Detecting pedestrian destinations from ubiquitous digital footprint
Walking is the key for efficient multimodal transport systems. Pedestrian infrastructures, such as railway stations, face congestion and need to understand flows. Data is required for models that can help managers in their infrastructure planning. In this context, data from communication network are more suitable than mobile phone data. We propose to use existing WiFi traces. Due to the poor quality of WiFi localization, a Bayesian approach is proposed. We generate candidate lists of destinations and compute the likelihood of observing these traces in the pedestrian network. We made a proof of concept on a campus, using class schedule and walking distance.
Keywords: Pedestrian ; Walking ; WiFi ; Traces ; Pedestrian network ; Pedestrian map ; Class schedules ; Train schedules ; Concert schedules ; Activity ; Destination ; Location choice model ; Bayesian
Record created on 2013-01-22, modified on 2017-02-16