Motion control and monitoring of robot arm via feedback error learning and statistic learning

点击翻译按钮获取中文摘要

The objectives of this research are 1) to expand/deepen feedback error learning as a scheme for adaptive learning control, and 2) to examine a statistic learning technique, which has made a lot of progress recently, from a dual viewpoint of feedback error leaning, thereby making full use of the both approaches complementary in order to attain motion control/monitering of robot arms. The former is successful in that seperating control/learning laws has enabled to delete the effect to closed-loop and that generalization to a single filter to a filter bank has lead to a powerful learning control scheme that can treat nonlinearity. On the other hand, concerning the latter, our simulation result has not given what we had expected in advance. We are currently investigating reasons why this does not work well.