Efficient Pedestrian and Bicycle Traffic Flow Estimation Combining Mobile-Sourced Data with Route Choice Prediction

Abstract

Accurate estimate of traffic flow measures like annual average daily traffic (AADT) is vital to making decisions about roadway planning, safety, maintenance, operation etc. Methodology to inexpensively obtain an accurate estimate of traffic flow especially for pedestrian and bicyclist traffic is lacking in the literature. High expenses of conducting household surveys and setting up traffic monitoring stations to collect data motivated us to look for cheaper solutions. In this study, we develop a methodology to inexpensively obtain a good estimate of pedestrian and bicyclist traffic flow from mobile data sources while avoiding the privacy issues associated with models based on household survey data. However, the accuracy of mobile data is unknown and may vary in different locations. To deal with erroneous data sources we use different techniques to estimate and keep improving an origin-destination (OD) matrix from the observed link flows to ultimately get the actual link flows. In our model, we enforce the consistency between the number of productions and attractions of trips for different regions with the OD-matrix. Using the network topology, we use trip distribution based on the gravity model to generate an initial OD-matrix. Then we use an optimization formulation to improve the initial OD-matrix so that the link flow obtained using the improved OD-matrix matches with the partially observable link flows. Furthermore, we present the performance of the solution algorithm for the Twin Cities' bicycle and pedestrian networks. We also compare the accuracy of our estimate with manually collected traffic flow data for the real networks.