Identifying and Handling Uncertainties in the Feedback Control Loop

In the feedback control loop, uncertainty is associated to different sources (e.g., the environment), and appears in different forms (e.g., as noise in variables or imperfections in techniques being used).

In the MAPE-K control loop, uncertainty is normally handled by the decision maker at the Plan stage.

However, depending on the complexity of the stages of the MAPE-K control loop, uncertainties need also to be handled at other stages, depending on the uncertainties associated with that stage.

Moreover, uncertainties may also propagate between the stages of the control loop, which might affect how uncertainties are handled.

In this position paper, we claim that uncertainties should be identified and handled at the different stages of the feedback control loop.

We propose an approach for the identification of internal and external sources of uncertainty for a given stage, and we promote error propagation analysis as a method for analysing the propagation of uncertainties between stages.

In terms of tradeoff analysis, which can take place at any stage of the MAPE-K control loop, such an approach provides a clear benefit since it leads to a more accurate estimation of the system quality attributes because uncertainties are handled in the context where they arise.