Constructing A Hybrid Algorithm for Tracking and Following People using a Robotic Dog

In order to create a social robot, able to interact with people in a natural way, one of the things it should be able to do is to follow a person around an area. This master’s research is aimed at enabling a Sony AIBO robot dog to watch people closely and follow them around the house. This is done using the visual data provided by the robot’s nose-mounted camera. By analysing the data with respect to colour distributions and salient features, a tracker is conceived that is able to track and follow a person in the robot’s field of view. By interfacing the tracker with a motion feedback system, the tracker can guide the robot around the house. This can even be done considering the erratic movement pattern of an AIBO robot dog walking around. To gain a robust object tracker, a fusion algorithm is created which uses background estimation as its initialisation method. Tracking is done using a combination of an Expectation Maximization (EM)-based colour histogram tracker and the Kanade-Lucas-Tomasi feature point tracker. This hybrid algorithm returns an ellipsoidal tracking kernel which is an estimate of the current position and shape of the person being followed. By using the tracking kernel’s position in the current frame perceived by the robot, the real-world location of the person relative to the robot can be roughly estimated. When the robot is controlled to keep the kernel in the centre of its field of view using head motion, the head position can be used to estimate whether the distance between the person and the robot is above or below a certain threshold. Together with the robot’s behaviour to turn when the person approaches the edge of its field of view, this will result in direct interaction with the person being tracked. As soon as the person starts moving away from the robot for example, the centre of the tracking kernel will be shifted towards the bottom of the robot’s view. This results in the robot lowering its head, which activates the robot’s behaviour to start moving forward.