The systems architecture of embedded software-electronics control systems (SECS) design is the focus of this paper. Since time-to-market requirements drive the development of products in this engineering area, there is no chance for companies to ignore that speed of execution, quality, productivity, and ability to adapt to business conditions are essential to survive economically. Hellestrand’s writing style is colloquial and familiar, maybe too much so, but the experimentation techniques for obtaining optimal architectures are explained.
The paper presents a framework for ad hoc and systematic engineering experimentation. The author claims that when ad hoc unstructured experiments are performed during the early phases of engineering, the consequences on the architecture (even in expert designs) are profound. This kind of experimentation is far from the scientific method, consisting of selecting hypotheses from structured and statistically significant experiments. Structured and systematic experiments offer a way to efficiently reduce the effort of building new optimized architectures for embedded SECS. The ability to support data-driven decision making in the engineering process requires building models of the intended system that are accurate and have a high standard of performance. The experiments that need to be performed require accurate models of successive physical systems known as virtual prototypes (VPs). Data is collected from probes inserted into hardware models and the actual software during experiments. These VPs are the foundations of empirical electronic systems architecture.
Hellestrand concentrates on the empirical process underlying data-driven architectural decision making, and the capabilities enabled when quasi-optimal architecture becomes an executable model driving the remainder of the systems engineering process. The paper uses generalized industrial examples to illustrate this empirical approach.