Runtime Personalization of In-Vehicle Assistants using Vehicle Context and Driver Profiles

In-vehicle assistants (IVAs) are increasingly central to enhancing driver experience, yet most existing systems rely heavily on cloud-based personalization, which poses latency, connectivity, and privacy concerns. This project addresses the challenge of delivering runtime personalization in IVAs through a lightweight, onboard adaptive intelligence layer. The proposed system modifies speech, visual overlays, and prompts dynamically, using real-time inputs such as vehicle context (e.g., speed, driving mode) and driver profiles (e.g., user role, behavior patterns). A modular rule-based engine combined with embedded machine learning allows adaptation without external server dependency, enabling consistent performance even in low-connectivity scenarios. The architecture supports three personalization tiers—trip-type, vehicle-state, and user-role—and demonstrates a latency reduction of up to 42% compared to conventional cloud-reliant methods. Early prototype testing in a simulated vehicle environment indicates a 23% improvement in driver engagement scores and smoother human-machine interaction. This approach bridges infotainment usability with local intelligence, paving the way for future personalization strategies that prioritize responsiveness, contextual relevance, and data sovereignty.