Identifying travel mode from GPS trajectories through fuzzy pattern recognition

Although GPS-based travel survey has been studied by many, automated travel mode detection still remains a technical challenge. This paper proposed and tested a fuzzy approach to travel mode recognition from the GPS travel data collected from 32 volunteers for 142 days in Shanghai. Four speed-related fuzzy variables were selected to characterize five movement patterns (walk, bike, bus, rail, and rest) in the urban daily traffic. Fuzzy sets and membership functions were constructed for the patterns using self-reported sample data. A procedure of data cleaning and trip segmentation was performed to divide GPS trajectories into mode stages. The final step involved determining the travel mode of each mode stage through a min-max fuzzy operation. Evaluation results indicated that the approach handled the data uncertainty and vagueness rather well. It properly incorporated partial information from the fuzzy variables into the mode recognition process for accuracy enhancement.