I have the following problem: I have a SIMO system and I want to apply subspace identification algorithm (MOESP in the specific case) using the Matlab function n4sid (which performs subspace identification of state space models) of the System Identification Toolbox. Due to the large amount of data and outputs is not possible to feed the algorithm with all the data at once but something more gradually is more suitable.
Consider I have just 2 outputs with one single input: I would like to perform subspace identification of the 2 SISO systems independently and then concatenate them to obtain the final and complete model having the same order of previous models: is this possible with the default system identification toolbox of Matlab? Do you have any reference? I have already asked on the Matlab forums but I get no answer plus I read User's guide of the System Identification Toolbox.
// Example: in = excitation signal. out1, out2 = output signals
// Ts = sampling interval The signals are 0 mean
// putting data into iddata format
DAT1 = iddata(out1,in,Ts);
DAT2 = iddata(out2,in,Ts);
//Considering a model order of 10
M1 = n4sid(DAT1,10); // M1 = SISO model with output out1 and input in
M2 = n4sid(DAT2,10); // M2 = SISO model with output out2 and input in
// Estimating the model with 2 outputs and 1 input:
DAT = iddata([out1 out2], in,Ts);
M = n4sid(DAT,10); // M = SIMO model with outputs out1 and out2 and input in
Looking at the example: how to get M from M1 and M2 without performing estimation again?
I have also a related question: the subspace algorithms are based on the L-Q decomposition of a matrix. Suppose I have 2 matrices to be decomposed for example A1, A2. Is there a way in Matlab to get the decomposition of the matrix obtained by concatenating A1 A2 as
A =[A1, A2]
from the single decompositions?(Especially I am interested in the L matrix) Thanks in advance
