The engineered systems , such as autonomous self-driving vehicles, that we (want to) design and build, are characterized by an ever increasing complexity , offering ever more advanced functionality and comfort. At the same time, the demands on energy efficiency and cost, but also on safety and reliability of those systems, become more stringent, in a quest for some form of optimal, fit-for-purpose designs. Furthermore, in a circular economy, we wish to take into account not only the product, but an ecosystem, spanning entire families of related products, over their entire life-cycle, including production, maintenance, and recycling. The fact that such advanced systems can be built today is largely thanks to the ubiquitous use of models . Models, encoding (for reuse) our knowledge about various aspects of a system or system component, can namely be used for "virtual experimentation" : to perform computer simulations to answer "what if" questions. Such questions allow us to explore different design alternatives. It is this capability that is fueling the 4 th industrial revolution. Models in complex engineered systems vary widely in nature and purpose. They may describe structure and behaviour of systems in different domains such as mechanical, electrical, software, and networks, or different views on the systems such as the stability/control view, the safety view, and the cost/efficiency view, at different levels of abstraction/detail/fidelity. The may also be used to describe and even prescribe (for automation purposes) the complex, concurrent development processes. Process models can be used for "what if" analysis of the engineering processes themselves, leading not only to optimal products, but also to optimal time-to-market. When "what if" analysis is automated , exploring billions of alternatives efficiently in a computer, reaching optimal products/production designs can be accelerated , taking a matter of days or weeks on a cloud computing infrastructure as opposed to the decades required for organic convergence over generations of human engineering improvements. Engineering is however hitting a wall, keeping us from a truly exponential leap in complex systems development . Though advanced computer support exists in the form of modelling languages, model management tools, simulators, etc. for "what if" analysis, managing the meaningful and correct (re)use of models is still a mostly human enterprise, for which no rigorous foundations nor advanced tooling exist. Being constrained by human capabilities, it is costly, slow, and error prone. In some important, yet restricted, areas such as Electronic Design Automation, such foundations and tooling do exist (and fuel a thriving billion $ market). For truly complex, multi-domain systems, knowledge is scattered, often either in experts' minds, or in the best case in text documents and spreadsheets. In this project, we propose to develop a foundational framework as well as prototype tooling for the computer-assisted/automated meaningful (re)use of models . The key to our approach is that we will "eat our own dog food" : we will now apply advanced modelling language engineering, model transformation, property specification, modelling and simulation techniques we have helped develop over the last decades, to explicitly model and reason about the context in which models can be meaningfully (re)used. We call such models "frames" after the original, but incomplete "experimental frames" idea proposed by Bernard Zeigler in the 1980s. Concretely, we will start by using our experience with the modelling language Modelica (for physical systems) and DEVS (for discrete-event modelling of software and networks) to develop the theoretical foundations and application of frames, initially on a representative autonomous vehicle case .