Intersection Coordination For Automative Vehicles Based On Intelligent Driving

With the rapid development of traffic during urbanization nowadays,the tremendous number of vehicles lead to huge exhaust emission,traffic congestion,accidents and many other social problems.Although intersections constitute only a small proportion of the entire road,they play important roles in modern transportation coordination systems.Based on the presented report,nearly 20% of the fatal accidents takes place around the intersections in EU and US.Meanwhile,the development of autonomous vehicles has brought significant eco nomic benefits to the society.In combination with the development of wireless communications,environmental awareness as well as intelligent control and deci sion-making algorithms,opportunistic vehicle coordination becomes promising,especially for autonomous vehicles,which has attracted more attention from both academic and industrial areas.This thesis mainly aims at the problem that manned vehicles is prone to traffic accidents and traffic jams at intersection,and proposes a variety of safe and efficient coordination schemes.To start with,the mathematical model is established based on a real intersection.With the basis of this model,the centralized and distributed coordination schemes are proposed.The centralized scheme can be formulated as an optimization problem which results in a global optimal solution.As for the distributed scheme,the MARL(Multi-Agent Reinforcement Learning)is adopted to solve this problem.In the MARL,in order to solve the dimension explosion problem,the divide and conquer Q learning algorithm is introduced to improve the convergence of the algorithm.Besides,the impact of positioning accuracy on the coordination algorithm is considered in the schemes.Moreover,in some existing coordination algorithms,the complexity is too high to achieve real time solutions,that restricts their applications.Therefore,aiming at a classic dynamic coordination strategy of autonomous vehicles,the corresponding task model and communication model are established based on the practical intersection and the particularity of coordination algorithm.A variety of task offloading strategies are proposed according to different evaluation index,w hich reduces the resource consumption in the intersection coordination.The simulation results show that the coordination strategies proposed in this thesis have a significant improvement in coordination efficiency under the assurance of safety.In the complex dynamic coordination strategy,the task offloading strategies proposed in this thesis can fully reduce the resources needed by the intersection coordination and meet the real-time requirements.