BFS 및 K-Means 기반 결합 초기화를 이용한 다중 로봇 커버리지의 DARP 초기 위치 결정 및 수렴 효율 향상

초록

DARP is a widely used partitioning algorithm for grid-based multi-robot coverage that balances workload while preserving region connectivity. However, in indoor environments, unfavorable initial robot placements can increase cell reassignments, leading to long runtimes or convergence failure. This study quantitatively analyzes DARP’s sensitivity to initial robot positions and mitigates the effect of initial partition imbalance on convergence efficiency. To this end, a BFS–K-Means hybrid initialization with an imbalance-rate-based selection strategy is proposed. After an initial DARP run, an imbalance rate based on Jain’s fairness index is computed from the number of cells assigned to each robot. Only when this value exceeds a predefined threshold, BFS constructs candidate regions reflecting free-space connectivity, and K-Means rearranges the initial centers before DARP is rerun. Experimental results show that the proposed method maintains fairness and connectivity while reducing iterations by up to about 75% in complex environments and 60% in normal environments compared with random initialization.

키워드