Engineering the Smarts : An Illustration of the Disconnect between Control Engineering and AI
conference paper
The design and analysis of smart system behavior needs to bridge between worlds – both with regard to engineering methods and to the technologies central to adaptivity and thus a
core functionality of smart systems-of-systems. We illustrate this in our domain, smart buildings, where AI technologies like rule based or probabilistic reasoning set the strategies that define the building’s adaptive behavior, but control loops firmly set in control engineering implement it. However, we observe an apparent disconnect between these approaches. Using demand response (DR, the change of consumption patterns in response to global demand) as an example, we illustrate consequences of such mono-disciplinary thinking that both academic researchers and engineering practitioners are prone to. Specifically, we show how adaptive behavior couples previously independent dynamics and how reacting to these dynamics may lead to emerging unintended system state changes and non-acceptable performance. We argue for stringent links between system design and analysis based on probabilistics and simulation, as adaptivity in cyber physical systems needs the best from both AI and control engineering.
core functionality of smart systems-of-systems. We illustrate this in our domain, smart buildings, where AI technologies like rule based or probabilistic reasoning set the strategies that define the building’s adaptive behavior, but control loops firmly set in control engineering implement it. However, we observe an apparent disconnect between these approaches. Using demand response (DR, the change of consumption patterns in response to global demand) as an example, we illustrate consequences of such mono-disciplinary thinking that both academic researchers and engineering practitioners are prone to. Specifically, we show how adaptive behavior couples previously independent dynamics and how reacting to these dynamics may lead to emerging unintended system state changes and non-acceptable performance. We argue for stringent links between system design and analysis based on probabilistics and simulation, as adaptivity in cyber physical systems needs the best from both AI and control engineering.
TNO Identifier
551109
ISBN
978-1-4673-8727-9
Publisher
IEEE
Source title
11th System of Systems Engineering Conference (SoSE), 2016.
Collation
5 p.
Files
To receive the publication files, please send an e-mail request to TNO Repository.