Optimization and Learning for Rough Terrain Legged Locomotion
Abstract¶
This paper presents a novel approach to legged locomotion over rough terrain rooted in optimization. The approach relies on a hierarchy of fast, anytime algorithms to plan a set of footholds and the dynamic body motions required to execute them. Components within the planning framework coordinate to exchange plans, cost-to-go estimates, and certificates that ensure the output of an abstract high-level planner can be realized by lower layers of the hierarchy. The burden of careful engineering of cost functions is mitigated by a simple inverse optimal control technique, and robustness is achieved by real-time re-planning of the full trajectory, augmented by reflexes and feedback control. We demonstrate the approach guiding the LittleDog quadruped robot over various rough terrains.
Summary¶
Proposes a hierarchical optimization framework for quadruped locomotion over rough terrain, combining anytime footstep planning with dynamic body motion optimization. It uses inverse optimal control to reduce hand-tuning of costs and real-time replanning with reflexes and feedback to robustly guide the LittleDog robot across varied terrain.