Middle-mile logistics through the lens of goal-conditioned reinforcement learning

TL;DR

This paper introduces a goal-conditioned reinforcement learning approach using graph neural networks to optimize middle-mile logistics routing through hub networks.

Contribution

It presents a novel method combining graph neural networks with model-free RL for efficient parcel routing in middle-mile logistics.

Findings

Effective routing strategies learned through the proposed method.

Improved efficiency in parcel delivery routing tasks.

Potential for scalable logistics optimization.

Abstract

Middle-mile logistics describes the problem of routing parcels through a network of hubs linked by trucks with finite capacity. We rephrase this as a multi-object goal-conditioned MDP. Our method combines graph neural networks with model-free RL, extracting small feature graphs from the environment state.