Flip Stunts on Bicycle Robots using Iterative Motion Imitation
This enables agile acrobatic behaviors for bicycle robots, representing an incremental advance in robot motion control.
The paper tackled the problem of performing front-flips on bicycle robots by proposing Iterative Motion Imitation (IMI), a reinforcement learning method that iteratively imitates infeasible reference motions to train policies, resulting in successful ground-to-ground and ground-to-table flips with higher success rates compared to single-shot imitation.
This work demonstrates a front-flip on bicycle robots via reinforcement learning, particularly by imitating reference motions that are infeasible and imperfect. To address this, we propose Iterative Motion Imitation(IMI), a method that iteratively imitates trajectories generated by prior policy rollouts. Starting from an initial reference that is kinematically or dynamically infeasible, IMI helps train policies that lead to feasible and agile behaviors. We demonstrate our method on Ultra-Mobility Vehicle (UMV), a bicycle robot that is designed to enable agile behaviors. From a self-colliding table-to-ground flip reference generated by a model-based controller, we are able to train policies that enable ground-to-ground and ground-to-table front-flips. We show that compared to a single-shot motion imitation, IMI results in policies with higher success rates and can transfer robustly to the real world. To our knowledge, this is the first unassisted acrobatic flip behavior on such a platform.