An Approach Based on a Programmable Intelligent Space for Human-robot Interaction

• Luiz Carlos Cosmi Filho UFES
• Matheus Dutra de Oliveira UFES
• Miquelly Nicolini Lucas UFES
• Lucas Fehlberg Follador UFES
• Raquel Frizera Vassallo UFES

Resumo

Gesture recognition is a challenging problem in the area of human-robot interaction, which includes dealing with a complex environment and gesture perception. This paper proposes a set of services within a Programmable Intelligent Space (PIS) architecture, enabling human 3D pose estimation, gesture recognition, mobile robot navigation and human-robot interaction. We leverage advanced methods for 3D pose estimation and tracking with a multi-camera calibrated system, and integrate a simple rule-based gesture classifier to enable an elementary interaction with a mobile robot. Also, the robot sensors and the cameras installed in the intelligent space are both used for a robust localization method, which allows the robot to navigate not only in areas covered by cameras, but throughout the entire mapped area of the environment. We present simple human-robot interaction experiments, showing potential to be applied across various settings, such as factories, hospitals and museums.

Palavras-chave: Three-dimensional displays, Navigation, Robot vision systems, Pose estimation, Human-robot interaction, Gesture recognition, Cameras, Mobile robots, Robots, Visual odometry, intelligent spaces, pose estimation, pose reconstruction, pose tracking, human-robot interaction, mobile robots