Mingliang Zhou

Xiaomi (China), Beijing Institute of Technology

Papers

Total Citations

H-Index

About

Mingliang Zhou is a roboticist pushing the boundaries of legged locomotion and dynamic manipulation. His research focuses on enabling agile, real-world capabilities for quadrupedal and humanoid robots, bridging the gap between simulation and physical deployment. Zhou’s major contributions include the development of **State Estimation Transformers (SET)**, a novel architecture that accurately predicts a robot’s privileged states—such as terrain contact and body dynamics—allowing quadrupeds to perform advanced skills like jumping in unstructured outdoor environments. He also pioneered a control pipeline for dynamic object-catching, enabling a quadruped to run and catch a thrown object (e.g., a frisbee) using stereo vision, expanding robots beyond pure locomotion into interactive tasks. Earlier work on the BHR-5 humanoid addressed impact motion control (running, jumping) via an energy integral method, tackling the challenge of high-force dynamic actions. Though early in his career, Zhou’s work has already garnered attention, with his 2024 SET paper and 2023 run-to-catch study each accumulating 3 citations, signaling growing impact in the field. His research is a key step toward robots that move with the agility and versatility of animals.