High Throughput Dynamic Vehicle Coordination For Intersection Ground Traffic

With the rapid development of sensing technology and information processing technology,more comfortable and intelligent vehicle driving has become a necessary trend of the times.In recent years,some progresses have been made in unmanned driving technology.But there is still a long way to go before the technology can be fully applied in daily life.Through the V2 V and V2 I communication,it is still difficult on how to make a rational use of the perception information of AV(Automatic Vehicle)in order to realize the safe and efficient driving.As a typical scenario where vehicles and road infrastructures interact frequently,the intersection is a hot topic in the field of unmanned vehicles in recent years.Through vehicle networking,AVs can obtain context-aware information beyond the horizon,significantly improving the safety hazards caused by unexpected events.However,when realizing the AV coordination at intersections,the coordination system will confront with the following problems.As an area where multiple roads merge,there are often a large number of vehicles to be coordinated at intersections,which requires the AV scheduling system to have a lower solution complexity.The system needs to dynamically handle scheduling problems based on real-time traffic conditions considering the randomness of the arrival of AVs in practical driving scenarios.In addition,the resources available at intersections are limited,and the system needs to pay attention to the efficiency of the scheduling process to fully utilize the advantages of AVs.In this thesis,aiming at the problem of road congestion and low road resource utilization in the AVs coordination network,an efficient coordination scheme based on AV position and speed information is proposed.The appropriate mathematical model is established for the arrival,queuing and scheduling process of the AV,and the optimal solution is given by the effective solution algorithm.Simultaneously,the phased solution is proposed for the non-convexity of the model,which reduces the complexity of the scheduling problem.In order to solve the problem of excessive congestion in some lanes,the stochastic optimization theory is used to flexibly allocate the intersection resources,make full use of the scheduling system resources,and realize the stability of the lanes under the premise of satisfying the high efficiency.According to the numerical results,compared with the traditional traffic light scheduling,the dynamic scheduling algorithm proposed in this thesis has greatly improved the throughput and stability.Furthermore,the algorithm can make full use of the sensory information of the AV,make analysis and decision of the traffic situation in real time,and obtain the upper bound of the system throughput under the premise of ensuring the stability of the scheduling system.Compared with the static AV scheduling algorithm,the proposed algorithm significantly reduces the complexity of the problem.In summary,this thesis presents a stable AV intersection scheduling scheme in a complex environment,which improves the resource efficiency of intersection resources.