Abstract | ||
---|---|---|
Practical application of H[infinity] robust control relies on system identification of a valid model-set, described by a norm-bounded differential inclusion, which explains all possible behavior for the control plant. This is usually approximated by measuring the plant repeatedly and finding a model that explains all observed behavior. Typical modern approaches must anticipate the uncertainty-shaping aspects of the final model in order to maintain tractability. This paper offers a linear matrix inequality constrained optimization for the MIMO model fitting problem that does not require such knowledge. We do this with a novel Quadric Inclusion Program which replaces the least squares problem in traditional model identification---however rather than linear equation models, it returns linear norm-bounded inclusion models. We prove several key properties of this algorithm and give a geometric interpretation for its behavior. While we stress that the models are outlier-sensitive by design, we offer a method to mitigate the effect of measurement noise. The paper includes an example of how the theory could be applied to frequency domain data. Time-domain data could also be used, provided a state vector is constructed from measured signals and their derivatives to use as regressors for a vector of maximal derivatives. |
Year | Venue | Field |
---|---|---|
2018 | arXiv: Optimization and Control | H-infinity methods in control theory,Differential inclusion,Linear equation,Mathematical optimization,State vector,Robust control,System identification,Mathematics,Linear matrix inequality,Constrained optimization |
DocType | Volume | Citations |
Journal | abs/1802.07695 | 1 |
PageRank | References | Authors |
0.37 | 5 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Gray C. Thomas | 1 | 48 | 6.31 |
Luis Sentis | 2 | 574 | 59.74 |