Adaptive control of multiped robot

Abstract

In the paper, a logical-probabilistic method of adaptive control of modular systems is presented. It base on the functional similarity of modules, the logical-probabilistic algorithm of directed search of rules and joint training of control modules. Starting with the search for common rules for all modules of the control system then it upload them with a more specific ones in accordance with the ideas of semantical-probabilistic inference. With the use of an interactive 3D-simulator, successful experiments were conducted with virtual multiped robot mode. Experimental studies have shown that the proposed approach is quite effective and can be used to manage modular systems with a large number of degrees of freedom.