Proactive Multi-Robot Path Planning via Monte Carlo Congestion Prediction in Intralogistics

Abstract

The increasing deployment of autonomous robots in intralogistics has significantly enhanced operation efficiency and throughput. However, dynamic and uncertain environments introduce challenges like congestion, collisions, and deadlocks. Current multi-robot path finding algorithms are often reactive, addressing congestion only after it occurs and neglecting dynamic environmental factors. To overcome these limitations, this study presents an environment-independent and scalable dynamic multi-robot path planning method. The method utilizes Monte Carlo-based probabilistic congestion predictor integrated with a congestion-aware path planning algorithm. This predictor anticipates congestion points based on task assignments, robot locations and environmental factors. By integrating these models with real-time data, the system can dynamically update congestion density fields enabling autonomous robots to modify their routes instantaneously to steer clear of congested areas and diminish the likelihood of deadlocks. Experimental findings validate the efficacy of this strategy by showcasing a reduction in congestion and deadlock occurrences compared to conventional methods highlighting its capacity to enhance efficiency and resilience, in dynamic intralogistics settings.