Multimodal Locomotion and Operation for Biomimetic Legged Robots

This project studies heterogeneous reconfiguration and multimodal locomotion for biomimetic legged robots.

My work focuses on multimodal control algorithm development. I expanded the command space for multimodal motion, designed implicit state estimation with a variational autoencoder, trained reinforcement-learning policies for multimodal locomotion, and supported sim-to-real deployment.

The current capability targets complex-terrain walking, fall recovery, quadruped crawling, vehicle riding and separation, and other mode transitions on legged robot platforms.