An AI-Interactive Shopping Navigation Assistance System for Visually Impaired People via Passive Kinesthetic Guidance
概要
Visually impaired individuals face two main challenges when shopping independently: accessing product information and moving safely. Natural-language interfaces can solve the first problem by allowing users to easily ask about the desired item. However, ensuring safe physical mobility in a complex store environment remains a challenge. For this topic, we propose an integrated shopping assistance system combining an LLM interface with a servo-brake-based passive navigation mechanism. The system operates through an integrated four-step pipeline: first, the AI conversational interface facilitates intuitive target selection; next, ArUco marker detection provides accurate self-localization and spatial awareness; then, a kinematically-aware planner generates a smooth, navigable trajectory; and finally, the passive braking mechanism executes physical route guidance, ensuring safe navigation without compromising the user’s inherent sense of control. We evaluated the feasibility and effectiveness of the proposed system. Fifty shopping experiments were conducted in a mock supermarket environment. The results demonstrate that the system successfully guided participants to their target destinations with an 88% success rate, and the subjects felt safe and independent, with user-friendly shopping experiences.
