Research On Trajectory Planning For Intelligent Vehicles In Uncertain Environments

In recent years,the rapid development of computer technology has injected new vitality into the automotive industry,and the wave of intelligentization in the automotive industry continues to advance.However,due to the complex and changeable traffic environment and the inability to anticipate the intentions of traffic participants in advance,achieving fully automated driving will face enormous challenges.Specifically,in scenarios with high uncertainty,such as unsignalized intersections,reasonable planning of the main vehicle trajectory has become an extremely challenging task.Therefore,studying vehicle trajectory planning in uncertain environments has significant practical significance.The objective of this study is to develop a method for autonomous vehicles to plan collision-free and comfortable trajectories in scenarios with high uncertainty,such as unsignalized intersections.The primary focus of this research paper includes the following:(1)Generate a rough path for the main vehicle based on deep learning methods.First,data mining rules are developed to mine and clean the Argoverse 2 trajectory prediction dataset,and obtain cases with good quality for the scenarios required for this study.Second,annotation rules are developed to implement batch direction annotation of the mined scenarios.Then,after preprocessing the scenario data,a deep learning model was constructed,and the model was trained using the direction labels obtained from batch annotation.Finally,the output results of the model on the test set are computed,and the results show that the model constructed in this study can output reasonable paths for the main vehicle.(2)Optimize the initial path of the main vehicle and construct a reference line.First,a nonlinear constraint problem is constructed to optimize the initial path generated by the deep learning model,and smoother path points of the main vehicle are obtained.Then,the smooth path points of the main vehicle are interpolated using the cubic spline interpolation method to obtain a continuous path,and the Frenet coordinate system of the main vehicle is established using the path curve as the reference line.(3)Match the optimized path of the main vehicle with speed.First,the construction process of ST graph is clarified.Second,the problem of solving the speed of the main vehicle in the ST graph is modeled as a Dynamic Programming(DP)problem,and the DP algorithm is used to solve the rough speed of the vehicle.Then,in the convex space created by the DP process,considering the continuity constraints,velocity constraints,and acceleration constraints of the velocity curve,the Quadratic Programming(QP)method is used to smooth the velocity curve.Finally,the coupling between the path and velocity curve of the vehicle is achieved under the Frenet coordinate system,which is constructed based on the reference line of the main vehicle.Consequently,the trajectory of the main vehicle is obtained.(4)Verify the effectiveness of the algorithm.First,the trajectory planned by the algorithm proposed in this study is compared with the real-world trajectory from the Argoverse 2 dataset,which preliminarily proves the effectiveness of the planning algorithm;Then,dynamic simulation scenarios of unsignalized intersections and T-intersections are built in SUMO,and the map and vehicle trajectory data in the scenario are extracted to construct the input of the deep learning model.And,A bicycle model is constructed as the vehicle model of the main vehicle,and a Stanley trajectory tracking controller is constructed to track the trajectory planned in this study.Finally,by using the combined simulation of SUMO and Python,the main vehicle is simulated in scenarios such as straight,left,and right turns at unsignalized intersections,and straight ahead at T-intersections.The results show that the algorithm proposed in this study can plan a reasonable driving trajectory,and ensure the safety and comfort of the vehicle driving in a highly uncertain environment.