Abstract— Admittance control allows the user to input a force and translates that force into motion. The positive features associated with admittance control include: force amplification, intuitive control, and proportional control. These features were implemented to develop a single DOF exoskeleton to provide grasp assistance through force application. I. INTRODUCTION Robotic control paradigms rely on two types of control systems: impedance control and admittance control. Admittance control allows the user to input a force, translating the force into motion. This motion can be represented by target velocity or a position. The positive features associated with an admittance control system for upper extremity rehabilitation include: force amplification, back-drivability, intuitive control and proportional control. Under admittance control, the user senses only the inertia of the small virtual mass which can be very small compared to the inertia and load capabilities of the actual robot actuators. This makes it useful for individuals with significant muscle weakness to employ their residual motor control to effectively drive a powerful robot [1]. These features reflect essential characteristics that could lead to the development of complementary control systems for rehabilitation. Rehabilitation robots rely on interactive control systems that react to inputs from the subject [1]. An admittance control system was implemented in a single DOF exoskeleton to provide grasp assistance. II. METHODS A single DOF finger exoskeleton for grasp was designed and printed in ABS plastic, Fig 1. To implement admittance control, a single DOF strain gauge load cell (Phidget Force Sensor) was installed on the index finger thimble aligned with the direction of finger flexion-extension. The solution to the differential equation defines the position and velocity associated with the applied force (F(x)) and applied damping B, for a virtual mass, M. The force data is collected and processed in MATLAB in real time at 100 Hz. The position is computed by solving the differential equation (Equation 1) using CVode (Ordinary Differential Equation Solver) developed at Eindhoven University [2]. Calculated position values are converted to motor positions and sent to a Robotis Dynamixel MX-64 motor to perform the desired command using onboard 1000 Hz PID controller. ሺሻ  ሺሻ   ሺሻ  Equation 1: Differential Equation for Position: F(x)= Force (N), I=Inertia (kg  ) , B = damping (Ns/m), x’(t)=velocity (m/s), x’’(t)=acceleration(m/  ) Research supported by Gustavus and Louise Pfeiffer Research Foundation. Kevin Abbruzesse, Kiran Karunakaran, Hao Xu and Richard Foulds are with the New Jersey Institute of Technology, Newark, NJ 07102 USA Figure 1: Top Left: Grasp Exoskeleton Top Right: Force sensor on dorsal side Bottom: Actual vs desired trajectory for 3 movements of 180 ° III. RESULTS AND DISCUSSION The outer admittance loop for the control of the movement runs at 100 Hz. The actual motor position closely followed by the desired position with a 10ms lag. The movement amplitude error was computed and it was found to increase with applied force. The increase in error is due to the motor not being able to reach the desired position within the 10ms time duration. This error is attributed to running the motor at a constant velocity (currently at half the rated velocity).This does not affect performance as the user cannot distinguish between operations as these errors are of small magnitude. Small forces were able to generate desired trajectories, demonstrating rehabilitation application for individuals with extreme muscle weakness. IV. CONCLUSION One DOF gripper was constructed and tested. The goal is to build a 4 DOF extension for the haptic master, a 3 DOF haptic device with admittance control. This would provide upper limb rehabilitation with the ability to provide distal as well as proximal rehabilitation with integrated haptics. REFERENCES [1] Hu, Xiaoling. "A Novel Continuous Intention-Driven Rehabilitation Robot and its Training Effectiveness." Biomechatronics in Medicine and Healthcare. Singapore: Pan Stanford Publishing, 2011. [2] Van Riel, Natal. "Speeding up simulations ofODE modelsin Matlab using CVode and MEX files." Eindhoven University of Technology (2012): A 1-DOF Admittance Control Hand Exoskeleton for Grasp Kevin M. Abbruzzese, Kiran K. Karunakaran, Hao Xu, Richard Foulds Ph.D