Volume IV-4/W7 | Copyright
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-4/W7, 3-10, 2018
https://doi.org/10.5194/isprs-annals-IV-4-W7-3-2018
© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

  20 Sep 2018

20 Sep 2018

SHORT TERM URBAN TRAFFIC FORECASTING USING DEEP LEARNING

G. Albertengo and W. Hassan G. Albertengo and W. Hassan
  • Dept. of Electronics and Telecommunications, Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129 Turin, Italy

Keywords: Deep Learning, traffic flow prediction, urban traffic forecasting

Abstract. In today’s world, the number of vehicles is increasing rapidly in developing countries and China and remains stable in all other countries, while road infrastructure mostly remains unchanged, causing congestion problems in many cities. Urban Traffic Control systems can be helpful in counteracting congestion if they receive accurate information on traffic flow. So far, these data are collected by sensors on roads, such as Inductive Loops, which are rather expensive to install and maintain. A less expensive approach could be to use a limited number of sensors combined with Artificial Intelligence to forecast the intensity of traffic at any point in a city. In this paper, we propose a simple yet accurate short-term urban traffic forecasting solution applying supervised window-based regression analysis using Deep Learning algorithm. Experimental results show that is it possible to forecast the intensity of traffic with good accuracy just monitoring its intensity in the last few minutes. The most significant result, in our opinion, is that the machine can generate accurate predictions even with no knowledge of the current time, the day of the week or the type of the day (holiday, weekday, etc).

Download & links