遺伝アルゴリズムによるロボットマニピュレータのモデル化誤差推定とその設定パラメータに関する一考察

概要

論文の詳細を見る
In this paper, we propose a tuning method of parameters for genetic algorithm in an estimation algorithm of modeling errors. First, we explain compensating nonlinear terms that can be determined by a nominal model of a multi-degree-of-freedom manipulator. Then we construct a linearized model. The modeling errors are defined by difference of input torques between a real manipulator and the linearized model. Second, we explain an estimation algorithm of the modeling errors. It consists of the following two processes : one is disturbance obserber process (process A) using an inverse model of a nominal model and the other is genetic algorithm process (process B) using the results obtained by process A. Third, A problem of process B is a tuning method of parameters that are weight matrix of fitness function, crossover and mutation probabilities of a genetic algorithm. We propose a tuning method of these parameters using numerical simulations. Finally, the effectiveness of the proposed the tuning method of parameters is confirmed by numerical simulations of the 2-degree-of-freedom manipulator.
岡山理科大学の論文