<ul><li>Reinforcement learning has shown potential in training legged robots for complex locomotion behaviors.</li><li>Policies trained in simulation struggle to transfer to real-world robots due to unrealistic assumptions.</li><li>Traditional methods penalize aggressive motions, but require extensive tuning.</li><li>This work proposes Spectral Normalization as an efficient method to enforce Lipschitz continuity and reduce memory usage.</li></ul>

Spectral Normalization for Lipschitz-Constrained Policies on Learning Humanoid Locomotion

Discover more