A Dynamic Multi-Risk Management based on Cooperative Optimization Architecture for CAVs at Unsignalized Intersection

He, Shan; Adouane, Lounis · 2024 · Crossref

DOI: 10.1016/j.ifacol.2024.09.012

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 Connected Autonomous Vehicles (CAVs) at unsignalized intersections, aiming to improve traffic efficiency and safety while avoiding the computational bottlenecks of centralized control and the local minima issues of existing decentralized methods. The authors propose a Dynamic Multi-Risk Management based on Cooperative Optimization (D-MRMCO) architecture. This approach extends the previously developed MRMCO-PIDP method, which uses Predicted Inter-Distance Profiles (PIDP) to assess collision risks, by integrating Simulated Annealing (SA) to escape local optima and adapt to dynamic environments. The methodology relies on a decentralized cooperative optimization framework where CAVs communicate and negotiate speed profiles. The core risk assessment metric is the excess PIDP ($ePIDP$), defined as the difference between the minimum predicted inter-vehicle distance and a safety threshold. To resolve conflicts, the algorithm evaluates acceleration and deceleration strategies, selecting the profile that maximizes $ePIDP$ or minimizes state changes. To prevent the algorithm from settling in suboptimal local solutions, the authors incorporate the Metropolis Criterion from Simulated Annealing. This stochastic optimization is governed by a novel intersection model divided into five zones: Core, Stubborn, Feasible, Optimizing, and Buffer. The architecture dictates that deterministic MRMCO-PIDP is used in the Feasible Zone to ensure rapid collision avoidance, while the stochastic SA algorithm is employed in the Optimizing Zone when vehicles have sufficient time to search for globally better solutions. Vehicles in the Stubborn Zone maintain their committed strategies to ensure execution feasibility. Simulation results conducted in MATLAB with 3 to 5 CAVs demonstrate that the D-MRMCO architecture successfully generates collision-free trajectories. The system maintains positive $ePIDP$ values, indicating no collision risk, while optimizing speed profiles to reduce crossing times. Comparative analysis shows that D-MRMCO achieves lower average crossing times than the static MRMCO-PIDP method. Furthermore, the proposed architecture exhibits greater robustness as the number of vehicles increases, with the impact of additional vehicles on computational time and performance being smaller than in the non-dynamic counterpart. The average computational time per iteration remains low (e.g., 0.0168 seconds for 5 vehicles), confirming the method's suitability for real-time dynamic environments. The significance of this work lies in providing a scalable, decentralized solution for unsignalized intersection management that balances safety, efficiency, and computational feasibility. By combining deterministic risk assessment with stochastic global optimization within a structured spatial model, the D-MRMCO architecture effectively handles multi-vehicle collision risks and dynamic changes. This approach offers a viable alternative to centralized controllers, which suffer from high complexity, and pure stochastic methods, which lack real-time responsiveness, thereby advancing the field of cooperative autonomous driving.

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.

• induced exposure