Abstract:
Navigating a robot in an new environment without predefined graphs presents significant chal- lenges in perception, planning, and adaptability. However, traditional approaches rely on struc- tured maps, which limits their flex- ibility in dynamic and unexplored environments. Therefore, we propose a foundational framework that enables robots to navigate and execute tasks through human interaction in natural language. By leveraging Generative AI and multimodal learning, our system allows robots to dynamically adapt to new environments without requiring prede- fined graphs
