Mingjie Pan
Papers
3
Total Citations
70
H-Index
2
About
Mingjie Pan is a rising researcher at the forefront of embodied AI, with key contributions spanning 3D scene perception and generalizable robotic manipulation. Pan’s most impactful work, *RenderOcc* (2024, 59 citations), introduces a paradigm shift in vision-centric 3D occupancy prediction. By replacing costly 3D voxel annotations with 2D rendering supervision, this method dramatically reduces annotation expense while maintaining high-fidelity semantic scene understanding—a critical enabler for autonomous driving and robot perception. Building on this foundation, Pan’s *OmniManip* (2025) tackles the grand challenge of general robotic manipulation in unstructured environments. This work bridges the gap between high-level reasoning in Vision-Language Models and the fine-grained 3D spatial understanding required for precise manipulation, proposing object-centric interaction primitives as spatial constraints. Together, these contributions demonstrate a coherent research arc: from efficient 3D scene representation to actionable robotic control. With an emerging citation footprint and work appearing at top venues, Pan is establishing a reputation for solving fundamental bottlenecks in embodied perception and manipulation, making these papers essential reading for researchers in robotics and autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3