A Simulator for Testing Planar Upper Extremity Rehabilitation Robot Control Algorithms

Abstract

In this study, we took advantage of the emergence of accurate biomechanical human hand models to develop a system in which the interaction between a human arm and a rehabilitation robot while performing a planar trajectory tracking task can be simulated. Seven biomechanical arm models were based on the 11-degree-of-freedom Dynamic Arm Simulation model and implemented in OpenSim. The model of the robot was developed in MatlabSimulink and interaction between the arm and robot models was achieved using the OpenSim API. The models were tested by simulating the performance of each model while moving the end effector of a simulated planar robot model through an elliptical trajectory with an eccentricity of 0.94. Without assistance from the robot, the average root-mean-square error (RMSE) for all subjects was 3.98 mm. With the simulated robot providing assistive torque, the average RMSE error reduced to 2.88 mm. The test was repeated after modifying the length of the robot links, and an average RMSE of 2.91 mm recorded. A single-factor ANOVA test revealed that there was no significant difference in the RMSE for the two different robot geometries (p-value = 0.479), revealing that the simulator was not sensitive to robot geometry.