A Real-Time Planning and Control Framework for Robust and Dynamic Quadrupedal Locomotion

摘要

Abstract Legged locomotion poses significant challenges due to its nonlinear, underactuated and hybrid dynamic properties. These challenges are exacerbated by the high-speed motion and presence of aerial phases in dynamic legged locomotion, which highlights the requirement for online planning based on current states to cope with uncertainty and disturbances. This article proposes a real-time planning and control framework integrating motion planning and whole-body control. In the framework, the designed motion planner allows a wider body rotation range and fast reactive behaviors based on the 3-D single rigid body model. In addition, the combination of a Bézier curve based trajectory interpolator and a heuristic-based foothold planner helps generate continuous and smooth foot trajectories. The developed whole-body controller uses hierarchical quadratic optimization coupled with the full system dynamics, which ensures tasks are prioritized based on importance and joint commands are physically feasible. The performance of the framework is successfully validated in experiments with a torque-controlled quadrupedal robot for generating dynamic motions.