Behaviour based on decision matrices for a coordination between agents in a urban traffic simulation

Mandiau, René; Champion, Alexis; Auberlet, Jean-Michel; Espié, Stéphane; Kolski, Christophe · 2008 · Crossref

DOI: 10.1007/s10489-007-0045-3

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of coordinating autonomous vehicles (agents) in urban traffic simulations, specifically focusing on resolving conflicts at intersections to prevent deadlocks and accidents. As autonomous vehicle technology advances, simulating the complex, dynamic interactions between agents becomes critical for assessing infrastructure and driver behavior. The authors identify that traditional mathematical traffic models are insufficient for capturing individual behavioral nuances, while standard multi-agent systems often struggle with "deadlocks" (complete blockages) or "livelocks" (indecision) at complex junctions. The research aims to develop a realistic, distributed coordination mechanism that mimics human driver behavior by balancing individual objectives (crossing the intersection) with collective safety constraints. The proposed method models intersection scenarios as non-cooperative games using decision matrices. The authors define three basic two-player situations based on priority relationships: no conflict, one agent has priority, and dual priority (conflict). Each agent selects between "Go" and "Stop" strategies based on payoff matrices that reflect gains for successful crossing and costs for conflicts or unnecessary stopping. To handle the ambiguity in dual-priority situations, the model introduces discriminants (such as license date or random selection) to break ties, ensuring a deterministic outcome. For complex intersections involving $n$ agents, the approach aggregates pairwise two-player matrices into an $n$-player game. Since agents have limited perception and incomplete information about the global state, they construct their own local decision matrices based on perceived interactions. The authors analyze the efficiency of this distributed approach, noting that while deadlock probability increases as the number of agents rises and information decreases, the system remains effective for typical intersection sizes. The study validates the mechanism using the ARCHISIM simulation tool, which supports behavioral modeling of heterogeneous traffic including autonomous vehicles and human drivers. The authors detail the mathematical constraints required to ensure the decision matrices yield optimal solutions—specifically, avoiding unresolved conflicts and minimizing deadlocks. They demonstrate that for two-agent scenarios, the derived solutions align with standard driving norms: agents proceed when there is no conflict, yield when lacking priority, and resolve dual-priority conflicts via discriminants. The paper also examines a specific three-agent case to illustrate the aggregation process. The results indicate that the distributed coordination mechanism successfully prevents blockages and replicates realistic traffic flow, even when agents operate with partial information. The significance of this work lies in providing a scalable, behavior-based coordination framework for multi-agent traffic simulations. By grounding the coordination logic in game theory and decision matrices, the approach ensures that simulated agents exhibit realistic, conflict-avoidant behaviors without requiring centralized control. This is particularly valuable for long-duration traffic trials used to evaluate infrastructure changes, as it prevents costly simulation invalidations caused by deadlocks. The findings suggest that modeling local, pairwise interactions is sufficient to achieve global coordination stability in urban traffic environments, offering a robust foundation for future research into autonomous vehicle navigation and mixed-traffic scenarios.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• traffic density

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model