Trade-Offs in Assigning Active Travel GPS Trajectories to a Geospatial Infrastructure Network

GPS trajectories, which reveal travellers’ preferences, can be utilised in various fields of transportation research. However, since GPS trajectories contain errors, they need to be matched to the links that were actually used in the real world. This task is particularly challenging for active travel trajectories, as they often have irregular movement patterns. To overcome this challenge, it is necessary to identify the spatial influences and characteristics of these trajectories. Therefore, this study compares a range of geometric matching methods using buffers and length overlapping proportion and assesses the trade-offs between these methods. The analysis was based on a case study centred on Winchester, UK and 4589 trajectories were assigned to an enhanced routable active travel infrastructure network. Buffers were set from 10 to 60 m in 5 m increments and 70% length overlapping proportion was applied. On overall matching results, as buffer sizes increased, the number of matched trajectories also increased linearly. The smaller buffer sizes showed that buffers smaller than 15 m are more likely to be affected by errors or to capture each side of a link separately. Meanwhile, the larger buffer sizes capture trajectories on adjacent links, so that the number of trips on a link tends to be overestimated. For further analysis, bearing criteria were also applied and the trend of the results was similar to the results obtained without bearing rules. Overall, the results suggest that a 30–35 m buffer is able to provide bi-directional coverage for large active travel GPS trajectories, while buffers smaller than 15 m are able to match trajectories to one side of a link but with a higher likelihood of error. The findings help understand the spatial trade-offs in assigning large active travel GPS trajectories and could be used to set the spatial extent of buffers in relevant research.