Passive Velocity Field Control of a Forearm-Wrist Rehabilitation Robot Ahmetcan Erdogan and Aykut Cihan Satici and Volkan Patoglu Faculty of Engineering and Natural Sciences Sabancı University, ˙ Istanbul, Turkey Email:{ahmetcan,acsatici,vpatoglu}@sabanciuniv.edu Abstract—This paper presents design, implementation and control of a 3RP S-R exoskeleton, specifically built to impose targeted therapeutic exercises to forearm and wrist. Design of the exoskeleton features enhanced ergonomy, enlarged workspace and optimized device performance when compared to previous versions of the device. Passive velocity field control (PVFC) is implemented at the task space of the manipulator to provide assistance to the patients, such that the exoskeleton follows a desired velocity field asymptotically while maintaining passivity with respect to external applied torque inputs. PVFC is aug- mented with virtual tunnels and resulting control architecture is integrated into a virtual flight simulator with force-feedback. Experimental results are presented indicating the applicability and effectiveness of using PVFC on 3RP S-R exoskeleton to deliver therapeutic movement exercises. I. I NTRODUCTION Robotic devices designed for physical rehabilitation are becoming ubiquitous, since they decrease the cost of repet- itive movement therapies, enable qualitative measurement of progress and promise development of novel rehabilitation protocols. Early robotic devices for upper limb rehabilitation had primarily focused on proximal joints such as shoulder and elbow, while recently the attention has been shifting towards more distal joints, such as the wrist and the hand. In the literature, several robotic devices have been developed to target wrist rehabilitation exercises. The most commonly used wrist rehabilitation devices are developed as extension modules of task-space arm rehabilitation systems. Once such device is the wrist extension module of the MIT-Manus system [1], [2]. This wrist module comprises of an actuated cardan joint coupled to a curved slider and allows for 3 degrees-of-freedom (DoF) forearm-wrist movements. Another wrist module exists as a part of the Robotherapist upper- extremity rehabilitation support system [3]. This system is capable of controlling all forearm-wrist rotations utilizing ER actuators [4]. Another task-space rehabilitation device, haptic knob, has been proposed by Dovat et al. to target com- bined wrist-hand therapy [5]. Haptic knob is a 2 DoF back- driveable mechanism, with one rotation assigned for wrist movements [6]. Even though task-space arm rehabilitation systems are practical and simpler to implement, these devices cannot guarantee decoupled actuation and measurement of human joint motions. Exoskeleton type rehabilitation devices are relatively more complex but can be effectively used for the implementation and measurement of targeted joint movements. There exist several upper-extremity rehabilitation systems that include forearm-wrist rotations. Armin and IntelliArm are two ex- oskeleton type full-arm therapy systems, which allow for forearm supination/pronation as well as the palmar/dorsal flexion of the wrist [7], [8]. These systems are also equipped with a multi-axis force sensors to collect force/torque data during therapy. RiceWrist is another exoskeleton designed to target physical rehabilitation of forearm-wrist motions [9], [10]. This device is of 3RP S-R kinematical structure and possesses 4 DoF [11]. With RiceWrist, all forearm and wrist motions can be independently controlled over their rotational axes. RiceWrist has also been extended to deliver full arm rehabilitation therapy, through synchronized control of this device with the MIME system [9]. Earlier implementations of rehabilitation robots relied on stiff position controllers that impose predetermined trajectories to patients [12]. In these controllers patient forces were viewed as disturbances and counteracted. However, clinical studies provided strong evidence that active participation of patients is crucial for increasing efficacy of robotic rehabilitation. Consequently, impedance/admittance control techniques and more recently patient-cooperative methods have been proposed to allow active participation of patients in robotic therapy [13], [14], [15], [16]. The main idea in patient-cooperative methods is to adjust the assistance provided to the patient based on patient’s performance. For instance, in [17], time-optimal trajectories are calculated for point-to-point reaching tasks and tracking errors are penalized adaptively based on the deviation from the optimal trajectory. This way rehabilitation robots can “assist-as-needed”, where the main contribution for a successful completion of the task is left to the patient. Unfortunately, many of the existing implementations of assist- as-needed protocols rely on minimization of trajectory tracking error, which cannot ensure that patients do not significantly deviate from the pre-determined path [18], [19]. Virtual tunnel approach is an alternative way to provide freedom to patients by allowing them to move freely as long as they do not violate the bounds defining forbidden regions. Since by merely implementing virtual tunnels cannot ensure that the tunnel is traced within a reasonable amount of time, virtual tunnels are generally augmented with a moving window that pushes the patient forward in the tunnel if he/she falls behind the pre- determined timing along the tunnel [20], [21], [22]. However, 2011 IEEE International Conference on Rehabilitation Robotics Rehab Week Zurich, ETH Zurich Science City, Switzerland, June 29 - July 1, 2011 978-1-4244-9862-8/11/$26.00 ©2011 IEEE