Why Solving Multi-agent Path Finding with Large Language Model has not Succeeded Yet (2401.03630v2)
Abstract: With the explosive influence caused by the success of LLMs (LLM) like ChatGPT and GPT-4, there has been an extensive amount of recent work showing that foundation models can be used to solve a large variety of tasks. However, there is very limited work that shares insights on multi-agent planning. Multi-agent planning is different from other domains by combining the difficulty of multi-agent coordination and planning, and making it hard to leverage external tools to facilitate the reasoning needed. In this paper, we focus on the problem of multi-agent path finding (MAPF), which is also known as multi-robot route planning, and study the performance of solving MAPF with LLMs. We first show the motivating success on an empty room map without obstacles, then the failure to plan on the harder room map and maze map of the standard MAPF benchmark. We present our position on why directly solving MAPF with LLMs has not been successful yet, and we use various experiments to support our hypothesis. Based on our results, we discussed how researchers with different backgrounds could help with this problem from different perspectives.
- Evaluating Multi-Agent Coordination Abilities in Large Language Models. arXiv preprint arXiv:2310.03903.
- Teaching large language models to self-debug. arXiv preprint arXiv:2304.05128.
- Scalable Multi-Robot Collaboration with Large Language Models: Centralized or Decentralized Systems? arXiv preprint arXiv:2309.15943.
- PRIMAL$_2$: Pathfinding Via Reinforcement and Imitation Multi-Agent Learning - Lifelong. IEEE Robotics Autom. Lett., 6(2): 2666–2673.
- Improving Factuality and Reasoning in Language Models through Multiagent Debate. CoRR, abs/2305.14325.
- Complexity-based prompting for multi-step reasoning. arXiv preprint arXiv:2210.00720.
- The capacity for moral self-correction in large language models. arXiv preprint arXiv:2302.07459.
- DDM: Fast Near-Optimal Multi-Robot Path Planning Using Diversified-Path and Optimal Sub-Problem Solution Database Heuristics. IEEE Robotics Autom. Lett., (2): 1350–1357.
- Planning with diffusion for flexible behavior synthesis. arXiv preprint arXiv:2205.09991.
- Camel: Communicative agents for” mind” exploration of large scale language model society. arXiv preprint arXiv:2303.17760.
- MAPF-LNS2: Fast Repairing for Multi-Agent Path Finding via Large Neighborhood Search. In Proceedings of the Thirty-Sixth Conference on Artificial Intelligence (AAAI), 10256–10265. AAAI Press.
- Symmetry-breaking constraints for grid-based multi-agent path finding. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, 6087–6095.
- Moving Agents in Formation in Congested Environments. In Harabor, D.; and Vallati, M., eds., Proceedings of the Thirteenth International Symposium on Combinatorial Search, SOCS 2020, Online Conference [Vienna, Austria], 26-28 May 2020, 131–132. AAAI Press.
- Llm+ p: Empowering large language models with optimal planning proficiency. arXiv preprint arXiv:2304.11477.
- Dynamic LLM-Agent Network: An LLM-agent Collaboration Framework with Agent Team Optimization. arXiv preprint arXiv:2310.02170.
- Roco: Dialectic multi-robot collaboration with large language models. arXiv preprint arXiv:2307.04738.
- A Language Agent for Autonomous Driving. arXiv preprint arXiv:2311.10813.
- Demystifying GPT Self-Repair for Code Generation. arXiv preprint arXiv:2306.09896.
- High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10684–10695.
- PRIMAL: Pathfinding via Reinforcement and Imitation Multi-Agent Learning. IEEE Robotics Autom. Lett., 4(3): 2378–2385.
- Schlangen, D. 2023. Dialogue games for benchmarking language understanding: Motivation, taxonomy, strategy. arXiv preprint arXiv:2304.07007.
- Conflict-based search for optimal multi-agent pathfinding. Artif. Intell., 40–66.
- Reflexion: an autonomous agent with dynamic memory and self-reflection. arXiv preprint arXiv:2303.11366.
- Automatic Prompt Augmentation and Selection with Chain-of-Thought from Labeled Data. arXiv preprint arXiv:2302.12822.
- Learn to Follow: Decentralized Lifelong Multi-agent Pathfinding via Planning and Learning. arXiv preprint arXiv:2310.01207.
- Multi-Agent Pathfinding: Definitions, Variants, and Benchmarks. In Proceedings of the Twelfth Annual Symposium on Combinatorial Search (SoCS), 151–159. AAAI Press.
- Efficient SAT Approach to Multi-Agent Path Finding Under the Sum of Costs Objective. In 22nd European Conference on Artificial Intelligence (ECAI), volume 285 of Frontiers in Artificial Intelligence and Applications, 810–818. IOS Press.
- Language conditioned traffic generation. arXiv preprint arXiv:2307.07947.
- Large Language Models Still Can’t Plan (A Benchmark for LLMs on Planning and Reasoning about Change). CoRR, abs/2206.10498.
- On the Planning Abilities of Large Language Models–A Critical Investigation. arXiv preprint arXiv:2305.15771.
- Prompt a robot to walk with large language models. arXiv preprint arXiv:2309.09969.
- Unleashing Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration. CoRR, abs/2307.05300.
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. In NeurIPS.
- Large language models as optimizers. arXiv preprint arXiv:2309.03409.
- React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629.
- In-context instruction learning. arXiv preprint arXiv:2302.14691.
- Building cooperative embodied agents modularly with large language models. arXiv preprint arXiv:2307.02485.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.