Combining Automated Planning and Hybrid Control: A Quadruped Bouncing Gait

This proposed thesis topic aims to combine automated planning and hybrid control techniques. In this paper we describe a particular case study in this direction: the development of a quadruped bouncing gait with four qualitative states, quad-stance, touch-off, flight, and touchdown. Our approach is novel in that gait achievement is defined in terms of flexible constraint windows in statespace and time. This key feature enables the use of a controller that chooses from entire sets of optimal joint motion trajectories at execution time instead of tracking a single pre-planned optimal trajectory with high-impedance. In simulation, we demonstrate the robustness of this approach by maintaining a steady-state quadruped bouncing gait despite significant disturbances at execution time. The focus of this thesis work will be to investigate techniques to interleave and concurrently execute such plans in order to achieve complex and abstract goal specifications.