Trajectory Prediction And Safety Decision For Vehicle Lane Change

As an important safety means to avoid vehicle accidents,vehicle safety technology has been increasingly applied today.Compared with the passive safety technology,the active safety technology can warn the driver before the collision accident,and provide the safety decision for the subsequent vehicle driving to avoid the accident.Trajectory prediction and safety decision are important components of active safety technology.Among them,the vehicle trajectory prediction technology can use current and historical vehicle operating states to make predictions about vehicle position,speed and other motion parameters in the future period.The prediction result can provide support for judging the safety of the vehicle during the prediction period.Safety decision-making is based on the vehicle’s state and surrounding environment to make subsequent safety judgments on vehicle operations and behaviors.If there is a possibility of collision,the driver could get a warning to adjust subsequent vehicle operations.Through the trajectory prediction and safety decision of the active safety technology,the safety level of vehicle operation can be greatly improved to avoid collision accidents.Compared with vehicles driving straight,lane-changing behavior as a common traffic behavior is more complicated and variable,and more factors need to be considered.These factors make trajectory prediction and safety decision making for lane change relatively difficult.In order to solve the problem of vehicle trajectory prediction and safety decision-making during the lane change process,this paper uses high-precision GPS instruments and VLP-16 laser radar to build a experimental platform,and collect the positions,speeds and surrounding environment information on the platform.Through the vehicle experiments on urban roads,a large amount of real data about the lane change trajectory of the vehicle and its surrounding movement environment were collected.On the basis of absorbing trajectory prediction and safety decision research results at home and abroad,this paper studies the trajectory prediction and safety decision of vehicle lane change.The specific research contents are as follows:1.The original experimental data collected by high-precision GPS instruments and VLP-16 laser radar were processed and cleaned.And the lane change trajectory data and the corresponding surrounding point cloud data were extracted.For lane change trajectory data,a trajectory clustering algorithm based on improved Hausdorff distance is established,which implements clustering of lane change trajectory and provides data support for subsequent trajectory prediction;For the surrounding environment data of the vehicle,the segmentation and clustering algorithm have been used to get the vehicle point cloud data.And according to the representation of vehicle point cloud data in space,the clustering results are identified to extract the vehicle target and relative coordinates.In order to solve the problem of vehicle point cloud association in consecutive frames,a neighboring vehicle point cloud matching method based on KM algorithm is proposed.The same vehicle point cloud in adjacent frames is correlated and matched to establish a surrounding vehicle data set which provide a data foundation for subsequent security decisions.2.A vehicle lane change trajectory prediction model based on a generated confrontation network is established,which includes a trajectory generation network and a trajectory discrimination network.Aiming at the deficiency of traditional neural network in sequence prediction,this paper establishes a trajectory generation network based on the encoder-decoder structure,and uses a MLP perceptron to establish a trajectory discrimination network.This paper uses multiple loss functions to improve the loss function of the adversarial network to adapt the requirements of vehicle trajectory prediction.By using real lane change trajectory data,the model is tested and error is analyzed.The prediction results show that the model can perform sequence prediction on lane change trajectories during the forecast period,and its prediction accuracy is improved compared with existing models.3.The lane change safety decision model based MCDM is established.First,this paper classifies the vehicle lane-changing behavior according to its causes,and discusses the conditions for the lane-changing behavior under the free lane-changing.Then,according to the movement state of the vehicle,the whole process of lane change is divided into different stages,and the types of collisions that may occur in different stages are studied.The safetydecision indicators at different stages are analyzed,and the calculation of the safety decision indicators is improved by combining the lane change prediction trajectory and the lane change decision plan.Finally,the decision model was verified and analyzed by numerical simulation and calculation.The results show that the decision model can fully consider each safety decision index and select the best scheme for lane change under the premise of ensuring the safety of vehicle.And this model can improves lane change efficiency and safety.Thus,from what has been discussed above,this paper is mainly study the trajectory prediction and safety decision of the lane change.For lane change trajectory prediction and safety decision-making,a lane change trajectory prediction model based on generated adversarial network and safety decision model based on MCDM are established respectively.In order to solve the problem of indefinite sequence prediction in trajectory prediction,a trajectory generation network based on encoder-decoder and a trajectory discrimination network based on the generation of adversarial networks were established,and the proposed model was tested and analyzed.The safety decision indicators in vehicle lane change are calculated and analyzed,and the safety decision model based on MCDM is established.The research results of this paper can make people further understand the vehicle lane changing process,and can enrich or improve the theoretical research results of the existing vehicle active safety technology.At the same time,this research can also provide a certain theoretical supplement for advanced intelligent transportation and traffic safety.