Max Robotics — Global Service Robot Directory + US Certification

About

Ziming He is a robotics researcher advancing the frontier of autonomous skill acquisition, with a focus on unsupervised reinforcement learning and human-robot interaction. His core contributions lie in developing algorithms that enable robots to discover and master locomotion skills without explicit task supervision—a critical capability for creating truly intelligent, adaptable machines. In his highly cited work “Robotic Locomotion Skill Learning Using Unsupervised Reinforcement Learning With Controllable Latent Space Partition” (11 citations), He introduces a method for learning task-agnostic skills that can be efficiently fine-tuned for complex, long-horizon tasks. He further pushes this paradigm with PDRL (Progressive Diversity Reinforcement Learning), which encourages robots to explore “deeper states” requiring extended action sequences, thereby unlocking more sophisticated behaviors. Complementing these theoretical advances, He also tackles practical imitation learning in “Behavior Cloning and Replay of Humanoid Robot via a Depth Camera” (4 citations), streamlining the process of teaching robots through observation. His work bridges the gap between unsupervised exploration and real-world robotic deployment, with recent investigations into segmented motion synthesis for task-oriented locomotion.