Prediction Of Pedestrian Trajectory In Mixed Traffic Conditions And Anti-collision Control Of Autonomous Vehicle

With the rapid development of intelligent driving technology,intelligent vehicles are developing towards the direction of autonomous driving with higher level and wider application scenarios.As one of the key technologies in the field of active safety,vehicle active collision avoidance system needs stronger performance.As a vulnerable group of traffic participants,improving the safety of pedestrian has become a research focus in the field of vehicle active safety.Due to the uncertainty of pedestrian trajectory caused by pedestrian's subjective intention,the traditional active collision avoidance system has some limitations.Therefore,this paper proposes a method of pedestrian trajectory prediction based on the analysis of pedestrian motion features,and designs an algorithm of pedestrian active collision avoidance system,in order to improve the safety and comfort of active collision avoidance method.Firstly,this paper analyzed the typical motion features of pedestrians and reveals the dynamic feature transfer rules of pedestrians under different trajectories.The 7-layer convolutional neural network model is used to identify the typical motion features of pedestrians,and four kinds of pedestrian intention changes are simulated based on the change of motion features,which represent the difference between the change time point of actual trajectory and pedestrian intention simply.Then,the kinematics characteristics of crossing pedestrian are analyzed,and the influence of the change of crossing pedestrian trajectory on vehicle control strategy is revealed.Combined with the comprehensive consideration of pedestrian walking intention,this paper proposes a pedestrian trajectory prediction algorithm based on motion feature analysis to predict the change of pedestrian trajectory in advance.Next,considering the influence of Pedestrian factor,Distance factor and Speed factor on the active collision avoidance control strategy,the PDS predicted safety distance model is established for different trajectory conditions.On this basis,this paper designs a pedestrian active collision avoidance algorithm,and the braking strategy is planned by means of trajectory intersection time comparison and velocity error judgment,so as to guarantee the pedestrian safety and the smoothness of the braking process.Finally,the Carsim/Simulink co-simulation model was built to compare and analyze the braking distance and deceleration values of the traditional safety distance model and the PDS predicted safety distance model,indicating the advantages of the PDS predicted safety distance model in the pedestrian active collision avoidance algorithm.At the same time,five kinds of pedestrian trajectory changing conditions are simulated to verify the effectiveness of the proposed algorithm.The safety and reliability of the pedestrian active collision avoidance system are further proved by the real vehicle experiments.