Application of distributed artificial intelligence in complex modular critical applications

Abstract

This report provides an overview of the emerging technology Distributed Artificial Intelligence, in particular in the area of Distributed Problem Solving (DPS). DPS refers to coarse-grained (task-level) problem decomposition resulting in a number of expert or knowledge-based systems, generally called agents of which each exhibits some intelligence. The DPS technology has features that may reduce system design complexity through a highly modular approach and, consequently, may reduce life cycle costs through improved maintainability. These problems of complexity and maintenance are often faced with the design of complex critical applications (including many aerospace applications). DPS can provide a more natural solution with respect to system design, development, and maintenance. This report surveys DPS methods and techniques that have potential benefit for these critical applications. The two main approaches in DPS are discussed: blackboard systems and multi-agent systems. Further, the technology is evaluated along a number of criteria relevant for the envisaged applications. Based on this evaluation it is recommended to consider DPS technology in complex modular (decomposable) critical systems and let it be a driving technology for the overall system architecture.