LEARNING BASED CONTROL SYSTEM OF FOUR-LEGGED ROBOT

Abstract cze:
Abstrakt: Possible discretization technique oj the continuous state space oj Jour-Iegged robot using simultaneous compositions oj behaviors is described in this contribution. Compositions are generated by the instances oj two basic controllers. The aim is to automatically develop the gait policy. Possible composition strategies are implemented through undeterministic state machine. In the machine design stage the number oj both machine states and transitions could be essetially reduced using Jormal considerations only. Furthermore machine transitions are evaluated by selected alternative oj Q-Iearning to incrementally improve the estimation oj the Juture benefit Jrom single transitions. This technique results in simultaneous activations oj controller compositions instances with the highest estimatio n oj the Juture benefit.

Abstract eng:
Abstract: Possible discretization technique oj the continuous state space oj Jour-Iegged robot using simultaneous compositions oj behaviors is described in this contribution. Compositions are generated by the instances oj two basic controllers. The aim is to automatically develop the gait policy. Possible composition strategies are implemented through undeterministic state machine. In the machine design stage the number oj both machine states and transitions could be essetially reduced using Jormal considerations only. Furthermore machine transitions are evaluated by selected alternative oj Q-Iearning to incrementally improve the estimation oj the Juture benefit Jrom single transitions. This technique results in simultaneous activations oj controller compositions instances with the highest estimatio n oj the Juture benefit.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket