Li, Shihua (Tongji University), ZHANG, Lei (Tongji University), Ou, Dongxiu (Tongji University), NING, Zheng (Tongji University, School of Transportation Engineering)

Autonomous Operation Trajectory Planning for Maglev Train: A Case Based on LLM-Agent

Scheduled for presentation during the Invited Session "S16c-Control, Communication and Emerging Technologies in Smart Rail Systems" (TH-LA-T16), Thursday, November 20, 2025, 16:40−17:00, Southport 1

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

This paper proposes an agent-driven autonomous operation control system architecture for maglev transportation. To address the inherent contradiction between the uncertainty of intelligence and strict security of system, a hierarchical decision-making architecture is presented. It comprises three layers: organization, interface and action, where the decision-making domain of maglev agent is confined to the first two layers. Further, to validate both the architectures, a framework simulating agent decision-making process is constructed. It is implemented by the integration of three core components, including large language model (LLM), domain toolset and maglev train dynamics module. The results from testing autonomous operation planning task, demonstrate that maglev agent has established the cognition of operation task and achieve appropriate decision-makings by leveraging domain-specific knowledge, tools and instructions, and information interaction. Ultimately, the proposed architectures transform the operation task expressed in natural language semantics into executable train operation strategy through multi-layer decision conversion.