S3: Architecting intelligent systems
Is a paradigm shift needed in architecting intelligent systems?
The examples of artificial intelligent systems, such as self-driving cars, autonomous drones or autonomous weapons, are all well-known, hyped may be. From architecting perspective, does it make a difference to call them artificial intelligent and does that influence the way architects have to do their job? Is a paradigm shift needed to architect these systems? Can architects delegate their work to smart digital assistants? These and many other questions are being discussed in this session, with a focus on the impact of AI on systems architecting.
Moderation: Richard Doornbos, ESI (TNO)
The possibilities to apply artificial intelligence in our systems poses new challenges and opportunities for system architects. The intelligent system will adapt to new unforeseen circumstances. What architecting skills are needed to ‘guarantee’ a well behaved system? We’ll discuss the challenges in this track.
Architecting Intelligent Cyber Physical Systems
There is an ongoing trend on Cyber Physical Systems (CPS) to become more and more complex, due to the relative low cost of the always increasing computing power. Currently this also includes the addition of some flavor of intelligence. The main task of the System Architect is to match business value proposition and market opportunities with technology, to come to the most optimal product definition, balancing all relevant aspects. When a cyber physical systems becomes intelligent, how does it affect this way of reasoning? Are available reasoning frameworks like CAFCR still useful?
Can AI create patent-worthy designs and novel system architectures?
A novel design of a complex engineering system depends on creativity, intuition and a deep technical understanding of its designer - which often is also secured by a patent. Such systems have in most cases a multi-physical nature and are rarely understood by one single engineer. An example for such a design problem is the design of hybrid automotive drivetrain. The (inter)-dependency of each component on each other is hard to grasp intuitively or analytically in simple mathematical models. The only remedy is a systems thinking approach, were design decisions are based on abstractions, behavior simulations, multiple domain-experts and were also alternative designs are taken into account. Especially the alternative solutions for one specific engineering problem often decide on the success of a product.
In the presentation a methodology, generative engineering, for automatically generating systems architectures will be presented. The method will be demonstrated on various examples from automotive and aerospace. The methodology allows generating architectures from a declarative system model. This declarative model captures the designer's intent and is formal to be treated mathematically. The question about patent-worthiness will answered by reflecting on existing patents and creating them alongside with new designs which outperform the existing ones. This will be done a realistic engineering examples, like hybrid powertrains or urban air mobility solutions