Gc Vansteenkiste

Sub-optimal mask search in SAPS.

Abstract In this paper, a new and very simple algorithm for mask search in SAPS (System Approach Problem Solver) is introduced to illustrate a sub-optimal approach. It is intended to be used as an alternative of the optimal mask optimization present in previous versions of SAPS. The algorithm described in this paper illustrates the general use of non-optimal search strategies. It is based on the hill-climbing strategy and as such it is not exponential as in all known previous versions of SAPS, but polynomial in nature. This results in an improved executing speed; very complex systems that were intractable with previous versions of SAPS can be handled with additional flexibility. The previous (exponential) algorithm is also implemented. Hence, comparisons and benchmarks are possible because of the same object-oriented environment. The main focus in this paper will lie on the justification for sub-optimal mask search techniques and its illustration with the algorithm

DEVELOPMENT OF AN AUTOMATIC OBJECT-ORIENTED CONTINUOUS SIMULATION ENVIRONMENT

The construction of automatic control and modelling environments has been attempted using shallow reasoning expert systems. The inadequacy of this approach for real-life systems has become apparent, and the need for deeper knowledge—which can only be obtained by extensive simulation—is now acknowledged. The simulation process, which consists of model building and model selection, followed by the generation and execution of a software simulator, can be partially automated if an Object-Oriented methodology is adopted. In this article, a methodology is presented, which is presently under investigation in the DESiRE (Dynamic Expert Systems in Robotic Experimentation)project In the modelling phase, a hierarchical, uniform way of describing and manipulating continuous and discrete models is needed, if the highly desirable reusability of submodels is to be achieved. This is only possible if a clear distinction is made between bare models (and information about their intrinsic coupling) and causal simulation expe...