Supporting a Human-Aware World Model Through Sensor Fusion

Recent research in robotics aims at combining the abilities of humans and robots through human-robot collaboration. Robots must overcome additional challenges to handle dynamic environments within shared workspaces. They especially must perceive objects and the working progress to synchronize with humans in shared tasks. Due to unpredictable human interaction, local information about objects detected by eye-in-hand cameras and stored within a world model falls in value as soon as respective objects get out of sight. Our contribution is an approach to making world models aware of human influences and thus allowing robots to decide, whether information is still valid. To this end, we annotate pieces of information with certainty values encoding how trustworthy they are. Certainty is adapted over time according to additional knowledge about human presence within the workspace, provided by a global sensor. Thus, we achieve human-awareness through fusion of local and global sensor data. Our concept is validated through a prototype implementation and experiments that regard certainty of objects in different scenarios of human presence.