Yizhe Wu

University of Oxford, Technical University of Munich, Robotics Research (United States)

Papers

Total Citations

H-Index

About

Yizhe Wu is a robotics researcher whose work sits at the intersection of unsupervised learning, 3D scene understanding, and bipedal locomotion. His key contributions focus on enabling robots to perceive and interact with the world through object-centric representations, moving beyond traditional synchronized control methods. In his highly cited work "APEX: Unsupervised, Object-Centric Scene Segmentation and Tracking for Robot Manipulation" (8 citations), Wu demonstrated how probabilistic latent-variable models can be used for unsupervised object detection and tracking, a significant step toward more adaptable robotic manipulation. His earlier research on "Time-variable, event-based walking control for biped robots" (8 citations) challenged conventional phase-based walking controllers, introducing biologically-inspired, event-driven strategies that make humanoid robots more resilient to uneven terrain. Most recently, with "DreamUp3D: Object-Centric Generative Models for Single-View 3D Scene Understanding and Real-to-Sim Transfer" (4 citations), Wu is pushing the boundaries of 3D scene understanding by combining real-time inference with accurate 6D pose estimation and object reconstruction. This work directly addresses the critical challenge of transferring learned skills from simulation to the real world, a key bottleneck in modern robotics.