Rolling in the Deep -- Hybrid Locomotion for Wheeled-Legged Robots using Online Trajectory Optimization

Wheeled-legged robots have the potential for highly agile and versatile\nlocomotion. The combination of legs and wheels might be a solution for any\nreal-world application requiring rapid, and long-distance mobility skills on\nchallenging terrain. In this paper, we present an online trajectory\noptimization framework for wheeled quadrupedal robots capable of executing\nhybrid walking-driving locomotion strategies. By breaking down the optimization\nproblem into a wheel and base trajectory planning, locomotion planning for high\ndimensional wheeled-legged robots becomes more tractable, can be solved in\nreal-time on-board in a model predictive control fashion, and becomes robust\nagainst unpredicted disturbances. The reference motions are tracked by a\nhierarchical whole-body controller that sends torque commands to the robot. Our\napproach is verified on a quadrupedal robot with non-steerable wheels attached\nto its legs. The robot performs hybrid locomotion with a great variety of gait\nsequences on rough terrain. Besides, we validated the robotic platform at the\nDefense Advanced Research Projects Agency (DARPA) Subterranean Challenge, where\nthe robot rapidly mapped, navigated and explored dynamic underground\nenvironments.\n