Research On Modeling Of Driver-Vehicle-Road System And And Trajectory Planning Based On Kinematic Vehicle Model

"Intelligence,network,electric and sharing" is the current trend of the automotive industry.Autonomous vehicles and semi-autonomous vehicles are the research focus of many technology companies and automobile manufacturers at home and abroad.Intelligent vehicles should be able to exchange and share information with X(Vehicle,Road,Human,etc.),and have the functions of complex environment perception,intelligent decision-making,and cooperative control,etc.Its research goal is to replace human drivers to drive the vehicle autonomously,and to improve the driving safety and efficiency.However,the fully automated driving technology that liberates drivers from boring driving is not achieved overnight,but a step by step process.Compared with fully autonomous driving,semi-autonomous driving is now more likely to be realized,which has become a research hotspot in the field of automobile.In the process of semi-autonomous driving,the control authority of the vehicle is shared between the human driver and the automatic control system,and the drivers with different driving characteristics will show different vehicle handling behaviors in face of the same driving task.At this time,how to take into account the driver’s differential driving characteristics in the automatic control system,ensure the driving safety,at the same time,make the car have excellent drivable performance,to achieve the human driver’s easy and comfortable driving behavior,is a crucial step.In this paper,the single-point/double-point driver model is established based on the visual preview characteristics of the drivers on different roads,and the driver-vehicle-road(DVR)closed-loop system is further established based on the kinematic vehicle-road model.A multi-objective model predictive trajectory planner is designed to help drivers with individual driving characteristics to plan different safety trajectories,greatly reduce the physiological and psychological load of drivers,and improve the drivers’ driving comfort,which has important theoretical reaearch significance and engineering practice value.The main contents of this paper include:(1)Modeling of Driver-Vehicle-Road system and the design of collision avoidance path planner applicable for straight and other roads with small curvature.A DVR closed-loop system is established by integrating the kinematics vehicle-road model and the single-point visual preview driver model considering driver’s delay time,preview time and the steering proportional gain.The effectiveness of the proposed system is verified by comparing with the nonlinear whole vehicle model in Carsim through the path tracking effects of a Single lane change condition.The artificial potential field method and the circle decomposition method are used to construct the virtual potential field mountain of the obstacle vehicle and the lane boundary to describe the collision constraints.Using multi-objective linear time-varying model predictive control(MPC)method,considering the key factors such as collision security constraints,drivers’ individual control characteristics,global reference path tracking and actuator constraints,a path planner is designed to help different drivers plan out the safe trajectories.The effectiveness of the controller is verified by the simulation of stationary and moving obstacles collision avoidance.(2)DVR system modeling,that is applicable for large curvature roads.In order to solve the defect that the single-point preview DVR system in the previous chapter is only suitable for straight and small curvature roads,the feature information of the road ahead can be obtained from the far point on the lane boundary to prepare for the steering behavior at the next moment,which serves as the driver’s feedforward control.The lateral position deviation and heading angle deviation of the path tracking are eliminated by the information of the preview point on the target path,as the feedback control,a two-point visual preview driver model is established,which includes the expected steering gain,compensated steering gain,differential time constant,preview time,brain response and neuromuscular delay time.Combining with the kinematics vehicle-road model,the DVR closed-loop system with two-point visual preview is established,and the effectiveness of the system is verified by comparing the tracking effects of a 100 m radius large curvature road with the nonlinear full-vehicle model in Carsim.(3)Design of path planner for collision avoidance in large curvature roads.The artificial potential field method and circle decomposition method are used to increase the longitudinal influence range of obstacle vehicles on the road with large curvature by adding different numbers and different radii of touchable constraint circles in front and rear of the obstacle vehicle.And the virtual potential field mountain suitable for obstacle vehicles and lane boundaries on large curvature roads are constructed as soft constraints for collision avoidance.The multi-objective linear time-varying MPC method is adopted,and the key factors such as collision security constraints,driver’s individual control characteristics,the global reference path tracking and actuators constraints are considered to design the multi-objective model predictive path planner,which can plan the safe pathes for different drivers.The validity of the planner is verified by the simulation of 100 m radius large curvature roads collision avoidance.In order to solve the problem that the vehicle’s lateral acceleration is too large when returning to the original lane after obstacle avoidance,which will affect the driver’s comfort.A heuristic method is proposed,which is more in line with the driver’s driving habit,and the effectiveness of the method is verified through simulation test.(4)Experimental research on Miniature intelligent model vehicle.The high-precision GNSS/INS and 16-line Lidar is used as the sensing system,Minipc and Single-chip microcomputer(SCM)as the control system,a Miniature intelligent model vehicle experimental platform is built.To solve the problem of poor tracking effect due to the fluctuation of GNSS data,the least-square method is used to optimize the initial reference path.The effectiveness and environmental adaptability of the designed path tracking controller is verified by experiments under various working conditions.Considering that in the actual random time-varying complex traffic environment,it is necessary for intelligent vehicles to avoid the encountered obstacles.The key factors such as safety performance,real-time performance,tracking error and actuator limitation are considered,using the cubic spline curve method,and the size and position information of obstacles are combined,to solve the obstacle avoidance replanning trajectory.The pure tracking method is used to calculate the required front wheel steering angle and track the replanned trajectory.The effectiveness of the designed local path planner is verified by the stationary obstacle collision avoidance experiment in continuous curves.