Gait Optimization through Search

Abstract

We present a search-based method for the generation of a terrain-adaptive optimal gait of a six-legged walking machine. In this, several heuristic rules have been proposed to reduce the search effort. We identify the useful support states of the machine and form a table to indicate for each of these states the list of other states to which a transition can be made. This helps in converging to and maintaining a periodic gait through a limited search while retaining adequate options to deviate from such a gait as and when needed. The criterion for optimization is coded into a function that evaluates the promise of a node in the search graph. We have shown how this function may be designed to generate the common periodic gaits like the wave gait, the equal phase gait, and the follow-the-leader gait. The purpose is to demonstrate that the proposed method is sufficiently general and can cater to a wide range of optimizing requirements.