IDAS: Intelligent Driving Assistance System Using RAG

Abstract

In the fast-growing automotive technology sector, it has become increasingly clear that there is a need for cars with smarter and more interactive systems. This article presents the Intelligent Driving Assistance System (IDAS), an artificial intelligence system that enables the driver to use voice commands to access various features of a car. The primary component of IDAS is a Large Language Model (LLM), which, through retrieval augmented generation (RAG), can efficiently read and understand the car manual for immediate context-based aid. In addition, this system incorporates speech recognition and speech synthesis capabilities, it can understand commands given in multiple languages, improving user experiences among diverse driver communities. Our results show a minimum response time of one second for the pipeline using GPT-4o-mini and Mistral Nemo.