Smart Vehicle Surveillance and Predictive Maintenance System

Abstract

This study presents the design and implementation of an end-to-end intelligent vehicle monitoring system that simultaneously supports predictive maintenance and driving behaviour analysis. The proposed framework integrates on-board vehicle diagnostics data (OBD-II/CAN bus) with smartphone-based sensing, including inertial measurement unit (IMU) and GPS signals, to capture comprehensive real-time vehicle and driver information. A scalable streaming data pipeline processes continuous sensor inputs and feeds machine learning models for anomaly detection, remaining useful life (RUL) estimation, and driver scoring. The anomaly detection module identifies abnormal engine and system patterns, while time-series predictive models forecast component degradation to enable proactive maintenance. In addition, driver behaviour profiling evaluates acceleration, braking, cornering, and speed dynamics to generate safety and eco-driving insights. The architecture follows a privacy-first design by enabling on-device (edge/mobile) inference with optional cloud integration for advanced analytics and fleet-level monitoring. Experimental evaluation demonstrates the system’s effectiveness in improving fault prediction accuracy, reducing unexpected breakdowns, and enhancing driving safety. The proposed solution offers a scalable, cost-efficient, and intelligent framework for modern connected vehicle ecosystems..