Navigation Planning for Legged Robots

Abstract

As legged robots gain the abilities to walk, run, and balance on more than just flat, obstacle-free floors, they grow closer to fulfilling the potential of legged locomotion shown by biological systems. To truly fulfill this potential these robots must successfully traverse complicated, rough terrain, requiring the robots to step onto or over various features of the environment. Furthermore, when a robot must spend a significant amount of time supported by a single foot, the contact that foot makes with the ground is very important for stability, requiring that the robot properly assess foot placement to maintain balance during locomotion. This thesis addresses the problem of navigation for legged robots by using a global planning approach built on top of existing walking and running controllers. The planning process reasons about foot placement and contact configurations within the terrain, and the connectivity of those contact configurations. This thesis provides a general global navigation strategy for a wide range of legged robots through complicated environments, utilizing the advantages of their legs as well as ensuring their safety and stability during locomotion