ADFA-LSTM: An Abnormal Trajectory Prediction Method Based on Bionic Neural Network

Yan Wang, Gang Cui, Jun Qiang Zhou, Zhi heng Han

Abstract:

In order to detect the state of vehicle trajectory, an abnormal prediction method of vehicle trajectory based on bionic neural network, called ADFA-LSTM, is proposed according to the temporal characteristics and position relationship of vehicle trajectory. The long-short memory neural network (LSTM) is used to predict the trajectory in the short term based on the characteristics of the vehicle’s historical trajectory, and the trajectory prediction results are analyzed to identify the abnormal trajectory. In the process of trajectory prediction, the improved
biomimetic firefly algorithm (FA) is used to optimize the prediction the parameters of LSTM model, which improves its processing efficiency and enhances its fitting ability. Through a lot of experimental analysis, the detection effect of the algorithm is verified.