RESEARCH · RESEARCH · #849
Affordable offline AI-integrated smart cane on Raspberry Pi Zero 2W (arXiv:2609.22277v1)
This arXiv paper presents an $88 offline AI-integrated smart cane prototype running on a Raspberry Pi Zero 2W that fuses RGB vision and Time-of-Flight distance sensing, uses an INT8-quantized SSD MobileNet V1 for obstacle recognition, and provides distance-aware vibrotactile and audio alerts; reported performance includes macro F1 0.82, mean latency 330 ms, peak power 2.8 W, and a preliminary 12-participant usability study (SUS 78.5).
KEY POINTS
- This arXiv paper presents an $88 offline AI-integrated smart cane prototype running on a Raspberry Pi Zero 2W that fuses RGB vision and Time-of-Flight distance sensing, uses an INT8-quantized SSD MobileNet V1 for obstacle recognition, and provides distance-aware vibrotactile and audio alerts; reported performance includes macro F1 0.82, mean latency 330 ms, peak power 2.8 W, and a preliminary 12-participant usability study (SUS 78.5).
- Demonstrates the feasibility of low-cost, privacy-preserving, edge-native assistive AI for mobility support without cloud dependence or expensive hardware.
- An Affordable AI-Integrated Smart Cane for Multimodal Mobility Assistance of Visually Impaired Users
WHY IT MATTERS
Demonstrates the feasibility of low-cost, privacy-preserving, edge-native assistive AI for mobility support without cloud dependence or expensive hardware.