Mixed Traffic Flow Modeling And Operation Efficiency Evaluation For Freeway Under Low Visibility

With the maturity and popularization of the connected and autonomous vehicles（CAVs）,the highway traffic flow is increasingly moving away from the human driven vehicles（HDVs）towards the CAVs.Meanwhile,mixed running will become a normal state and bring new challenges to highway traffic planning and management.Moreover,the traditional traffic flow theories and methods are gradually losing effectiveness as the CAVs constantly join high traffic.Especially in the low-visibility environments,how to bring the perceptive ability of autonomous vehicles into full play to improve traffic safety and efficiency and maximize the resource utilization is very critical.And it will well reflect the superiorities of the CAVs in applications.Therefore,it’s particularly important and urgent to study the theories and methods of mixed traffic flow modelling to provide valid and reliable decision support for highway management and control under low visibility.In this paper,we present the driving behavior models of HDVs and CAVs under low-visibility mixed running conditions,in which drivers’ behavior characteristics and traffic safety and efficiency are well considered.And the performance of models under different scenarios is reasonably evaluated based on a self-built simulation platform.The main contents of this paper are as follows.（1）The influencing factors and behavioral characteristics of mixed traffic flow under low visibility are analyzed and quantified based on an online questionnaire survey firstly.And then two models are separately proposed.The first one is an improved full velocity difference（FVD）model for the HDV based on the trust factor,which considers the differences in drivers’ trust for the CAVs ahead.And the other one is the car following model based on the Adaptive Cruise Control（ACC）strategy,which considers the position and communication performance of the CAVs.（2）A rule-based decision model for the lane changing behavior of HDVs and a lane changing execution model based on the Control Lyapunov Function-Control Barrier Function-Quadratic Programming under low visibility are proposed,in which we consider the fact that drivers rarely change lanes under low visibility.Meanwhile,the mandatory lane change decision model based on the reinforcement learning and the reward-based discretionary lane change decision model for CAV are proposed in view of the driver’s responses and safety requirements for lane changing.Finally,the CAV lane changing execution model is proposed based on the polynomial and Model Predictive Control（MPC）.（3）A simulation platform is built based on the construction and analog of the microscopic vehicle models under low-visibility mixed traffic conditions firstly.And then the performance of all the models proposed in this paper are all evaluated.The influence of different scenarios on the operation efficiency of highway traffic flow is also effectively analyzed based on the criteria including maneuverability,stability,fuel consumption and some main emission indexes simultaneously.