Model And Verification Of Vehicle Adaptive Cruise Control Based On Cooperative Vehicle Infrastructure System

The autonomous driving of automobiles is the current development trend of the automobile industry and the transportation industry.The structure of intelligent connected vehicles(ICV)consists of three parts: perception system,decision-making system and execution system.The advanced in-vehicle sensors,communication and networking modules,and chassis control modules carried by ICVs enable vehicles to have excellent road information capture and processing capabilities.Compared with traditional vehicles,ICVs can reduce road traffic accidents,reduce the driver’s fatigue driving rate,and improve road utilization.With the technological updates and iterations in the fields of computer and communication in recent years,ICVs have also developed to a great extent,which is of great practical significance for the realization of autonomous driving and intelligent transportation.Only relying on the intelligence of the vehicle itself,it is a long process to realize automatic driving.The complementation of intelligent vehicles and roadside facilities to form a cooperative vehicle infrastructure system(CVIS)is an effective measure to shorten this process.CVIS is to connect the three terminals of vehicle,road and cloud through wireless communication technology and computer technology to realize real-time interaction of road information.CVIS conducts all-round detection of road information through sensors,and provides the vehicle with information on blind spots and beyond-the-horizon distance through the Internet of Vehicles communication,which is of great significance to the driving strategy of autonomous vehicles.CVIS jointly support driving intelligence,which provides the possibility for the cooperative adaptive cruise control method of group vehicles.In recent years,learning-based methods have been gradually applied to decision-making systems for autonomous vehicles,including directly supervised deep learning and deep reinforcement learning.Existing learning-based vehicle control methods mainly consider ICVs,rarely apply vehicle-road coordination systems,and do not pay attention to the impact of non-intelligent vehicle driving behaviors on ICV behavior decisions.In view of this,this paper proposes a group vehicle platoon cooperative adaptive cruise control model based on reinforcement learning.The model controls the longitudinal acceleration of the vehicle based on the three evaluation indicators of safety,driving efficiency and comfort during driving,and forms a learning-based vehicle driving control strategy.Secondly,establish a custom simulation environment to simulate and verify the established cooperative adaptive cruise control model,and perform comparative experiments by changing the simulation initialization settings(initial spacing,initial speed and the number of cars downstream)to compare and analyze the model operation results;social vehicles are taken into account in the influencing factors of the model control strategy,with the number of vehicles downstream of the queue as a variable,comparative analysis can evaluate the average speed of the queue and the average distance between the heads of the queue for the model control strategy.Finally,a physical vehicle-road coordination system was built on the test site using physical equipment.By collecting the driving trajectory of real vehicle,the simulated downstream vehicle was associated with the test system,and the hardware-in-the-loop test of the proposed collaborative adaptive cruise control model was performed.The analysis of the model training results shows that the cooperative adaptive cruise control model proposed in this study can complete the platooning task with driving behaviors that reflect safety,driving efficiency and comfort;Experiments show that the platoon can use the road information beyond visual range to optimize its own driving strategy in the environment of CVIS.CVIS built in this paper is able to collect real vehicle trajectories for hardware-in-the-loop simulation experiments to verify the validity of the model.The analysis of the model training results shows that the cooperative adaptive cruise control model proposed in this study can output effective vehicle control strategies,and can reflect better driving safety,driving efficiency and driving comfort.The input content of the algorithm shows that the platoon can use the road information beyond visual distance to optimize its own driving strategy in the vehicle-road cooperative environment.The effectiveness and control performance of the model in the experiment can help solve the problem of cooperative adaptive cruise control.The CVIS built can collect real vehicle trajectories for hardware-in-the-loop experiments to verify the validity of the model.