TAR: Teacher-Aligned Representations via Contrastive Learning for Quadrupedal Locomotion

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Mousa, Amr, Karavis, Neil, Caprio, Michele, Pan, Wei, Allmendinger, Richard
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914174199660544
author Mousa, Amr
Karavis, Neil
Caprio, Michele
Pan, Wei
Allmendinger, Richard
author_facet Mousa, Amr
Karavis, Neil
Caprio, Michele
Pan, Wei
Allmendinger, Richard
contents Quadrupedal locomotion via Reinforcement Learning (RL) is commonly addressed using the teacher-student paradigm, where a privileged teacher guides a proprioceptive student policy. However, key challenges such as representation misalignment between privileged teacher and proprioceptive-only student, covariate shift due to behavioral cloning, and lack of deployable adaptation; lead to poor generalization in real-world scenarios. We propose Teacher-Aligned Representations via Contrastive Learning (TAR), a framework that leverages privileged information with self-supervised contrastive learning to bridge this gap. By aligning representations to a privileged teacher in simulation via contrastive objectives, our student policy learns structured latent spaces and exhibits robust generalization to Out-of-Distribution (OOD) scenarios, surpassing the fully privileged "Teacher". Results showed accelerated training by 2x compared to state-of-the-art baselines to achieve peak performance. OOD scenarios showed better generalization by 40% on average compared to existing methods. Moreover, TAR transitions seamlessly into learning during deployment without requiring privileged states, setting a new benchmark in sample-efficient, adaptive locomotion and enabling continual fine-tuning in real-world scenarios. Open-source code and videos are available at https://amrmousa.com/TARLoco/.
format Preprint
id arxiv_https___arxiv_org_abs_2503_20839
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TAR: Teacher-Aligned Representations via Contrastive Learning for Quadrupedal Locomotion
Mousa, Amr
Karavis, Neil
Caprio, Michele
Pan, Wei
Allmendinger, Richard
Robotics
Machine Learning
Systems and Control
Quadrupedal locomotion via Reinforcement Learning (RL) is commonly addressed using the teacher-student paradigm, where a privileged teacher guides a proprioceptive student policy. However, key challenges such as representation misalignment between privileged teacher and proprioceptive-only student, covariate shift due to behavioral cloning, and lack of deployable adaptation; lead to poor generalization in real-world scenarios. We propose Teacher-Aligned Representations via Contrastive Learning (TAR), a framework that leverages privileged information with self-supervised contrastive learning to bridge this gap. By aligning representations to a privileged teacher in simulation via contrastive objectives, our student policy learns structured latent spaces and exhibits robust generalization to Out-of-Distribution (OOD) scenarios, surpassing the fully privileged "Teacher". Results showed accelerated training by 2x compared to state-of-the-art baselines to achieve peak performance. OOD scenarios showed better generalization by 40% on average compared to existing methods. Moreover, TAR transitions seamlessly into learning during deployment without requiring privileged states, setting a new benchmark in sample-efficient, adaptive locomotion and enabling continual fine-tuning in real-world scenarios. Open-source code and videos are available at https://amrmousa.com/TARLoco/.
title TAR: Teacher-Aligned Representations via Contrastive Learning for Quadrupedal Locomotion
topic Robotics
Machine Learning
Systems and Control
url https://arxiv.org/abs/2503.20839