Traffic Condition Detection in Freeway by using Autocorrelation of Density and Flow

Traffic conditions vary over time, and therefore, traffic behavior should be modeled as a stochastic process. In this study, a probabilistic approach utilizing Autocorrelation is proposed to model the stochastic variation of traffic conditions, and subsequently, predict the traffic conditions. Using autocorrelation of the time series samples of density and flow which are collected from segments with predefined specifications is the main technique to detect the trend in flow and density changes if exist. A table of possibilities for flow and density changes in two sequential segments will help to detect congestion or any other abnormal traffic events.
In this study proposes a stochastic approach to predict the traffic situation in freeway. The dynamic changes of freeway traffic conditions are addressed with state transition probabilities. For sequence trends of density and flow change, using autocorrelation of speed and flow series will estimate the most likely sequence of traffic states. This is the novelty in this paper that introduces a robust method to recognize the traffic state in a segmented freeway. According to the model definitions 3-state traffic pattern prediction implemented as No Risk (NR), Risk (R) and High risk (HR). We evaluated the proposed method using different data sources of real traffic scenes from Tehran-Qom freeway, Iran. A total of 480 minutes, which corresponds to interstate highways, are chosen for testing. The number of passed vehicle and mean speed are collected by six traffic counter every 1 minute. The estimation rate of this model is 95% over a short time period for the month of July 2014.