A Physical Imitation Learning Pipeline for Energy-Efficient Quadruped Locomotion Assisted by Parallel Elastic Joint

2026-04-01

AI Summary

Traditional robotic control often suppresses intrinsic body dynamics, hindering energy-efficient, animal-like locomotion. This work proposes Physical Imitation Learning (PIL), a novel approach that, for the first time, distills reinforcement learning policies into passive body behaviors executable by parallel elastic joints, which operate in concert with residual motor control—enabling brain-body co-design without joint optimization. Evaluated on a simulated quadrupedal robot, the method demonstrates that up to 87% of mechanical power on flat terrain and 18% on rough terrain can be passively provided by elastic joints, substantially reducing overall energy consumption. These results establish PIL as a new paradigm for achieving highly efficient biomimetic locomotion.

Abstract

Due to brain-body co-evolution, animals' intrinsic body dynamics play a crucial role in energy-efficient locomotion, which shares control effort between active muscles and passive body dynamics -- a principle known as Embodied Physical Intelligence. In contrast, robot bodies are often designed with one centralised controller that typically suppress the intrinsic body dynamics instead of exploiting it. We introduce Physical Imitation Learning (PIL), which distils a Reinforcement Learning (RL) control policy into physically implementable body responses that can be directly offloaded to passive Parallel Elastic Joints (PEJs), enabling therefore the body to imitate part of the controlled behaviour. Meanwhile, the residual policy commands the motors to recover the RL policy's performance. The results is an overall reduced energy consumption thanks to outsourcing parts of the control policy to the PEJs. Here we show in simulated quadrupeds, that our PIL approach can offloads up to 87% of mechanical power to PEJs on flat terrain and 18% on rough terrain. Because the body design is distilled from -- rather than jointly optimised with -- the control policy, PIL realises brain-body co-design without expanding the search space with body design parameters, providing a computationally efficient route to task-specific Embodied Physical Intelligence applicable to a wide range of joint-based robot morphologies.

Problem

• Energy-Efficient Locomotion
• Embodied Physical Intelligence
• Parallel Elastic Joints
• Quadruped Robot
• Body Dynamics

Innovation

• Physical Imitation Learning
• Parallel Elastic Joints
• Embodied Physical Intelligence
• Energy-Efficient Locomotion
• Reinforcement Learning Distillation