A challenge in GPS-based travel behavior surveys is obtaining as much knowledge as possible on travel mode while minimizing respondent burden. GPS data recording offers the promise of reducing effort for survey participants, but in current practice is primarily supplemented with written diaries, memory joggers, or prompted recall for capturing travel mode. Accelerometer data collected simultaneously with GPS data has proven promising in accurate detection of travel mode, but accelerometers are not always deployed as part of GPS data logger systems. Furthermore, such information is not stored with previously collected and archived data for major metropolitan regions. Given that GPS data loggers generally capture more trips than are recorded in diaries, mode imputation must be enacted at some level to ensure completion of the data sets. The distinction between bicycle and personal vehicle traffic has been in some cases indeterminable based on travel speeds alone, since vehicles in heavier traffic tend to mimic bicycle speeds. This is one of the key difficulties that must be resolved before GPS can eliminate the need for participants to personally indicate travel mode.
This study examines the characteristics of GPS-recorded bicycle travel in several major metropolitan regions and determines how acceleration, directional variations, elevation, and road type can be used to statistically differentiate between bicycle trips and other modes of travel. Figures of intrinsic statistical variations by geography for each mode of travel will be provided in the presentation. The utility of maximizing correct representation of alternate mode traffic in travel demand models lies in planning improvements for bicycle and other transportation infrastructure. Reducing the need for accelerometer data in travel surveys can trim deployment costs, data integration time, respondent burden, and comparison challenges when the data do not match between technologies, overall producing a cost savings for tight budget surveys.