Research On Active Disturbance Rejection Longitudinal Control Algorithm For Driverless Vehicle Based On State Parameter Observers

Longitudinal control of driverless vehicles is an important part to achieve driverless technology and its essence is the control of velocity.However,the system nonlinearity,parameter variation and disturbance from environment influence the control accuracy and pose a challenge to control strategy design.Therefore,a longitudinal controller for driverless vehicle with high adaptive ability to all the disturbance listed above is becoming an urgent issue in current engineering filed.In this paper,an active disturbance rejection longitudinal control algorithm for driverless vehicle based on real-time state observers is put forward.The main research work includes:First,a mass observer based on recursive least square(RLS)and a slope observer based on extended state observer(ESO)are designed and the observation results will be updated to the model feedforward control.Then a co-simulation platform of CarSim and Simulink is built to verify the observation performance.The simulation results show that the average error of vehicle mass observation is about 2.03%,and the average error of road slope observation is 2.70%,which are accurate enough to be used in feedforward control.Second,a longitudinal model feedforward control(MFC)algorithm is established based on coupled observer.The mass and slope values observed by the coupled observer are updated to the model feedforward controller and then the controller structure parameters are changed in real time to compensate control amounts.Then the target acceleration and the resistance demand acceleration are calculated by the upper pre-imager to design acceleration/brake switching logic strategy,which determines the control output of acceleration pedal or braking torque.Then,a longitudinal control algorithm based on the Active Disturbance Rejection Control method(ADRC)feedback is designed.First,the extended state observer is proposed to estimate and then compensates the total disturbances in the control input design,which can theoretically simplify the system into a pure integrating element.Meanwhile,the feedback control is designed.As a result,only such three parameters as b0,observer bandwidth and controller bandwidth has to be tuned,which simplifies the parameter tuning process to some extend.The model-in-the-loop simulation(MIL)results show that the PID control error increases from 3.31% to 9.16% when the vehicle condition is switched from a mass of 1800 kg and a slope of 0° to a full load of 2100 kg and a slope of 6°.The control error of the control algorithm consisting of ADRC and MFC is changed from 0.86% to 1.26% under the same condition.Therefore,the control quality of the algorithm proposed in this article is more adaptive than that of PID.Lastly,hardware-in-the-loop(HIL)testing and vehicle test are carried out.An XCU controller for driverless vehicle is developed and HIL testing based on dSPACE has been completed.Then,the vehicle test is completed on the driverless vehicle BYD.The vehicle test results show that the control algorithm can control the driverless vehicle from idle speed to the target speed of 15 km/h or 30 km/h in about 4～5 s with good responsiveness and stability.Furthermore,the adaptive ability of the control algorithm is validated by tests with changes in road curvature and target speed while parameters of the controller unchanged.That is to say,both the real-time performance and effectiveness of the longitudinal control algorithm are verified.In summary,the active disturbance rejection longitudinal control algorithm based on model feedforward and ADRC feedback can effectively resist internal and external disturbances such as vehicle parameters and road environment changes with much more satisfying adaptive performance.