A Demonstration of Voice-Interactive AI Agents for Vehicles Utilizing Multiple LLMs

Voice user interfaces (VUIs) based on large language models (LLMs) can significantly enhance the user experience for vehicle drivers. However, relying on a single, centrally-located LLM does not ensure high-quality and diverse functionalities across different locations. In this demonstration, we propose a voice-interactive AI agent system for connected vehicles that utilizes multiple LLMs, integrated within an end-to-end infras-tructure that spans in-vehicle devices, edge data centers, and public cloud services. This proposed system enables drivers to fulfill various needs through natural conversations by selecting the most suitable AI agent for their intended purposes. AI agents are deployed in the optimal execution environment, whether in the in-vehicle device, edge data center, or cloud, based on each AI agent's characteristics. With this system, it is possible to achieve multiple scenarios that were challenging with traditional VUIs.