Predicting Vehicle Lane-changing Behavior with Awareness of Surrounding Vehicles Using LSTM Network

Abstract

Autonomous vehicles often need to predict the trajectories of surrounding vehicles for planning and decision making. We propose a socially-aware long short-term memory (LSTM) algorithm in real lane changing scenario to solve the intention recognition and trajectory prediction problems, taking advantage of both the past trajectories of host vehicle and the states of its neighbors. The proposed method is verified, tested and analyzed through the open and useful NGSIM dataset. We compare our approach with the existing traditional motion prediction methods, and our results show an improvement in terms of RMSE values of prediction error.