Deadlock-Free Method for Multi-Agent Pickup and Delivery Problem Using Priority Inheritance with Temporary Priority

This paper proposes a control method for the multi-agent pickup and delivery problem (MAPD problem) by extending priority inheritance with backtracking (PIBT) to make it applicable more general environments. PIBT is an effective algorithm that introduces each agent, at timestep, agents, in descending order of priority, decide their next neighboring locations timestep through communications only local agents. Unfortunately, environments are modeled as bi-connected area, if contains dead-ends, such tree-shaped paths, may cause deadlocks. However, real-world environment, there many dead-end paths shelves where materials stored well loading/unloading transportation trucks. Our proposed enables MAPD tasks be performed some without deadlock while preserving feature; does this allowing agents have temporary priorities restricting agents’ movements trees. First, we demonstrate can always reach location deadlock. experiments indicate very efficient, even not applicable, comparing them those obtained using well-known token passing baseline.