Solving Delivery Assignment in Hybrid-Transit Network Using Multi-agent Reinforcement Learning

Abstract

A robust logistics delivery infrastructure is an essential part in our life. The increasing demand of delivery service requires an optimization in the operation to reduce delivery cost and time. In this paper, the logistic delivery problem is modeled as a task-assignment problem. The problem is then solved using multi-agent reinforcement learning approach, particularly using graph convolutional reinforcement learning algorithm. The goal is to deliver the packages to their respective destinations using least possible fuel, which is the shared resource. Our results show that by encouraging cooperative between the couriers, which act as the agents, the couriers are able to discover ways to preserve resource while completing the delivery tasks.