Autonomous Gait Planning for a Hexapod Robot in Unstructured Environments based on 3D Terrain Perception

Legged robots have extremely important applications in unstructured environments, e.g. accomplishing rescuing and detecting tasks in the earthquake and nuclear accidents instead of human beings. In such environments, legged robots must have the potential to plan gaits autonomously based on the site environment without detailed instructions from a human operator. In this paper, we present an autonomous gait planning method for the hexapod robot “Octopus”, which will be applied in unstructured environments. The method is based on the terrain map perceived by a 3D vision system. Further, two indices are defined to describe terrain conditions. An important part of the method is a foothold selection function, which considers the influence of terrain conditions. The detailed walking process of Octopus on the rough terrain is discussed too. At last, we apply our method to Octopus, and show that it allows Octopus to pass through the uneven terrain autonomously.