LLM-PCPP: A novel robot prioritized coverage path planning approach for complex environments

This study introduces a novel prioritized coverage path planning method that combines the efficiency of Large Language Models (LLMs) with traditional optimization techniques to enhance the identification and traversal of key points of interest (POIs) on semantic maps of complex environments. By leveraging LLMs, the proposed approach dynamically identifies POIs, addressing inaccuracies and inefficiencies associated with manual selection. The LLM-A* algorithm is then utilized to compute the shortest paths between POIs, significantly reducing operational and storage parameters and thereby minimizing computational overhead. Finally, the traversal problem is formulated as a Traveling Salesman Problem for optimal route planning. This method demonstrates substantial improvements in computational efficiency, particularly benefiting resource-constrained robotic systems, and offers a scalable solution for efficient path planning in complex environments.