Estimation of modeling errors for robot manipulators using genetic algorithm

摘要

In this paper, we propose an entimation algorithm of modeling errors for robot manipulators using a genetic algorithm. First, we compensate nonlinear terms that can be determined by a nominal model of a robot manipulator. Then we construct a linearized model. However, this linearized model includes modeling errors, because a nonlinear compensator does not suppress variations of physical parameters and an unknown nonlinearity of the actuator in a general. The estimation algorithm of modeling errors consists of the disturbance observer process (process A) and genetic algorithm process (process B). Supposing that both processes A and B are stable, the real trajectory approaches the reference trajectory upon repetition of these processes. We can obtain the modeling errors from the result of the repetition. Finally, the effectiveness of the proposed algorithm is confirmed by numerical simulations and from experimental results of a 2-degree-of-freedom manipulator.