Decentralized Multi-Agent Goal Assignment for Path Planning using Large Language Models

2025-10-27

AI Summary

This work addresses the target assignment problem in decentralized multi-agent pathfinding. To overcome performance bottlenecks of conventional heuristic approaches, we propose a negotiation-free target assignment framework leveraging large language models (LLMs): agents generate preference-ordered target lists solely from local structural environment information; structured prompting and non-iterative, deterministic conflict-resolution rules enable fully distributed decision-making. Our method integrates grid-environment encoding, scenario-aware prompt engineering, and a dual evaluation mechanism comparing greedy versus optimal assignment outcomes. Experiments in fully observable environments demonstrate that, under well-designed prompts, the approach achieves near-optimal task completion time—substantially outperforming classical heuristics. Crucially, this is the first study to empirically validate both the effectiveness and scalability of LLMs for distributed target assignment without explicit inter-agent communication or iterative negotiation.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Distributed Problem Solving
• Search and Optimization: Heuristic Search

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Coordinating multiple autonomous agents in shared environments under decentralized conditions is a long-standing challenge in robotics and artificial intelligence. This work addresses the problem of decentralized goal assignment for multi-agent path planning, where agents independently generate ranked preferences over goals based on structured representations of the environment, including grid visualizations and scenario data. After this reasoning phase, agents exchange their goal rankings, and assignments are determined by a fixed, deterministic conflict-resolution rule (e.g., agent index ordering), without negotiation or iterative coordination. We systematically compare greedy heuristics, optimal assignment, and large language model (LLM)-based agents in fully observable grid-world settings. Our results show that LLM-based agents, when provided with well-designed prompts and relevant quantitative information, can achieve near-optimal makespans and consistently outperform traditional heuristics. These findings underscore the potential of language models for decentralized goal assignment in multi-agent path planning and highlight the importance of information structure in such systems.

Problem

• Decentralized goal assignment for multi-agent path planning in shared environments
• Agents independently generate goal preferences without negotiation or iterative coordination
• Comparing LLM-based agents with traditional heuristics for conflict resolution performance

Innovation

• LLM-based agents generate goal preferences
• Agents exchange rankings without negotiation
• Fixed deterministic rule resolves assignment conflicts