Decentralized Multi-Drone Coverage Path Planning with Adaptive Path Deconfliction

Decentralized multi-UAV coverage approaches often remain inefficient due to weak inter-agent coordination and the lack of mechanisms that promote coherent global organization under limited communication. To address these challenges, we present a fully decentralized coverage framework that integrates frontier-based local exploration with implicit region allocation emerging from decentralized map fusion and repeated frontier competition. This process induces balanced and non-overlapping exploration behavior without explicit partition computation or centralized supervision. A Kalman-based trajectory prediction module is incorporated directly into the planning process to anticipate inter-agent interactions, enabling proactive deconfliction and safer multi-robot navigation using only local information. Experiments in indoor and outdoor environments demonstrate an average reduction of over 27% in unvisited-cell rate, a 91% decrease in path overlap and redundancy in decentralized flights, and 58% faster coverage compared to classical decentralized baselines.