Vehicle trajectory prediction in urban traffic networks using deep learning

Abstract

Our work focuses on vehicles positions prediction which is fundemental problem that can be deployed in many important smart vehicles technologiesand services related.The main objectives are two folds. First we investigate Deep Learning (DL) based models to generate prediction .Second objectif concerned wether we can relay on artificial (simulated ) data to generate accurate models we needed. To do that we have designed two Deep Learning (DL) approches LSTM and CNN. In the experiments we have used a set of tools; keras and several python librairies. concerning the data we have used San Francisco mesured data and Sumo simulator to generate artificial data. The results showed that LSTM model is a little more efficient than CNN, also result showed that artificial data has a suitable degree of reliability.