@article{ref1,
title="Enhancing lane change prediction at intersections with spatio-temporal adequacy information",
journal="Journal of big data analytics in transportation",
year="2022",
author="Fafoutellis, Panagiotis and Plymenos-Papageorgas, Jason and Vlahogianni, Eleni I.",
volume="4",
number="1",
pages="73-84",
abstract="Lane changing has been highlighted as one of the main causes of near misses and road accidents both in urban road networks and highways. During the last decade, a significant effort has been made by researchers to model and predict the exact time a driver starts a lane changing maneuver and to investigate the factors that may affect the above decision. However, lane changing at intersections remains an under-researched area due to the complexity of the maneuvers and the need for very detailed trajectory data. In this paper, a methodology for modeling the likelihood of lane changing at intersections is proposed based on a comprehensive feature extraction mechanism, which attempts to quantify the favorability of the surrounding environment towards lane changing and simple LSTM modeling structures. The models are trained on data coming from a highly congested intersection in the city of Athens (Greece). <br><br>FINDINGS indicate that the proposed lane changing model trained to predict the occurrence of a lane changing maneuver within 1 and 5 s ahead, achieves an accuracy of about 96% for both applications (single step and multiple step prediction). These results show strong evidence that, regardless of the complexity of the problem at hand, simple predictive deep learning architectures enhanced with meaningful spatio-temporal representation and related features can achieve a high level of performance.<p /> <p>Language: en</p>",
language="en",
issn="2523-3556",
doi="10.1007/s42421-022-00055-6",
url="http://dx.doi.org/10.1007/s42421-022-00055-6"
}