Iterative Motion Imitation enables bicycle robots to perform unassisted front-flips by learning from initially 'impossible' reference motions.
March 31, 2026
Original Paper
Flip Stunts on Bicycle Robots using Iterative Motion Imitation
arXiv · 2603.27944
The Takeaway
It demonstrates that RL can bridge the gap between kinematically infeasible target motions and real-world dynamics. This is the first unassisted acrobatic behavior on a bicycle robot, showing a path toward extremely agile, dynamic robotics.
From the abstract
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-Mobilit