Computer Simulation-Based Synthesis of Learning-Control Law of Robots

The paper deals with computer simulation and synthesis of learning trajectory-tracking robot control based on the explicit form dynamic equations. Robots with inaccurate inertia parameters are considered. Two learning operators are proposed. The simplified one is a scalar function identical to the minimal eigenvalue of the inertial matrix of the dynamic equations. The complicated one is the inertial matrix itself. The learning operators are the basis of two learning-con trol laws, both with a feedback controller attached. For comparison of the results, a computer model of a virtual (test) robot with precise inertia parameters is assumed. The convergence of the learning procedure for both control laws proposed is examined by computer dynamics simulation or the resulting trajectory-tracking motion of the virtual robot. In spite of significant initial trajectory-tracking errors, good convergence of the learning procedures is achieved. The results are promising when the learning operator is in compliance with the robot inertial matrix.