Feiqiang Lin
Papers
4
Total Citations
18
H-Index
3
About
Feiqiang Lin is a leading researcher in reinforcement learning (RL) for autonomous navigation, with a particular focus on creating safe, real-world-ready robotic systems. His work addresses a critical gap in the field: most RL-based navigation assumes perfect localisation, which is unrealistic for real-world deployment. Lin’s major contributions lie in developing "localisation-safe" RL frameworks that integrate sensor-based state estimation—such as LiDAR and visual odometry (VO)—directly into the learning process. His 2021 paper on RL-based mapless navigation with fail-safe localisation (7 citations) laid the foundation for this approach. Building on this, his 2022 work (4 citations) removed the assumption of ground-truth poses, while his 2024 VO-Safe RL for drone navigation (4 citations) specifically prevents drones from flying into visually poor areas that could cause tracking loss. Most recently, his 2024 hierarchical RL method (3 citations) tackles the challenge of local minima in unstructured environments like long corridors. Collectively, Lin’s work is pioneering the next generation of robust, perception-aware navigation systems that can operate safely in the real world.
Research Focus
Key Achievements
Top Papers
- 1
- 2VO-Safe Reinforcement Learning for Drone Navigation4 citations · 2024
- 3Localisation-Safe Reinforcement Learning for Mapless Navigation4 citations · 2022
- 4