Neural dynamic programming in reactive navigation of wheeled mobile robot

Abstract

In the article a new approach to a reactive navigation of a wheeled mobile robot (WMR), using a neural dynamic programming algorithm (NPD), is presented. A proposed discrete hierarchical control system consists of a trajectory generator and a tracking control system. In the trajectory generator we used a sensor-based approach to path design for the WMR in an unknown 2-D environment with static obstacles. The main part of the navigator is an action dependant heuristic dynamic programming algorithm (ADHDP), that generates control signals used to design a collision-free trajectory, that makes reaching a goal possible. ADHDP is the discrete algorithm of actor-critic architecture, that works on-line and does not require a preliminary learning or a controlled system knowledge. The tracking control system realises the generated trajectory, it consists of dual-heuristic dynamic programming (DHP) structure, PD controller and the supervisory term derived from the Lyapunov stability theorem. Computer simulations have been conducted to illustrate the performance of the algorithm. © 2012 Springer-Verlag Berlin Heidelberg.