Learning Emergent Gaits with Decentralized Phase Oscillators: on the role of Observations, Rewards, and Feedback

Jenny Zhang, Steve Heim, Se Hwan Jeon, Sangbae Kim
2/13/2024
cs.ROcs.LG

No official code found. Best implementation: mit-biomimetics/ORCAgym (52 stars, last commit 2024-03-07).

Abstract

We present a minimal phase oscillator model for learning quadrupedal locomotion. Each of the four oscillators is coupled only to itself and its corresponding leg through local feedback of the ground reaction force, which can be interpreted as an observer feedback gain. We interpret the oscillator itself as a latent contact state-estimator. Through a systematic ablation study, we show that the combination of phase observations, simple phase-based rewards, and the local feedback dynamics induces policies that exhibit emergent gait preferences, while using a reduced set of simple rewards, and without prescribing a specific gait. The code is open-source, and a video synopsis available at https://youtu.be/1NKQ0rSV3jU.

View on arXivView PDF

Code Implementations(1)

mit-biomimetics/ORCAgymCommunity100%

phase-based Observations, Rewards, Coupling Ablation

5211Shell, MATLABSep 15, 20232 years agoNOASSERTION

Similar Papers

Identifying Important Sensory Feedback for Learning Locomotion Skills

82% match2023

DeepTransition: Viability Leads to the Emergence of Gait Transitions in Learning Anticipatory Quadrupedal Locomotion Skills

82% match2023

Learning Gait-Aware Quadruped Locomotion with Temporal Logic Specifications

82% match2026

Combining model-predictive control and predictive reinforcement learning for stable quadrupedal robot locomotion

81% match2023

Reinforcement Learning on Cost-Constrained Quadrupedal Hardware

80% match2026

Gait in Eight: Efficient On-Robot Learning for Omnidirectional Quadruped Locomotion

80% match1 repo142025

Cite this paper

@article{zhang2024learning,
  title  = {Learning Emergent Gaits with Decentralized Phase Oscillators: on the role of Observations, Rewards, and Feedback},
  author = {Jenny Zhang and Steve Heim and Se Hwan Jeon and Sangbae Kim},
  year   = {2024},
  eprint = {2402.08662},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2402.08662v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2402.08662.svg)](https://codeofpaper.com/paper/2402.08662)

Discussion