Sequence Pathfinder for Multi-Agent Pickup and Delivery in the Warehouse

2025-09-28

AI Summary

To address the challenges of limited local observations, high explicit communication overhead, and low global coordination efficiency in Multi-Agent Pickup and Delivery (MAPD) tasks within narrow-aisle environments (e.g., warehouses), this paper proposes a sequence-based implicit coordination path planning framework. It jointly encodes multi-agent trajectories into a permutation-invariant sequence and leverages Transformer architectures to enable implicit information exchange without explicit inter-agent communication. We theoretically establish that the resulting sequential policy possesses order-invariance optimality, reducing decision complexity from exponential to linear in the number of agents. Integrating distributed execution with imitation learning ensures both real-time responsiveness and strong generalization. Experiments demonstrate that our method significantly outperforms existing learning-based approaches across multiple MAPF benchmarks and their variants, while maintaining robust high performance in unseen complex environments—thereby enhancing both coordination efficiency and global situational awareness.

Technology Category

• Multiagent Systems: Multiagent Planning
• Planning, Routing, and Scheduling: Planning with Markov Models (MDPs, POMDPs)
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Multi-Agent Pickup and Delivery (MAPD) is a challenging extension of Multi-Agent Path Finding (MAPF), where agents are required to sequentially complete tasks with fixed-location pickup and delivery demands. Although learning-based methods have made progress in MAPD, they often perform poorly in warehouse-like environments with narrow pathways and long corridors when relying only on local observations for distributed decision-making. Communication learning can alleviate the lack of global information but introduce high computational complexity due to point-to-point communication. To address this challenge, we formulate MAPF as a sequence modeling problem and prove that path-finding policies under sequence modeling possess order-invariant optimality, ensuring its effectiveness in MAPD. Building on this, we propose the Sequential Pathfinder (SePar), which leverages the Transformer paradigm to achieve implicit information exchange, reducing decision-making complexity from exponential to linear while maintaining efficiency and global awareness. Experiments demonstrate that SePar consistently outperforms existing learning-based methods across various MAPF tasks and their variants, and generalizes well to unseen environments. Furthermore, we highlight the necessity of integrating imitation learning in complex maps like warehouses.

Problem

• Addresses poor performance in warehouse MAPD with narrow pathways
• Reduces computational complexity from exponential to linear
• Ensures effective path-finding with order-invariant optimality properties

Innovation

• Uses Transformer for implicit agent communication
• Models pathfinding as sequence with order-invariant optimality
• Reduces complexity from exponential to linear