Detection of driving style and road profile with trained classification algorithms using vehicle can bus data

As the number of vehicles used increases day by day, drivers are experiencing more accidents due to their driving styles and road profiles. Official records show that in parallel with the increasing number of accidents, there has been a significant increase in the loss of life and property of people. This situation has led researchers working on smart traffic systems to conduct new studies to reduce the accident rates that may occur or to inform drivers in advance of the reasons that may lead to an accident. The aim of this thesis is to find the profile of the roads on which the vehicles are driven by detecting the driver's driving style by using sensor data via CANBus. In this study, the most used classifier algorithms traditionally accepted in the literature, such as Artificial Neural Networks, Nearest Neighborhood, Support Vector Machines, C4.5 Algorithm, Naive Bayes and LSTM, which is the Deep Learning algorithm, as well as the ones with the highest success performance of today's age and the Deep Learning algorithm, are used. Learning classification algorithm was used. These algorithms used data received from the vehicle OBD II socket via CANBus to detect the driver's aggressive/calm driving style and the smooth/rough road profiles on which the vehicles are driven. The data in the study were collected by different drivers consciously driving in aggressive and calm driving styles accepted in the literature, and the CAN Bus sensor data taken from the vehicle was converted into training data labeled as "aggressive" and "calm". In addition, sensor data collected from various drives on rough and smooth roads were used as training data for the road profile. In this thesis, it has been shown that driving style and road profile can be determined with high accuracy by using classification algorithms, including deep learning methods. As a result, it has been determined that the LSTM model is more successful in classification of driving style and road profile than others. It is envisaged that the results obtained will provide feedback to both drivers and authorities in order to prevent accidents that may occur.