Research On The Scenario Analysis And Decision-making Method Of The Lane Change During Fast Driving

With the rapid development of autopilot and its broad application scenarios,so that the performance of intelligent driving technology have higher requirements,Behavioral decisionmaking system as a key component of which is particularly important.Considering the stage of mass production of intelligent driver assistance vehicles mainly for the action in the lane,it has not been involved in the car lane changing decisions.For this reason,this paper mainly studies the analysis and decision-making of lane changing scenarios when the car is driving fast.Based on the information provided by existing sensors,a lane changing scenario analysis model is designed,including traffic information classification,traffic information processing,and lane changing intention generation.Considering comfort and safety,the lane change decision framework of this paper is designed,including lane change trajectory planning,safety detection,and traffic lane change intention recognition.Finally,the trajectory following control is designed based on the planned lane change trajectory.In the analysis of lane changing scenarios,a large amount of traffic scene information is given according to the sensors.First,a traffic information classification model is used to classify all the information and propose the movement and location information required by the traffic participants.However,some of the above information cannot be directly obtained through sensors,and a traffic information processing model is designed to process the sensor information,record the historical trajectory of traffic vehicles,and divide the driving area.Next,the processed information was used to screen and analyze the scenarios that may generate lane changing intent,and corresponding lane changing intent generating methods were designed for different scenarios.At the same time,it is simulated to verify the rationality of its intention.To realize the lane change intention,a lane change decision algorithm is designed in this paper.Considering the driver’s comfort during the lane change process,the longitudinal and lateral acceleration peaks are used as constraints,and the polynomial planning method is used to plan the lane change trajectory.Then,a safety test is performed on the trajectory,and the detection mainly determines whether the lane change trajectory is safe by calculating the longitudinal safety distance of the own vehicle and the traffic vehicle.And,the hidden Markov model is used to identify the lane change intentions of the traffic vehicles,and the traffic vehicle is divided into three lane change behaviors: lane keeping,lane change to the left and lane change to the right.The matching degree of each model to the same observation sequence is compared.The high matching degree is the lane changing behavior corresponding to the observation sequence.To realize the vehicle following lane change trajectory and desired speed,a lane change following model based on model predictive control and longitudinal vehicle speed control model are designed.First,a vehicle dynamic model is established to describe the vehicle’s movement.Then,the rolling time domain optimization problem is described,and the trajectory following problem is transformed into a constrained objective function solution problem.The Hilderain quadratic programming algorithm with a linear optimization problem is then used to quickly solve the optimal sequence of expected control variables.The control variable sequence is output to the EPS to realize the lane changing behavior.The longitudinal vehicle speed control model uses expected vehicle speed and actual vehicle speed to determine the desired acceleration to be applied to the vehicle through PI control.Finally,the trajectory following algorithm proposed in this paper is simulated and verified under two road conditions,straight and curved,and proves that it has a good following effect.As the construction of the virtual traffic scene,how to determine the movement of the traffic vehicle becomes a key factor.Therefore,a real vehicle test platform is set up in this paper to collect the traffic scene information of the actual road,and the information is filtered.Then,the data is used in the SCANeR software to establish a virtual scene that can reflect the real traffic.The platform tested the algorithm proposed in this paper.The experimental results show that the algorithm can effectively analyze the lane change intent scene in a typical lane change scenario,and can make a comfortable and safe lane change trajectory while accurately controlling.The vehicle follows the lane change track to complete the lane change behavior.