Tian Zheng

Tsinghua University, University College London, Technical University of Munich

Papers

Total Citations

404

H-Index

About

Tian Zheng is a researcher at the forefront of 3D computer vision and intelligent robotic control, whose work bridges the gap between perception and action in complex environments. His most impactful contribution is **OccuSeg**, an occupancy-aware 3D instance segmentation framework that has garnered over 250 citations. This work addresses a critical challenge in robotics and augmented reality by moving beyond 2D images to directly interpret metric 3D scenes, enabling robots to understand object-level geometry without the ambiguities of occlusion or scale. Zheng’s influence extends to reinforcement learning, where his work on **Expert Iteration (ExIt)**—a method combining deep learning with tree search—has been cited over 130 times for its novel approach to sequential decision-making in structured prediction and game playing. More recently, his research has advanced into practical robotics with a focus on robust, incremental model predictive control for flexible-joint robots, leveraging time-delay estimation to handle uncertain dynamics efficiently. By integrating deep learning, planning, and control, Zheng’s work provides a powerful toolkit for building autonomous systems that can both perceive and act in the real world.