Malec, J. (1994). On Implementing Behaviours Using a three-layered Architecture. Technical Report LiTH-IDA-R-94-21, Department of Computer and Information Science, Linköping University, Sweden. Accepted to Intelligent Robotic Systems'94, Grenoble, France, 11-15 July 1994. (bibtex),
Abstract: The problem of designing reactive autonomous systems has been the subject of much attention in recent years. The research has focused on designing systems which would possess the following attributes: 1. reactivity, in order to allow the system to cope with unpredictable changes in the dynamic environment while performing its mission; 2. robustness, meaning the ability to function in a variety of situations, including failure of some of its subsystems; 3. selectivity of attention, in order to effectively use existing resources of the system, such as computing power, or sensory equipment; 4. ability to pursue goals either defined by the designer, or by the system itself. One can distinguish several approaches to this problem. One of them is called ``behaviour-based'', because behaviour is the basic structure used for specifying control for an autonomous system. Another one is based on layered control architecture, where usually each layer has its own, well-defined task(s) to perform, often some software tools supporting implementation of those tasks, and some appropriate mechanism for passing control between the layers (or distributing it among them). Those two approaches are usually opposed to each other: ``Behaviourists'' claim that a rigid architecture requires excessive computational power and enforce too many constraints on the actual design and implementation. The other group claims that behaviour-based approach is unsuitable for applications requiring either sophisticated control algorithms, or symbolic reasoning (as e.g. planning) procedures, or both. Both criticisms are to some extent justified. In this paper we present an approach to implementing behaviour-based control using a three-layered architecture. In this way we intend to combine the advantages of both approaches.
CS Dept TR Overview