Kunyang Lin

Papers

Total Citations

H-Index

About

Kunyang Lin is a researcher advancing the frontier of embodied AI and robot navigation, with a focus on enabling agents to understand and act within complex, human-centric environments. His work bridges the critical gap between language understanding and physical movement, particularly in vision-and-language navigation (VLN). In his highly cited 2022 paper, "Weakly-Supervised Multi-Granularity Map Learning for Vision-and-Language Navigation," Lin tackled the practical challenge of training robots to follow path descriptions containing object references, achieving efficient navigation without exhaustive annotation. This work, garnering 16 citations, introduced a novel multi-granularity map learning approach that significantly improves an agent's ability to ground language in visual surroundings. Further demonstrating his impact in multi-object navigation, his 2022 study "Learning Active Camera for Multi-Object Navigation" (8 citations) addressed the critical bottleneck of exploration with camera-only sensors. By developing a method for active camera control, Lin enabled robots to more efficiently locate and navigate to multiple targets. His contributions are pivotal for creating autonomous systems that can operate seamlessly in homes and workplaces, moving beyond static perception to dynamic, goal-driven interaction with the world.