ITSC 2025 Paper Abstract

Paper TH-EA-T22.3

Shi, Xiaoyu (KAIST), Chen, Sikai (University of Wisconsin-Madison), Jang, Kitae (KAIST), Chen, Tiantian (KAIST)

Conflict-Aware Platoon Scheduling at Unsignalised Intersections Via Traveling Salesman Problem: A Large Language Model-Guided Heuristic Solution

Scheduled for presentation during the Invited Session "S22b-Emerging Trends in AV Research" (TH-EA-T22), Thursday, November 20, 2025, 14:10−14:30, Coolangata 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

Keywords Traffic Management for Autonomous Multi-vehicle Operations, Cooperative Driving Systems and Vehicle Coordination in Multi-vehicle Scenarios, Multi-vehicle Coordination for Autonomous Fleets in Urban Environments

Abstract

Managing vehicle movements at unsignalised intersections is a critical challenge in urban transportation, especially with the increasing use of connected and automated vehicles (CAVs). The purpose of this study is to introduce a novel conflict-aware scheduling approach by transforming the intersection management problem into a Traveling Salesman Problem (TSP). It is envisioned that directional platoons of vehicles are represented by nodes in a directed graph, with edges representing empirically derived travel costs that are informed by conflict. To solve this optimization effectively, we use the Evolution of Heuristics (EoH) framework guided by Large Language Models (LLMs). Through iterative evolution, EoH develops specialized scheduling heuristics that significantly reduce computational complexity while maintaining the quality of the solution. In simulation experiments, the proposed EoH-based TSP method outperforms conventional signal-based and fixed-order scheduling strategies in terms of reducing vehicle delays and waiting times. While EoH's median performance metrics are similar to those of OR-Tools optimization, EoH exhibits greater variability, indicating a greater potential for optimal solutions under certain circumstances. The proposed approach also results in smoother and more synchronized vehicle flows, underscoring its applicability and benefits in autonomous intersection management.