Integrating Action Knowledge and LLMs for Task Planning and Situation Handling in Open Worlds
27 May 2023
Proposes COWP, a framework integrating classical planning and LLMs for open-world task planning
Enables robots to handle unforeseen situations by augmenting action knowledge with common sense
Evaluated on a dataset of 1,085 real-world situations across 12 tasks
Outperformed baselines on task completion rate and situation handling
Demonstrated on a mobile manipulator performing delivery tasks
Using AI to Help Robots Plan Tasks and Handle Unexpected Situations
This paper proposes a framework to enable robots to generate plans for long-horizon tasks and adapt when unexpected situations arise during execution. It integrates classical AI planning with large language models to leverage both structured action knowledge and flexible commonsense reasoning.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper