Xingyuan Sun

Shanghai Jiao Tong University, Princeton University, Massachusetts Institute of Technology

Papers

Total Citations

214

H-Index

About

Xingyuan Sun is a roboticist whose research lies at the intersection of perception, manipulation, and multi-agent coordination. A central theme in Sun’s work is enabling robots to perceive and interact with the physical world more robustly. In their highly cited work on 3D shape perception (118 citations), Sun tackled the fundamental challenge of inferring accurate object geometry from limited visual data by integrating monocular vision with tactile sensing and shape priors, overcoming the inherent ambiguities of 2D-to-3D reconstruction. Sun also advanced mobile manipulation by introducing “Spatial Action Maps” (75 citations), a framework that moves beyond simple steering commands to learn spatially-aware, pixel-level action policies for navigation and manipulation. This work was extended to multi-agent settings with “Spatial Intention Maps,” enabling decentralized robots to communicate their goals for improved coordination. More recently, Sun has explored novel non-prehensile manipulation, using a mobile blower to pneumatically move scattered objects—a chaotic, contact-rich task requiring adaptive control. Across these contributions, Sun’s research consistently pushes toward more perceptive, adaptable, and collaborative robotic systems, with a strong focus on real-world physical interaction.