Insight Into Driver Behavior and Usage of ADAS Functions Using Machine Learning

Abstract

Advanced Driver Assistance Systems (ADAS) is a technology that prevents road accidents and improves driving safety. Front-Short Range Assist is a newly released function that alerts drivers of any vulnerable road users in the front area of the truck when it is moving at low speed.

This thesis aims to analyze and predict the deactivation of this function by the drivers, using recorded vehicle data from trucks and external weather and road type data. Several models were utilized to answer these questions: CNN, LSTM, a hybrid CNN-LSTM, and a Bi-LSTM autoencoder. The hybrid model achieved the highest performance. However, all models yielded poor predictive power, making it unclear whether the given dataset has a discernible pattern to predict and understand the reasons behind the deactivation.