Landmark-based robotics navigation for indoor environments – an extended reality perspective

This thesis chronicles the research undertaken to create and deploy an indoor robot navigation system with limited sensors. The project was completed in three distinct phases, with each stage undergoing independent real-time assessments using the NAO humanoid robot. During the initial phase of this research study, the challenge of landmark-based navigation is tackled through an innovative approach using augmented landmark vision-based Extended Kalman Filter (EKF) and Ellipsoidal-based SLAM techniques. The primary goal was to enable seamless robot navigation using a monocular camera while ensuring precise SLAM solutions. In the second phase, we introduce a novel SLAM method called HoloSLAM, utilizing Microsoft HoloLens and mixed reality. HoloSLAM capitalizes on mixed reality to offer virtual landmarks for mapping and navigation when real landmarks are unavailable or undetectable. This approach creates a more immersive, interactive, and engaging environment for robots. In the concluding phase, a "Follow me" audio-assisted robot indoor navigation system is introduced. The system incorporates modules from earlier stages while introducing new components for speaker recognition systems and virtual map construction. The comprehensive system undergoes evaluation in a real-world setting using a Nao humanoid robot.