Abstract
Cyber-physical systems (CPS) are engineered systems that have a tight integration between the cyber part (computation and networking) and its physical components. Examples include but are not limited to industry 4.0, automotive and aerospace. To allow decisions to be made in a CPS (strategic control, tactical control and, low-level control), decision models are used. These models use input from sensors, but also from other supporting processes, e.g. predictions over the state of its contexts, to come to a control decision. The decision processes are implemented in software that runs on embedded hardware and is commonly real-time constrained, meaning that the time at which the decision is taken, is as import as the decision itself. In literature several techniques are available to reduce the computational cost of executing models by using abstraction and approximation (e.g. surrogate modelling). This reduced cost would make the process to come to a decision easier (scheduling) and would require less computational resources. However, we still need to be sure that the decision process is robust against approximations and uncertainties in these models. Furthermore, an approximated and/or abstracted model is most probable not valid in all the different contexts the system will be in. To enable this, the system should be able to switch at run-time between different abstractions and approximations. Therefore, this project will create the foundations to reason about dynamically adapting the decision models and prediction models with different abstractions and approximations depending on the context of the system. The project will result in a framework with supporting modelling languages, methods and proof-of-concept tools to reason on the trade-off between uncertainty (from the approximation) and the real-time behavior of the system.
Researcher(s)
Research team(s)
Project type(s)