Personal Guides: Heterogeneous Robots Sharing Personal Tours in Multi-Floor Environments

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• GidaBot facilitates multi-robot guided tours in multi-floor buildings, requiring seamless collaboration.
• Ensuring visitor privacy and preventing interference from unknown individuals is crucial for personalized tours.

Purpose of the Study:

• To develop and integrate a robust face re-identification and verification module into GidaBot.
• To enhance the reliability and security of GidaBot's guided tour services by accurately identifying visitors.

Main Methods:

• Offline evaluation of state-of-the-art face detectors and embedding-based distance metrics.
• Online implementation using a parallel fusion of multiple face detectors to mitigate alignment bias.
• Decision-making based on a minimum distance criterion in face embedding space.

Main Results:

• The fused face detection approach significantly outperforms individual methods.
• The integrated face re-identification module demonstrably improves the real-time reliability of GidaBot.
• Successful deployment and testing of the system with real robots at the Faculty of Informatics.

Conclusions:

• A fused face recognition system effectively addresses challenges in identifying casual visitors for robotic tour guides.
• The developed module enhances GidaBot's operational security and user experience by managing visitor interactions.
• This research advances the practical application of computer vision in heterogeneous robotic systems.