LLMs for Robotics and Autonomy: Safety Perspective

LLMs for Robotics and Autonomy: Safety Perspective

Sayan Mitra, Coordinated Science Laboratory Department of Electrical and Computer Engineering , University of Illinois, Urbana-Champaign

Researchers and engineers in robotics and autonomous systems have recognized the potential of large language models (LLMs). LLMs can provide natural language support for human robot interactions, they can enhance various components in the autonomy stack such as perception, planning, and control, and they could generate code for these components. In this project, we are exploring how Large Language Models can be used for solving planning problems and how they can be used to test and improve existing planners. We focus on the Reach-Avoid planning task in which a dynamic agent, such as a car, has to reach a goal from all possible starting conditions and avoid obstacles.