Model-Based Product-Oriented Certification

Abstract

Future space missions such as the Mars Science Laboratory and Project Constellation suggest the engineering of some of the most complex man-rated software systems. The present process-oriented certification methodologies employed by NASA are becoming prohibitively expensive when applied to systems of such complexity. The process of software certification establishes the level of confidence in a software system in the context of its functional and safety requirements. Providing such certification evidence may require the application of a number of software development, analysis, and validation techniques. We define product-oriented certification as the process of measuring the system's reliability and efficiency based on the analysis of its design (expressed in models) and implementation (expressed in source code). In this work we introduce a framework for model-based product-oriented certification founded on the concept of source code enhancement and analysis. We describe a classification of the certification artifact types, the development and validation tools and techniques, the application domain-specific factors, and the levels of abstraction. We demonstrate the application of our certification platform by analyzing the process of model-based development of the parallel autonomic goals network, a critical component of the Jet Propulsion Laboratory's Mission Data System (MDS). We describe how we identify and satisfy seven critical certification artifacts in the process of model-driven development and validation of the MDS goal network. In the analysis of this process, we establish the relationship among the seven certification artifacts, the applied development and validation techniques and tools, and the level of abstraction of system design and development.