Selecting and commanding individual robots in a multi-robot system

In this thesis, we present a novel real-time computer vision-based system for facilitating interactions between a single human and a multi-robot system: a user first selects an individual robot from a group of robots, by simply looking at it, and then commands the selected robot with a motion-based gesture. We describe a novel multi-robot system that demonstrates the feasibility of using face contact and motion-based gestures as two non-verbal communication channels for human-robot interaction. Robots first perform face detection using a well-known face detector. The resulting "score" of the detected face is used in a distributed leader election algorithm to estimate which robot the user is looking at. The selected robot then derives a set of motion features, based on blurred optical flow, which is extracted from a user-centric region. These motion cues are then used to discriminate between gestures (robot commands) using an efficient learned classifier.