In-Vehicle AI Agent (LLM-based)
- Scope:
- LLM-based conversation
- Multi-turn dialog & context memory
- Mapping intent → vehicle function control
- Hybrid edge–cloud execution
- Examples:
- Mercedes-Benz MBUX + ChatGPT beta in ~900k vehicles (US) Mercedes-Benz Group+2autofutures.tv+2
- Volkswagen IDA + ChatGPT in ID models, Golf, Tiguan, Passat Volkswagen Group+1
- Kia AI Assistant (EV3, K4) – generative-AI voice assistant with open-ended Q&A and vehicle control Maeil Economic Daily+3Kia Global Media Center+3Hyundai Motor Group+3
Transformer / LLM Models (architecture level)
- Use: Underlying model type for the above assistants (ChatGPT, Grok, Alexa LLM, Google Automotive AI Agent, etc.).
- Example: Mercedes (ChatGPT via Azure OpenAI), VW (ChatGPT via Cerence), Tesla’s planned integration of Grok into vehicles. The Verge+3Mercedes-Benz Group+3Volkswagen Newsroom+3
Self-Attention & Multimodal Fusion
- Use: Fuse voice, vehicle state, map and sensor data inside those LLM-based assistants for context-aware answers (e.g., “How much range do I have to reach the next charger?”).
- Example: Described conceptually in OEM stories about generative-AI companions that know vehicle status and environment (Hyundai/Kia, Mercedes, VW). IEEE Spectrum+2hyundai.news+2
Text Generation & Summarisation
- Use: Explaining vehicle functions, summarising manuals, generating trip summaries or messages.
- Example: Hyundai and VW descriptions of generative-AI in-car companions that can explain features, draft messages and answer open questions while driving. hyundai.news+2speedcraftvw.com+2
Dialogue Systems – full conversational UI
- Use: Entire HMI front-end for the smart cockpit; persistent conversations, persona, proactive suggestions.
- Example: Next-generation Mercedes MBUX Virtual Assistant using Google’s conversational AI with personalities (“natural, predictive, personal, empathetic”) in the new CLA. The Verge+1