Appropriate traffic coordination at road intersections plays a crucial part in modern intelligent transportation systems.In this paper,we first try to extend the traditional single collision-set coordination strategy ...Appropriate traffic coordination at road intersections plays a crucial part in modern intelligent transportation systems.In this paper,we first try to extend the traditional single collision-set coordination strategy to multiple-collision-set strategies,by which the traffic throughput can be significantly improved.Unlike the existing centralized coordination methods,two low complexity coordination methods are proposed based on the multi-agents Q-learning frameworks.Numerical results show that,the proposed high throughput strategies are able to provide safe and efficient traffic coordination.Meanwhile,since only local information is required,the coordination complexity can be reduced,which is attractive in highly dynamic real time scenarios.展开更多
基金supported by the National Natural Science Foundation of China(Nos.91638204 and 61771159)Guangdong Natural Science Foundation(No.2017A030313392)Shenzhen Fundamental Research Project(No.JCYJ20170811153639780).
文摘Appropriate traffic coordination at road intersections plays a crucial part in modern intelligent transportation systems.In this paper,we first try to extend the traditional single collision-set coordination strategy to multiple-collision-set strategies,by which the traffic throughput can be significantly improved.Unlike the existing centralized coordination methods,two low complexity coordination methods are proposed based on the multi-agents Q-learning frameworks.Numerical results show that,the proposed high throughput strategies are able to provide safe and efficient traffic coordination.Meanwhile,since only local information is required,the coordination complexity can be reduced,which is attractive in highly dynamic real time scenarios.
