Comparative Study of Density over Time by Several Approaches Using Individual and Sample Data in the Mixed Traffic

Abstract

At the macroscopic perspective, traffic analysis requires the knowledge of Fundamental Diagram, which involves the relationship between the variables of density, flow, and speed. As one of the macroscopic traffic flow variables, density can be derived by several approaches. At first, the density of traffic was measured over space, which difficult to be collected mainly in the long section of the road. Therefore, the density variable was simply derived from the fundamental relation of macroscopic traffic flow variable. By this method, the individual speed and flow variable are required in the local measurement. Both traffic density and flow will apply the concept of PCU, which refers to the Indonesian Highway Capacity Manual, 1997 to consider the different characteristic of the vehicle. In the mixed traffic of developing countries, providing traffic data was difficult due to the limitation of the traffic sensor infrastructures. Frequently, providing density variable relies on the sample data for speed analysis. In the present study, the estimation of density will focus on the local measurement over a time interval. By using individual data, density is proposed to be measured directly over time, in which the equation can be modified to utilize the sample data. The number of sample for speed analysis will be varied to know the accuracy and the performance of each approach in the density estimation of mixed traffic. Several approaches for density estimation will be summarized and compared each other. Theoretically, the estimated density which measured over time and space by using individual data can provide the most appropriate result. So, this estimated density will be established as an actual density throughout the study. Then, the performance of each approach either using individual or sample data will be evaluated upon the actual traffic density by mean absolute percentage error (MAPE). The result shows by using the same trap length to measure the speed, the existing and the proposed approaches provide a good estimation of density either by utilizing individual data or sample data of the vehicle speed. This result was indicated by the MAPE value, which obtained under ten percent. Based on the further evaluation of the MAPE value, the performance of each approach was changed by utilizing the different category of data. In addition, estimation of traffic density which utilizes the sample data of vehicle speed has good reliability.