Efficient Morphology-Aware Policy Transfer to New Embodiments

2025-08-05

AI Summary

This study addresses the poor zero-shot transfer performance of morphology-aware policies to novel robotic morphologies. To avoid costly end-to-end fine-tuning, we propose a lightweight, parameter-efficient fine-tuning (PEFT)-based transfer method. Methodologically, we introduce— for the first time in morphology-aware policy transfer—input-learnable adapters, prefix tuning, and submodule-wise weight tuning, integrated within a multi-agent pretraining framework. Our approach updates fewer than 1% of the model parameters yet consistently outperforms zero-shot baselines. Experiments demonstrate substantial reductions in the number of samples required per target morphology and superior policy performance on unseen morphologies. The method achieves an effective balance between generalization capability and deployment efficiency, establishing a scalable new paradigm for cross-morphology policy reuse in embodied intelligence.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: Embodied AI
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Morphology-aware policy learning is a means of enhancing policy sample efficiency by aggregating data from multiple agents. These types of policies have previously been shown to help generalize over dynamic, kinematic, and limb configuration variations between agent morphologies. Unfortunately, these policies still have sub-optimal zero-shot performance compared to end-to-end finetuning on morphologies at deployment. This limitation has ramifications in practical applications such as robotics because further data collection to perform end-to-end finetuning can be computationally expensive. In this work, we investigate combining morphology-aware pretraining with parameter efficient finetuning (PEFT) techniques to help reduce the learnable parameters necessary to specialize a morphology-aware policy to a target embodiment. We compare directly tuning sub-sets of model weights, input learnable adapters, and prefix tuning techniques for online finetuning. Our analysis reveals that PEFT techniques in conjunction with policy pre-training generally help reduce the number of samples to necessary to improve a policy compared to training models end-to-end from scratch. We further find that tuning as few as less than 1% of total parameters will improve policy performance compared the zero-shot performance of the base pretrained a policy.

Problem

• Improve zero-shot performance of morphology-aware policies
• Reduce parameters needed for policy specialization
• Compare parameter-efficient finetuning techniques for online adaptation

Innovation

• Morphology-aware policy learning for data aggregation
• Parameter efficient finetuning (PEFT) techniques
• Tuning less than 1% of parameters improves performance