Multi-Agent Systems — Dynamic Intersection Control

Context: Research Project — Opening to Research Module
Supervisor: Prof. Samir Aknine — Université Claude Bernard Lyon 1
Date: February 2026

Summary

Design and implementation of two approaches for dynamic intersection control using autonomous vehicle platoons:

• Centralized approach (C-CSG-Plan): algorithm combining optimal coalition structure generation (CSG) and integrated temporal planning with spatio-temporal constraints
• Distributed approach: V2V/V2I negotiation protocols · argumentation · reinforcement learning · statistical validation (ANOVA/Wilcoxon)

Key Contributions

• Formal unification of platoon formation as a CSG problem with spatio-temporal constraints
• Novel C-CSG-Plan algorithm merging coalition generation and temporal scheduling
• Comparative evaluation: centralized vs distributed vs traffic light baseline
• Statistical rigor: 30 runs per scenario, ANOVA/Wilcoxon significance testing

Tech Stack

Python · SUMO · TraCI · NetworkX · ZeroMQ · NumPy · SciPy · Matplotlib

Full Report

📄 Read the research report (PDF)

Keywords

multi-agent-systems coalition-structure-generation autonomous-vehicles distributed-systems planning reinforcement-learning