Real-time Cooperative Vehicle Coordination at Unsignalized Road Intersections

Abstract: Cooperative coordination at unsignalized road intersections, which aims to improve the driving safety and traffic throughput for connected and automated vehicles, has attracted increasing interests in recent years. However, most existing investigations either suffer from computational complexity or cannot harness the full potential of the road infrastructure. To this end, we first present a dedicated intersection coordination framework, where the involved vehicles hand over their control authorities and follow instructions from a centralized coordinator. Then a unified cooperative trajectory optimization problem will be formulated to maximize the traffic throughput while ensuring the driving safety and long-term stability of the coordination system. To address the key computational challenges in the real-world deployment, we reformulate this non-convex sequential decision problem into a model-free Markov Decision Process (MDP) and tackle it by devising a Twin Delayed Deep Deterministic Policy Gradient (TD3)-based strategy in the deep reinforcement learning (DRL) framework. Simulation and practical experiments show that the proposed strategy could achieve near-optimal performance in sub-static coordination scenarios and significantly improve the traffic throughput in the realistic continuous traffic flow. The most remarkable advantage is that our strategy could reduce the time complexity of computation to milliseconds, and is shown scalable when the road lanes increase.

• The paper demonstrates that a TD3-based DRL framework computes near-optimal trajectories at unsignalized intersections within milliseconds.
• It reformulates the coordination problem as a model-free MDP to enable safe, efficient, and simultaneous intersection occupancy.
• Simulation results show enhanced traffic throughput and improved scalability under continuous flows and increased lane numbers.

Real-time Cooperative Vehicle Coordination at Unsignalized Road Intersections

Introduction

The paper "Real-time Cooperative Vehicle Coordination at Unsignalized Road Intersections" (2205.01278) addresses significant challenges in traffic management, particularly at unsignalized road intersections where traditional signal systems are inadequate for handling dynamic traffic flows effectively. With the proliferation of connected and automated vehicles (CAVs), there is a need for innovative intersection coordination frameworks that enhance driving safety and traffic throughput while minimizing computational complexity.

Intersection Coordination Framework

This paper proposes a centralized coordination framework at unsignalized road intersections where involved CAVs relinquish control authority to a centralized coordinator. The coordination node, typically a roadside unit (RSU), collects vehicle state information through V2I communication and applies coordination algorithms to maximize traffic throughput and ensure safety. By leveraging deep reinforcement learning (DRL), the authors aim to solve the cooperative trajectory planning problem formulated as a non-convex sequential decision-making issue.

Methodology

To address computational challenges, the paper reformulates the problem into a model-free Markov Decision Process (MDP) and employs a Twin Delayed Deep Deterministic Policy Gradient (TD3)-based strategy within the DRL framework. The TD3 algorithm is selected for its efficacy in reducing function approximation error in agent policies, which is crucial for real-time decision-making with reduced latency.

The problem transformation involves defining state space, action space, and reward functions that allow the centralized coordinator to learn near-optimal disjoint trajectories in the XYT domain. This approach enables simultaneous intersection occupancy by vehicles without reservation constraints, thus optimizing the use of intersection spatial resources.

Results

Extensive simulation and experimental results demonstrate that the TD3-based strategy achieves near-optimal performance under static coordination scenarios and significantly enhances traffic throughput under continuous traffic flow. The strategy's remarkable advantage is its ability to compute decisions within milliseconds, effectively scaling up with increased road lanes and traffic loads. This computational efficiency outperforms existing centralized strategies that face exponential increases in complexity with vehicle and lane numbers.

Implications and Future Directions

From a practical standpoint, the proposed system offers promising enhancements to urban traffic management by reducing congestion, minimizing collision risk, and improving overall efficiency in real-time operations. Theoretically, the integration of DRL into intersection coordination presents a scalable solution that could adapt to various traffic patterns and infrastructure configurations.

Future research could expand this framework by incorporating re-planning operations to handle system imperfections and improve trajectory accuracy. Additionally, considering turning radii and offering alternative path references could further optimize throughput. Exploration into optimal dynamic coordination strategies, particularly in adapting to changing traffic conditions, remains an open avenue for advancing practical deployments of this system.

Conclusion

The paper presents a sophisticated centralized coordination strategy that leverages deep reinforcement learning to address the complexities of vehicle coordination at unsignalized intersections. By achieving high throughput and low computational latency, this work lays a foundation for future intelligent transportation systems aiming for seamless integration of CAVs within complex urban environments.