REAL-TIME TRAFFIC DENSITY ESTIMATION UNDER HETEROGENEOUS MIXED TRAFFIC CONDITION

Abstract

Traffic density is one of the essential macroscopic parameters and it is of prime importance when a facility is evaluated from the perspective of both users and planners. It is a major congestion indicator. Road traffic congestion is a severe issue with many detrimental implications for the economy and the environment. Traffic conditions are predicted as a result of recent advances in Intelligent Transportation Systems (ITS). Implementing traffic control strategies using ITS requires an accurate real-time estimate of traffic characteristics. One of the key features of ITS applications is predicting real-time traffic density. The present study estimates real-time traffic density using several Machine-Learning (ML) regression models and Kalman Filter (KF). Macroscopic traffic flow models are formulated and using this model, a model-based estimation scheme are designed based on the Kalman filtering technique. The data required for implementing this method in the field requires traffic flow at entry and spot speeds at entry and exit locations. The proposed method is validated using the input-output method of density estimation. The suitable method is identified for predicting traffic density for the given study stretch. It is found that Machine learning methods are more efficient in predicting traffic density when compared to Kalman Filter. Among different ML regression models, Random Forest is found to be more accurate in density prediction