Optimization of operative decisions ; computer analysis

When entering the age of ubiquitous computing we observe a rapidly increasing number of applications in which - in real time - operative (control or management) decisions are made based on optimization. To implement repetitive optimization of these decisions we need to use process models, models or directly available forecasts of free, uncontrolled, inputs, methods for state and parameter estimation, model adaptation techniques, and, finally, suitable optimization techniques. Analysis and validation of such optimization based systems require - prior to actual real-life implementation - the carefully planned and performed computer experiments. Theoretical methods are always welcome to assist in this analysis - if possible - but it is mainly the computer based simulation experiment which allows us to investigate, verify and tune a complicated decision mechanism and then control system as a whole before this system is authorized to be put to real-life test. Success in design of a system with on-line computed optimization-based operative decisions is therefore dependent upon many factors, in particular - on having good models, efficient and reliable optimization techniques, good software tools for computer based simulation and on proper, real-life-like conditions under which the computer experiments are performed. These key issues concerning development of such control systems are discussed in this paper.