|Ph.D Student||Toledo-Crespin Galit|
|Subject||Analysis and Modeling of Driving Behavior Using In-Vehicle|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Yoram Shiftan|
|Professor Shalom Hakkert|
|Full Thesis text|
Driver behavior and errors are a major cause of car crashes. Driving behavior is related to the driver’s character and socio-economic background. Still, it may be influenced by various means to create positive change in driving behavior.
This study used newly available data of detailed observations of driving patterns to develop a modeling framework of driving behavior. For this purpose, In Vehicle Data Recorders (IVDR) were installed in about 100 vehicles during a one year experiment in an Israeli air force base. The IVDR continuously recorded driving parameters such as speed, acceleration, braking and location. The system processed these observations to identify various driving events.
A general theoretical framework of driver behavior was developed. This framework incorporates available measuring tools, which express the relations among the different factors of driving behavior. In order to estimate models based on this framework, IVDR data were collected, as well as data about driving styles, characteristics, attitudes and perceptions from traditional self-reported questionnaires, safety supplementary data and vehicle operational cost data.
The experiment was designed in three stages. Initially drivers were not exposed to the system and no feedback was provided to them. Next, drivers were informed about the IVDR and some drivers received periodic feedback from the safety fleet officer based on the information collected by the IVDR. In the last stage, all drivers received periodic individual feedback reports.
The various IVDR events and supplementary data were used to create risk indices that may be used to evaluate drivers' risk of crash involvement and fuel consumption. These indices were developed such that they best explain the risk of crash involvement and fuel consumption as expressed by crash and fuel consumption records. The IVDR data were also evaluated against the traditional alternative of self-reported driving behavior and socio-demographic characteristics. Finally, the influence of the feedback provided to the drivers based on the IVDR data was evaluated.
These models may help improve our understanding of the characteristics associated with different patterns of driving behavior and will support the development of methods and tools to modify such behaviors in an effort to improve road safety.