Lane Change Risk Prediction And Trajectory Planning For Intelligent Vehicles In Multi-vehicle Environment

It will be a long time before cars are fully unmanned,so unmanned intelligent vehicles and manned vehicles will be on the road together for a long time.In this stage,the intelligent vehicle will face the complexity and uncertainty of the surrounding traffic environment when driving on a road with multiple lanes and involving multiple surrounding vehicles.When the intelligent vehicle performs the lane change action in this environment,it faces a greater risk of collision,which forms a higher requirement for the safety and reliability of the trajectory planning algorithm.Therefore,aiming at the multi-vehicle traffic environment of structured roads,it is of great significance to carry out the research on lane change risk prediction and trajectory planning methods.To predict the risk of lane changing behavior,an objective and reasonable risk quantification method should be established for lane changing trajectory.Based on the traffic conflict theory,the lane changing process is divided into three stages,and the potential conflict risks of the lane changing process are analyzed by stages.Based on the stopping distance index and lost energy index,combined with the fault tree analysis method,a comprehensive trajectory risk quantification method considering both risk exposure level and risk severity level was established.Based on the open source natural driving data,the comprehensive trajectory risk quantification method is applied and analyzed,and the rationality and objectivity of the method are verified.Considering the long-term prediction demand of lane change risk,the LSTM neural network is used to build the trajectory prediction model of side vehicles.According to the longitudinal motion characteristics of the side vehicle at different positions,the method of data classification and model subdivision was adopted,and the longitudinal velocity and lateral displacement which were more consistent with the driver’s control intentions were selected as the output characteristics of the model to train the trajectory model,so as to achieve the trajectory prediction effect of the side vehicles with higher accuracy and better robustness.A quintic polynomial model with the optimal comfort is selected and a lane change trajectory planning method considering the predicted risk is designed.Based on the alternative lane sampling trajectory generated by the trajectory model,a target lane decision-making method was proposed by calculating the risk mean of the sample trajectory.The selection criteria of trajectory feasibility were established from three aspects of collision feasibility,kinematics feasibility and stability feasibility.The loss terms of lane changing efficiency,risk and comfort were derived and the trajectory evaluation function was constructed to select the optimal trajectory,so as to realize safe lane changing in multi-vehicle environment and maintain good comfort at the same time.The real scene data were collected,and the co-simulation platform of Simulink and Car Sim was built to realize the simulation verification of lane change trajectory planning function considering the predicted risk under two actual real scenes,urban expressway and expressway.The results show that the lane change trajectory planning method can effectively realize the human-like decision of the target lane and the safe lane change trajectory planning of the vehicle in the real multi-vehicle environment,and the vehicle can meet the requirements of stability and comfort in the process of lane change.