Adaptive Controller Algorithm for 2-DOF Humanoid Robot Arm

摘要

A computational model of human motor control for a nonlinear 2 degrees-of-freedom (DOF) robot arm to mimic humanlike behavior is developed and presented in this paper. The model is based on a simple mathematical model of a 2-segment compound pendulum which mimics the human upper arm and forearm. Using the Lagrangian and Euler-Lagrange equations, the 2-DOF dynamic equations were successfully derived and solved using Euler's method. Two types of controllers; a feedback Proportional-Derivative (PD) controller and a feedforward controller, were combined into the model. The algorithm exhibited learning of the necessary torque required in performing the desired Position Control via Specific Trajectory (PCST) rehabilitative task via feedback control and using it as the feedforward torque in subsequent trial motions. After 30 trials, the mean absolute error with respect to the desired motion of the upper arm, showed a decrease from 0.09533 to 0.005859, and the forearm motion from 0.3526 to 0.006138. This decrement trend in mean absolute errorwith increase in number of trials is consistent with the adaptive control strategy of the human arm known as the Feedback Error Learning (FEL) strategy.