This research demonstrates how pre-trained large language models can be leveraged for efficient multi-agent decision making in transportation networks, specifically for taxi routing and passenger pickup optimization.

• Zero-shot LLM performance is surprisingly strong for complex spatial planning tasks
• Proper prompting enables LLMs to solve multi-agent coordination problems efficiently
• The approach minimizes passenger waiting times in taxi routing scenarios
• Results show potential for LLMs to enhance optimization in transportation systems

Engineering Impact: This work bridges the gap between language models and spatial planning, offering a new approach to solving complex transportation optimization problems without requiring specialized algorithms or extensive training data.

Data-Efficient Multi-Agent Spatial Planning with LLMs