Study On CPS-based Modeling And Control Of Longitudinal Following With Humanvehicle Cooperation

With the increasing of the level of automobile automation,the driving automation system plays a more and more important role in vehicle control,and the driver gradually changes from an operator to a supervisor or user.However,considering the complexity of traffic problems,the reliability of sensing equipment and the restriction of price and cost,it is very difficult for the real self-driving vehicle to be realized in a short term.For a long time to come,intelligent vehicles will be in the stage of human vehicle co-piloting.In the human vehicle co-piloting mode,due to the essential differences in the sensing method,decision-making mechanism and control execution mode of human driver and driving automation system,poor human-vehicle cooperation will inevitably cause problems such as human-vehicle interaction imbalance,control conflicts,and vehicle instability and out of control.From the perspective of cyber physical system,the interweaving and integration of discrete cyber system and continuous physical system of the vehicle are more complex,and it becomes more obvious that the impact on stable driving of the characteristics of cyber physical integration.Compared with the characteristics of human driver and driving automation system,it can be found that driver has unique advantages in scene learning and trend judgment.Compared with human driver,driving automation system has the advantages of quick response,precise control and good behavior consistency,although its learning ability is weak and scene adaptability is poor.Human diver and driving automation system have strong complementarity.Therefore,to make full play of the advantages of human intelligence and machine intelligence,developing the human-machine cooperative control system and optimizing the human-vehicle cooperative scheme can not only effectively solve the problem of driver not in the loop and the challenge of takeover safely faced by the automated driving,but also provide a certain degree of reference for the development and application of human-machine hybrid intelligent theory.Therefore,in this paper,combined with the development frontier of human vehicle co-piloting,taking the typical longitudinal car following process as the object,from the perspective of cyber physical system,longitudinal car-following models for human drivers and automated vehicles are constructed with consideration of cyber physical factors such as: vehicle inertia,communication conditions,sensor acquisition characteristics,and human factors such as: perception insensitivity,response delay,fuzzy decision-making.Then the modeling and control of human-machine cooperative schemes are studied systematically from the idea of human-machine time-division cooperation,human-machine work-division cooperation and human-machine hierarchy-division cooperation to make full play of the advantages of human driver and driving automation system,reduce the driver's workload and improve the driving safety and stability of intelligent vehicle.Specifically,the research work of this paper mainly includes the following aspects:(1)Taking the typical longitudinal car-following process as the object,this paper deeply analyzes the characteristics and differences of human drivers and driving automation system in perception,decision-making and execution mode.Considering the driver's insensitivity to distance perception and the bounded randomness of response delay,a new longitudinal car-following model for human drivers is established.Through the analysis of nonlinear stability based on descriptive function method,the stability conditions of the new model with dead-zone and dead-zone relay parallel nonlinearity are obtained.Considering the fast and accurate characteristics of data acquisition of the automatic driving sensor,based on the short-term prediction of the preceding vehicle movement,a longitudinal car-following model for automated vehicles is established.Through local stability analysis,linear stability analysis and non-linear stability analysis,the stability conditions of the new model and the influence of prediction error of autoregressive model on car-following performance are obtained.(2)In order to solve the problem that the driver out-of-the-loop caused by automated driving,based on the longitudinal car-following model for human drivers and automated vehicles developed in(1),a man-machine cooperative switching driving control scheme based on model predictive control is proposed from the idea of human machine timedivision cooperation,and a man-machine cooperative switching driving model is constructed.Based on Laypunov stability analysis and linear matrix inequality(LMI)method,sufficient conditions for the switched system are obtained.By minimizing the overall indexes consisting of consisted of the operation workload,takeover risk,tracking errors and comfort,the overall optimization of driving authority switching signal is achieved.The simulation results show that the overall performance index of the manmachine cooperative switching driving control scheme is smaller than that of the manual driving and the automated driving alone.At the same time,the optimization algorithm takes less time and can meet the actual requirements of the project.(3)To solve the problems that driver out of the loop can not be avoid completely and the irregular switching affects driving experience,a human-machine hybrid intelligence cooperative control scheme is proposed from the idea of human-machine work-division cooperation,in which the driving automation system and human driver are assigned to be responsible for the velocity tracking and headway adjusting respectively.For the goal of velocity tracking,a feedforward-feedback control strategy was designed firstly,then an H-infinity suboptimal control method was developed to optimize the controller parameters according to the desired performance index,finally the controller was further fine-tuned based on the idea of human-simulated intelligent control(HSIC)to improve the performance of the velocity tracking.For the human driver's headway adjusting task,the stability analysis method based on the Lyapunov function proved that the variable gain proportional feedback control can be executed by the driver to ensure the system stability under the cooperation of automated velocity tracking.The experimental results of longitudinal car-following based on driving simulator show that the human-machine hybrid intelligence cooperative control scheme can reduce the tracking error of vehicle distance and keep the driver in the control loop with a small operating load.(4)Aiming at the problem of human-machine interaction conflict in man-machine hybrid intelligence cooperative control scheme,a human-machine double closed-loop cooperative car following control strategy is proposed from the idea of human-machine hierarchy-division cooperation.In the outer loop,driver adjusts the vehicle distance according to the vehicle velocity,and control effects of the driver and the velocity of the preceding vehicle synthetically generate the referenced velocity in the inner loop.In the inner loop,driving automation system controls the vehicle speed to realize the fast tracking of the referenced velocity.Considering the uncertainty of vehicle dynamics parameters,the controller parameters design based on H-infinity state feedback control is realized in the inner loop.Considering the fuzzy characteristics of driver's perception and response to vehicle distance error,the design of spacing adjustment strategy based on the feedback control idea and the corresponding stability proof are realized in the outer loop.The simulation results show that the strategy can improve the stability,safety and comfort of the intelligent vehicle,and can effectively avoid the human-machine interaction conflict because human and machine are in different control loops.In summary,based on the perspective of cyber physical system,this paper establishes longitudinal car-following model for human drivers and automated vehicles separately,and deeply studies the method of man-machine cooperative longitudinal carfollowing modeling and control from the idea of human-machine time-division cooperation,human-machine work-division cooperation and human-machine hierarchy-division cooperation.The research results can not only provide theoretical guidance for optimizing human-vehicle cooperation strategies and improve the overall performance of intelligent vehicles,but also meet the long-term layout of the national development of artificial intelligence.They can also provide reference for alleviating traffic congestion and promoting the theoretical development and application of CPS.