RideGuide
A customizable multimodal conversational tour guide for autonomous vehicle passengers, supporting location-aware engagement and spatial learning.
- Designed and implemented a multimodal conversational user interface combining voice input, touch, and real-time map synchronization.
- Architected event-driven pipeline: continuous speech transcription → multi-layered LLM intent detection and data extraction → response generation → text-to-speech
- Evaluated in a 270° lab simulator (n=12): UEQ-S 1.10, NASA-TLX 32.4%, 67% would choose AI over human driver, 3.3 landmarks recalled per route.
