From boarding counting error to error in O-D matrix estimates: how to improve the reliability of passive data?
Résumé
The past decades have been marked by several transformations that impact mobility behaviors. In this context, using technologies such as WiFi and Bluetooth allows one to gather passive mobility data, useful for ensuring the sustainable development of transport infrastructures. Compared to smart-card data or other passive data collections, these technologies have the additional advantage of providing information about the entire trips. As WiFi data collections are quite novel in transport modelling, their results still need to be confronted to ground data sources. Unfortunately, while we may dispose of data on boarding passengers (smart-card data) or on boarding passengers and alighting passengers separately (optical counts) on a regular basis, Origin-Destination (O-D) pairs are much more difficult to get. This paper proposes to quantify the error between WiFi data and O-D surveys data in order to better appreciate the relevance of WiFi data. By using a theoretical model, we present a way to estimate the error on O-D pairs when the error on volumes per boarding (respectively alighting) is available. This method will allow analysts to control the quality of O-D pairs produced by WiFi data, using count data, which is continuously available. The method can also give insight into the main sources of error in the WiFi-based reconstruction of O-D matrices.