Comparison of different driving style analysis approaches based on trip segmentation over GPS information

Abstract

Over one billion cars interact with each other on the road every day. Each driver has his own driving style, which could impact safety, fuel economy and road congestion. Knowledge about the driving style of the driver could be used to encourage "better" driving behaviour through immediate feedback while driving, or by scaling auto insurance rates based on the aggressiveness of the driving style. In this work we report on our study of driving behaviour profiling based on unsupervised data mining methods. The main goal is to detect the different driving behaviours, and thus to cluster drivers with similar behaviour. This paves the way to new business models related to the driving sector, such as Pay-How-You-Drive insurance policies and car rentals. Driver behavioral characteristics are studied by collecting information from GPS sensors on the cars and by applying three different analysis approaches (DP-means, Hidden Markov Models, and Behavioural Topic Extraction) to the contextual scene detection problems on car trips, in order to detect different behaviour along each trip. Subsequently, drivers are clustered in similar profiles based on that and the results are compared with a human-defined groundtruth on drivers classification. The proposed framework is tested on a real dataset containing sampled car signals. While the different approaches show relevant differences in trip segment classification, the coherence of the final driver clustering results is surprisingly high.

Research on Classification and Recognition of Driving Styles Based on Feature Engineering

“…With the development of data mining and modern communication technologies, more and more driving data can be collected and numerous machine learning algorithms are employed to classify driving styles with improvement of rationality and accuracy [12], [13]. In [14], [15], the driver's behavioral characteristics are studied by collecting information from on-board GPS sensors and applying three different approaches, i.e., the DP-means algorithm, hidden Markov model (HMM), and behavioral topic extraction. As such, the contextual scene detection is conducted and different behaviors in each trip are identified.…”

Driving Performances Assessment Based on Speed Variation Using Dedicated Route Truck GPS Data

“…These studies often use pattern recognition techniques [12] such as supervised or unsupervised classification, and dimension reduction. These techniques are based on measuring the differences of the patterns, which can be expressed as a distance matrix using similarity measure [13] or other measures [14]. Zhu et al studied smartphone GPS data from 12 test drivers when they are traveling [15].…”

“…Brambilla et al studied 27 trips from GPS data in order to extract recurrent driving patterns from trips to detect different behaviors [14]. Three variables were used: acceleration, speed and the difference in yaw.…”

“…In such a context, the amount of GPS traced vehicle samples used in previous driver behavior studies is very limited. Most of these studies focus on establishing complex models (14), analyzing a few individual cases collected from experiments [14], [15], or studying the distribution of travel behavior such as trip generation and purpose [18]. Therefore, although the use of GPS data is promising, there is still a need for new methodologies for extracting valuable information from the large GPS datasets, and a great room for research on driver behavior patterns based on GPS data.…”

“…This measure assumes that the greater the frequency of speed variations, the more unstable the driving behavior, thus the greater the risk of crashes. The above proportion of frequency in speed variation within different levels (DS 1 = 8 km/h and DS 2 = 16 km/h) are combined to establish each driver's risk profile, as calculated in equation (14). The calculation results are summarized in table 1.…”