Multi-Agent Transport Planning

We discuss a distributed transport planning problem with competitive autonomous actors that carry out time-constrained pick-up delivery orders from customers. The agents have to find conflict-free routes to execute a series of orders they have accepted. Hatzack and Nebel [2] were the first to suggest that finding such a conflict-free schedule can be reduced to solving a job shop scheduling problem. In this paper, we extend their approach in several ways. First of all, we use an iterative traffic-aware planning procedure to determine feasible routes and schedules, called HNZ-1, using a combination of route and schedule finding heuristics. That is, instead of using a static routing approach, we allow agents to replan their route if a preplanned route becomes unfeasible. Moreover, we adapt this (re)route and scheduling heuristic, to take into account incidents that might occur during the execution of orders. Two parameters used in this this HNZ-1 heuristic turn out to be important for the performance: the agent selection function and the resource block size. Results indicate that giving priority to agents that have more strict deadlines as well as scheduling with smaller resource block sizes improve the performance of the agents.