Full Professor, Xiamen University
3 papers at NeurIPS 2025
This work is the first to address the challenge of large-scale (over 100km^2) urban cross-view LiDAR-based place recognition with high-resolution remote sensing imagery.
This paper proposes GTR-Loc, the first text-assisted LiDAR localization framework that integrates geospatial text regularization into an SCR network to reduce localization ambiguities.
We propose PlanU, a method that enhances LLM-based decision-making under uncertainty by modeling value distributions via quantile regression and guiding MCTS exploration using a novel Upper Confidence Bounds with Curiosity (UCC) score.
