Terrain characterisation and gait adaptation by a hexapod robot

Abstract

Legged robots are especially suited for traversing unstructured terrain outdoors. To maintain efficient locomotion, it is necessary to use appropriate gait parameters informed by terrain interactions such as slipping and sinking. Humans and legged animals inherently use this skill when walking. Effectively characterising the terrain with proprioceptive sensing provides information to the robot control system that can be used to inform a gait optimisation algorithm. This paper presents a novel methodology for any legged robot to characterise the terrain it is traversing in real time. This is achieved by introducing the concept of a timeinvariant gait phase. We also develop a method for on-line adaptation of gait parameters such as leg height and stride length based on these terrain characteristics. An experimental evaluation of the real time terrain characterisation methodology and gait adaptation system is presented. The experimental results show that through gait adaptation, the cost of transport can be reduced by up to 69.9% on a carpeted surface and 41.66% on a mulch-covered surface.