Flip Stunts on Bicycle Robots using Iterative Motion Imitation
On the map: Deep reinforcement learning methods → RL for robot locomotion and manipulation
Abstract (arXiv)
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.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in RL for robot locomotion and manipulation, one of 15 directions in Deep reinforcement learning methods. Builds on 41 of its 53 references that are in the map. First posted on arXiv 30 Mar 2026.
Builds on
41 of 53 references.
Closest ancestors · 3
Same direction, newest first
- 2025High-Performance Reinforcement Learning on Spot: Optimizing Simulation Parameters with Distributional Measures24 Apr
- 2024CaT: Constraints as Terminations for Legged Locomotion Reinforcement Learning27 Mar
- 2024Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal Locomotion Control30 Jan
Other references · 38
- 2017Proximal Policy Optimization Algorithmscited by 15,215
- 2017Deep reinforcement learning from human preferencescited by 3,912
- 2019AMASS: Archive of Motion Capture as Surface Shapescited by 1,236
- 2020Learning Quadrupedal Locomotion over Challenging Terraincited by 604
- 2019Learning agile and dynamic motor skills for legged robotscited by 504
- 2023Orbit: A Unified Simulation Framework for Interactive Robot Learning Environmentscited by 295
- 2019Extrapolating Beyond Suboptimal Demonstrations via Inverse Reinforcement Learning from Observationscited by 260
- 2020Learning Agile Robotic Locomotion Skills by Imitating Animalscited by 251
- 2021Robust Motion In-betweeningcited by 229
- 2023Perpetual Humanoid Control for Real-time Simulated Avatarscited by 218
- 2024OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learningcited by 189
- 2022Rapid Locomotion via Reinforcement Learningcited by 168
26 more
- 2023Extreme Parkour with Legged Robotscited by 164
- 2023ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robotscited by 152
- 2023Robot Parkour Learningcited by 149
- 2025ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skillscited by 130
- 2019Imitation Learning from Imperfect Demonstrationcited by 114
- 2018Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flowcited by 105
- 2025TWIST: Teleoperated Whole-Body Imitation Systemcited by 101
- 2025GMT: General Motion Tracking for Humanoid Whole-Body Controlcited by 83
- 2023DTC: Deep Tracking Controlcited by 56
- 2025KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skillscited by 52
- 2024Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robotscited by 48
- 2022Learning Agile Skills via Adversarial Imitation of Rough Partial Demonstrationscited by 45
- 2025Attention-Based Map Encoding for Learning Generalized Legged Locomotioncited by 40
- 2024Opt2Skill: Imitating Dynamically-feasible Whole-Body Trajectories for Versatile Humanoid Loco-Manipulationcited by 39
- 2022Advanced Skills through Multiple Adversarial Motion Priors in Reinforcement Learningcited by 37
- 2022OPT-Mimic: Imitation of Optimized Trajectories for Dynamic Quadruped Behaviorscited by 35
- 2021Learning from Imperfect Demonstrations from Agents with Varying Dynamicscited by 27
- 2023RL + Model-based Control: Using On-demand Optimal Control to Learn Versatile Legged Locomotioncited by 26
- 2025High-speed control and navigation for quadrupedal robots on complex and discrete terraincited by 19
- 2025SAIL: Faster-than-Demonstration Execution of Imitation Learning Policiescited by 18
- 2023Imitating and Finetuning Model Predictive Control for Robust and Symmetric Quadrupedal Locomotioncited by 10
- 2024Spatio-Temporal Motion Retargeting for Quadruped Robotscited by 8
- 2025Unsupervised Skill Discovery as Exploration for Learning Agile Locomotioncited by 6
- 2026System Design of the Ultra Mobility Vehicle: A Driving, Balancing, and Jumping Bicycle Robotcited by 4
- 2023Reinforcement Learning for Legged Robots: Motion Imitation from Model-Based Optimal Controlcited by 3
- 2025Feature-Based vs. GAN-Based Learning from Demonstrations: When and Whycited by 2
Built on by
Citing papers per year: 2026: 2.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2024Path Following and Stabilisation of a Bicycle Model using a Reinforcement Learning Approach
- 2024I-CTRL: Imitation to Control Humanoid Robots Through Constrained Reinforcement Learning
- 2024Fusion Dynamical Systems with Machine Learning in Imitation Learning: A Comprehensive Overview
- 2026Learning Sidewalk Autopilot from Multi-Scale Imitation with Corrective Behavior Expansion
- 2022Human Motion Control of Quadrupedal Robots using Deep Reinforcement Learning
- 2025The Mini Wheelbot: A Testbed for Learning-based Balancing, Flips, and Articulated Driving
- 2026Learning Quadruped Walking from Seconds of Demonstration
- 2019Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learning