Collision Risk Assessment Model On Autonomous Tram Based On Trajectory Prediction

The tram has the characteristics of large passenger capacity,low pollution and convenient driving,and has been put into use in tourist attractions such as Chengdu and Dalian.Unmanned technology has aroused the attention of many researchers,and various types of applications have also been developed on trams.In the study of unmanned driving combined with trams,safety is the most important thing for its promotion.Active safety of vehicles usually uses kinematic models to predict vehicle trajectories,and uses predicted trajectories and kinematic formulas to determine the degree of real-time collision risk between vehicles.However,the use of kinematic models to predict vehicle trajectories has the problem that the scene is relatively simple,and it is impossible to analyze complex road conditions and poor prediction results.Therefore,this paper combines the track prediction method based on(long short term memory,LSTM)neural network with the multi particle collision model of the tram,and proposes the collision risk assessment model of the self-driving tram based on the track prediction.The main work is as follows:(1)Considering that the trajectory data has the characteristics of time series in the process of moving objects,the LSTM deep learning algorithm with strong ability to describe the time series is used to construct the trajectory prediction model.The characteristics of time series trajectory data are extracted by LSTM to get the predicted trajectory,and the influence of different parameters on the accuracy of trajectory prediction is analyzed.The experimental results show that the mean square error of trajectory prediction is 0.094,which shows that LSTM is very effective for vehicle trajectory prediction.(2)Considering that the collision risk is inversely proportional to the relative position of the surrounding vehicles and proportional to the speed during the running process of the tram.Based on the simplification of the tram into a multi particle model,a collision risk assessment model for driverless trams is proposed.The traditional collision evaluation model simplifies the vehicle to the elemental point model and calculates the relative distance between the particle and the particle for the collision evaluation.However,due to the long body of tram,the single particle model will lead to the unreal results of collision evaluation.After the tram is simplified to a multi-particle model and the predicted trajectory is combined with the time series to analyze whether the tram is in collision at each moment and the degree of collision risk at each moment.The feasibility of the collision risk assessment model is verified in the simulation experiment.After realizing the multi particle tram collision risk assessment model,the real-time collision situation of trams can be predicted,and the real-time risk degree between trams and surrounding vehicles can be analyzed.Based on the real-time risk degree,the speed of the tram can be controlled to ensure the safety of the tram.