Stigmergic Graph Memory: An Environment-Aware Approach for Many-to-Many Multi-Agent Pickup and Delivery

2026-07-16

AI Summary

This work addresses the multi-agent pickup and delivery (MAPD) problem, where existing methods often assign depots to requests with unspecified endpoints without considering recent traffic conditions, leading to congestion and reduced efficiency. To overcome this limitation, the authors propose an environment-aware mechanism that jointly optimizes endpoint selection and path planning through a Stigmergic Graph Memory (SGM)—a graph-structured, finite memory layer with decay characteristics that records recent execution signals on nodes and edges to dynamically guide feasible endpoint choices and path preferences. Notably, this approach integrates execution history into task assignment without modifying collision constraints or the underlying planner. Extensive experiments across five layouts, three load levels, and 25 random seeds demonstrate that SGM consistently outperforms baseline methods in all 15 scenarios, achieving throughput improvements of 20.5%–36.7%.

Abstract

Automated fulfillment warehouses must continuously assign and execute pickup-and-delivery work while avoiding congestion. In many-to-many Multi-Agent Pickup and Delivery (MAPD), a request specifies a stock-keeping unit rather than fixed endpoints, requiring the controller to select an agent, source, and destination before path planning. Existing graph-guidance methods primarily influence routing after goals are fixed, leaving endpoint instantiation uninformed by recent traffic. We introduce Stigmergic Graph Memory (SGM), a bounded, decaying memory layer that records recent execution signals on warehouse nodes and directed edges to rank feasible endpoints and route preferences without altering collision constraints or planner validity. Across paired request streams on five layouts, three load levels, and 25 seeds per condition, SGM outperforms two reconstructed many-to-many allocation baselines in all 15 map-load conditions, with paired throughput gains of 20.5-36.7%. These results show that recent execution memory can improve warehouse throughput by shaping which feasible goals enter the planner, not only how agents travel to already fixed goals.

Problem

• Multi-Agent Pickup and Delivery
• stigmergy
• warehouse automation
• congestion avoidance
• goal assignment

Innovation

• Stigmergic Graph Memory
• Multi-Agent Pickup and Delivery
• Environment-Aware Planning
• Execution Memory
• Warehouse Automation