contexts. Moreover, the introduction of controlled mobility has opened up the applications of sparse sensing [158] which can be exploited for energy efficient, non-redundant sensing. 24 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Table 5: Summary of Multi-Robot Sensing Existing Works • Multi-agent sensing and coverage algorithms from WSN [147, 148] • Distributed coverage control in distributed robotics [149–151] • Combined optimization of sensing coverage placement and efficient data routing and collection [152–161, 163] Potential Future Directions • Sensor data collection abstraction (e.g., publish-subscribe model) for multiple dissimilar robotic systems • Sparse sensing for energy efficient non-redundant sensing • More algorithms on combined optimization of robotic path planning, sensing, and communication quality 3.4.2 Robotic Router Positioning Robotic router/relay placement is a cutting edge topic of research in the field of RWSN. It mostly concerns the second trend in RWSN research i.e., the use of a robotic network to form a temporary communication backbone or sup- port an existing backbone to improve performance. The problem of robotic router placements is complex and involves direct relations with many other research pieces of RWSN such as connectivity maintenance, communication link modeling, and localization. A robotic router/relay is a device with wire- less communication capabilities and controlled mobility. Such devices can be employed to form temporary/adaptable communication paths and to ensure robust information flow between a set of nodes that wish to communicate but lack direct links between each other. Note that, we use ‘robotic router/relay’ to refer to the robotic nodes helping in setting up communication and ‘com- munication endpoints’ to refer to the nodes willing to communicate. Robotic relay/router nodes relay messages between such communication endpoints. The communication endpoints might have certain communication require- ments such as min achievable data rate, high throughput, and lower delay. Moreover, the communication endpoints can be mobile or the environment can be dynamic with changing communication link properties. The objective of a robotic router placement/positioning algorithm is to place the relays in an optimal manner such that the communication requirements are fulfilled throughout the deployment time, and to adapt with the network dynamics. Before moving on, we present a commonly used term in such contexts called ‘flow’ . A ‘flow’ is defined as the communication path between a pair of com- munication endpoints via a set of robotic routers/relays. In other words, a set of robotic routers assigned to a flow are dedicated to form and support the communication path between the respective pair of communication end- points. Based on the objectives as well as network dynamics, the allocation of a set of robotic routers among different flows may also change over time. Some of the major components of robotic router placements algorithms are: proper positioning and movement planning of these robotic router, allocation and reallocation of robots among flows as they arrive or disappear, and connec- Robotic Wireless Sensor Networks 25 tivity maintenance. In this section, we present the state-of-the-art techniques on robotic router placements. What is already out there? The earlier relevant state-of-the-arts on robotic router placement/movement algorithms are linked to the connectiv- ity maintenance problem. A major focus of such methods was to keep a moving target/node connected to a static base station via a set of robots with the assumption of an initially connected network. The work of Stump, Jadbabaie, and Kumar [164] is mentionable in this context where either the transmitter node is fixed and the receiver node is moving or vice-versa. They developed a framework to control a team of robots to maintain connectivity between a sender and a receiver in such cases. Among other state-of-the arts, Tekdas, Yang, and Isler [165] focused on the connectivity of a single user to the base station and proposed two different models based motion planning algorithms. One is based on known user motion (user trajectory algorithm) and the other is for unknown-random, adversarial motion of the user (ad- versarial user trajectory algorithm). However, this class of works do not deal with the qualities of the communication links as well as the end-to-end per- formance. De Hoog, Cameron, and Visser [166–168] have also proposed some techniques for maintaining connectivity to a command center in the context of exploration of unknown environments. In their tree like role based network formulation, the leaf nodes are the ‘explorers’ that explores new frontiers, the root of the tree is the ‘base-station’, and the rest of the nodes are ‘relays’ to keep connectivity between the ‘explorers’ and the ‘base-station’. Over last couple years, a handful of researchers have started exploring the problem with more realistic communication models. Yan and Mostofi [169, 170] are among these handful of researchers to work on the robotic router problem. They extended the concept of connectivity maintenance to formu- late an optimization problem which considers true reception quality expressed in terms of bit-error rate. The goal was to minimize bit-error rates of the re- ceivers for two scenarios of multi-hop and diversity. They also demonstrated that the Fiedler eigenvalue optimization based approach results in a perfor- mance loss. They used an extension of the channel modeling technique in- troduced in [45, 46]. In [171], Dixon and Frew presented a gradient based mobility control algorithm for a team of relay robots with the goal of forma- tion and maintenance of an optimal cascaded communication chain between endpoints. Rather than considering the relative positions of the neighbors, they used SNR of communication links between neighbors. They presented an adaptive extremum seeking (ES) algorithm which is employed for op- erating a distributed controller. Goldenberg et al. [172] presented another distributed, self-adaptive technique for mobility control with the goal of im- proving communication performance of information flows. In their work, they tried to design and analyze a simple system to address three issues: appli- cation dependency, distributed nature, and self-organization. A solution to the problem of computing motion strategies for robotic routers in a simply- 26 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari connected polygon environment is presented in [173]. For a summary of all the relevant references, a reader is referred to Table 6. As mentioned earlier, the main goals of robotic router placements are to fulfill certain communication requirements such as supporting a set of flows [174], guaranteeing certain performance criterion (say, data rate) to the customers [175], or fixing holes [176]. In [174], Williams, Gasparri, and Krishnamachari presented a hybrid architecture called INSPIRE, with two separate planes called Physical Control Plane (PCP) and Information Control Plane (ICP). Their goal was to support a flow based network between multiple pairs of senders and receivers using a group of robots and optimize the overall packet reception rate (PRR) (or the expected number of transmissions, ETX). In [113], Wang, Gasparri, and Krishnamachari presented a method called robotic message ferrying, where a set of robots literally travel from a source to a sink/destination to deliver data. The main objective of this work was the allocation of such robotic router nodes among a set of senders and the optimization of the communication performance such as throughput. Tuna, Nefzi, and Conte [176] also proposed a centralized method of fixing routing holes (due to absence of nodes or failure of nodes) using a group of robotic routers. In the proposed method, all nodes communicate to a central server to send the sensed data, which in turn controls the positioning and deployment of a set of UAVs to fix routing holes. This algorithm employs geographical routing and Bellman Ford routing algorithms to find the missing nodes/links. In [53], Fink, Ribeiro, and Kumar focused on guaranteeing a minimum end- to-end rate in a robotic wireless network. They model the communication channels via a Gaussian process model learned with a set of initially collected data. They proposed a stochastic routing variable to calculate an end-to-end rate estimate, which is exploited to find a slack in the achieved rate and the required rate. This estimated slack is further used for proper mobility control. Fida, Iqbal, and Ngo [177] proposed a new metric, called reception probability, and a throughput based route optimization method that employs the new metric. They used Particle Swarm Optimization (PSO) [178] for finding the optimal configuration due to non-convexity of the problem. Gil et al. [175], also proposed a method of robotic router placements where the communication demands (in terms of data rates) of a set of clients are fulfilled by another set of robotic routers. The demands are modeled in terms of effective SNR (ESNR) to represent the required rate. Each client is serviced by the router closest to it while the router to router communications are assumed to have a very high capacity. They used a synthetic aperture radar (SAR) [179] concept based directional signal strength (both amplitude and phase) estimation method and a Mahalanabis distance based cost function for the positioning and path planning of the routers. Some preliminary works on optimizing SINR of the links, i.e, minimizing the effect of interference is presented in [180]. The work of Wang, Krishnamachari, and Ayanian [54] on robotic router placements in cluttered environments, is also related to this context. On a related note, Ghosh and Krishnamachari [181] showed that Robotic Wireless Sensor Networks 27 there exists a bound on the number of robotic routers we need to deploy to guarantee certain communication requirements in terms of SINR. They also proposed a method of estimating worst case interference and SINR in a flow based robotic router deployment context. What would be the potential future research directions? To our understanding, there exist a lot of research opportunities in this field of re- search. First, there is a lack of a physical robotic system based experimenta- tion of the existing works. One potential direction is to implement some of the promising algorithms and concepts on a real system and perform thor- ough analysis. To this end, there is a lack of academic open-source robotic network testbeds. Thus building a generic, scalable, adaptable robotic net- work testbed is another potential direction of research. Furthermore, most of the solutions are centralized and need to be converted to decentralized meth- ods. Interference among the robotic routers as well as the effect of CSMA or any other channel access are also unaccounted for in most of the existing works. Nonetheless, to our understanding, the main focus of future should be on developing scalable, adaptable physical systems to test out the developed algorithms. Table 6: Summary of Robotic Router Placements Related Works Existing Works • Tethering a moving object to a base station [164, 165] • Robot assisted static relay placements [182, 183] • Single communication chain formation with performance guaran- tee in terms of bit-error rate, number of hops, end-to-end-rate, or SINR [53, 169–173, 177, 184–186] • Multiple flow based robotic network [174, 180] or mesh robotic net- work placement and performance optimization [175, 176] • Use controlled mobility to carry the message from a source to a destination [113] • Integrated framework for network goal oriented mobility control [54, 174, 184] Potential Future Directions • Hardware implementation and evaluation of the existing algorithms • Include the effects of interference and channel access protocols in the placement optimization problem • More decentralized method developments 3.5 Localization In this section, we present a survey on the state-of-the-arts on localization techniques in the context of RWSNs. The problem of localization is very well-known in the contexts of sensor networks and distributed robotics. The state-of-the-arts on localization are very mature and require a complete sep- arate survey [34–37]. In this section, we provide a very brief overview of the 28 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari existing localization works and point out the works most relevant to the field of RWSN. Note that, the concept of ‘localization’ is to locate a node in a deployment arena with respect to a reference frame or a reference location. A commonly used system called the Global Positioning System (GPS) lo- calizes objects in terms of their latitudes and longitudes. However, GPS is known to not work properly in cluttered or indoor environments. Therefore, most of the target application contexts of RWSN require an alternate and efficient localization scheme for indoor environments such as RF based lo- calization. Moreover, while absolute locations are much important, a relative localization between the nodes in the network is sufficient in many contexts of RWSN. For example, consider a scenario where a group of robotic routers are employed to connect a moving target with a base station. In such contexts, the robots form a chain where each robot positions itself with respect to its neighboring nodes only. Relative positions with respect to the neighboring nodes are enough for a node’s movement control decisions in this context. In this section, we present a summary of the existing literature on the relevant localization techniques. Vision and Range Based System: As mentioned earlier, localization has been a very active field of research in the domain of distributed robotics. The most popular localization systems in the field of robotics employ cameras and range finders. With the help of efficient sampling and filtering algorithms such as Particle Filtering or Kalman Filtering, the camera based systems locate the object in its field of view while range finders provide depth/distance information [187–190] . In order to deal with the movements as well as errors, some researchers use temporal snapshot of the targets [191, 192]. However, any camera/vision based approach has many limitations such as the visibility requirement, limited field of vision of traditional cameras, larger form factor of the robots, and costly image processing software requirements. On the other hand, while the laser range based methods do not suffer from the visibility problem, they are limited to direct line of sight between the target and the tracker and require complicated processing. RF Based System: As an alternative to GPS and vision based local- ization systems, wireless sensor network researchers have proposed a variety of Radio Frequency (RF) based localization systems. As mentioned earlier, there exist a range of survey papers with the sole focus on such RF localiza- tion techniques [34, 35]. Briefly speaking, the existing localization techniques employ either of the following aspects: Direction of Arrival (DoA), Time of Arrival (ToA), Time Difference of Arrival (TDoA), Received Signal Strength (RSS), and proximity. The typical underlying technologies used to realize these techniques are RFID, WLAN, Bluetooth, and ZigBee. Liu et al. [34] provided a great outline and classifications of general wireless localization techniques and systems. They outline most of the performance metrics used in traditional RF based localization as follows: accuracy (mean error), pre- cision (variance, or distribution of accuracy), complexity, robustness, scala- bility, and cost. With the recent trends of networked robots, it is of interest Robotic Wireless Sensor Networks 29 to the research community to look at these performance metrics from the perspectives of RWSN. The difference between an RWSN and a WSN with static and mobile actuating sensor nodes is that robots are more dynamic in position and require greater performance and flexibility in localization due to the larger range of tasks the robots are expected to perform. RFID Based System: Over last two decades, many researchers have peered into the use of RFID tags because of their low cost and power (or zero power for passive tags) . A confined deployment arena for a team of robots can be fitted with a mass deployment of RFID tags. Then, we can localize any RFID/RF device carrying robots in that arena [193, 194]. Zhou et al. present a survey of existing research and deployments of RFID tags for localization in [195] which shows its usefulness in robotics. However, one large quirk of RFID tags is the static nature of the tag placements and limited tag functions and information. The benefits of RFID tags manifest when meticulous planning or post-deployment positioning (using a robot) is done. In [196], a multi-robot exploration of an unknown graph describing a set of rooms connected by opaque passages is considered. In particular, the authors demonstrate how in this framework, which is appropriate for scenarios like indoor navigation or cave exploration, communication can be realized by bookkeeping devices, such as RFID tags, being dropped by the robots at explored vertices, the states of which are read and changed by further visiting robots. As an alternative, RFID tags can be replaced or complemented by WSNs, which have greater capabilities and flexibility. Wireless AP Based System: With the ubiquity of WLAN access points and wide availability of wireless sensor nodes, the research community has in- vestigated the use of a network infrastructure to position and navigate robots. The work of Ladd et al. [197] illustrated the feasibility of using commercial off-the-shelf radios and radio signal strength measurements as a robust lo- cater of robots. To enhance the overall performance of RSS based localization, researchers have investigated RF scene analysis or fingerprinting as a viable option for indoors. Ocana et al. [198] presented such a robot localization system that starts with a semi-autonomous method to fingerprint indoor en- vironments using a robot. Many other research teams followed up with works concentrated on RF surveying with directional antennas on a robot, such as in [199–203]. We encourage readers to also refer to our section on RF mapping (Section 3.1) to complement the RF scene analysis. Distributed and Cooperative System: Distributed and cooperative network and RF based localization in a dynamically moving network of robots is still not a fully understood area. Wymeersch [204] showed the positive im- pacts of cooperative localization on achieving more complex tasks with a team of robots. This further motivates the need for a better understanding of cooperative network and RF based localization to see if we can enhance the functionality of RWSNs. In [205], Koutsonikolas et al. delved into the prob- lem of cooperative localization, but the authors assume there are a subset of robots that carry extra sensors, including GPS, to aid in the overall system. 30 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari The rest of the robots carry 802.11 radios, only using beacons to determine proximity. On that node, the work of Zickler and Veloso [206]focus on rel- ative localization and tethering between two robots based on the received signal strengths. They opt for a discrete grid based Bayesian probabilistic approach. In their system, a locator node moves to different relative posi- tions with respect to the node being localized to collect multiple RSS values while they communicate their odometer readings. In [207], Filoramo et al. describe an RSSI-based technique for inter-distance computation in multi- robot systems. In particular, for a team of robots equipped with Zigbee radio transceivers, they propose a data acquisition technique which relies on spatial and frequency averaging to reduce the effect of multi-path for both indoor and outdoor environments. Furthermore, they show how the proposed data acquisition technique can be used to improve a Kalman Filter-based localiza- tion approach. Another RSSI based relative localization system is proposed by Oliveira et al. [208] which also relies on pairwise RSSI measurements between the robots. To improve the performance and accuracy of the local- ization, they apply Kalman filter and the FloydWarshall algorithm. One of the most recent significant work on relative localization is presented in [209] which applies MIMO based system for a single node based localization. These methods are particularly relevant to the field of RWSN. What would be the potential future research directions? Most of the current robotics network related research are performed in artificial fixed indoor environments as there exist costly camera based solutions (such as VICON system [210]) to provide millimeter level accuracy required for the experimentation. We believe that RF localization will be able to help extend such experiments to truly indoor cluttered environments with much lower cost than deploying camera systems. However, most of the existing RF solu- tions are also not portable e.g., they require either a fixed infrastructure with RF beacons, a map of the environment, etc. Therefore, the future direction of localization relation research should be focused on developing portable scalable at-least centimeter level accurate RF localization systems that can be quickly deployed on demand and can be remove easily. Moreover, the frameworks should be portable to deploy in real-world applications such as in firefighting. In that context, another potential direction is towards ‘self sufficient’ RWSN i.e., the network will not require help from any existing in- frastructure to perform the assigned tasks. Further, researchers should study different localization properties such as accuracy, complexity, and communi- cation requirements in the context of RWSN where a robot’s performance directly relies on localization e.g., in the context of tracking and following a firefighter while providing connectivity to a command center. Robotic Wireless Sensor Networks 31 Table 7: Summary of Localization Related Works Localization Algorithm Type/Class References Comments Vision and Range Based System [187–192] • Popular in robotics • High accuracy • Many off-the-shelf solutions are available • Requires heavy computation • Requires line of sight • Large form factor RF Based System [34, 35] • Popular in WSN community • Uses RF signal properties (like signal strength or time of arrival) • Lower accuracy compared to camera based system • Low cost and low computation • No line of sight requirement • Small form factor RFID Based System [193–196] • Low cost and power (or zero power for pas- sive tags) • Requires static placements of the tags in the deployment arena • Requires pre-deployment of the localization infrastructure Wireless AP Based System [197–203] • Use pre-deployed or existing WLAN access points • Use commercial off-the-shelf radios • Requires pre-mapping or fingerprinting of the deployment region Distributed and Cooperative System [204–209] • A subset of anchor robots have GPS or other localization capabilities • Rest of the robots localize themselves relative to the anchor nodes • Sometimes odometer readings are combined for better accuracy [206] • Some types of filtering can be involved to improve performance Future Directions • Co-optimizations of localization accuracy and communication goal • Develop cheap, scalable, high accuracy system • Study the trade-offs among accuracy, complexity, and cost 4 RWSN Network Stack Layer Analysis: In the last section, we categorized the exiting state-of-the-arts in the field of RWSN according to the key research problems in focus. The majority of the works in RWSN are focused on RSS modeling and mapping , connectiv- ity maintenance , routing algorithms , communication aware robot placements , connectivity maintenance , and localization . However, one key thing missing in the last section is how all these works fit in the contexts of traditional networking concepts. Traditionally, the protocols relevant to a network (say 32 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari sensor network) are developed by following the well-known layered network stack architectures such as OSI model or Internet model. Ideally, we would want the same for RWSN. However, we discover that the layered structure in an RWSN device does not follow the traditional norm. Rather, it mostly relies on inter-dependencies between layers. To understand the layering re- quirements, we first analyze the existing works explained in the last section according to the five layered Internet protocol stack: Physical layer, MAC Layer, Network layer, and Transport layer, Application layer. Next, we dis- cuss about some potential unified architectures for RWSN. 4.1 Internet Model for Network 4.1.1 Physical Layer The physical layer in a traditional network deals with the physical com- munication between nodes. The function of physical layer includes but not limited to representation of bits, controlling data rate, synchroniza- tion between transmitter and receiver, defining the communication inter- face, and controlling the mode of communication such as simplex or du- plex. Therefore, research on topics such as communication hardware, phys- ical medium and communication technologies (e.g., RF and Bluetooth), and modulation-demodulation of signal falls under this category. There ex- ist many sorts of wireless communication technologies such as RF, Blue- tooth, and Sonar for communication among robots. For obvious reasons, the dominant technologies for above-ground communication are RF communica- tion techniques [211–214]. For short range communication, some networks of robots use bluetooth [215] and infrared. The above-ground radio frequency based methods are not applicable for acoustic communication (e.g., underwa- ter communication) due to reasons like high loss exponents and fading due to turbulence. The most common acoustic communication techniques are spe- cial RF communication [216] and sonar. While RF based communication still remains the mainstream for RWSN system, the introduction of robots and controlled mobility have also opened up some new but relatively unexplored communication methods. On that node, Ghosh et al. [217] have presented a proof-of-concept of a new method of communication that employs the loca- tion and the motion pattern of a communicating robot as the communication signal. Most of the works on RSS measurements, RF mapping, and position con- trol to improve link qualities fall under the purview of the physical layer since these works are directly linked to the signal strengths variations over the area of interest. On a similar note, most of the RF based localization techniques use properties (such as signal strength) of the communication signals (RF or ultrasound) and, thus, also fall under the purview of the physical layer. Robotic Wireless Sensor Networks 33 Among the emerging research domains related to physical layer of an RWSN, the concept of distributed MIMO implementation using robots is promising. In [218], Zhang et al. presented a cooperative MIMO like communication structure in mobile sensor/robotic networks. They subdivide a network into twin sub-networks where each transmitter node pairs itself with another node for transmitting cooperatively in a MIMO like fashion. Another class of work that also partly falls under the purview of physical layer lies within the communication aware robot positioning and movement control related works where the robots adapt their positions to optimize the quality of the communication links. In such cases, the robots employ the physical layer information such as RSSI [180] or data rate [175] to control the movements of the robots acting as relays/routers/sensors. In summary, we can state that a major focus of most of the existing state- of-the-arts on RWSN has been towards the physical layer. 4.1.2 Media Access Control Layer Briefly speaking, Media Access Control (MAC) protocols deal with proper distributed access of the physical medium among nodes, node to node com- munication, framing, and error corrections. While most of the classical MAC layer modules and protocols [5] are applicable to RWSN, to our knowledge, there exist only a few MAC layer protocols designed specifically for net- working between robots. Related to media access protocols, CSMA/CA is an obvious choice for RWSN due to its ubiquitous properties such as random- ness and scalability. On the other hands, TDMA and FDMA systems can also be modified for RWSN with the extra feature of controlled mobility. On that note, Hollinger et al. [219] presented a MAC protocol for robotic sensor networks in acoustic environments. They proposed a three-phased method based on TDMA with acknowledgments. The phases are Initiation, Schedul- ing, and Data transfer, respectively. There also exist works related to mobile WSN which use predicted mobility patterns, such as pedestrian mobility or vehicular mobility, of both source and sink to design efficient application con- text specific MAC protocols. Some examples of such MAC protocols are: MS- MAC [220], M-MAC [221], M-TDMA [222], MA-MAC [223], MobiSense [224], and MCMAC [225]. These works show that many difficulties arises as a re- sult of mobility (mainly uncontrolled mobility) such as random variations in link quality and frequent route changes. To deal with these problems of mobility, researchers proposed a class of methods like negotiation based rate- adaptation and handover by transmitter. For more detailed survey on such techniques, the reader is referred to [226]. In contrast, the mobility of the nodes in RWSN are controlled and, thus, can be exactly known or predicted with higher accuracy. This opens up a new domain of research where the mobility controller and the MAC protocol could work in an integral manner to optimize the utilization of the radio resources. 34 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari In summary, while the existing MAC protocols are applicable to RWSN, there is a lack of MAC layer protocols specifically designed and optimized for RWSN. As mentioned earlier, one of the future directions would be to incorporate mobility control with MAC access protocols for better perfor- mance. Another future direction would be to use the knowledge of the MAC protocol to estimate signal properties such as Interference and SINR [181], and to control the link properties such as interference [227]. 4.1.3 Network Layer Network Layer in a traditional wireless network deals with the packetization of data as well as routing of the packets from source nodes to respective des- tination nodes. One of the main application contexts of RWSN is to maintain a temporary communication backbone to support data flow between commu- nication endpoints. Thus, a major focus of RWSN related works till date has been on developing network layer protocols and combining controlled mobility with routing of packets. All the routing related works presented in Section 3.2 directly fall under this category for obvious reasons. Similarly, all the works on communication aware robot placements (discussed in section 3.4) also fall under the purview of network layer. However, as mentioned earlier, the concept of layering in RWSN is slightly vague as it relies on cross-layer de- pendencies such as dependencies between the physical layer and the network layer. This is mostly apparent in the contexts of robotic router placement algorithms that deal with the placements of the robotic nodes (based on the physical layer information) to optimize the end-to-end path from a source to destination as well as to optimize each link [81, 174]. All connectivity related works presented in Section 3.3 also fall under this category as the key goal in such connectivity maintenance algorithms is to guarantee the existence of a communication path between every pairs of nodes in the network. Without connectivity, the network might be segregated into smaller sub-networks and won’t be able to fulfill the routing goals. 4.1.4 Transport Layer In the field of RWSN, the researchers are yet to significantly focus on the transport layer protocols. Till date, researchers employed traditional trans- port layer protocols such as TCP, UDP, or some MANET transport layer pro- tocols for robotic networks. However, unlike MANET, robotic networks have an extra feature of controllable mobility which provides an extra dimension. But, there is no significant work on controlling mobility for improved perfor- mance of transport layer protocols. Conversely, controllability requires highly reliable, low delay, and error free communication between robots, which is not possible using original TCP or UDP and requires some special transport layer Robotic Wireless Sensor Networks 35 protocol. In this section, we present a brief overview of the existing transport layer protocols for RWSN. Among the state-of-the-arts, the work of Douglas W. Gage on the MSSMP Transport layer protocol (MTP) [211] is mentionable. This protocol is based on the Reliable Data Protocol (RDP). The RDP is much more effective and appropriate service model for mobile robot applications than TCP. There are many features of RDP which are more useful in RWSN such as more com- munications bandwidth than TCP and simpler in terms of implementation. But the interface to the application layer in MTP is similar to TCP, i.e., based on socket like API. There is also couple of congestion control protocol designed for tele-operation of robots such as Trinomial method[228], Real- Time Network Protocol (RTNP) [229], and Interactive Real-Time Protocol (IRTP) [230]. All these methods are not directly related to the RWSN but can be ported. Similarly, there exists a group of works on Transport layer for MANET [231, 232]. For detailed overview on existing congestion control pro- tocols of mobile ad-hoc network, an interested reader is referred to [233]. In summary, the transport layer related research on RWSN requires significant attention in future with a major focus on reliability and delay performances. 4.1.5 Application Layer Many of the existing works related to RWSN are actually related to the ap- plication layer. In some of the target application contexts, the robots in an RWSN need to make informed movement and communication decisions in order to work in a cooperative manner. For example, in order to form a communication relay path between a pair of communication endpoints, the relay robots need to process a combination of information such as neighboring node locations, flow endpoint requirements (say, a minimum data rate), cur- rent link status, and expected interference power to adaptably and optimally position the relays. According to the layered hierarchy, all these processing and decision making should be done in the application layer to keep the sys- tem modular. For example, in the route swarm work [174], the Information Control Plane (ICP) takes care of making decision regarding state changes of the robots as well as the allocation of robots among different flows. Thus, the ICP in that architecture is mostly implemented in the application layer. In the same manner, all the robotic router related works are partly/fully dependent upon the application layer. Another mentionable field related to the application layer is cloud robotics. Cloud robotics basically uses an application layer abstraction of a heteroge- neous network of robots to perform a group of tasks. Cloud robotics provides a unified scalable control platform for a group of heterogeneous robots. The work of Du et al. [29] is among the most promising works in this field of re- search. Quintas et al. [30] also proposed a related architecture. In [31], Kamei et al. presented a detailed study of the advantages, concerns, and feasibility in 36 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari Cloud Networked Robotics. In this paper, we do not delve into cloud robotics as it does not directly fall under the RWSN research domain. Table 8: Summary of Relevant Keywords for the RWSN Layering Architecture Layer References Related Keywords Physical Layer [211–218] RSS Measurement, RF Mapping, Localization, Dis- tributed MIMO, Connectivity, Robotic Router Place- ment, Communication Aware Robot Positioning MAC Layer [5, 181, 219– 227] Scheduling, CSMA, TDMA, FDMA Network Layer Routing, Robotic Router Placement, Communication Aware Robot Positioning, Connectivity Transport Layer [211, 228–233] Delay, Real-time communication, UDP, TCP Application Later [29–31, 174] Connectivity, Positioning, Robotic Router Placement, Communication Aware Robot Positioning, Cloud Robotics In summary, there is no clean way of classifying the existing works into the layered architecture. Rather, each of the problems and solutions belong to multiple layers. Moreover, we need a new layer/module to deal with the mobility control. All these lead us to believe that maybe we need a new architecture for RWSN that builds upon the existing layered network stacks, discussed in the next section. Note that we present a summary of our network protocol stack related discussion in Table 8. 4.2 An Unified System Architecture For RWSN Based on our analysis in Section 4.1, we find that the existing works in RWSN do not fit well in the Internet model of networking stack. Rather, most of the solutions in RWSN require inter-layer dependencies. For example, a robotic router placement algorithm relies on the physical layer estimation models which in turn rely on some knowledge about the relay node positions and the network graphs. Moreover, we need to have a control layer to combine the network goals with the movement of the robots. Thus, we require a new sys- tem hierarchy for RWSN where the existing network architecture can be kept intact to the most extent. On that note, Williams, Gasparri, and Krishna- machari [174] have proposed an architecture with two planes: the Information Control Plane (ICP) and the Physical Control Place (PCP) where the ICP takes care of the networking as well as high level movement decisions and the PCP takes care of the movements. There also exist software architec- ture solutions for a system of multiple robots such as ALLIANCE[234] and Robotic Wireless Sensor Networks 37 CLARAty [235]. The works of Arkin and Balch[236] and Stoeter et al. [237] are also relevant in this context. However, the concept of a unified system architecture is one of the rela- tively unexplored domain of research in RWSN. Most of the existing literature emphasize only certain aspects of system challenges instead of focusing on the networked robot system as a whole. Thus, there is a need of a base, decen- tralized, realistic system framework using realistic communication models that can autonomously control an individual robot as well as a group of robots in any kind of RWSN application contexts. The term framework here refers to a collective set of system modules such as movement control, sensors and connectivity maintenance with necessary in- terconnections, as illustrated in Figure 4. This base framework should have plug & play flexibility as well, i.e., any extra module pertinent to a specific requirement can be added or removed. In Figure 4, we present a sample, illustrative architecture for RWSN, based on our understanding. Fig. 4: Illustration of a Unified Architecture 5 Collaborative Works on Networked Robots There are many projects on collaborative robotics with different goals in fo- cus. Among them, Ubiquitous Robotics network system for Urban Settings (URUS) project (http://urus.upc.es) [238], Japan’s NRS project [239], Phys- ically Embedded Intelligent Systems Ecology [240] project, DARPA LAN- droids program [241], Mobile Autonomous Robot Systems (MARS) [242], Mo- bile Detection Assessment and Response System (MDARS) [211] are the im- portant ones. Among other projects, the swarm-bot project by cole Polytech- 38 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari nique Fdrale de Lausanne (EPFL) ([243], [244]), the NECTAR Project [245] by Filippo Arrichiello and Andrea Gasparri, and I-Swarm project ([246], [247]) by Karlsruhe are the significant ones. Since we are mainly interested in network related research in RWSN, the main project that falls in our category is the the DARPA LANdroids program. This is one of the recent projects undertaken on RWSN. Tactical communi- cation enhancement in urban environments is the main goal of this program [241]. Towards this goal, the researchers tried to develop pocket-sized intel- ligent autonomous robotic radio relay nodes that are inexpensive. One of the serious communications problems in urban settings is multipath effect. LANdroids are envisioned to mitigate the problem by acting as relay nodes, using autonomous movements and intelligent control algorithms. LANdroids will also be used to maintain network connectivity between dismounted war- fighters and higher command by taking advantage of their co-operative move- ments. On the other hand, there are some industrial projects that also fall under the purview of RWSN such as Facebook’s Aquila project [248] and Google’s project Loon [249]. There are also some open source projects on swarm robotics such as Swarmrobot [250] and Swarm-bots [251]. Swarming Micro Air Vehicle Network (SMAVNET) is a related project by EPFL where a swarm of UAVs are envisioned to be used to create temporary communica- tion networks. 6 Summary and Conclusion The main aim of this chapter was to identify and define a new field of re- search, RWSN, and provide a starting point to the new researchers. Briefly speaking, a RWSN consists of a group of controllable robots with wireless capabilities that is deployed with the goal of improving/providing a portable wireless network infrastructure in application with need of sudden and tem- porary wireless connectivity such as in a search and rescue mission or in a carnival. While there exist a range of relevant state-of-the-arts, the applica- tion of controlled mobility to the advantage of wireless communication is still an open area of research. However, like every new field of research there are some challenges in RWSN research. Some of the known challenges are: (1) lack of programmable, scalable RWSN testbeds for implementing and vali- dating concepts, (2) lack of good venues to publish research (there is only a handful of new workshops and conferences that focus on RWSN), (3) because of the inter-disciplinary nature of this field it requires the researchers to have knowledge on a vast range of topics such as robotics, control, communication, embedded systems, and networks. Nonetheless, based on all the discussions in this chapter, it is evident that RWSN is an emerging and promising piece of Robotic Wireless Sensor Networks 39 technology with limitless possibilities. Some of the known promising ongoing and future directions of RWSN related research can be listed as follows. Systems  Build a full fledged low-power reusable RWSN testbed  Implement and analyze promising theoretical concepts on a real system  Extensive measurements in real environments (such as mines), identify the RF properties, and formulate communi- cation models and emulators Modeling and Mapping  Build a mathematical or systemic model for Interference and SINR estimation in a RWSN  Incorporate the effects of MAC protocols such as CSMA into interference estimations  RF based online mapping of an unknown environment Routing  Develop routing algorithms with guaranteed lower delay but higher reliability  Apply existing MANET protocols in the context of RWSN  Include controllability of the nodes in the routing decision where a bad link can be potentially improved via small move- ments Connectivity maintenance  Realistic communication model based connectivity control Robotic Router  Optimization of router placements with realistic SINR mod- els  Guaranteed communication performance such as min achievable data rate by placing robotic routers between TX- RX pairs  Robot based communication link repair  Robotic message ferrying related research with more fo- cus on the tradeoff between movement energy consumption cost and the payoff from good performance or timely message delivery Localization  Build a portable RF based localization system with at least centimeter level accuracy  Focus more on relative localization than absolute localiza- tion Network Stack  Multiple robot based co-operative MIMO  MAC protocols with mobility control, engineered specifi- cally for RWSN  Transport layer protocols (alternate to UDP or TCP) en- gineered for RWSN  Unified system architecture for RWSN Table 9: Ongoing and Future Research Directions 40 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari References [1] Jacques Penders, Lyuba Alboul, Ulf Witkowski, Amir Naghsh, Joan Saez-Pons, Stefan Herbrechtsmeier, and Mohamed El-Habbal. A robot swarm assisting a human fire-fighter. Advanced Robotics , 25(1-2):93– 117, 2011. [2] Robin R. Murphy. Trial by fire [rescue robots]. IEEE Robotics & Automation Magazine , 11(3):50–61, 2004. [3] Veysel Gazi and Kevin M Passino. Swarm Stability and Optimization . Springer, 2011. [4] Erol Aahin and Alan Winfield. Special issue on swarm robotics. Swarm Intelligence , 2(2-4):69–72, 2008. [5] Theodore S Rappaport. Wireless communications: principles and prac- tice . prentice hall PTR New Jersey, 1996. [6] A Ollero, J Alc ́ azar, F Cuesta, F L ́ opez-Pichaco, and C Nogales. Heli- copter teleoperation for aerial monitoring in the comets multi-uav sys- tem. In Proceedings of the 3rd IARP Workshop on Service, Assistive and Personal Robots , 2003. [7] Sebastian Thrun, Scott Thayer, William Whittaker, Christopher Baker, Wolfram Burgard, David Ferguson, Dirk Hahnel, D Montemerlo, Aaron Morris, and Zachary Omohundro. Autonomous exploration and map- ping of abandoned mines. IEEE Robotics & Automation Magazine , 11(4):79–91, 2004. [8] Robin R Murphy, Jeffery Kravitz, Samuel L Stover, and Rahmat Shoureshi. Mobile robots in mine rescue and recovery. IEEE Robotics & Automation Magazine , 16(2):91–103, 2009. [9] Manoja D Weiss, James Peak, and Thomas Schwengler. A statistical radio range model for a robot manet in a subterranean mine. IEEE Transactions on Vehicular Technology , 57(5):2658–2666, 2008. [10] Jonathan Baber, Julian Kolodko, Tony Noel, Michael Parent, and Ljubo Vlacic. Cooperative autonomous driving: intelligent vehicles sharing city roads. IEEE Robotics & Automation Magazine , 12(1):44– 49, 2005. [11] Robert Nagel, Stephan Eichler, and Jorg Eberspacher. Intelligent wire- less communication for future autonomous and cognitive automobiles. In Proceedings of the IEEE Intelligent Vehicles Symposium (IV) , 2007. [12] Vicente Milan ́ es, Javier Alonso, Laurent Bouraoui, and Jeroen Ploeg. Cooperative maneuvering in close environments among cybercars and dual-mode cars. IEEE Transactions on Intelligent Transportation Sys- tems , 12(1):15–24, 2011. [13] Naixue Xiong, Athanasios V Vasilakos, Laurence T Yang, Witold Pedrycz, Yan Zhang, and Yingshu Li. A resilient and scalable flock- ing scheme in autonomous vehicular networks. Mobile networks and applications , 15(1):126–136, 2010. Robotic Wireless Sensor Networks 41 [14] Lynne E Parker. Alliance: An architecture for fault tolerant multi- robot cooperation. I EEE Transactions onRobotics and Automation , 14(2):220–240, 1998. [15] Peter Ibach, Nikola Milanovic, Jan Richling, Vladimir Stantchev, An- dre Wiesner, and Miroslaw Malek. Cero: Ce robots community. IEE Proceedings-Software , 152(5):210–214, 2005. [16] Tam ́ as Kov ́ acs, Attila P ́ asztor, and Zolt ́ an Istenes. Connectivity in a wireless network of mobile robots doing a searching and collecting task. In Proceedings of the IEEE International Symposium on Applied Computational Intelligence and Informatics (SACI) , 2009. [17] Luc Steels. Cooperation between distributed agents through self- organisation. In Proceedings of the IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) , 1990. [18] Daniel J Stilwell and John S Bay. Toward the development of a material transport system using swarms of ant-like robots. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 1993. [19] Hoa G Nguyen, Narek Pezeshkian, Michelle Raymond, Anoop Gupta, and Joseph M Spector. Autonomous communication relays for tactical robots. Technical report, DTIC Document, 2003. [20] M Ani Hsieh, Anthony Cowley, James F Keller, Luiz Chaimowicz, Ben Grocholsky, Vijay Kumar, Camillo J Taylor, Yoichiro Endo, Ronald C Arkin, and Boyoon Jung. Adaptive teams of autonomous aerial and ground robots for situational awareness. Journal of Field Robotics , 24(11-12):991–1014, 2007. [21] James K Erickson. Living the dream-an overview of the mars explo- ration project. IEEE Robotics & Automation Magazine , 13(2):12–18, 2006. [22] David Calkins. An overview of robogames [competitions]. IEEE Robotics & Automation Magazine , 18(1):14–15, 2011. [23] JB Petelin, ME Nelson, and J Goodman. Deployment and early ex- perience with remote-presence patient care in a community hospital. Surgical endoscopy , 21(1):53–56, 2007. [24] Seung-Ho Baeg, Jae-Han Park, Jaehan Koh, Kyung-Wook Park, and Moon-Hong Baeg. Robomaidhome: A sensor network-based smart home environment for service robots. In Proceedings of the IEEE Inter- national Symposium on Robot and Human interactive Communication (RO-MAN) , 2007. [25] Seung-Ho Baeg, Jae-Han Park, Jaehan Koh, Kyung-Wook Park, and Moon-Hong Baeg. Building a smart home environment for service robots based on rfid and sensor networks. In Proceedings of the IEEE International Conference on Control, Automation and Systems (IC- CAS) , 2007. [26] JR De La Pinta, Jose Maria Maestre, Eduardo F Camacho, and I Gon- zalez Alonso. Robots in the smart home: a project towards interop- 42 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari erability. International Journal of Ad Hoc and Ubiquitous Computing , 7(3):192–201, 2011. [27] Nikolaus Correll, Jonathan Bachrach, Daniel Vickery, and Daniela Rus. Ad-hoc wireless network coverage with networked robots that cannot lo- calize. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2009. [28] Maxim A Batalin and Gaurav S Sukhatme. Coverage, exploration and deployment by a mobile robot and communication network. Telecom- munication Systems , 26(2-4):181–196, 2004. [29] Zhihui Du, Weiqiang Yang, Yinong Chen, Xin Sun, Xiaoying Wang, and Chen Xu. Design of a robot cloud center. In Proceedings of the Inter- national Symposium on Autonomous Decentralized Systems (ISADS) , 2011. [30] J Quintas, P Menezes, and J Dias. Cloud robotics: Towards context aware robotic networks. In Proceedings of the 16th IASTED Interna- tional Conference on Robotic , 2011. [31] Koji Kamei, Shuichi Nishio, Norihiro Hagita, and Miki Sato. Cloud networked robotics. IEEE Network , 26(3):28–34, 2012. [32] Francesco Bullo, Jorge Cortes, and Sonia Martinez. Distributed Control of Robotic Networks . Applied Mathematics Series. Princeton University Press, 2009. Electronically available at http://coordinationbook.info. [33] Peng Yang, Randy A Freeman, Geoffrey J Gordon, Kevin M Lynch, Siddhartha S Srinivasa, and Rahul Sukthankar. Decentralized esti- mation and control of graph connectivity for mobile sensor networks. Automatica , 46(2):390–396, 2010. [34] Hui Liu, Houshang Darabi, Pat Banerjee, and Jing Liu. Survey of wireless indoor positioning techniques and systems. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews , 37(6):1067–1080, 2007. [35] Guolin Sun, Jie Chen, Wei Guo, and KJ Ray Liu. Signal processing techniques in network-aided positioning: a survey of state-of-the-art positioning designs. IEEE Signal Processing Magazine , 22(4):12–23, 2005. [36] Ismail Guvenc and Chia-Chin Chong. A survey on toa based wireless localization and nlos mitigation techniques. IEEE Communications Surveys & Tutorials , 11(3):107–124, 2009. [37] Isaac Amundson and Xenofon D Koutsoukos. A survey on localization for mobile wireless sensor networks. In Mobile Entity Localization and Tracking in GPS-less Environnments , pages 235–254. Springer, 2009. [38] Yasamin Mostofi and Pradeep Sen. Compressive Cooperative Mapping in Mobile Networks. In Proceedings of the American Control Conference (ACC) , 2009. [39] Alejandro Gonzalez-Ruiz, Alireza Ghaffarkhah, and Yasamin Mostofi. An Integrated Framework for Obstacle Mapping with See-Through Ca- Robotic Wireless Sensor Networks 43 pabilities using Laser and Wireless Channel Measurements. IEEE Sen- sors Journal , 14(1):25–38, January 2014. [40] Alejandro Gonzalez-Ruiz, Alireza Ghaffarkhah, and Yasamin Mostofi. A comprehensive overview and characterization of wireless channels for networked robotic and control systems. Journal of Robotics , 2011, 2012. [41] Yasamin Mostofi and Pradeep Sen. Compressed mapping of communi- cation signal strength. In Proceedings of the Military Communications Conference (MILCOM) , 2008. [42] Emmanuel J Cand` es. Compressive sampling. In Proceedings of the international congress of mathematicians , volume 3, pages 1433–1452. Madrid, Spain, 2006. [43] Yasamin Mostofi, Alejandro Gonzalez-Ruiz, Alireza Gaffarkhah, and Ding Li. Characterization and modeling of wireless channels for net- worked robotic and control systems-a comprehensive overview. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2009. [44] Yasamin Mostofi. Compressive cooperative sensing and mapping in mo- bile networks. IEEE Transactions on Mobile Computing , 10(12):1769– 1784, 2011. [45] Mehrzad Malmirchegini and Yasamin Mostofi. On the spatial pre- dictability of communication channels. IEEE Transactions on Wireless Communications , 11(3):964–978, 2012. [46] Yasamin Mostofi, Mehrzad Malmirchegini, and Alireza Ghaffarkhah. Estimation of communication signal strength in robotic networks. In Proceedings of the IEEE International Conference on Robotics and Au- tomation (ICRA) , 2010. [47] Yuan Yan and Yasamin Mostofi. Co-optimization of communication and motion planning of a robotic operation under resource constraints and in fading environments. IEEE Transactions on Wireless Commu- nications , 12(4):1562–1572, 2013. [48] Yasamin Mostofi. Cooperative Wireless-Based Obstacle/Object Map- ping and See-Through Capabilities in Robotic Networks. IEEE Trans- actions on Mobile Computing , January 2012. [49] Saandeep Depatla, Lucas Buckland, and Yasamin Mostofi. X-Ray Vi- sion with Only WiFi Power Measurements Using Rytov Wave Models. IEEE Transactions on Vehicular Technology , 64:1376–1387, April 2015. [50] Mong-ying A Hsieh, Vijay Kumar, and Camillo J Taylor. Constructing radio signal strength maps with multiple robots. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2004. [51] M Ani Hsieh, Anthony Cowley, Vijay Kumar, and Camillo J Taylor. Maintaining network connectivity and performance in robot teams. Journal of Field Robotics , 25(1-2):111–131, 2008. 44 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari [52] Jonathan Fink and Vijay Kumar. Online methods for radio signal mapping with mobile robots. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2010. [53] Jonathan Fink, Alejandro Ribeiro, and Vijay Kumar. Robust control of mobility and communications in autonomous robot teams. IEEE Access , 1:290–309, 2013. [54] Shangxing Wang, Bhaskar Krishnamachari, and Nora Ayanian. The optimism principle: a unified framework for optimal robotic network deployment in an unknown obstructed environment. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2015. [55] Pradipta Ghosh and Bhaskar Krishnamachari. Interference power bound analysis of a network of wireless robots. CoRR , abs/1608.08261, 2016. [56] Fikadu T Dagefu, Gunjan Verma, Chirag R Rao, L Yu Paul, Jonathan R Fink, Brian M Sadler, and Kamal Sarabandi. Short- range low-vhf channel characterization in cluttered environments. IEEE Transactions on Antennas and Propagation , 63(6):2719–2727, 2015. [57] Nicolas Boillot, Dominique Dhoutaut, and Julien Bourgeois. Large scale mems robots cooperative map building based on realistic simula- tion of nano-wireless communications. Nano Communication Networks , 6(2):51–73, 2015. [58] Khalid A Qaraqe, Serhan Yarkan, Sabih G ̈ uzelg ̈ oz, and H ̈ useyin Arslan. Statistical wireless channel propagation characteristics in underground mines at 900mhz: A comparative analysis with indoor channels. Ad hoc networks , 11(4):1472–1483, 2013. [59] Samir R Das, Elizabeth M Belding-Royer, and Charles E Perkins. Ad hoc on-demand distance vector (aodv) routing. 2003. [60] Ian D Chakeres and Elizabeth M Belding-Royer. Aodv routing proto- col implementation design. In Proceedings of the IEEE International Conference on Distributed Computing Systems Workshops , 2004. [61] Stephen E Deering. Internet protocol, version 6 (ipv6) specification. 1998. [62] David Johnson, Yin-chun Hu, and David Maltz. The dynamic source routing protocol (dsr) for mobile ad hoc networks for ipv4. Technical report, RFC 4728, 2007. [63] Thomas Clausen and Philippe Jacquet. Optimized link state routing protocol (olsr). Technical report, RFC 3626, 2003. [64] David Johnson, Ntsibane Ntlatlapa, and Corinna Aichele. Simple prag- matic approach to mesh routing using batman. 2008. [65] Saumitra M Das, Y Charlie Hu, CS George Lee, and Yung-Hsiang Lu. Mobility-aware ad hoc routing protocols for networking mobile robot teams. Journal of Communications and Networks , 9(3):296–311, 2007. Robotic Wireless Sensor Networks 45 [66] David B Johnson and David A Maltz. Dynamic source routing in ad hoc wireless networks. Kluwer International Series in Engineering and Computer Science , pages 153–179, 1996. [67] Charles E Perkins and Elizabeth M Royer. Ad-hoc on-demand dis- tance vector routing. In Proceedings of the IEEE Workshop on Mobile Computing Systems and Applications , 1999. [68] Sung Ju Lee, William Su, and Mario Gerla. On-demand multicast routing protocol in multihop wireless mobile networks. Mobile Networks and Applications , 7(6):441–453, 2002. [69] TK Abishek, KR Chithra, and MV Ramesh. Aer: Adaptive energy efficient routing protocol for network of flying robots monitoring over disaster hit area. In Proceedings of the Wireless and Optical Commu- nications Conference (WOCC) , 2012. [70] Akihiro Matsumoto, Hajime Asama, Yoshiki Ishida, Koichi Ozaki, and Isao Endo. Communication in the autonomous and decentralized robot system actress. In Proceedings of the IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) , 1990. [71] Alexander Tiderko, Thomas Bachran, Frank Hoeller, and Dirk Schulz. Rose–a framework for multicast communication via unreliable net- works in multi-robot systems. Robotics and Autonomous Systems , 56(12):1017–1026, 2008. [72] Edsger W Dijkstra. A note on two problems in connexion with graphs. Numerische mathematik , 1(1):269–271, 1959. [73] Alfredo Weitzenfeld, Luis Mart ́ ınez-G ́ omez, Juan Pablo Francois, Ale- jandro Levin-Pick, Katia Obraczka, and Jay Boice. Multi-robot sys- tems: Extending robocup small-size architecture with local vision and ad-hoc networking. In Proceedings of the IEEE 3rd Latin American Robotics Symposium (LARS) , 2006. [74] Hai Liu, Amiya Nayak, and Ivan Stojmenovi ́ c. Localized mobility con- trol routing in robotic sensor wireless networks. In Mobile Ad-Hoc and Sensor Networks , pages 19–31. Springer, 2007. [75] Hisayoshi Sugiyama, Tetsuo Tsujioka, and Masashi Murata. Qos rout- ing in a multi-robot network system for urban search and rescue. In Proceedings of the IEEE International Conference on Advanced Infor- mation Networking and Applications , 2006. [76] Chunhung Richard Lin and Jain-Shing Liu. Qos routing in ad hoc wireless networks. IEEE Journal on Selected Areas in Communications , 17(8):1426–1438, 1999. [77] Charles E Perkins and Pravin Bhagwat. Highly dynamic destination- sequenced distance-vector routing (dsdv) for mobile computers. ACM SIGCOMM Computer Communication Review , 24(4):234–244, 1994. [78] Kurt Konolige, Charles Ortiz, Regis Vincent, Andrew Agno, Michael Eriksen, Benson Limketkai, Mark Lewis, Linda Briesemeister, Enrique Ruspini, and Dieter Fox. Large-scale robot teams. Multi-Robot Systems: From Swarms to Intelligent Autonoma , 2:193–204, 2003. 46 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari [79] Bhargav Bellur and Richard G Ogier. A reliable, efficient topology broadcast protocol for dynamic networks. In Proceedings of the IEEE International Conference on Computer Communications (INFOCOM) , 1999. [80] Richard Ogier, Fred Templin, and Mark Lewis. Topology dissemination based on reverse-path forwarding (tbrpf). Technical report, IETF RFC 3684, 2004. [81] Benjamin Sliwa, Daniel Behnke, Christoph Ide, and Christian Wietfeld. Bat mobile: Leveraging mobility control knowledge for efficient routing in mobile robotic networks. arXiv preprint arXiv:1607.01223 , 2016. [82] Brad Karp and Hsiang-Tsung Kung. Gpsr: Greedy perimeter stateless routing for wireless networks. In Proceedings of the ACM International conference on Mobile computing and networking (MobiCom) , 2000. [83] Dongjin Son, Ahmed Helmy, and Bhaskar Krishnamachari. The effect of mobility-induced location errors on geographic routing in mobile ad hoc sensor networks: analysis and improvement using mobility prediction. IEEE Transactions on Mobile Computing , 3(3):233–245, 2004. [84] Sunder Aditya Rao, M Pai, Mounir Boussedjra, and Joseph Mouzna. Gpsr-l: Greedy perimeter stateless routing with lifetime for vanets. In Proceedings of the IEEE International Conference on ITS Telecommu- nications (ITST) , 2008. [85] Martin Mauve, Holger F ̈ ußler, J ̈ org Widmer, and Thomas Lang. Position-based multicast routing for mobile ad-hoc networks. ACM SIGMOBILE Mobile Computing and Communications Review , 7(3):53– 55, 2003. [86] Christian Lochert, Hannes Hartenstein, Jing Tian, Holger Fussler, Dag- mar Hermann, and Martin Mauve. A routing strategy for vehicular ad hoc networks in city environments. In Proceedings of the IEEE Intelli- gent Vehicles Symposium , 2003. [87] Brad Karp. Challenges in geographic routing: Sparse networks, obsta- cles, and traffic provisioning. In Presentation at the DIMACS Workshop on Pervasive Networking , 2001. [88] Martin Mauve, Jorg Widmer, and Hannes Hartenstein. A survey on position-based routing in mobile ad hoc networks. IEEE Network , 15(6):30–39, 2001. [89] Aruna Balasubramanian, Brian Neil Levine, and Arun Venkatara- mani. Replication routing in dtns: a resource allocation approach. IEEE/ACM Transactions on Networking (TON) , 18(2):596–609, 2010. [90] Aruna Balasubramanian, Brian Levine, and Arun Venkataramani. Dtn routing as a resource allocation problem. In ACM SIGCOMM Com- puter Communication Review , volume 37, pages 373–384. ACM, 2007. [91] John Burgess, Brian Gallagher, David Jensen, and Brian Neil Levine. Maxprop: Routing for vehicle-based disruption-tolerant networks. In Proceedings of the IEEE International Conference on Computer Com- munications (INFOCOM) , 2006. Robotic Wireless Sensor Networks 47 [92] Thrasyvoulos Spyropoulos, Konstantinos Psounis, and Cauligi S Raghavendra. Spray and wait: an efficient routing scheme for inter- mittently connected mobile networks. In Proceedings of the 2005 ACM SIGCOMM workshop on Delay-tolerant networking , 2005. [93] Thrasyvoulos Spyropoulos, Konstantinos Psounis, and Cauligi S Raghavendra. Spray and focus: Efficient mobility-assisted routing for heterogeneous and correlated mobility. In Proceedings of the IEEE In- ternational Conference on Pervasive Computing and Communications Workshops (PerCom) , 2007. [94] Xu Li, Wei Shu, Minglu Li, Hongyu Huang, and Min-You Wu. Dtn rout- ing in vehicular sensor networks. In Proceedings of the Global Telecom- munications Conference (GLOBECOM) , 2008. [95] Kevin Fall and Stephen Farrell. Dtn: an architectural retrospective. IEEE Journal on Selected Areas in Communications , 26(5):828–836, 2008. [96] Kevin Fall. A delay-tolerant network architecture for challenged inter- nets. In Proceedings of the Conference on Applications, technologies, architectures, and protocols for computer communications , 2003. [97] Leandros Tassiulas and Anthony Ephremides. Stability properties of constrained queueing systems and scheduling policies for maximum throughput in multihop radio networks. IEEE Transactions on Au- tomatic Control , 37(12):1936–1948, 1992. [98] Scott Moeller, Avinash Sridharan, Bhaskar Krishnamachari, and Om- prakash Gnawali. Routing without routes: The backpressure collection protocol. In Proceedings of the ACM/IEEE International Conference on Information Processing in Sensor Networks , 2010. [99] JG Dai and Wuqin Lin. Asymptotic optimality of maximum pressure policies in stochastic processing networks. The Annals of Applied Prob- ability , 18(6):2239–2299, 2008. [100] Devavrat Shah and Damon Wischik. Optimal scheduling algorithms for input-queued switches. In Proceedings of the IEEE International Conference on Computer Communications (INFOCOM) , 2006. [101] Mohammad Naghshvar, Hairuo Zhuang, and Tara Javidi. A general class of throughput optimal routing policies in multi-hop wireless net- works. IEEE Transactions on Information Theory , 58(4):2175–2193, 2012. [102] Reza Banirazi, Edmond Jonckheere, and Bhaskar Krishnamachari. Heat-diffusion: Pareto optimal dynamic routing for time-varying wire- less networks. In Proceedings of the IEEE International Conference on Computer Communications (INFOCOM) , 2014. [103] Reza Banirazi, Edmond Jonckheere, and Bhaskar Krishnamachari. Dirichlet’s principle on multiclass multihop wireless networks: minimum cost routing subject to stability. In Proceedings of the ACM Interna- tional Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems , 2014. 48 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari [104] Jose Nunez-Martinez, Josep Mangues-Bafalluy, and Marc Portoles- Comeras. Studying practical any-to-any backpressure routing in wi-fi mesh networks from a lyapunov optimization perspective. In Proceed- ings of the IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS) , 2011. [105] Majed Alresaini, Maheswaran Sathiamoorthy, Bhaskar Krishna- machari, and Michael J Neely. Backpressure with adaptive redundancy (bwar). In Proceedings of the IEEE International Conference on Com- puter Communications (INFOCOM) , 2012. [106] Srikanth Nori, Suvil Deora, and Bhaskar Krishnamachari. Backip: Backpressure routing in ipv6-based wireless sensor networks. usc ceng technical report ceng-2014-01. [107] Pradipta Ghosh, He Ren, Reza Banirazi, Bhaskar Krishnamachari, and Edmond A. Jonckheere. Empirical evaluation of the heat-diffusion col- lection protocol for wireless sensor networks. CoRR , abs/1609.03289, 2016. [108] Omprakash Gnawali, Rodrigo Fonseca, Kyle Jamieson, David Moss, and Philip Levis. Collection tree protocol. In Proceedings of the ACM Conference on Embedded Networked Sensor Systems , 2009. [109] Federico Ferrari, Marco Zimmerling, Lothar Thiele, and Olga Saukh. Efficient network flooding and time synchronization with glossy. In Proceedings of the ACM/IEEE International Conference on Informa- tion Processing in Sensor Networks , 2011. [110] Nicolas Burri, Pascal Von Rickenbach, and Roger Wattenhofer. Dozer: ultra-low power data gathering in sensor networks. In Proceedings of the ACM/IEEE International Conference on Information Processing in Sensor Networks , 2007. [111] Federico Ferrari, Marco Zimmerling, Luca Mottola, and Lothar Thiele. Low-power wireless bus. In Proceedings of the ACM Conference on Embedded Networked Sensor Systems , 2012. [112] Tim Winter et al. Rpl: Ipv6 routing protocol for low-power and lossy networks. March 2012. [113] Shangxing Wang, Andrea Gasparri, and Bhaskar Krishnamachari. Robotic message ferrying for wireless networks using coarse-grained backpressure control. IEEE Transactions on Mobile Computing , PP(99):1–1, 2016. [114] Reza Olfati-Saber and Richard M Murray. Consensus problems in net- works of agents with switching topology and time-delays. IEEE Trans- actions on Automatic Control , 49(9):1520–1533, Sept 2004. [115] Lin Xiao, Stephen Boyd, and Sanjay Lall. A scheme for robust dis- tributed sensor fusion based on average consensus. In Proceedings of the ACM/IEEE International Conference on Information Processing in Sensor Networks , 2005. Robotic Wireless Sensor Networks 49 [116] Mauro Franceschelli and Andrea Gasparri. Gossip-based centroid and common reference frame estimation in multiagent systems. IEEE Transactions on Robotics , 30(2):524–531, April 2014. [117] Zhiyun Lin, Bruce Francis, and Manfredi Maggiore. Necessary and sufficient graphical conditions for formation control of unicycles. IEEE Transactions on Automatic Control , 50(1):121–127, Jan 2005. [118] Peng Yang, Randy A Freeman, and Kevin M Lynch. Multi-agent coor- dination by decentralized estimation and control. IEEE Transactions on Automatic Control , 53(11):2480–2496, Dec 2008. [119] Meng Guo, Michael M Zavlanos, and Dimos V Dimarogonas. Control- ling the relative agent motion in multi-agent formation stabilization. IEEE Transactions on Automatic Control , 59(3):820–826, March 2014. [120] Miroslav Fiedler. Algebraic connectivity of graphs. Czechoslovak Math- ematical Journal , 23(2):298–305, 1973. [121] Bojan Mohar and Y Alavi. The laplacian spectrum of graphs. Graph theory, combinatorics, and applications , 2:871–898, 1991. [122] Christopher David Godsil, Gordon Royle, and CD Godsil. Algebraic graph theory , volume 207. Springer New York, 2001. [123] Dimos V Dimarogonas and Kostas J Kyriakopoulos. Connectedness Preserving Distributed Swarm Aggregation for Multiple Kinematic Robots. IEEE Transactions on Robotics , 24(5):1213–1223, 2008. [124] Giuseppe Notarstefano, Ketan Savla, Francesco Bullo, and Ali Jad- babaie. Maintaining limited-range connectivity among second-order agents. In Proceedings of the American Control Conference (ACC) , 2006. [125] Ketan Savla, Giuseppe Notarstefano, and Francesco Bullo. Maintaining limited-range connectivity among second-order agents. SIAM Journal on Control and Optimization , 48(1):187–205, 2009. [126] Demetri P Spanos and Richard M Murray. Robust connectivity of networked vehicles. In Proceedings of the IEEE Conference on Decision and Control (CDC) , 2004. [127] Zhenwang Yao and Kamal Gupta. Backbone-based connectivity con- trol for mobile networks. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2009. [128] Tove Gustavi, Dimos V Dimarogonas, Magnus Egerstedt, and Xiaoming Hu. Sufficient conditions for connectivity maintenance and rendezvous in leader–follower networks. Automatica , 46(1):133–139, 2010. [129] Maria Carmela DeGennaro and Ali Jadbabaie. Decentralized control of connectivity for multi-agent systems. In Proceedings of the IEEE Conference on Decision and Control (CDC) , 2006. [130] Dimos V Dimarogonas and Karl H Johansson. Bounded control of network connectivity in multi-agent systems. IET control theory & applications , 4(8):1330–1338, 2010. [131] Lorenzo Sabattini, Cristian Secchi, Nikhil Chopra, and Andrea Gas- parri. Distributed control of multirobot systems with global connec- 50 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari tivity maintenance. IEEE Transactions on Robotics , 29(5):1326–1332, Oct 2013. [132] Andrea Gasparri, Lorenzo Sabattini, and Giovanni Ulivi. Bounded con- trol law for global connectivity maintenance in cooperative multirobot systems. IEEE Transactions on Robotics , 33(3):700–717, 2017. [133] Michael M Zavlanos, Alejandro Ribeiro, and George J Pappas. Network integrity in mobile robotic networks. IEEE Transactions on Automatic Control , 58(1):3–18, 2013. [134] Michael M Zavlanos and George J Pappas. Distributed connec- tivity control of mobile networks. IEEE Transactions on Robotics , 24(6):1416–1428, 2008. [135] Michael M Zavlanos and George J Pappas. Potential fields for maintain- ing connectivity of mobile networks. IEEE Transactions on Robotics , 23(4):812–816, 2007. [136] Michael M Zavlanos, Magnus B Egerstedt, and George J Pappas. Graph-theoretic connectivity control of mobile robot networks. Pro- ceedings of the IEEE , 99(9):1525–1540, 2011. [137] Florian Knorn, Rade Stanojevic, M Corless, and Robert Shorten. A framework for decentralised feedback connectivity control with applica- tion to sensor networks. International Journal of Control , 82(11):2095– 2114, 2009. [138] Michael Schuresko and Jorge Cort ́ es. Distributed motion constraints for algebraic connectivity of robotic networks. Journal of Intelligent and Robotic Systems , 56(1-2):99–126, 2009. [139] Sabato Manfredi. An algorithm for fast rendezvous seeking of wireless networked robotic systems. Ad Hoc Networks , 11(7):1942–1950, 2013. [140] Stephanie Gil, Dan Feldman, and Daniela Rus. Communication cover- age for independently moving robots. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2012. [141] Ryan K Williams, Andrea Gasparri, Gaurav S Sukhatme, and Giovanni Ulivi. Global connectivity control for spatially interacting multi-robot systems with unicycle kinematics. In Proceedings of the IEEE Interna- tional Conference on Robotics and Automation (ICRA) , 2015. [142] Yasamin Mostofi. Decentralized communication-aware motion plan- ning in mobile networks: An information-gain approach. Journal of Intelligent and Robotic Systems , 56(1-2):233–256, 2009. [143] Matthew Powers, Tucker Balch, et al. Value-based communication preservation for mobile robots. In 7th International Symposium on Distributed Autonomous Robotic Systems , 2004. [144] Stuart O Anderson, Reid Simmons, and Dani Golberg. Maintaining line of sight communications networks between planetary rovers. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , volume 3, 2003. Robotic Wireless Sensor Networks 51 [145] Danilo Tardioli, Alejandro R Mosteo, Luis Riazuelo, Jos ́ e Luis Villar- roel, and Luis Montano. Enforcing network connectivity in robot team missions. The International Journal of Robotics Research , 29(4):460– 480, 2010. [146] Boda Ning, Jiong Jin, Jinchuan Zheng, and Yee Wei Law. Connectivity control and performance optimization in wireless robotic networks: Is- sues, approaches and a new framework. In Proceedings of the 6th Inter- national Conference on Modelling, Identification & Control (ICMIC) , 2014. [147] Bang Wang. Coverage control in sensor networks . Springer Science & Business Media, 2010. [148] Jorge Cortes, Sonia Martinez, Timur Karatas, and Francesco Bullo. Coverage control for mobile sensing networks. IEEE Transactions on Robotics and Automation , 20(2):243–255, 2004. [149] Mac Schwager, Brian J Julian, and Daniela Rus. Optimal coverage for multiple hovering robots with downward facing cameras. In Proceed- ings of the IEEE International Conference on Robotics and Automation (ICRA) , 2009. [150] Howie Choset. Coverage for robotics–a survey of recent results. Annals of mathematics and artificial intelligence , 31(1):113–126, 2001. [151] Noam Hazon and Gal A Kaminka. On redundancy, efficiency, and robustness in coverage for multiple robots. Robotics and Autonomous Systems , 56(12):1102–1114, 2008. [152] Yiannis Kantaros and Michael M Zavlanos. Communication-aware cov- erage control for robotic sensor networks. In Proceedings of the IEEE Conference on Decision and Control (CDC) , 2014. [153] Yuan Yan and Yasamin Mostofi. Communication and path planning strategies of a robotic coverage operation. In Proceedings of the Amer- ican Control Conference (ACC) , 2013. [154] Alireza Ghaffarkhah and Yasamin Mostofi. Optimal motion and com- munication for persistent information collection using a mobile robot. In Proceedings of the IEEE Globecom Workshops (GC Wkshps) , 2012. [155] Alireza Ghaffarkhah and Yasamin Mostofi. Path planning for net- worked robotic surveillance. IEEE Transactions on Signal Processing , 60(7):3560–3575, 2012. [156] Alejandro Gonzalez-Ruiz and Yasamin Mostofi. Cooperative robotic structure mapping using wireless measurements—a comparison of ran- dom and coordinated sampling patterns. IEEE Sensors Journal , 13(7):2571–2580, 2013. [157] Alejandro Gonzalez-Ruiz, Alireza Ghaffarkhah, and Yasamin Mostofi. An integrated framework for obstacle mapping with see-through capa- bilities using laser and wireless channel measurements. IEEE Sensors Journal , 14(1):25–38, 2014. 52 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari [158] Yasamin Mostofi. Cooperative wireless-based obstacle/object mapping and see-through capabilities in robotic networks. IEEE Transactions on Mobile Computing , 12(5):817–829, 2013. [159] Jerome Le Ny, Alejandro Ribeiro, and George J Pappas. Adaptive communication-constrained deployment of unmanned vehicle systems. IEEE Journal on Selected Areas in Communications , 30(5):923–934, 2012. [160] Ryan K Williams and Gaurav S Sukhatme. Cooperative multi-agent inference over grid structured markov random fields. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2011. [161] Andrea Gasparri, Bhaskar Krishnamachari, and Gaurav S. Sukhatme. A framework for multi-robot node coverage in sensor networks. Annals of Mathematics and Artificial Intelligence , 52(2):281–305, 2008. [162] Urs Hunkeler, Hong Linh Truong, and Andy Stanford-Clark. Mqtt-s—a publish/subscribe protocol for wireless sensor networks. In Proceedings of IEEE Conference on Communication systems software and middle- ware and workshops, (COMSWARE) , 2008. [163] Wen Jiang and Milos Zefran. Coverage control with information aggre- gation. In Proceedings of the IEEE Conference on Decision and Control (CDC) , 2013. [164] Ethan Stump, Ali Jadbabaie, and Vijay Kumar. Connectivity manage- ment in mobile robot teams. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2008. [165] Onur Tekdas, Wei Yang, and Volkan Isler. Robotic routers: Algorithms and implementation. The International Journal of Robotics Research , 29(1):110–126, 2010. [166] Julian De Hoog, Stephen Cameron, and Arnoud Visser. Dynamic team hierarchies in communication-limited multi-robot exploration. In Pro- ceedings of the IEEE International Workshop on Safety Security and Rescue Robotics (SSRR) , 2010. [167] Julian De Hoog, Stephen Cameron, and Arnoud Visser. Selection of ren- dezvous points for multi-robot exploration in dynamic environments. In Proceedings of the International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS) , 2010. [168] Julian De Hoog, Stephen Cameron, and Arnoud Visser. Role-based au- tonomous multi-robot exploration. In International Conference on Ad- vanced Cognitive Technologies and Applications (COGNITIVE) , 2009. [169] Yuan Yan and Yasamin Mostofi. Robotic router formation-a bit error rate approach. In Proceedings of the Military Communications Confer- ence (MILCOM) , 2010. [170] Yuan Yan and Yasamin Mostofi. Robotic router formation in real- istic communication environments. IEEE Transactions on Robotics , 28(4):810–827, 2012. Robotic Wireless Sensor Networks 53 [171] Cory Dixon and Eric W Frew. Maintaining optimal communication chains in robotic sensor networks using mobility control. Mobile Net- works and Applications , 14(3):281–291, 2009. [172] David Kiyoshi Goldenberg, Jie Lin, A Stephen Morse, Brad E Rosen, and Y Richard Yang. Towards mobility as a network control primitive. In Proceedings of the ACM International Symposium on Mobile ad hoc networking and computing (MobiHoc) , 2004. [173] Onur Tekdas, Patrick A Plonski, Nikhil Karnad, and Volkan Isler. Maintaining connectivity in environments with obstacles. In Proceed- ings of the IEEE International Conference on Robotics and Automation (ICRA) , 2010. [174] Ryan K Williams, Andrea Gasparri, and Bhaskar Krishnamachari. Route swarm: Wireless network optimization through mobility. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2014. [175] Stephanie Gil, Swarun Kumar, Dina Katabi, and Daniela Rus. Adap- tive communication in multi-robot systems using directionality of signal strength. The International Journal of Robotics Research , 34(7):946– 968, 2015. [176] Gurkan Tuna, Bilel Nefzi, and Gianpaolo Conte. Unmanned aerial vehicle-aided communications system for disaster recovery. Journal of Network and Computer Applications , 41:27–36, 2014. [177] Adnan Fida, Muhammad Iqbal, and Trung Dung Ngo. Communication- and position-aware reconfigurable route optimization in large-scale mo- bile sensor networks. EURASIP Journal on Wireless Communications and Networking , 2014(1):1–20, 2014. [178] James Kennedy. Particle swarm optimization. In Encyclopedia of ma- chine learning , pages 760–766. Springer, 2011. [179] John C Curlander and Robert N McDonough. Synthetic aperture radar . John Wiley & Sons New York, NY, USA, 1991. [180] Pradipta Ghosh, Raktim Pal, and Bhaskar Krishnamachari. Towards controllability of wireless network quality using mobile robotic routers. CoRR , abs/1607.07848, 2016. [181] Pradipta Ghosh and Bhaskar Krishnamachari. Interference power bound analysis of a network of wireless robots. In Proceedings of the International Conference on Communication Systems and Networks (COMSNETS) , 2016. [182] Arpan Chattopadhyay, Marceau Coupechoux, and Anurag Kumar. Se- quential decision algorithms for measurement-based impromptu deploy- ment of a wireless relay network along a line. IEEE/ACM Transactions on Networking , 24(5):2954–2968, 2016. [183] Avishek Ghosh, Arpan Chattopadhyay, Anish Arora, and Anurag Ku- mar. As-you-go deployment of a 2-connected wireless relay network for sensor-sink interconnection. In Proceedings of the International Con- ference on Signal Processing and Communications (SPCOM) , 2014. 54 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari [184] Michael M Zavlanos, Alejandro Ribeiro, and George J Pappas. A frame- work for integrating mobility and routing in mobile communication net- works. In Conference Record of the Forty Fifth Asilomar Conference on Signals, Systems and Computers (ASILOMAR) , 2011. [185] Nicola Bezzo, Rafael Fierro, Ashleigh Swingler, and Silvia Ferrari. A disjunctive programming approach for motion planning of mobile router networks. International Journal of Robotics and Automation , 26(1):13, 2011. [186] Harris Chi Ho Chiu and Wei-Min Shen. Anchor-self-configuring robotic network. In Proceedings of the IEEE International Conference on Robotics and Biomimetics (ROBIO) , 2010. [187] Boyoon Jung and Gaurav S Sukhatme. Detecting moving objects using a single camera on a mobile robot in an outdoor environment. In Proceedings of the International Conference on Intelligent Autonomous Systems (IAS) , 2004. [188] Dirk Schulz, Wolfram Burgard, Dieter Fox, and Armin B Cremers. Tracking multiple moving targets with a mobile robot using particle filters and statistical data association. In Proceedings of the IEEE In- ternational Conference on Robotics and Automation (ICRA) , 2001. [189] Nikolaos P Papanikolopoulos, Pradeep K Khosla, and Takeo Kanade. Visual tracking of a moving target by a camera mounted on a robot: A combination of control and vision. IEEE Transactions on Robotics and Automation , 9(1):14–35, 1993. [190] M. Lindstr ̈ om and J. Eklundh. Detecting and tracking moving objects from a mobile platform using a laser range scanner. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2001. [191] Ivan Markovi ́ c, Fran ̧ cois Chaumette, and Ivan Petrovi ́ c. Moving object detection, tracking and following using an omnidirectional camera on a mobile robot. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2014. [192] Erwin Prassler, Jens Scholz, and Alberto Elfes. Tracking multiple moving objects for real-time robot navigation. Autonomous Robots , 8(2):105–116, 2000. [193] Kentaro Yamano, Kanji Tanaka, Mitsuru Hirayama, Eiji Kondo, Yoshi- hiko Kimuro, and Michihito Matsumoto. Self-localization of mobile robots with rfid system by using support vector machine. In Proceed- ings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2004. [194] Dirk Hahnel, Wolfram Burgard, Dieter Fox, Ken Fishkin, and Matthai Philipose. Mapping and localization with rfid technology. In Proceed- ings of the IEEE International Conference on Robotics and Automation (ICRA) , 2004. Robotic Wireless Sensor Networks 55 [195] Junyi Zhou and Jing Shi. Rfid localization algorithms and applications—a review. Journal of Intelligent Manufacturing , 20(6):695–707, 2009. [196] Peter Brass, Flavio Cabrera-Mora, Andrea Gasparri, and Jizhong Xiao. Multirobot tree and graph exploration. IEEE Transactions on Robotics , 27(4):707–717, Aug 2011. [197] Andrew M Ladd, Kostas E Bekris, Algis Rudys, Lydia E Kavraki, and Dan S Wallach. Robotics-based location sensing using wireless ethernet. Wireless Networks , 11(1-2):189–204, 2005. [198] MSJNM Ocana, LM Bergasa, MA Sotelo, J Nuevo, and R Flores. In- door robot localization system using wifi signal measure and minimizing calibration effort. In Proceedings of the IEEE International Symposium on Industrial Electronics , 2005. [199] Dezhen Song, Chang-Young Kim, and Jingang Yi. Monte carlo simul- taneous localization of multiple unknown transient radio sources using a mobile robot with a directional antenna. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2009. [200] Sriram Venkateswaran, Jason T Isaacs, Kingsley Fregene, Richard Rat- mansky, Brian M Sadler, Joao P Hespanha, and Upamanyu Madhow. Rf source-seeking by a micro aerial vehicle using rotation-based angle of arrival estimates. In Proceedings of the American Control Conference (ACC) , 2013. [201] Ravishankar Palaniappan, Piotr Mirowski, Tin Kam Ho, Harald Steck, Philip Whiting, and Michael MacDonald. Autonomous rf surveying robot for indoor localization and tracking. In International Conference on Indoor Positioning and Indoor Navigation (IPIN) , 2011. [202] Karthik Dantu, Prakhar Goyal, and Gaurav Sukhatme. Relative bear- ing estimation from commodity radios. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , 2009. [203] Travis Deyle, Hai Nguyen, Matthew Reynolds, and Charles C Kemp. Rf vision: Rfid receive signal strength indicator (rssi) images for sen- sor fusion and mobile manipulation. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2009. [204] Henk Wymeersch. The impact of cooperative localization on achieving higher-level goals. In Proceedings of the IEEE International Conference on Communications Workshops (ICC) , 2013. [205] Dimitrios Koutsonikolas, Saumitra M Das, Y Charlie Hu, Yung-Hsiang Lu, and CS George Lee. Cocoa: Coordinated cooperative localization for mobile multi-robot ad hoc networks. In Proceedings of the IEEE In- ternational Conference on Distributed Computing Systems Workshops (ICDCS) , 2006. [206] Stefan Zickler and Manuela Veloso. Rss-based relative localization and tethering for moving robots in unknown environments. In Proceed- 56 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari ings of the IEEE International Conference on Robotics and Automation (ICRA) , 2010. [207] Carlo Filoramo, Andrea Gasparri, Federica Pascucci, Attilio Priolo, and Giovanni Ulivi. A rssi-based technique for inter-distance computation in multi-robot systems. In Proceedings of the Mediterranean Conference on Control Automation (MED) , 2010. [208] Luis Oliveira, Hongbin Li, Luis Almeida, and Traian E Abrudan. Rssi- based relative localisation for mobile robots. Ad Hoc Networks , 13:321– 335, 2014. [209] Deepak Vasisht, Swarun Kumar, and Dina Katabi. Decimeter-level lo- calization with a single wifi access point. In Proceedings of the USENIX Symposium on Networked Systems Design and Implementation (NSDI) , 2016. [210] VICON. https://www.vicon.com/. [211] Douglas W Gage. Network protocols for mobile robot systems. In Intelligent Systems & Advanced Manufacturing , pages 107–118. Inter- national Society for Optics and Photonics, 1998. [212] CC Fung, H Eren, and Y Nakazato. Position sensing of mobile robots for team operations. In Proceedings of the IEEE Instrumentation and Measurement Technology Conference (IMTC) , 1994. [213] Peter Wilke and Thomas Br ̈ aunl. Flexible wireless communication net- work for mobile robot agents. Industrial Robot: An International Jour- nal , 28(3):220–232, 2001. [214] Elizabeth A Thompson, Charles McIntosh, James Isaacs, Eric Harmi- son, and Ross Sneary. Robot communication link using 802.11 n or 900mhz ofdm. Journal of Network and Computer Applications , 52:37– 51, 2015. [215] Yeong Che Fai, Shamsudin HM Amin, Nb Fisal, and J Abu Bakar. Bluetooth enabled mobile robot. In Proceedings of the IEEE Interna- tional Conference on Industrial Technology (ICIT) , 2002. [216] John Heidemann, Wei Ye, Jack Wills, Affan Syed, and Yuan Li. Re- search challenges and applications for underwater sensor networking. In ’Proceedings of the IEEE Wireless Communications and Networking Conference (WCNC) , volume 1, pages 228–235. IEEE, 2006. [217] Pradipta Ghosh, Nachikethas A Jagadeesan, Pranav Sakulkar, and Bhaskar Krishnamachari. Loco: A location based communication scheme. In MadCom: New Wireless Communication Paradigms for the Internet of Things, 2017 . [218] Qingquan Zhang, Woong Cho, Gerald E Sobelman, Liuqing Yang, and Richard Voyles. Twinsnet: A cooperative mimo mobile sensor network. In Ubiquitous Intelligence and Computing , pages 508–516. Springer, 2006. [219] Geoffrey A Hollinger, Sunav Choudhary, Parastoo Qarabaqi, Christo- pher Murphy, Urbashi Mitra, G Sukhatme, Milica Stojanovic, Hanu- mant Singh, and Franz Hover. Communication protocols for underwater Robotic Wireless Sensor Networks 57 data collection using a robotic sensor network. In Proceedings of the IEEE GLOBECOM Workshops (GC Wkshps) , 2011. [220] Huan Pham and Sanjay Jha. An adaptive mobility-aware mac protocol for sensor networks (ms-mac). In Proceedings of the IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS) , 2004. [221] Muneeb Ali, Tashfeen Suleman, and Zartash Afzal Uzmi. Mmac: A mobility-adaptive, collision-free mac protocol for wireless sensor net- works. In Proceedings of the IEEE International Performance, Com- puting, and Communications Conference (IPCCC) , 2005. [222] Arshad Jhumka and Sandeep Kulkarni. On the design of mobility- tolerant tdma-based media access control (mac) protocol for mobile sensor networks. In Distributed Computing and Internet Technology , pages 42–53. Springer, 2007. [223] Tang Zhiyong and Waltenegus Dargie. A mobility-aware medium access control protocol for wireless sensor networks. In Proceedings of the IEEE GLOBECOM Workshops (GC Wkshps) , 2010. [224] Antonio Gonga, Olaf Landsiedel, and Mikael Johansson. Mobisense: Power-efficient micro-mobility in wireless sensor networks. In Proceed- ings of the IEEE International Conference on Distributed Computing in Sensor Systems and Workshops (DCOSS) , 2011. [225] Majid Nabi, Milos Blagojevic, Marc Geilen, Twan Basten, and Teun Hendriks. Mcmac: An optimized medium access control protocol for mobile clusters in wireless sensor networks. In Proceedings of the IEEE Communications Society Conference on Sensor Mesh and Ad Hoc Com- munications and Networks (SECON) , 2010. [226] Qian Dong and Waltenegus Dargie. A survey on mobility and mobility- aware mac protocols in wireless sensor networks. IEEE Communica- tions Surveys & Tutorials , 15(1):88–100, 2013. [227] Boda Ning, Jiong Jin, Jinchuan Zheng, and Huan Zhang. Minimiz- ing network interference through mobility control in wireless robotic networks. In Proceedings of the International Conference on Control Automation Robotics & Vision (ICARCV) , 2014. [228] Peter X Liu, MQ-H Meng, Polley R Liu, and Simon X Yang. An end-to- end transmission architecture for the remote control of robots over ip networks. IEEE/ASME Transactions on Mechatronics , 10(5):560–570, 2005. [229] Yutaka Uchimura and Takahiro Yakoh. Bilateral robot system on the real-time network structure. Industrial Electronics, IEEE Transactions on , 51(5):940–946, 2004. [230] Li Ping, Lu Wenjuan, and Sun Zengqi. Transport layer protocol recon- figuration for network-based robot control system. In Proceedings of the IEEE International Conference on Networking, Sensing and Control , 2005. [231] Gavin Holland and Nitin Vaidya. Analysis of tcp performance over mobile ad hoc networks. Wireless Networks , 8(2/3):275–288, 2002. 58 Pradipta Ghosh*, Andrea Gasparri, Jiong Jin and Bhaskar Krishnamachari [232] Khaled A Harras and Kevin C Almeroth. Transport layer issues in delay tolerant mobile networks. In NETWORKING 2006. Network- ing Technologies, Services, and Protocols; Performance of Computer and Communication Networks; Mobile and Wireless Communications Systems , pages 463–475. Springer, 2006. [233] Christian Lochert, Bj ̈ orn Scheuermann, and Martin Mauve. A survey on congestion control for mobile ad hoc networks. Wireless Communi- cations and Mobile Computing , 7(5):655–676, 2007. [234] Lynne E Parker. Alliance: An architecture for fault tolerant, coop- erative control of heterogeneous mobile robots. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 1994. [235] Richard Volpe, Issa Nesnas, Tara Estlin, Darren Mutz, Richard Pe- tras, and Hari Das. The claraty architecture for robotic autonomy. In Proceedings of IEEE Aerospace Conference , 2001. [236] Ronald C Arkin and Tucker Balch. Cooperative multiagent robotic systems. Artificial Intelligence and Mobile Robots. MIT/AAAI Press, Cambridge, MA , 1998. [237] Sascha A Stoeter, Paul E Rybski, Michael D Erickson, Maria Gini, Dean F Hougen, Donald G Krantz, Nikolaos Papanikolopoulos, and Michael Wyman. A robot team for exploration and surveillance: Design and architecture. In Proceedings of the International Conference on Intelligent Autonomous Systems (IAS) , 2000. [238] Alberto Sanfeliu and Juan Andrade-Cetto. Ubiquitous networking robotics in urban settings. In Proceedings of the Workshop on Network Robot Systems. Toward Intelligent Robotic Systems Integrated with En- vironments , 2006. [239] Alberto Sanfeliu, Norihiro Hagita, and Alessandro Saffiotti. Network robot systems. Robotics and Autonomous Systems , 56(10):793–797, 2008. [240] Alessandro Saffiotti, Mathias Broxvall, Marco Gritti, Kevin LeBlanc, Robert Lundh, Jayedur Rashid, Beom-Su Seo, and Young-Jo Cho. The peis-ecology project: vision and results. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2008. [241] Mark McClure, Daniel R Corbett, and Douglas W Gage. The darpa landroids program. In SPIE Defense, Security, and Sensing , 2009. [242] L Chaimowicz, A Cowley, D Gomez-Ibanez, B Grocholsky, MA Hsieh, H Hsu, JF Keller, V Kumar, R Swaminathan, and CJ Taylor. Deploying air-ground multi-robot teams in urban environments. In Multi-Robot Systems. From Swarms to Intelligent Automata Volume III , pages 223– 234. Springer, 2005. [243] Marco Dorigo, Elio Tuci, Roderich Groß, Vito Trianni, Thomas Halva Labella, Shervin Nouyan, Christos Ampatzis, Jean-Louis Deneubourg, Robotic Wireless Sensor Networks 59 Gianluca Baldassarre, Stefano Nolfi, et al. The swarm-bots project. In Swarm Robotics , pages 31–44. Springer, 2005. [244] Francesco Mondada, Michael Bonani, Andr ́ e Guignard, St ́ ephane Mag- nenat, Christian Studer, and Dario Floreano. Superlinear physical per- formances in a swarm-bot. In Advances in Artificial Life , pages 282– 291. Springer, 2005. [245] NECTAR: NEtworked Cooperative Teams of Autonomous Robots. http://webuser.unicas.it/nectar/. [246] J ̈ org Seyfried, Marc Szymanski, Natalie Bender, Ramon Estana, Michael Thiel, and Heinz W ̈ orn. The i-swarm project: Intelligent small world autonomous robots for micro-manipulation. In Swarm Robotics , pages 70–83. Springer, 2005. [247] Heinz Woern, Marc Szymanski, and Joerg Seyfried. The i-swarm project. In Proceedings of the IEEE International Symposium on Robot and Human interactive Communication (RO-MAN) , 2006. [248] Aquila Project. https://code.facebook.com/posts/268598690180189. [249] Project Loon. https://x.company/loon/. [250] Swarmrobot. http://www.swarmrobot.org/. [251] Swarm-bots. http://www.swarm-bots.org/.