Intelligent transportation systems have gained a significant momentum in recent years due to the advancements in computing hardware, sensors and related infrastructure. A key component of the aforementioned systems is traffic flow prediction; that is, the ability to forecast the evolution of traffic volume in future time windows, either short-term or long-term. Benefits of traffic flow forecasting on city, regional, trans-regional roads and highways include the ability of drivers to better plan their journey in order to avoid traffic congestion, the capability of regulation authorities to apply proactive policies such as diverting traffic to alternative routes so as to avoid foreseen bottlenecks or even the option of highway concessionaires to adjust their tariff policies based on the predictive volume of traffic. In this work, a novel traffic flow prediction system, based on deep recurrent neural networks, that has been developed from scratch, is presented. The proposed system is trained on traffic data collected at the entrance toll stations of Olympia Odos, a major highway in Greece, and is subsequently deployed to the infrastructure of Olympia Odos S.A, the concessionaire of the highway, with predictions being publicly available via a web interface and mobile applications. The overall architecture exhibits more than satisfactory performance both in normal traffic conditions and in cases of increased traffic (weekends, public holidays etc).
Deep Recurrent Neural Networks for Highway Traffic Flow Prediction at Toll Stations
2024-05-14
975644 byte
Conference paper
Electronic Resource
English
Optimization of lane configuration of highway toll stations considering short-time traffic flow
British Library Conference Proceedings | 2023
|Highway Toll Management and Traffic Prediction Using Data Mining
Springer Verlag | 2020
|Highway Short-Term Traffic Flow Prediction with Traffic Flows from Multi Entry Stations
British Library Conference Proceedings | 2020
|