Cycling has been increasingly popular in many cities over the past decades because of its benefits for both environment and human health. However, there is still lack of knowledge on the characteristics specific to this traveller group and recent promotion of bicycle use in transport policies has even expanded the demand for understanding cyclist behaviour and bicycle dynamics. It is believed that such understanding can further facilitate the evaluation and improvement of cycling safety as well as accessibility on the network. This paper therefore presents an essential methodological framework for processing and analysing naturalistic data collected by commuter cyclists in Stockholm equipped with portable GPS devices. On one hand, the GPS coordinates are filtered by the Kalman smoothing algorithm to obtain accurate and consistent estimates of cyclists’ position, speed and acceleration. On the other locally weighted regression is applied to abstract gradient profiles of cycling paths using data of both altitude and travel distance. After information estimation, the characteristics of cyclist acceleration behaviour are then analysed using statistical approaches. The results show that the acceleration profiles have a linear correlation with the total variance in speed during acceleration or deceleration. The data is finally applied to identify cyclist acceleration models proposed for the development of cycling simulation. (Author/publisher)
Abstract