  Abstract — CLEVERarm (Compact, Low-weight, Ergonomic, Virtual/Augmented Reality Enhanced Rehabilitation arm) is a novel exoskeleton with eight degrees of freedom supporting the motion of shoulder girdle, glenohumeral joint, elbow and wrist. O f the exoskeleton’s eight degrees of freedom,   six are active and the two degrees of freedom supporting the motion of wrist are   passive.   This   paper   briefly   outlines   the   design   of CLEVERarm and its control architectures.  I.   I NTRODUCTION  Stroke   affects   an   increasing   portion   of   the   aging population of world, leaving many of the survivors with different levels and forms of disability. There is a recent surge in use of robotic systems for rehabilitation purposes due to their inherent capabilities in producing high intensity, repeatable,   and   precisely   controllable   motions   [1].   End effector based systems [2] and exoskeletons [3] are the two category   of   the   robotic   systems   designed   to   provide automated therapy to stroke patients. CLEVERarm   is   a   novel   upper-limb   exoskeleton   for rehabilitation of stroke patients. CLEVER ARM has six active, and two passive degrees of freedom (DoF), allowing the motion of shoulder girdle,   glenohumeral (GH) joint, elbow, and wrist [4]. An active degree of freedom is used for assisting Flexion/Extension of the elbow, while the remaining five active degrees of freedom are used in the design of the device shoulder to improve the ergonomics of the device. The device is also equipped with visual technologies such as virtual and augmented reality to enable diverse, task specific and immersive training scenarios. Compactness and weight reduction are two main criteria in   development   of   CLEVERarm.   A   combination   of   3D printed   carbon-fiber   reinforced   plastic   and   machined aluminum was used to achieve a low weight design. This manuscript briefly reviews the techniques used to ensure viability of using a metal-plastic structure as the chassis of a robotic   device   considering   non-isotropic   properties   of carbon-fiber   reinforced   plastic.   Additionally,   this   paper provides brief details on kinematic structure of CLEVERarm, the embodiment of the design and the control architecture. II.   CLEVER   ARM CLEVER ARM has eight degrees of freedom supporting the motion of shoulder girdle, GH joint, elbow, and wrist. Fig. 1 shows the CAD model of the exoskeleton on a dummy human model, and the actual prototype:  *This research is sponsored by QNRF, NPRP N. 7-1685-2-626. R. Soltani-Zarrin A. Zeiaee, A. Eib and R. Langari are with Texas A&M University, College Station, TX USA (email: rana.soltani@tamu.edu). R. Tafreshi is with Texas A&M University at Qatar, Doha, Qatar.  Figure 1 CLEVERarm: (a) prototype, (b) CAD model  The motion of the GH joint and the inner shoulder are supported   by five   degrees of   freedom in   the   design   of CLEVER ARM. Three hinged joints, constituting a spherical linkage, is used to provide the three DoF required for the motion of GH joint. CLEVER ARM uses two active degrees of freedom (a revolute joint followed by a prismatic joint) to model the displacement of GH joint center in the frontal plane of human body which allows accurate tracking of GH joint center path on the frontal plane without approximating it as a circular path [4]. Denavit-Hartenberg   (DH)   convention   was   used   for kinematic   analysis.   Fig.   2   shows   the   assignment   of coordinate systems and the corresponding DH parameters where   p 1   through   p 6   are the physical parameters of system. While   p 1   and   p 2   are constants, other parameters can be changed to accommodate different patient body dimensions.  Figure 2.   (a) DH coordinate frame, (b) DH parameters  III.   P ROTOTYPE   D ESIGN AND   F ABRICATION  Weight reduction and compactness of the device are the two criteria in choosing components for motorization of design.   Electric   motors   coupled   with   strain   wave   gears (Harmonic Drive LLC), were used in rotary joint while the  CLEVERarm: A Novel Exoskeleton for Rehabilitation of Upper  Limb Impairments  Rana Soltani - Zarrin,   Amin Zeiaee, Andrew Eib,   Reza Langari , Reza Tafreshi  ( a )   ( b )
prismatic joint in the design of exoskeleton was realized by a direct drive linear actuator. To ensure reliability of joint level feedback, an incremental and an absolute encoder is used in each joint. Moreover, the two physical interfaces between the device and body are equipped with 6 axis force/torque sensors to enable achieving back-drivability. To minimize the weight of the device body, various choices   of   material   and   manufacturing   techniques   were studied and a combination of Aluminum and 3D printed Carbon Fiber (CF) reinforced plastic were selected. While carbon fiber possess many advantageous properties, design of fasteners for   metal/composite interfaces and ensuring sufficient rigidity in the structure is challenging due to the anisotropic properties of the composite materials. To address the   aforementioned   challenges,   extensive   finite   element analysis (FEA) were done to ensure that maximum deflection of the structure in worst case loading scenarios is within the acceptable range. Figure 3.a shows the achieved minimal deflection under worst-case static loading and figure 3.b shows the orientation of the laid carbon fibers achieving the desired strength properties.  Figure 3.   (a) FEA results Max Deflection , (b) Direction of the Carbon Fibers shown in blue in Eiger Software (Markforged Inc.)  Plastic/metal interfaces were designed by distributing the load on larger surfaces to avoid concentration of stress. The weight of the device body is 14 lbs., while the weight of the entire device is 27 lbs. IV.   C ONTROL   A RCHITECTURE  The overall control architecture of CELEVRarm is shown in figure 4. This control block diagram is implemented on  National Instruments’ CompactRio (cRio) RealT ime target and FPGA. The main control loop runs at 10 kHz while the force sensors data is acquired by the FPGA at 50 kHz rate. The   gravitational   model   used   to   cancel   the   weight   of exoskeleton is derived using the inertial properties of the system CAD model (SolidWorks, Dassault Systèmes) [5].  Figure 4.   CLEVERarm Control Architecture  Game environments are part of the control architecture of the   CLEVERarm   since   they   represent   the   final   desired position for the patient hand. Using the output of the game environment, reference generation block within the control architecture uses the algorithms developed by the authors for generating   human-like   motions   considering   the scapulohumeral rhythms [6,7]. Finally, friction compensation is achieved by admittance-based control. The interaction forces   measured   with   the   F/T   sensors   and   the   desired impedances are used to calculate the desired velocity for the interaction   ports   [8].   Using   the   system   Jacobian,   these desired angular velocities are derived and tracked with a PD control law [9]. Figure 5 shows two different impedance values rendered by the exoskeleton.  Figure 5.   Rendered Impedances by CLEVERarm  V.   C ONCLUSION  This paper details the design process and features of the CLEVER ARM, a compact and low weight upper-limb exoskeleton for rehabilitation of stroke patients. R EFERENCES  [1]   B. R. Brewer, et. al, "Poststroke upper extremity rehabilitation: a review of robotic systems and clinical results,"   Topics in stroke rehab.   2014. [2]   H. I. Krebs, et. al, "3-D extension for MIT-MANUS: a robot- aided neuro-rehabilitation workstation," in ASME IDETC/CIE, DETC2000/MECH-14151, Baltimore, MD , 2000 [3]   T. Nef, M. Guidali, and R. Riener, "ARMin III – arm therapy exoskeleton with an ergonomic shoulder actuation,"   Applied Bionics and Biomechanics,   vol. 6, pp. 127-142, 2009 [4]   A. Zeiaee, et. al,   " Design and kinematic analysis of a novel upper limb exoskeleton for rehabilitation of stroke patients," in in IEEE-RAS-EMBS International Conference on Rehabilitation Robotic s , London, 2017, pp. 759-764. [5]   M. W. Spong, et al.,   Robot modeling and control . Vol. 3. New York: Wiley, 2006.  [6]   R. Soltani-Zarrin, et. al, "Reference path generation for upper- arm exoskeletons considering scapulohumeral rhythms," in in IEEE-RAS-EMBS International Conference on Rehabilitation Robotic s , London, 2017, pp. 753-758. [7]   R.   Soltani-Zarrin,   et.   al.,   "A   Computational   Approach   for Human-like Motion Generation in Upper Limb Exoskeletons Supporting   Scapulohumeral   Rhythms",   in   2017   IEEE  International   Symposium   on   Wearable   &   Rehabilitation Robotics (WeRob), Houston, 2017. [8]   C.   Carignan,   et   al.   "A   configuration-space   approach   to controlling   a   rehabilitation   arm   exoskeleton." Rehabilitation Robotics,   2007.   ICORR   2007.   IEEE   10th   International Conference on. IEEE, 2007. [9]   C.   Carignan,   et.   al.   "Controlling   shoulder   impedance   in   a rehabilitation arm exoskeleton." Robotics and Automation, 2008. ICRA 2008. IEEE International Conference on. IEEE, 2008.