Generating probabilistic path observation from GPS data for route choice modeling
Map matching algorithms are the conventional way to generate path observations from GPS data for route choice models. The deterministic matching may introduce extra biases to parameters of route choice models if the matching is wrong. In this paper, a new methodology is proposed to probabilistically generate path representation from GPS location data and the underlying network. This methodology takes advantage of both spatial and temporal relationships existing in the location data and the network. The generated result includes a set of potential true paths, along with a probability of each proposed path to have been the actual path. An algorithm is designed and applied to a simulated trip.
Record created on 2010-09-30, modified on 2016-09-16