Performance Comparison of Filtering Techniques for Real Time Traffic Density Estimation Under Indian Urban Traffic Scenario

Real time traffic state estimation is important to facilitate better traffic management in urban areas and is a prime concern from a traffic engineer's viewpoint. Traffic density is a key traffic variable that can be used to characterize the traffic system and can be a valuable input to the functional areas of Intelligent Transportation Systems (ITS). However, measurement of density in the field is difficult due to several practical limitations. This creates a need for inferring density from other traffic variables that are easily measurable in the field. In this paper, model based approaches for the estimation of traffic density are discussed. The non-linear model equations are based on the conservation principle and the fundamental traffic flow. The technique used for recursive estimation of density in real time plays a key role in terms of estimation accuracy. The Extended Kalman Filter (EKF) is a common tool for recursive estimation for nonlinear systems. This study investigates the application of particle filter (PF) and Unscented Kalman Filter (UKF) as alternatives to (EKF) for nonlinear traffic state estimation in the context of traffic conditions in India. The estimated density values were corroborated using manually extracted field density values. The performance of these methods was also compared with a base model, where the fundamental traffic flow equation was used for calculating density. The convergence properties of these filters were also analyzed.