Research And Experimental Verification On Autonomous Lane Changing Decision Of Intelligent Vehicle

In recent years,with the increasingly serious traffic safety situation and the continuous warming of traffic congestion problems,vehicle intelligence has become the key research direction of the future development of the whole automobile industry.Intelligent vehicle can replace people to manipulate vehicle and realize autonomous driving.It plays an important role in improving traffic safety and efficiency,but its research and development and application are still facing many challenges.This topic is based on the development status of intelligent driving key technologies,relying on the Tianjin Science and Technology Innovation Platform Plan（the major projects of artificial intelligence science）to study the autonomous lane-changing decision-making method of intelligent vehicle.The main goal is to establish an autonomous lane-changing decision-making mechanism for intelligent vehicle to meet rationality,safety and comfort.Firstly,based on the analysis of human driver’s lane-changing behavior characteristics,this paper designs a typical lane-changing experimental scene,carries out the driver’s simulation driving experiment based on the simulation driving platform,constructs the sample database of lane-changing driving behavior and analyzes the main characteristic parameters.It is concluded that the free lane-changing time of the driver is usually 4～6 s,and the maximum lateral acceleration is about 1 m/s~2,which provides a basis for the formulation of lane-changing decision-making strategies.Secondly,this paper fully considers the driver’s lane changing intention generation process,and divides the autonomous lane changing decision-making problem into two parts:behavior decision-making and trajectory planning.In the specific behavior decision-making process,human driver driving behavior experience is learned based on deep learning LSTM method,so that the timing of the lane changing intention of the intelligent vehicle is the same or similar to the human driver,and the driver’s acceptance and satisfaction of the intelligent vehicle are improved.Aiming at the problem that the neural network model can achieve high accuracy after effective training,but it cannot guarantee the absolute correctness and reliability of decision-making,a feasibility judgment rule of lane changing considering the rationality and safety of Lane Changing is proposed,so that the behavior decision can not only fully reflect the characteristics of human drivers’driving behavior,but also ensure that only reasonable lane changing decisions are implemented.Then,in this paper,a variety of common lane changing trajectory planning models are compared and analyzed,and the Gaussian distribution-based trajectory planning method is selected to establish the lane changing trajectory planning model.Considering the safety and comfort requirements of lane changing,the best lane changing reference trajectory is selected.The trajectory tracking controller is designed through the model predictive control（MPC）theory to realize the trajectory tracking control of lane changing,and then the correctness of the decision instruction of lane changing behavior and the rationality of lane changing trajectory planning are verified.Finally,based on the co-simulation platform with Pre Scan/Simulink/Car Sim and the unmanned general vehicle platform,the simulation experiment and the semi-physical platform experiment are carried out respectively.The experimental results verify the rationality and effectiveness of the autonomous lane changing decision mechanism proposed in this paper.It is proved that the lane changing decision mechanism proposed in this paper can realize the autonomous lane changing of intelligent vehicle,and has good engineering application prospect.