Papers
6
Total Citations
46
H-Index
4
About
Changhe Tu is a researcher whose work bridges robotics, geometric computing, and artificial intelligence. His key research areas include multi-agent navigation, geometric arrangement analysis, and robotic grasping. Tu’s most notable contribution is a decentralized, learning-based solution for unlabeled multi-agent navigation in obstacle-rich environments, using graph neural networks to simultaneously solve goal assignment, collision avoidance, and path planning—a problem that has garnered 16 citations for its practical impact. He also developed a complete classification and efficient determination method for arrangements formed by two ellipsoids, a foundational tool for motion planning in CAD/CAM and robotics (12 citations). In robotic manipulation, Tu introduced the concept of “Caging Loops” in shape embedding space, enabling the synthesis of feasible caging grasps without relying on surface geometry (5 citations). His work on power grasp planning, reformulated as an infinite program under complementary constraints, further advances optimization-based grasp synthesis. With additional contributions to indoor visual localization (ReLoc) and hierarchical sitemaps, Tu’s research consistently addresses complex spatial reasoning problems, making him a notable figure in computational robotics and geometric design.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Caging Loops in Shape Embedding Space: Theory and Computation5 citations · 2018
- 4
- 5
- 6Caging Loops in Shape Embedding Space: Theory and Computation4 citations · 2018