TRIP/STOP Detection in GPS Traces to Feed Prompted Recall Survey

Abstract This paper presents two methods to extract trips and stops from GPS traces: the first one focuses on periods of non-movement (stops) and the second one tries to identify the longest periods of movement (trips). In order to assert the quality of both methods, the results are compared to stops and trips identified by the traveler; this was done by means of a visual tool aimed at alignment of manually reported periods in the diary to automatically recorded GPS coordinates. Several quality indicators are presented; they have been evaluated using sensitivity analysis in order to determine the optimal values for the detector's configuration settings. Person traces (as opposed to car traces) were used. Individual specific behavior seems to have a large effect on the optimal values for threshold settings used in both the trip and stop detector algorithms. Accurate detection of stops and trips in GPS traces is vital to prompted recall surveys because those surveys can extend over several weeks. Inaccurate stop detection requires frequent corrections by the users and can cause them to quit.