Big data based driving pattern clustering and evaluation in combination with driving circumstances

Paper #:
  • 2018-01-1087

  • 2018-04-03
Car driver’s behavior and its influence on driving characteristics play an increasing role in the development of modern vehicles, e.g. in view of efficient drive train control and the implementation of driving assistance functions. In addition, knowledge about the actual driving style can provide feedback to the driver and support efficient driving or even safety-related measures. Driving patterns are not caused by the driver himself only, but also influenced by road characteristics, environmental boundary conditions and other traffic participants. Thus, it is necessary to take the driving circumstances into account, when driving patterns are studied. This work proposes a methodology to cluster and evaluate driving patterns under consideration of vehicle-related parameters (e.g. acceleration and jerk) in combination with additional influencing factors, e.g. road style and inclination. Firstly, segmentation of the distance series trip is performed to generate micro cycles. Based on a set of measured data, a features matrix from the micro cycles is developed for clustering. Secondly, driving circumstances clustering is proceeded by using the k-means algorithm. Thirdly, driving patterns clustering under consideration of each cluster of driving circumstances is accomplished. The methodology is applied on the measured driving behavior of a fleet of electric cars. In this exemplary use-case, the driving circumstances are clustered into three groups (hilly roads, start/stop and turning cycles, and flat roads). The results show that, for example, ‘aggressive pattern’ is quantitatively described differently in the three groups. The study shows that the definition of driving patterns varies a lot from one type of driving circumstances to another and the same driver often change their behavior in the same and in different driving conditions. The discussed methodology can be applied to offer a set of exacter and more objective labels for a subsequent recognition of different driving patterns. Therefore, it can contribute to an enhanced development of future advanced driving assistant systems.
SAE MOBILUS Subscriber? You may already have access.
Attention: This item is not yet published. Pre-Order to be notified, via email, when it becomes available.
Members save up to 36% off list price.
HTML for Linking to Page
Page URL

Related Items

Technical Paper / Journal Article
Training / Education