The Impact Mechanism Of The Takeover Of Autonomous Vehicle On Traffic Oscillation In The Mixed Traffic Flow And The Coordinated Control Of Vehicles

With the rapid development of autonomated technology,when autonomous vehicles undergo road tests and enter the market,they will encounter driving scenarios that exceed the operational design domain of the advanced driving assistance system(ADAS)in some road environments.In this case,the ADAS system will fail and require the driver to take over the autonomous vehicle.What kind of takeover reaction behavior the driver will show when facing a takeover request,whether the driving behavior and driving safety of the takeover vehicle will be affected by the takeover process,what kind of the impact takeover request would cause to the operation of the mixed traffic flow composed of autonomous vehicles and human vehicles,and how to eliminate the adverse effect of takeover through collaborative control between vehicles are important issues that need to be resolved before commercial advocation of autonomous driving technology.This research aims to analyze the characteristics of driver’s takeover reaction behavior during the takeover process,build a full-cycle car-following model of autonomous vehicles that includes the takeover process,analyze the impact of takeover request on the generation and propagation of traffic oscillation in mixed traffic flow,and establish a collaborative control method to reduce the adverse impact of takeover on the stabilit y of the autonomous vehicle fleet.First,this paper builds a takeover driving simulation experiment platform and utilizes the UC-win Road traffic simulation software to make the takeover driving scenarios under the impact of road dangerous stimulus.We design the takeover experiment and collect drivers’ takeover reaction behavior data during the experiment.The outcome shows that if the driver is not engaged in the non-driving-related tasks(NDRTs),the average take-over reaction time is660ms;when the driver picks up NDRTs,the average take-over reaction time is 1450 ms.We make the takeover reaction time distribution curve under different takeover scenarios,and use Mann-Whitney U non-parametric test method to test the impact of diverse factors such as people,vehicles,roads,and environments on the takeover reaction time.Considering the impact of takeover scenario factors and driver characteristic factors,the takeover reaction time prediction model is established based on the classification and regression tree model(CART).Secondly,drivers’ braking and steering behavior characteristics during the takeover process are analyzed based on the collected takeover driving behavior data.The migration relationship of the vehicle driving mode between different car-following states before and after takeover process is characterized.The takeover process is subdivided into two phases: takeover preparation phase and driver takeover phase.For the takeover preparation phase,a headway control model is constructed;for the driver takeover phase,a car-following model that considers the driver’s takeover reaction delay effect and the driving ability recovery process is constructed.Then the vehicle trajectory data collected by the takeover experiment is used to calibrate the parameters of the car-following model in the driver’s driving ability recovery phase The results verify the effectiveness of the model in describing the driver’s takeover driving behavior.Numerical simulation method is utilized to analyze the changing of driving safety when takeover happens,the results show that the takeover process will increase the risk of collision between the takeover vehicle and the front and following vehicle.Then,under the current mixed traffic flow environment which composed of autonomous vehicles and human-driven vehicles,to solve the problem that whether the takeover request of autonomous vehicle will adversely affect the operation of the traffic flow.In this paper,the impact of the takeover process on the generation and propagation of traffic oscillation is analyzed through numerical simulation experiment in the mixed platoon.The results show that the takeover process will affect the driving characteristics(speed,acceleration and headway)of vehicles in the fleet.Similarly,the takeover process will also change the propagation characteristics of traffic oscillations.The speed of deceleration waves that propagate upstream of the road will increase significantly to 7.4m /s,and the propagation speed of acceleration waves will reduce to 3.7m/s,which means that the takeover process will increase the degree of traffic congestion.The traffic oscillation growth characteristics will also change significantly.Compared with the automated driving mode and manual driving mode,the oscillation amplitude will increase by 20% and 5.6%,respectively.The oscillation intensity will increase by 50% compared to the automated driving mode.At the same time,the influence of different takeover parameters(takeover trigger threshold,takeover duration,and penetration rate of autonomous vehicle)on the traffic oscillation is also studied.Through the traffic flow stability analysis method,the driving stability of an autonomous vehicle fleet is analyzed when it is disturbed by takeover.The results show that takeover disturbance will significantly damage the overall driving stability of the fleet.Finally,except that the takeover process will affect the stability of the traffic flow,there is a significant takeover time delay effect in the autonomous vehicle fleet.The analysis of driving behavior characteristics within the delay time shows that the takeover time delay effect will affect the driving stability of the vehicle to a certain extent.In order to realize the coordinated control between the takeover vehicle and following vehicle,the basic principle of Model Predictive Control(MPC)is introduced,and the framework of the coordinated control of takeover is built based on the MPC method.Firstly,the gray-scale prediction model is used to predict the driving speed of the takeover vehicle during the take-over process,and speed optimization model of following vehicle is established which takes the driving stability of following vehicle as the optimization objective.The optimization equation is solved to output the speed control value of the following vehicle.The outcome of control indicates that through MPC control,the following vehicle can follow the takeover vehicle more safely,and will not trigger the initiation of the takeover request of following vehicle,The MPC control method can reduce the impact of takeover vehicle on the driving stability of the following vehilce and suppress the adverse effects of the takeover transfer phenomenon on the traffic flow.