mlsanityData to 14 Oct 2026

Learning Quadruped Walking from Seconds of Demonstration

arXiv 2603.06961 ↗First posted 7 Mar 2026cs.LG cs.AIcited by 0

On the map: Offline RL and imitation learning from data → Real-world robot offline RL

Abstract (arXiv)

Quadruped locomotion provides a natural setting for understanding when model-free learning can outperform model-based control design, by exploiting data patterns to bypass the difficulty of optimizing over discrete contacts and the combinatorial explosion of mode changes. We give a principled analysis of why imitation learning with quadrupeds can be inherently effective in a small data regime, based on the structure of its limit cycles, Poincar'e return maps, and local numerical properties of neural networks. The understanding motivates a new imitation learning method that regulates the alignment between variations in a latent space and those over the output actions. Hardware experiments confirm that a few seconds of demonstration is sufficient to train various locomotion policies from scratch entirely offline with reasonable robustness.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Real-world robot offline RL, one of 13 directions in Offline RL and imitation learning from data. Builds on 12 of its 21 references that are in the map. First posted on arXiv 7 Mar 2026.

Builds on

12 of 21 references.

References · 12

  1. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,568
  2. 2021Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learningcited by 512
  3. 2020Learning Agile Robotic Locomotion Skills by Imitating Animalscited by 251
  4. 2024Deep Reinforcement Learning for Robotics: A Survey of Real-World Successescited by 176
  5. 2019Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Controlcited by 54
  6. 2024DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Offline Datasetscited by 49
  7. 2019Towards Learning to Imitate from a Single Video Demonstrationcited by 7
  8. 2024SDS -- See it, Do it, Sorted: Quadruped Skill Synthesis from Single Video Demonstrationcited by 4
  9. 2019From Bipedal Walking to Quadrupedal Locomotion: Full-Body Dynamics Decomposition for Rapid Gait Generationcited by 1
  10. 2024Activation-Descent Regularization for Input Optimization of ReLU Networkscited by 1
  11. 2025GRoQ-LoCO: Generalist and Robot-agnostic Quadruped Locomotion Control using Offline Datasetscited by 1
  12. 2025Simultaneous Contact Sequence and Patch Planning for Dynamic Locomotioncited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Offline Adaptation of Quadruped Locomotion using Diffusion Models
  2. 2025Learning to Walk With Less: A Dyna-Style Approach to Quadrupedal Locomotion
  3. 2025Gait in Eight: Efficient On-Robot Learning for Omnidirectional Quadruped Locomotion
  4. 2025RANDPOL: Parameter-Efficient End-to-End Quadruped Locomotion via Randomized Policy Learning
  5. 2022Human Motion Control of Quadrupedal Robots using Deep Reinforcement Learning
  6. 2024MetaLoco: Universal Quadrupedal Locomotion with Meta-Reinforcement Learning and Motion Imitation
  7. 2023A Real-World Quadrupedal Locomotion Benchmark for Offline Reinforcement Learning
  8. 2025Training Directional Locomotion for Quadrupedal Low-Cost Robotic Systems via Deep Reinforcement Learning