B.O.S.S – Belt of Super Sight

Independent indoor navigation remains a critical challenge for visually impaired individuals, particularly within complex academic environments where GPS-based positioning systems are entirely ineffective. This paper presents B.O.S.S (Belt of Super Sight), a wearable smart navigation assistant designed to enable visually impaired students to navigate safely and autonomously across the Byblos Campus of the Lebanese American University (LAU).

The proposed system integrates RGB-D Visual Simultaneous Localization and Mapping (SLAM), implemented through the RTAB-Map framework, to perform real-time mapping and localization without reliance on any external infrastructure. During an initial mapping phase, B.O.S.S constructs a detailed three-dimensional representation of campus corridors and stores it as a reusable database. Once deployed, the system localizes the user within this pre-mapped environment and guides them toward selected destinations using an offline speech-to-text module based on whisper.cpp for voice command input and an offline text-to-speech module based on Piper for spoken navigation guidance. Real-time obstacle awareness is provided by the OAK-D Lite stereo RGB-D camera, whose onboard Vision Processing Unit performs hardware-accelerated depth estimation partitioned into left, center, and right zones, translated into directional haptic feedback through two coin vibration motors driven by an L298N H-bridge module.

The system adopts a deliberate host and Docker container architectural split: ROS2, RTAB-Map, the DepthAI camera pipeline, the A* path-finding navigation daemon, the haptic planner, and the motor driver run inside a containerized environment, while Bluetooth audio and GPIO-driven push-to-talk interaction run directly on the Raspberry Pi 5 host, with the two sides communicating through a lightweight file-based bridge. A motion-based heading estimation strategy and automatic floor-plane detection eliminate the need for body-versus-camera calibration, making the system robust to variations in how the belt is worn. By combining infrastructure-free visual-inertial mapping, on-device speech interaction, A* path planning, and real-time depth-driven haptic feedback, B.O.S.S delivers a portable, low-cost, and fully offline assistive navigation solution that promotes safety, independence, and inclusion for visually impaired users in academic environments.

Project Details

[photo]