Hardware Conditioned Policies for Multi-Robot Transfer Learning

TL;DR

This paper introduces Hardware Conditioned Policies, a universal reinforcement learning approach that enables zero-shot transfer and efficient fine-tuning of robotic policies across diverse hardware configurations.

Contribution

The paper presents a novel hardware-conditioned policy framework that generalizes across different robot morphologies and dynamics, reducing training costs for new robots.

Findings

Zero-shot transfer to unseen robots using hardware encoding.

Fine-tuning is more sample-efficient than training from scratch.

Hardware embeddings improve policy generalization.

Abstract

Deep reinforcement learning could be used to learn dexterous robotic policies but it is challenging to transfer them to new robots with vastly different hardware properties. It is also prohibitively expensive to learn a new policy from scratch for each robot hardware due to the high sample complexity of modern state-of-the-art algorithms. We propose a novel approach called \textit{Hardware Conditioned Policies} where we train a universal policy conditioned on a vector representation of robot hardware. We considered robots in simulation with varied dynamics, kinematic structure, kinematic lengths and degrees-of-freedom. First, we use the kinematic structure directly as the hardware encoding and show great zero-shot transfer to completely novel robots not seen during training. For robots with lower zero-shot success rate, we also demonstrate that fine-tuning the policy network is significantly more sample-efficient than training a model from scratch. In tasks where knowing the agent dynamics is important for success, we learn an embedding for robot hardware and show that policies conditioned on the encoding of hardware tend to generalize and transfer well. The code and videos are available on the project webpage: https://sites.google.com/view/robot-transfer-hcp.