- Miyazawa, Y. and Dowell, E.H., Principal coordinate realization of state estimation and its application to order reduction,
J. Guid. Control Dyn. (USA), vol. 11 no. 3
pp. 286 - 8 .
(last updated on 2007/04/10)
Modeling of plant dynamics is an important problem, especially when designing a time-invariant optimal estimator. Since the optimal estimator is generally of the same order as the plant dynamics, the estimator becomes of high order if the plant is modeled accurately with a high-order system. In practice, however, such full-state estimation is often useless, and some sort of order reduction is possible and necessary. In order to make the iterative algorithm converge to the global minimum, an appropriate initial solution, or a suboptimal reduced-order estimator, is necessary. Furthermore, a quantitative index for measuring the accuracy of the order reduction is desirable. The paper proposes that a set of singular values be used as a quantitative index that shows estimation accuracy in terms of the state components. Such an approach introduces a unique realization of the plant model and the estimator. Since the singular values show not only estimation quality but also coupling intensities of each state, a principal coordinate realization can be used for the derivation of a reduced-order (or simplified) estimator. The proposed reduced-order estimator does not claim any optimality, nor guarantee better performance than other methods. However, with only a small computational demand, it gives reasonable results, especially when the system has several small singular values