Long-term prediction for high-resolution lane-changing data using temporal convolution network

Lane-changing is an important driving behaviour and unreasonable lane changes can potentially result in traffic accidents. Currently, the lane-changing data are often recorded with high resolution, which not appropriate for some common deep learning approaches. To capture stochastic time series of high-resolution behaviour, this study introduces a temporal convolutional network (TCN) to predict long-term trajectory behaviour. The dataset was collected by simulator at frequency 60 Hz. Prediction results show that TCN accurately shorter computational compared two benchmark models including neural (CNN) long short-term memory (LSTM). advantages rapid response accurate prediction, assistance advanced driver system.