Max Robotics — Global Service Robot Directory + US Certification

About

Pengxiang Ding is a researcher whose work spans computer vision, robotics, and embodied artificial intelligence, with a particular focus on human motion prediction, legged robot locomotion, and vision-language-action (VLA) models. His early contribution, *TrajectoryCNN* (2020), established his reputation in spatio-temporal feature learning by introducing a novel trajectory space for human pose prediction, accumulating over 113 citations and demonstrating his ability to bridge computer vision with real-world motion modeling. More recently, Ding has pivoted toward the frontier of generalist robotics, developing systems that enable quadruped robots to reason, adapt, and act in complex environments. His *QUAR-VLA* framework integrates vision-language models with robotic action policies, while *GeRM* leverages mixture-of-experts architectures for multi-task quadruped learning. Further work on *RL2AC* addresses robust locomotion through rapid online adaptive control, and *MoRE* and *PD-VLA* push scalability and efficiency in VLA systems. Collectively, Ding's research reflects a coherent trajectory from foundational motion understanding toward deployable, intelligent robotic systems—making his work essential reading for anyone exploring the intersection of large language models, reinforcement learning, and real-world robot deployment.