Shurjo Banerjee
Papers
3
Total Citations
37
H-Index
3
About
Shurjo Banerjee is a researcher advancing the frontiers of embodied AI, with a focus on how robots and agents can understand and interact with the world through natural language. His work sits at the intersection of vision-and-language navigation (VLN), reinforcement learning, and human-robot dialog. Banerjee’s most influential contribution is the introduction of **Iterative Vision-and-Language Navigation (IVLN)**, a paradigm that challenges agents to navigate persistent environments over multiple episodes—a stark departure from standard VLN benchmarks that reset memory at each start. This work, his most cited with 19 citations, tests long-term spatial reasoning and memory. He also pioneered **Floyd-Warshall Reinforcement Learning** (10 citations), a model-based approach enabling agents to learn from past experiences to reach new goals in static environments, offering a powerful alternative to model-free methods for multi-goal tasks like robotic pick-and-place. Additionally, Banerjee co-created the **RobotSlang benchmark** (8 citations), a dataset for studying dialog-guided robot localization and navigation, pushing toward more cooperative human-robot communication. His research is pivotal for developing robots that can follow complex, evolving instructions in real-world settings.
Research Focus
Key Achievements
Top Papers
- 1Iterative Vision-and-Language Navigation19 citations · 2023
- 2
- 3