Kasun Weerakoon
Papers
19
Total Citations
343
H-Index
7
About
Kasun Weerakoon is a robotics researcher specializing in autonomous robot navigation, terrain perception, and machine learning for unstructured outdoor environments. His work addresses one of robotics' most pressing challenges: enabling robots to safely and reliably traverse complex real-world terrains, from dense vegetation to uneven outdoor landscapes. Weerakoon's most influential contribution, GA-Nav (2022, 145 citations), introduced a pioneering group-wise attention mechanism for terrain segmentation from RGB images, enabling robots to identify navigable regions with remarkable efficiency. Building on this foundation, he developed TERP (67 citations), which leverages deep reinforcement learning and elevation maps for reliable navigation in uneven outdoor settings. His subsequent work demonstrates a natural progression toward multimodal sensing, with GrASPE fusing LiDAR, camera, and odometry data, while ProNav and AMCO integrate proprioceptive signals from legged robots for richer traversability estimation. More recently, Weerakoon has explored cutting-edge directions including vegetation-aware navigation (VERN), offline reinforcement learning (VAPOR), and Vision Language Models for context-aware navigation (CoNVOI). With over 300 cumulative citations and a consistently productive publication record, his research is shaping next-generation autonomous systems capable of operating confidently beyond structured environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10