On the improvement of an adaptive observer for multi-output systems

The paper proposes a design for an adaptive observer for multivariable systems and considers possible improvements for the identification of the unknown parameters. An orthogonalised projection algorithm which provides good convergence is described for identifying the unknown parameters of an aircraft. An adaptive observer which does not identify the plant´s initial state vector is also presented. Simulation studies are carried out using the data for the Japanese T-2 CCV aircraft to show that the orthogonalised projection method has better parameter convergence, and gives a better insight of how well the system is excited for parameter identification, than the least-squares method in ideal cases.