Copyright © 2001 IF AC IFAC Conference on New Technologies for Computer Control 19-22 November 2001, Hong Kong TOWARDS SELF-ORGANISING TIME-SENSITIVE CONTROL SYSTEM'S SOFTWARE Leo Motus and Merik Meriste# Tallinn Technical University, leo.mo/us(ii,dcc.ttu.ee tiTartu University, merik.meriste(wZlf.ee Estonia Abstract: The embedded computer applications are rapidly growing in numbers, in complexity, and in dependability. This trend has been accompanied by shift of systems building paradigm from individual algorithms to interacting computing entities (e.g. agents, components, modules, etc). A number of studies focus on self-organising capabilities of such entities (agents). Unfortunately for computer control systems, many essential issues - such as, forced con currency, time awareness of theories and software, and some others - have been almost neglected in the conventional theory. This paper provides a superficial survey of evolution of software-intensive systems towards multi- agent systems, lists subjectively some basic problems and achievements in developing time-critical systems. The authors suggest that sufficiently good theoretical and practical basis exists for building a test-bed that would assist in merging recent development trends in software engineering and artificial intelligence, and enable verification of time correctness of agent interaction. The test-bed would also assist in developing a theory for quantitative asscssment of some properties of the evolving structures, and enable step-wise extension and testing of the supported functions (as the theoretical basis develops). COp'vright @ 2001 IFAC Key words: multi-agent systems, time awareness, computational models, time models, test-bed. I. INTRODUCTION Everyday life in the contemporary world depends heavily on software-intensive time-, safety-, and security-critical systems. It is not easy to manage a day without using (the services) of the following devices, or systems - vehicles, telephones, audio- video equipment, security systems, medical devices, transport system, air conditioning systems. Some other examples are less visible, or not yet functioning - for instance, distributed computer control systems, mobile multi-robot systems, automatic motorways, human and computer system collaboration in problem sol ving. 236 Many of the listed software-intensive applications may have dynamically changing goals, should preferably be able to modify dynamically their configuration - most of these changes are invoked either by the environment they are embedded in, or by the subjective will of their users (or the goal function). Some systems must have fault-tolerant and/or safety-critical design and implementation and, in addition, must be (strictly) time-sensitive so as to cope with their goals and imposed constraints. The design and implementation methods have passed a remarkable evolution to enable the building of such