Integrated Task Assignment and Path Planning for Capacitated Multi-Agent Pickup and Delivery

Multi-agent Pickup and Delivery (MAPD) is a challenging industrial problem where team of robots tasked with transporting set tasks, each from an initial location to specified target location. Appearing in the context automated warehouse logistics mail sortation, MAPD requires first deciding which robot assigned what task (i.e., Task Assignment or TA) followed by subsequent coordination must be collision-free paths so as successfully complete its assignment Multi-Agent Path Finding MAPF). Leading methods this area solve sequentially: assigning then paths. In work we propose new coupled method choices are informed actual delivery costs instead lower-bound estimates. The main ingredients our approach marginal-cost heuristic meta-heuristic improvement strategy based on Large Neighbourhood Search. As further contribution, also consider variant can carry multiple tasks just one. Numerical simulations show that yields efficient timely solutions report significant compared other recent literature.