Models’ Validation for Complex Real-Time Systems

Models’ validation poses an interesting area of discussion in computer science and engineering as in other disciplines, which cannot easily be overlooked. No model of a complex system can be said to be absolutely correct, or a perfect description of the system, as such models are, and should be an abstraction of the actual system. This situation becomes even more glaring for real-time computer systems where stringent conditions are imposed on the system, for instance, timing and hence predictability constraints. In this thesis we shall explore different perspective on models’ validation and techniques in computer science and their attendant problems especially for complex real-time systems. Major open and philosophical issues on the correctness criteria for models validations that seek to answer such questions as: Can models be validated? Can ‘valid’ models yield information? Do correct models contain truth? and more shall be investigated as well. We will discuss the implications of our answers with respect to computer science and real-time systems in particular arguing vehemently that models can yield information. We shall briefly investigate validation of models in computational physics, and seek to draw some parallels between its techniques and methodologies with those in computer science. Finally, we will present results on our validation experiments for a robotic electric control motor running on VxWorks operating system and from our work deduce and propose guidelines for engineers and empirical scientists working on models. These guidelines suggest simple formats of determining the number of criteria and queries that could be used in validation, given different levels of abstraction and model confidence.