Research On Test And Evaluation Method Of Lane Change Control For Autonomous Vehicle

The application and development of autonomous vehicles requires sufficient testing and evaluation work.As one of the most basic driving behaviors,lane-changing behavior needs to be evaluated from the aspects of safety and efficiency,so as to avoid autonomous vehicles in Unnecessary accidents occur in mixed traffic flow environments.However,there is currently a lack of a relatively complete test and evaluation method for automatic driving lane change control,and there is also a lack of indicators for evaluating automatic driving control humanlike driving.To this end,this paper proposes a lane-changing test and evaluation method for autonomous vehicles based on reference trajectories.At the same time,a human-like index is added to the index,a prediction model of human lane-changing trajectory based on spatiotemporal attention mechanism is proposed,and an index calculation method is constructed.The main research contents include the following aspects:First,based on data requirements,compare the current natural driving trajectory data set,select the High D data set as the research sample,and perform preprocessing operations on the data set according to the follow-up research needs.In order to accurately extract the trajectory of the lane-changing execution process,a rule-based method for determining the start and end times of lane-changing is proposed.Second,build a highway lane changing trajectory prediction model based on the spatiotemporal attention mechanism and TCN and LSTM neural networks.By setting different network parameters,the prediction effect of the model is compared,and the optimal network hyperparameters are determined,and compared with the prediction effect of other deep learning models.The results of comparative experiments show that the proposed model achieves better accuracy in both the horizontal and vertical positions of trajectory prediction.Third,a reference trajectory-based test and evaluation method for autonomous vehicle lane change is proposed,which includes test site modeling,test item modeling,evaluation index selection,and evaluation criteria.The reference trajectory is generated by using the lanechanging trajectory prediction model,and the human-like index is calculated and analyzed by combining the actual trajectory and the reference trajectory,which is used for the evaluation needs of the mixed traffic flow environment.Finally,the feasibility and effectiveness of the proposed method are verified by virtual simulation verification with driving simulator software.