Robust Quadruped Locomotion via Evolutionary Reinforcement Learning

TL;DR

This paper demonstrates that combining evolutionary search with reinforcement learning enhances the robustness of quadruped locomotion policies, especially in unseen rough terrains, outperforming standard deep RL methods.

Contribution

It evaluates and compares evolutionary reinforcement learning methods with standard deep RL, showing improved transfer performance to rough terrains.

Findings

CEM-TD3 achieves the highest reward during training and testing.

Evolutionary variants retain capabilities on rough terrain better than standard RL.

Incorporating evolutionary search reduces overfitting and improves robustness.

Abstract

Deep reinforcement learning has recently achieved strong results in quadrupedal locomotion, yet policies trained in simulation often fail to transfer when the environment changes. Evolutionary reinforcement learning aims to address this limitation by combining gradient-based policy optimisation with population-driven exploration. This work evaluates four methods on a simulated walking task: DDPG, TD3, and two Cross-Entropy-based variants CEM-DDPG and CEM-TD3. All agents are trained on flat terrain and later tested both on this domain and on a rough terrain not encountered during training. TD3 performs best among the standard deep RL baselines on flat ground with a mean reward of 5927.26, while CEM-TD3 achieves the highest rewards overall during training and evaluation 17611.41. Under the rough-terrain transfer test, performance of the deep RL methods drops sharply. DDPG achieves -1016.32 and TD3 achieves -99.73, whereas the evolutionary variants retain much of their capability. CEM-TD3 records the strongest transfer performance with a mean reward of 19574.33. These findings suggest that incorporating evolutionary search can reduce overfitting and improve policy robustness in locomotion tasks, particularly when deployment conditions differ from those seen during training.