Vision based navigation for an office messenger robot

Abstract

Prototype office messenger robot with a new vision-based navigation/maneuvering system which implements the habitat constraint vision theory proposed by the Massachusetts Institute of Technology (MIT) has been built. The vision system is used for two objectives: to avoid collisions and find safe passage in a dynamically changing environment (maneuvering and short-range navigation); and to identify and follow landmarks to travel through its habitat using onboard topological map (long-range navigation). The process architecture for the robot follows the principles of subsumption architecture (SA) and the basic control program for the robot consists of a set of behaviors implemented in Behavior Language (BL). In addition, the bulk of the vision processing behaviors for guiding the robot and detecting landmarks in the office space we implemented on a vision processor board using neural networks. The behavior based processing of sensor inputs results in extremely efficient software with highly satisfactory operational characteristics. The hardware and software structures of the robot and results from preliminary experiments are described.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>