| By the end of 2020,the highways’ total length of our country reached 5.1981 million kilometers,of which 161,000 kilometers were expressways mileage,ranking first in the world.Building China’s strength in transportation accelerates transportation development in changing from recklessly expanding to focusing on high quality.In the context of the new infrastructure construction,combined with emerging technologies such as the Internet of Things,big data,and artificial intelligence,it has become an urgent task to realize automated driving with vehicle and infrastructure cooperation and to build digital,intelligent,and safe expressways.As a pioneer for new technologies,connected and autonomous vehicles’ exclusive lanes on expressways create better conditions for safe and efficient passage of Connected and Autonomous Vehicles(CAVs)and have broad development prospects.However,the research on the process of vehicles merging into CAV exclusive lanes has not been reported.Therefore,this paper focuses on the weave area of the CAV exclusive lane entrance and proposes the scenario setting and control strategy from the perspective of road designing,static facility planning,and dynamic traffic control,which has a valuable reference for the layout and management of expressway exclusive lanes.Firstly,the weave area scenario of the exclusive lane entrance was built.By studying the layout method of the High-Occupancy Vehicle lane entrance and expressway work zone,the entrance of the CAV dedicated lane was set near the on-ramp area,and the weave area of the entrance was divided into three sections: warning area,upstream transition area,and execution area.The vehicle driving targets of each area were formulated.By analyzing vehicles’ information needs,traffic signs and markings,signaling devices,sensor equipment,and roadside units were set on the roadside,so that CAVs could merge into the exclusive lane orderly and efficiently,and vehicles’ driving safety could be ensured.Secondly,vehicle micro-control models suitable for the weave area were established.Carfollowing models were established respectively for Human Driven Vehicles and CAVs.Based on analysis,it was found that vehicles may mandatorily change lanes in the rear section of the first lane in the warning area or on the acceleration lane.Thus,fuzzy control theory was introduced to model mandatory lane-changing decisions and simulate the vehicles’ lanechanging behavior in the weave area.Thirdly,the on-ramp signal control model based on the Deep Q-Learning(DQN)algorithm was built.To prevent traffic congestion caused by a large number of vehicles changing lanes in the weave area,the on-ramp control strategy was studied to optimize the traffic status of the mainline.To generate action decisions,the model considered the mainline and ramp’s traffic flow parameters and CAV market penetration rates as detection indicators,and then the green light period was acquired.The model used the weighted value of the average vehicles’ speed and the ramp queuing length as a reward function,which can prevent ramp vehicles from overflowing and make mainline vehicles pass quickly.Finally,the micro simulation platform was built using Python programming to conduct model and algorithm experimental verifications.To verify the proposed vehicle micro control models,the steady traffic flow was acquired through the cellular automata model with the periodic boundary condition.The simulation results show that the traffic basic graph and the vehicles’ spatio-temporal graph are reliable.Compared with the no-control and the D-ALINEA control scenario,this paper evaluated the DQN ramp control strategy,and drew the following conclusions: 1)The model does not significantly improve traffic efficiency,but it can effectively relieve mainline congestion,especially in the case of low CAV penetration or high traffic demand;2)The model can improve the success rate of changing to the exclusive lane compared with the no-control strategy;3)The model can mitigate safety risks of vehicles’ collisions. |
PDF Full Text Request