Decoupled Intelligence: A Multi-Agent LLM Framework for Controllable Traffic Scenario Generation in SUMO

2026-05-26

AI Summary

This work addresses the limitations of existing monolithic agent architectures in end-to-end traffic simulation, which suffer from reasoning failures, inconsistent parameterization, and inadequate state management, thereby hindering the generation of controllable, high-fidelity scenarios. To overcome these challenges, the authors propose a role-decoupled multi-agent LLM framework that decomposes the SUMO simulation pipeline into specialized modules—planning, construction, demand generation, execution, and analysis—orchestrated by a high-level reasoning engine. A Model Context Protocol (MCP) with state persistence enables seamless mapping from natural language intent to microscopic simulation. Experimental results demonstrate that the proposed approach significantly outperforms monolithic baselines in task success rate and parameter accuracy, efficiently translating natural language instructions into high-fidelity traffic simulations on real-world road networks.

Abstract

The integration of Large Language Models (LLMs) with microscopic traffic simulation offers a promising path toward autonomous urban planning and intelligent transportation analysis. However, existing monolithic agent architectures often struggle with the complexity of end-to-end simulation workflows, leading to reasoning failures, parameter inconsistency, and a lack of systematic state management. This paper proposes a novel multi-agent collaborative framework designed to automate the entire lifecycle of traffic simulation in SUMO (Simulation of Urban Mobility). Our approach decouples the simulation pipeline into specialized roles, including Planner, Builder, Demand, Runner, and Analyst, coordinated by a high-level reasoning engine. We introduce a state-persistent Orchestrator leveraging the Model Context Protocol (MCP) to ensure seamless data handover and environmental consistency across distributed agent actions. This architecture enables a robust closed-loop refinement process, where simulation outcomes are iteratively analyzed and optimized to satisfy user-defined Key Performance Indicators (KPIs). Experimental results through role ablation studies demonstrate that the proposed multi-agent framework significantly enhances task success rates and parameter accuracy compared to single-agent baselines. Furthermore, case studies on real-world network extraction and traffic optimization highlight the system's capability to bridge the gap between high-level natural language intent and low-level simulation execution.

Problem

• Large Language Models
• traffic simulation
• multi-agent systems
• SUMO
• state management

Innovation

• multi-agent LLM
• traffic simulation
• SUMO
• Model Context Protocol
• decoupled intelligence