Energy Management Through Footstep Selection For Bipedal Robots

Abstract

This work proposes a method of footstep placement that controls system energy to enable a dynamically-safe walking behavior. Contrasting many other works that treat rough terrain as a series of disturbances that need to be mitigated with control, we provide some insight into how energy-targeted foot placement is enough to allow a passive system to transit over rough terrain. This work explores the underlying complexities of one of the simplest walking models, the inverted pendulum, which, in its various forms, is the skeleton behind all bipedal robots, from Asimo to Atlas. Troubling all of these humanoids is the foot placement problem, especially when the terrain is not flat. This work uses analysis of the system energy to divide the feasible stepping area into regions that would either enable dynamic walking or cause a fall. Second we subdivide the walking region into sectors that promote the accumulation or dissipation of energy, stimulating or inhibiting future steps. Third, we introduce a method of global energy management using a moving reference point over rough terrain. We present results on how these concepts can be used to prevent falls, accumulate energy to cross gaps, and even enable a passive system to walk uphill.