| Abstract | This article discusses some of the associated difficulties in designing system architecture for the autonomous control of a mobile robot, in particular the problem associated with using strategic reasoning for selecting a set of "activities". An activity is designed as a stand alone sub-system capable of performing a particular focused task. Strategic reasoning is coded as a set of production rules and interaction between the activities and the strategic reasoning knowledgebase occurs through blackboard style architecture. The system architecture described in this article is known as a blackboard activity-based system and came about from a study of desired attributes for mobile robot system architecture. A definition of these attributes is given along with a qualitative comparison of several classical system architectures. The blackboard activity-based system is a hybrid system incorporating advantages of the three classical systems. It is designed with two hierarchical layers for strategic and reactive reasoning, a blackboard database for keeping track of the global state of the world, and a set of reactive behaviour-like activities. The feasibility of the approach is demonstrated by an actual implementation on a mobile platform which has the ability of real-time reactive navigation, detection of floor anomalies, and dynamic path planning. An example of strategic reasoning is demonstrated when information posted on the blackboard by the passageway activity is used to guide the floor anomaly activity. |
|---|
