Robust Path-following Control For Steer-by-wire Vehicle Considering Characteristics Of Human Driver

With the rapid development of sensor fusion,navigation,and communication technology,and the continuous improvement of road infrastructure,"intelligent vehicle,networked vehicle,electrified vehicle,shared vehicle"have become the current trend of the automotive industry.In recent years,various levels of self-driving car technologies have been rapidly developed since many automotive companies and Internet companies launch research and development of intelligent vehicles.For semi-autonomous vehicles,the driver-automation shared steering control enables the driver to share the steering authority of the vehicle with the steering assistance controller in a cooperative way,which contributes to improve the handing stability of the vehicle,and reduce the driver's workloads.This paper is mainly about the robust path-following control for steer-by-wire(SBW)vehicle considering different human drivers'characteristics.The SBW,as the most advanced steering technology,can directly correct the driver's steering operation by the shared controller.Note that different drivers possess different driving characteristics,the difference in characteristic parameters of the driver will make the design of the shared controller different.Therefore,by considering the driver's characteristics when designing the shared controller,not only the steering authority between the driver and the controller can be personalized,but also the steering conflict can be minimized during the human-machine interaction.The main contents and contributions of this paper are as follows.Firstly,a single-point preview driver-vehicle-road model used for path-following is presented.The driver-vehicle-road model is composed of a combination of a vehicle-road model and a driver model.Since the longitudinal motion of the vehicle is not considered,only the lateral and yaw motions of the vehicle in its own coordinate system and the lateral motion in the geodetic coordinate system are handled when building the vehicle-road model.As for the single-point driver model,it is established under the assumpation that the driver only takes steering actions through the road features observed at the preview point.Considering the driver's look-ahead behavior,brain response delay,neuromuscular delay on the arm,and PD adjustment to the vehicle's lateral motion,the driver model including preview time,delay time,and steering proportional gain is finally established.Secondly,since the single-point driver-vehicle-road model builded above is not suitable for tracking large-curvature path with"return characteristics",it is optimized by the two-point preview driver-vehicle-road model.The modeling work of the deviation of vehicle's directional angle and lateral position are added into the vehicle-road model.The feedforward and feedback control is adopted for the two-point preview driver model.The far preview point is used to obtain the features of the road ahead while the near preview point is used to eliminate the deviation of the directional angle and lateral position.Finally,the driver model including anticipatory steering gain,compensation steering gain,derivative time constant,preview time,and delay time is obtained.Thirdly,the uncertain driver's characteristic parameters are dealt with the polytope theory and fuzzy control,respectively.The nonlinear driver-vehicle-road models with uncertain parameters are transformed into linear parameter-varying(LPV)models,which greatly facilitate the design of the shared steering controller.Finally,output-feedback,robust,shared controllers based on linear matrix inequality(LMI)are designed,and the pole placement constraints are added.The proposed controllers can decrease the cost in sensors as it's unnecessary to measure the expensively obtainable vehicle lateral velocity,which makes the controller more valuable from the application perspective.In addition,we convert the multi-objective H_?performance into a single-objective one with a smaller interference suppression degress,and finally the problem is transformed into a convex optimization problem which can be solved by LMI toolbox in Matlab.At last,the effectiveness of the controllers is verified by the joint simulation of Matlab/Simulink-Carsim.