MAC-Planner: A Novel Task Allocation and Path Planning Framework for Multi-Robot Online Coverage Processes

Abstract

This paper presents a unified framework called MAC-Planner that combines Multi-Robot Task Allocation with Coverage Path Planning to better solve the online multi-robot coverage path planning (MCPP) problem. By dynamically assigning tasks and planning coverage paths based on the system's real-time completion status, the planner enables robots to operate efficiently within their designated areas. This framework not only achieves outstanding coverage efficiency but also reduces conflict risk among robots. We propose a novel task allocation mechanism. This mechanism reformulates the area coverage problem into a point coverage problem by constructing a coarse map of the target coverage terrain and utilizing $K$-means clustering along with pairwise optimization methods to achieve efficient and equitable task allocation. We also introduce an effective coverage path planning mechanism to generate efficient coverage paths and foster robot cooperation. Extensive comparative experiments against state-of-the-art (SOTA) methods highlight MAC-Planner's remarkable coverage efficiency and effectiveness in reducing conflict risks.