GRAND: Guidance, Rebalancing, and Assignment for Networked Dispatch in Multi-Agent Path Finding
This addresses the problem of congestion and efficiency in large-scale multi-agent path finding for logistics settings like warehouses, offering a practical and scalable solution with incremental improvements over existing methods.
The paper tackles task scheduling for large robot fleets in warehouses by proposing a hybrid method that combines learning-based global guidance with lightweight optimization, improving throughput by up to 10% over the 2024 winning scheduler on benchmarks with up to 500 agents while maintaining real-time execution within a 1 s compute budget.
Large robot fleets are now common in warehouses and other logistics settings, where small control gains translate into large operational impacts. In this article, we address task scheduling for lifelong Multi-Agent Pickup-and-Delivery (MAPD) and propose a hybrid method that couples learning-based global guidance with lightweight optimization. A graph neural network policy trained via reinforcement learning outputs a desired distribution of free agents over an aggregated warehouse graph. This signal is converted into region-to-region rebalancing through a minimum-cost flow, and finalized by small, local assignment problems, preserving accuracy while keeping per-step latency within a 1 s compute budget. On congested warehouse benchmarks from the League of Robot Runners (LRR) with up to 500 agents, our approach improves throughput by up to 10% over the 2024 winning scheduler while maintaining real-time execution. The results indicate that coupling graph-structured learned guidance with tractable solvers reduces congestion and yields a practical, scalable blueprint for high-throughput scheduling in large fleets.