Terrain Perception Using Wearable Parrot-Inspired Companion Robot, KiliRo

Area of Science:

• Robotics
• Computer Vision
• Human-Computer Interaction

Background:

• Falls are a leading cause of unintentional injury deaths globally.
• Increased risk factors include mobile phone use, impaired vision, and unexpected terrain changes.
• A proactive warning system is needed to mitigate fall-related injuries.

Purpose of the Study:

• To develop and validate a terrain perception system for a wearable robot.
• To alert users to imminent terrain changes and potential fall hazards.
• To enhance safety for individuals prone to falls.

Main Methods:

• Utilized a bio-inspired wearable robot, KiliRo, with a 180-degree camera range.
• Employed image processing for visual feature extraction from captured terrain images.
• Implemented K-nearest neighbor algorithm for terrain classification.

Main Results:

• The terrain perception system achieved over 95% accuracy.
• Successfully classified five distinct terrain types: pedestrian pathway, road, grass, interior, and staircase.
• Demonstrated the efficacy of the visual-based approach for real-time terrain analysis.

Conclusions:

• The KiliRo robot's terrain perception system effectively identifies diverse terrain types.
• This technology offers a promising solution for fall prevention by warning users of hazards.
• The system's high accuracy supports its potential for real-world application in enhancing mobility safety.