Solving Multi-Agent Pickup and Delivery Problems using Multiobjective Optimization

Multi-agent pickup and delivery is the problem of allocating tasks for the agents and finding short paths for agents without collisions. These tasks enter the system in different time steps. This article proposes new approaches to this problem based on genetic algorithms in order to optimize the allocation of tasks, minimizing the makespan. We also address this problem as multiobjective, where makespan and service time are minimized. Computational experiments were performed varying the number of agents in a simulated environment of a large-scale warehouse. The results obtained by the proposed approaches were compared with those from the literature and the proposals demonstrated improvements in both objectives.