Heshan Li
Papers
4
Total Citations
22
H-Index
3
About
Heshan Li is a robotics researcher specializing in visual place recognition (VPR) and autonomous navigation, with a focus on enabling mobile robots to operate reliably in large-scale, real-world environments. His work addresses critical challenges in VPR—such as appearance changes, computational efficiency, and map scalability—by developing innovative deep learning architectures and mapping pipelines. Li’s most cited paper, "LSDNet" (2022, 12 citations), introduces a lightweight self-attentional distillation network that balances high performance with model efficiency, a key contribution for resource-constrained robots. He further advanced the field with "AdaptSeqVPR" (2023, 4 citations), an adaptive sequence-based pipeline that improves loop closure detection, and "CAHIR" (2023, 2 citations), a co-attentive hierarchical framework that unifies global and local image descriptors for robust place recognition under severe appearance changes. In "C-TM" (2022, 4 citations), Li tackled practical deployment by developing a topo-metric mapping system for delivery robots on footpaths, intelligently storing expensive LIDAR data only at key locations. Collectively, his work bridges the gap between high-accuracy VPR algorithms and the real-world demands of lightweight, lifelong robot operation.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4