Travel Time Estimation and Prediction in Freeway Systems

Travel time is considered as an important performance measure for roadway systems, and dissemination of travel time information can help travelers to make travel decisions such as route choice or time departure. Since the traffic data collected in real time reflects the past or the current conditions on the roadway, a predictive travel time methodology should be used to obtain the information to be disseminated. However, an important part of the literature uses instantaneous travel time assumption, and sums the travel time of roadway segments at the starting time of the trip. The growing need for short-term travel time prediction also led to the development of forecasting algorithms. These methods can be broadly classified in two major categories; parametric methods (e.g. linear regression, time series models, Kalman filtering), non-parametric methods (neural network models, support vector regression, bayesian models, simulation models). This paper presents a predictive travel time methodology based on speed data at fixed loop detectors. However, in contrast to above mentioned existing methodologies, it benefits from the available traffic flow essentials (e.g. shockwave, bottlenecks). The proposed method makes use of both historical and real time traffic information to provide travel time prediction. First, an existing bottleneck identification algorithm is used to determine the location and spatial extent of the bottlenecks. In order to use the historical dataset in a useful and efficient manner, days with similar traffic patterns (i.e. speed profiles) should be identified. Since high number of detectors and time periods in a day lead to a large number of observations, Principal Component Analysis (PCA) is used to reduce the dimensions of the dataset. Then, Gaussian Mixture Model (GMM) is used to create clusters in the historical dataset. Optimal number of clusters can be determined by the use of average silhouette width and information criteria such as Akaike Information (AIC) and Bayesian Information (BIC) criteria. In this study, optimal number of clusters is also verified by a performance measure which indicates spatial and temporal distribution of congested regions detected by the bottleneck identification algorithm. Based on this distribution, a probability map of congestion can be created for each cluster and can be used to predict travel time for a day which belongs to that cluster. However, this approach would only work under recurrent traffic conditions. In case of non-recurrent congestion, bottleneck identification algorithm is implemented in real time and the congestion propagation is estimated using the shockwave speed …