AI-Powered Multimodal Diagnostic Assistant for Vehicle Fault Detection

Abstract

Vehicle maintenance poses real challenges for reg-ular drivers facing the growing complexity of today's cars, where OBD systems generate fault codes that demand expert knowledge to decipher, often resulting in avoidable trips to mechanics. This paper introduces a practical mobile solution-an AI-driven repair guide-that empowers non-experts by process-ing everyday inputs like spoken or typed problem descriptions, dashboard snapshots, and direct OBD-II data pulled over Blue-tooth. Through targeted natural language analysis of symptoms alongside decoded diagnostic codes, the system assesses issue severity via a conversational chatbot, offering clear DIY repair steps complete with tool lists and safety tips for minor fixes, while directing users to local workshops for anything serious. It further tracks full service histories and pushes timely alerts for routines like fluid checks or tire rotations to prevent future headaches. Deployed as a React Native app with a robust FastAPI backend for quick, reliable performance across phones, initial real-vehicle tests confirm its potential to cut down on unnecessary service calls and boost owner confidence in handling basics