Gait Motion Planning for a Six Legged Robot Based on the Associatron

Abstract

This research addresses a gait motion planning problem for a six-legged robot walking on an irregular field. In this proposal, we used a simplified neural network model called an Associatron that recalls total motion patterns sequentially from partial information. The Associatron is used here because it is more effective and adaptable than conventional methods. Using the proposed method, the robot is expected to walk in unknown fields. After verifying planning using an Open Dynamics Engine (ODE) by using simulations, we found that memorized patterns are recalled from developed patterns. We then conducted experiments using a real developed robot. Experiment results show that, when using the proposed planning method, the robot selects suitable gait motion patterns in the presence of an obstacle.