An optical-flow person following behaviour

摘要

Optical flow is a biologically inspired computation that has been widely used for tracking the movement of objects in a scene. This paper presents its novel application to an autonomous robot that is required to follow a person in an everyday environment. This robot skill is implemented as one of many concurrent behaviours and it is integrated in a general architecture called ERASMUS-Expert based Robot Architecture for Sensing and Moving in Unknown Surroundings. The details of the method are described as well as its integration in ERASMUS. Experimental results, evaluating both the performance and the reliability of our approach, will also be provided.