Bi-directional Motion Detection: A Neural Intelligent Model for Perception of Cognitive Robots  Abstract :   In this paper, a new neuronal circuit, based on the spiking neuronal network model, is proposed in order to detect the movement direction of dynamic objects wandering around cognitive   robots.   Capability   of   our   new   approach   in   bi- directional movement detection is beholden to its symmetric configuration of the proposed circuit. With due attention to magnificence of handling of blocking problems in neuronal networks   such   as   epilepsy,   mounting   both   excitatory   and inhibitory stimuli has been taken into account. Investigations upon applied implementation of aforementioned strategy on  PIONEER™ cognitive robot reveals that the   strategy leads to alleviation   of   potential   level   in   the   sensory   networks. Furthermore, investigation on intrinsic delay of the circuit reveals not only the noticeable switching rate which could be acquired but the high-efficient coupling of the circuit with the other high-speed ones.  Index   Terms   -   Cognitive   robots,   Braitenberg   vehicles, Movement detector, Neuronal networks  I.   INTRODUCTION  Investigation on mental and psychological properties of humans   and   animals   in   order   to   emulate   different operations of the brain have been taken into account from old   eras.   After   some   seminal   explorations   about construction of the neuronal cells in recent decades, many researches were commenced in order to map some simple sensory   concepts   of   the   animal   brains   into   neuronal circuits.   Analysis of excitatory and inhibitory stimuli could   be considered as denouements of Cajal’s primary  investigations on some neuronal systems of the human such as visual system [1]. Based on some breakthroughs and new accomplishments which were acquired in 1970 decade by neuroscientists and cyberneticians, especially Braitenberg great work [2], artificial neuronal network concept was recognized as the effective and practical approach   to   implement   all   ideas   for   simulation   of neuronal mechanisms. Different types of vehicles which were introduced by Braitenberg could be considered as an unique strategy to depict an artificial concept of different operations of the animal   brain,   an   imitation   of   rudimentary   mankind behaviors just based on simple stimuli acquired from circumferential   environment.   In   following   decades, various researches in different concepts related to basic Braitenberg   vehicles   were   done.   Robotics   could   be considered as one of the   PIONEER™   sciences which tried to employ these intrinsic smart creatures in mobile robots. It is simplicity itself that control of mobile robots is a crucial stuff which needs to complete knowledge about   operation   of   the   robot.   With   due   attention   to empirical attitude which was issued upon Braitenberg vehicles in most of researches, fuzzy control was taken into account. Beau frere’s viewpoint introduced in [3]  continued   in   next   researches.   A   sensor-interpreting approach   was   implemented   by   Maaref   et   al,   in   [4]. Combination of the fuzzy logic with different strategies is theme of the other researches. Application of genetic algorithms   and   adaptive   fuzzy-neural   networks cooperated with fuzzy logic were studied by Pratihar in [5] and Wang et al in [6], respectively. [7] investigated by Yang et al could be considered as a successful model for navigation of differential mobile robots. A solution on navigation problem in corridor is proposed by Lee et al in [8].  French’s attempts in switching between different types  led to introduction of the Neuromodulation concept in Braitenberg vehicles which is investigated in [9]. Finally, the most recent accomplishments in Braitenberg vehicles territory could be referred to Rano who has tried to extract some mathematical models for stimuli in order to reach a more investigated attitude regarding some basic vehicles like vehicle 3a, which is investigated in [10]. Neuronal circuits could be implemented in huge neuronal networks in order to syncretize various sensory operations . As Braitenberg’s attitude and based on the  fact   that   different   operations   need   different   neuronal structures, in other words neuronal circuits, optimization of sensory modules could be considered as a worthwhile consequence to acquire embedded neuronal systems as much   as   similar   to   nervous   construction   of   brain. Therefore, it sounds reasonable to segment our artificial brain   into   separate   parts.   Undoubtedly,   cognition   of dynamic objects around the vehicle is one of the most important senses which is desired for the creature. With taking the practical points into consideration, volume of circuit and delay of different synapses, in addition to some particular phenomena like epilepsy, must be taken into account. Detection   of   movement   direction,   as   a   necessary operation   in   cognition   process,   is   investigated   in  Braitenberg’s   great book and some primary designs are proposed   there.   Most   of   his   sensory-designs   are  Matin Macktoobian  Fault Detection & Identification Lab   (FDI)  Faculty of   Electrical   &   Computer Engineering  K. N. Toosi University of Technology  Tehran, Iran  matinking@hotmail.com
established based on activation of one or some neurons due to applied stimulating input. This paper discusses a novel movement detector which in one hand is able to distinguish both movements to the right and to the left. On the other hand, simultaneous utilization   of   excitatory   and   inhibitory   stimuli   has   a noticeable effect on decrement of possibility of epileptic seizures in the circuit. The organization of the rest of the paper is as follows, section   II   includes   s   a   brief   review   on   some   basic definitions of neuroscience which will be mentioned and investigated in the paper frequently. Section III describes the   primary   motion   detector   which   is   introduced   by Braitenberg.   Bi-directional   movement   detector   is presented   in   section   IV.   Section   V   shows   the comparisons   and   simulated   results.   Conclusions   and probable themes   for future works are investigated in section VI.  II.   SOME SEMINAL DEFINITIONS  Below terms could be evaluated as not only basic and general concepts of neuroscience but also the criteria to compare new accomplishments with prior designs. Stimulus: A burst of action potentials to activate sensory mechanisms. Epilepsy: In a population of elements in which excitatory connections abound, if the number of active elements reaches a certain critical level, chances are the remaining ones will also become activated. These elements, in turn, keep the first set active. A maximal condition of activity is then established and maintained until the supply of energy is exhausted. Excitatory:   Strengthening   effect   of   the   stimuli   on activation potentials. Inhibitory: Attenuating effect of the stimuli on activation potentials.  III.   BRAITENBERG ’S   MOVEMENT DETECTOR PROTOTYPE  Visual sy stem in Braitenberg’s for discri minating the movement   detection   view   is   constructed   entirely   of photocells. His primary 3-stage neuron al circuit for “left - to- right” detection is depicted as Fig.1.  Level (F) is responsible for receiving stimuli which are produced   due   to   object   movement.   Operation   of   the sensors mounted in this level is solely relied on object existence   in   active   scope   of   sensors.   Such   longer connections are in the neuronal circuit, more delay should be taken into account for those connections. Such delays are shown as simple digital filters in common notation of artificial neuronal network texts, as shown in Fig. 2. With due attention to applied delay on the left branch between each adjacent pair of sensors in level (D), exerted neuron on   right   branch   will   be   activated   on   time   so   that excitatory signals from both left and right branches will activate neuron of level (C). Object movement from left to right stimulates the left-most to the right-most sensor one by one. Therefore, activation of neurons of level (C) regarding the direction is the same with the movement direction of the object. Obviously, neuronal circuit which is   introduced   above   is   not   able   to   detect   movement direction from left to right. Such circuit could be acquired with displacing of delay connection and unit excitatory neuron from left branch to the right branch and vice versa,   regularly.   This   approach   not   only   leads   to   a sophisticated module but also epilepsy could be occurred with more probability. If the vehicle is surrounded by more than one dynamic object, it is possible that both sub-circuits of the module become active simultaneously. Considering a practical view, with increasing the number of the neurons in actual systems, epileptic seizures will be evaluated as a more crucial problem.  IV.   BI-DIRECTIONAL MOVEMENT DETECTOR  Simultaneous detection of movement directions to the right or to the left requires a symmetric construction for neuronal circuit that all activation time slots must sound entirely similar. The novel movement detector is depicted as Fig. 3. The circuit is consists of usual neurosensors and two distinctive parts for direction detection, upper one for right-to-left   and   the   other   one   for   left-to   right. Furthermore, each part is intrinsically constructed of two stages. First stage neurons, called simple neurons, are activated by just one stimulus whereas an excitatory- inhibitory set of neurotransmmiters is embedded in order to   connect   the   neuron   to   the   prior   and   posterior neurosensors.  Fig 2. Typical delay model in neur on al circuits  Fig 1. Braitenberg’s left to right motion detector [2]
Activation signals of each pair of successive neurons, as excitatory stimuli, are utilized to acvtivate second stage which   is   itself   constructed   of   neurons,   known   as combined   neurons,   activated   by   two   simultaneous stimuli. Unequal length of neurotransmmiters devised in both stages are necessary for overall valid operation of the circuit. With due attention to similar mechanism of both sub systems, investigation on operation of one of them could be verified the other subsystem. Here the left- to-right subcircuit is going to be explained. As a moving object approaches to a neurosensor, its activation signal will   be   transmitted   to   the   simple   neuron.   Due   to noticeable delay of excitatory connection, one can take some times into account for comprehension of activation signal by the simple neuron. When object comes rear enough to the next neurosensor and just after activation of simple neuron and transmission of its excitatory signal to ward   combined   sensor,   inhibitory   stimulus   which   is established by next neurosensor will deactivate the simple neuron. In addition, activation of next neurosensor also leads to activation of the anterior simple neuron with latency. Finally, shorter length of the right excitatory signal and its trivial latency will be syncretized with longer length of the left one and its noticeable latency in order   to   provide   two   desired   excitatory   signals   for activation of corresponding combined neuron. Movement direction of the object   could be evaluated by right- handed activation succession of the combined neurons. Above mechanism also could be valid for right-to-left subcircuit except the activation direction of the combined neurons which is to the left.  V.   AN INVESTIGATION ON IMPROVEMENTS  Proposed   bi-directional   movement   detector,   which   is analysed in previous section, could be assessed as a better generation of the movement detectors in comparison with Braitenberg prototype model. This claim could be proved with examination of two improved factors: circuit size and epilepsy inhibition. In one hand, Given the extent of Fig.   1   and   Fig.3   with   consideration   of   using   5  neurosensors, Braitenberg’s model needs 16 neurons for  direction   detection   with   mounting   of   two   neuronal circuits individually, Whereas, bi-directional movement detector just needs 14 neurons under aegis of one circuit for providing similar operation. Fig. 4 shows all the delay which must be taken into account for one circuit for providing similar operation. Fig. 4 shows all the delay which must be taken into account for all the delay which must be taken into account for regulation of the spikes. It is simplicity itself that delays must be controlled based in the velocity of the moving object. This quantity could be acquired by tuning of activation functions.  On the other hand, simultaneous utilization of excitatory and inhibitory connections stabilize the potential level of the system so that each set of neurons will be suppressed by reversed-type connections mounted to them.  VI.   DELAY TUNING OF CONNECTIONS  As   mentioned   before   and   based   on   the   convention illustrated   by   Fig.2,   longer   connections   are representatives   of   the   more   dilatory   ones.   Desired operation of the circuit depends on well-tuned latencies for combined neuron connections. It is possible to guess about important role of distance between each adjacent neurosensors and velocities of moving objects, which should be proved.  Proposition : If   and   are latency of right and left excitatory   signals   of   combined   neurons,   respectively and   is   distance   between   each   adjacent   neurosensors Fig 3. Bi-directional movement detection
and   is velocity of moving objects,   and   could be tuned by below equation:  Definitions :  : time origin as the moving object is once seen by the most-right or most-left neurosensor.  : activation time of right simple neuron corresponding to an unique combined neuron.  : activation time of left simple neuron corresponding to an unique combined neuron.  : right activation time of connection of a combined neuron.  : left activation time of connection of a combined neuron.  Proof : below equations could be acquired based on Fig.4 and above definitions:  Combined neuron could be become on if only if it both activation signals receive simultaneously which means below condition must be satisfied:  So,   with   fulfilling   above   condition,   the   proof   is completed:  VII.   SIMULATED DENOUEMENTS  Simulated   and   practical   tests   show   noticeable accomplishments   acquired   by   applying   bi-directional movement detector on neurorobots.   PIONEER™   mobile robot, shown in Fig.5, was utilized as a test bed for verifying the simulated results. It is equipped with 4 IR transducers as artificial neurosensors for this research. We   evaluated   the   performance   of   bi-directional movement   detection   on   it.   Embedded   IR   transducers transformed it into a Braitenberg-like machine. In these experiments, the stimuli of the moving objects in front of the vehicle were modelled as Gaussian-like signals which being applied in a determined range of distances from IR transducers. With due attention to desire for testing the detector under effect of different stimuli regarding to intensity of the stimulation, moving objects were allowed to cross in front of the vehicle with various velocities and distance   of   the   moving   object   from   vehicle   sensors, reactions   of   the   vehicle   are   recorded   as   numerical potential quantities. The coherent denouements after raw data   interpolation   and   computer   simulations   lead   to reaching a rough comparison between two circuits in epilepsy   rejection.   Obviously,   such   lower   the   circuit potential   within   the   experiment   toward   movement   of dynamic objects is, more security against critical bound of   epilepsy   could   be   taken   into   account.   Such investigation is depicted as Fig.7. With consideration the stimuli   shown   in   Fig.6,   our   bi-directional   movement detector works with lower potential level in comparison  with   Braitenberg’s   prototype   detector. Therefore,   less  number of active neurons of the bi-directional movement  Fig. 4. Illustarion of the de lays in one unit of the circuit
detector   in sampling time   interval   of the   experiment could be construed, reasonably. Spatial model of the bi-directional movement detector could be visualized as Fig. 8.  VIII.   CONCLUDING REMARKS AND FUTURE WORKS  Braitenberg   vehicles   as   one   of   the   most   inspiring concepts in various investigations on neurorobotics, have a long time to be progressed in a such complete form that different neuronal and psychological operations of the brain could be applied on them. So, robust perception of the surrounding environment could be considered as a seminal   requirement   for   reaching   to   this   great   goal. Presented   investigation   in   this   paper   shows   that   bi- directional movement detection not only can improve the size   of the   neuronal   circuit,   a   crucial   point   in   huge neuronal systems, but is able to ameliorate potential level of the different parts of the circuit far from danger of the epilepsy phenomenon. tuning of connections will be more complicated   with   consideration   of   multiple   moving objects   while   they   have   different   distances   from neurosensors.   Also,   hardware   implementation   of neurorobotics   systems   like   introduced   bi-directional movement   detector   on   some   architectures   like   ZISC could be taken into account.  REFERENCES  [1]   Cajal, S. R. y, “Structure of the optic chiasm and general theory of  the crossing visual pathways.   (Estructura del kiasma optico y  teoria general de los entrecruzamientos de las vias nervosas.)”,  quoted in Cajal, 1898. [2]   V. Braitenberg, “Vehicles Experiments in synthetic psychology”,  The MIT Press, 1984. [3]   Beaufrere, B, “A   mobile robot navigation method using a fuzzy  logic approach”, Robotica, No. 13, pp. 437– 448, 1995. [4]   Maaref, H. and Barret, C, “Sensor -based fuzzy navigation of an  autonomous mobile robot in an indoor environment”, Control Eng.  Pract., No. 8, pp. 747 – 768, 2000. [5]   Pratihar, D.K. and Dep, K., Ghosh, A, “A genetic -fuzzy approach  for mobile robot navigation among moving obstacles”,   Int. J.  Approx. Reason. 20, pp. 145 – 172, 1999. [6]   Wang, J.S. and Lee, C.S, “Self -adaptive recurrent neural-fuzzy control for an autonomou s underwater vehicle”, In Proceedings of  IEEE Int. Conf. on Robotics and Automation   2 , pp. 1095 – 1100, 2002. [7]   X. Yang, R. V. Patel, and M. Moallem, “A Fuzzy -Braitenberg  Navigation   Strategy   for   Differential   Drive   Mobile   Robots”,  Journal of Intelligent Robotic Systems, No. 47, pp. 101 – 124, 2006. [8]   K. Lee et al, “Topological Navigation of Mobile Robot in Corridor Environment   using   Sonar   Sensor”,   Proceedings   of   the   2006  IEEE/RSJ   International   Conference   on   Intelligent   Robots   and Systems, pp. 2760-2765, 2006. [9]   R.L.B . French and L. Canamero. “Introducing neuromodulation to a   Braitenberg   vehicle”,   In   Proceedings   of   the   2005   IEEE International Conference on Robotics and Automation, pp. 4188 –  4193, 2005. [10]   I. Rano, “On the Convergence of Braitenberg Vehicle 3a immersed  in P arabolic Stimuli”   2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 2346-2351, 2011.  Fig 6. Applied stimuli to the detectors  Fig 5.   PIONEER™   mobile robot   Fig 7. Potential va r iation of the detectors
Fig. 8. 3D visualization of the bi - directional movement detector