Figure 1
(a) The nominal model represents the current assumptions about the true system. The systems model is specified by its dynamic state variables and their interactions, here represented as vertices and edges of a graph. The systems border defines the distinction between internal states and exogenous inputs. The exogenous inputs u are assumed to be known. (b) In reality, the nominal model is embedded in a larger network outside the nominal systems border. The hidden dynamics of the exosystem interacts with the nominal system. In addition, some interactions between nominal state variables might be missing or misspecified in the nominal model. These model errors can potentially lead to discrepancies between model and experimental data. (c) Representation of model errors as hidden inputs to the nominal model. The dynamic elastic-net approach infers the hidden inputs from data and thereby corrects for the bias in the nominal state variables induced by model errors.
