Design and implementation of a model-based driving style estimator for real-rime applications

This thesis presents the design and implementation of a real-time system for detecting dangerous driving behavior using a model-based approach grounded in sensor data analysis. Unlike conventional systems that rely on GPS or georeferenced inputs, the proposed solution uses only inertial data, specifically, acceleration and gyroscopic measurements collected from a smartphone’s onboard IMU. This approach enables a low-cost, portable, and infrastructure-independent method for driving behavior evaluation. The methodology is based on a scientifically validated model for driving style classification, which was critically assessed and adapted. The initial phase involved validation within a controlled, virtual environment using Euro Truck Simulator 2, where acceleration data was gathered via a telemetry interface. This simulation environment enabled systematic tuning of behavior thresholds and classification rules under repeatable conditions. In the subsequent phase, real-world data was acquired using an Android smartphone. To address gravitational interference in raw accelerometer data, a sensor fusion algorithm sourced from existing literature was implemented. This algorithm combines accelerometer and gyroscope inputs to isolate true translational acceleration, ensuring accurate analysis. The full detection pipeline, including preprocessing, sensor fusion, and behavior classification, was developed in MATLAB and modeled using Simulink. The finalized system was deployed to a mobile platform via MATLAB’s Android hardware support package, enabling real-time data processing and user feedback on a smartphone. This work demonstrates a complete engineering cycle, from theoretical modeling and simulation to real-time embedded deployment, and delivers a functional, mobile-ready framework for driving behavior assessment. It offers a practical and accessible contribution to the fields of driver monitoring, smart mobility, and telematics.