Traffic Congestion Level Prediction Based on Recurrent Neural Networks

Abstract

In recent years, traffic congestion has become a global concern. Many researchers work on the development of Intelligent Transportation Systems (ITS) to reduce traffic congestion and improve transportation efficiency. Traffic Information such as vehicle speed, traffic volume, and inter-vehicle spacing are common indicators to determine traffic congestion situation. However, most existing studies only use a single indicator to estimate traffic congestion status. In this paper, we propose the Road-condition-based Congestion Prediction System (RCPS) that uses both traffic volume and vehicle speed to predict traffic congestion. The proposed solution collects real-time road images taken from camera drones to extract traffic volume and vehicle speed on the road. The extracted traffic indicators are then used to predict the congestion level in the future. Using two traffic indicators instead of one, the RCPS achieves high accuracy in congestion level prediction. The RCPS predictions can also been shown on the APP developed for it. It is expected that the road congestion level prediction provided by the RCPS will provide valuable information for drivers to choose the best route.