A Direct Method for State-Space Model and Observer/Kalman Filter Gain Identification

Minh Phan, Jason A. Solbeck, Laura Ray

AIAA Guidance, Navigation, and Control Conference and Exhibit · 2004 · 12 citations · 4 references

Concepts

Abstract

This paper presents a new direct method that identifies a linear state-space model of a system from input-output data. This is in contrast to existing indirect methods that first solve for the system Markov parameters, then use one of various realization algorithms to find a state-space representation. The direct method is very simple to implement and consists of solving only a few linear equations, both in the identification step and the model reduction step. When combined with the Observer/Kalman filter Identification (OKID) algorithm, the method also yields a simple way to produce an optimal Kalman filter gain for systems corrupted by process and measurement noise of unknown statistics. Both simulation and experimental results are provided to illustrate the direct identification method.

References

4