Simulating linear model, can't understand error

I have a linear system called "ss" identified using "velocity", "position", and "time_sec_" vectors imported from a CSV file.

Used time interval of 0.02, delay of 0.4, included disturbance.

The 3rd order system looks like this:

>> [A,B,C,D,K,x0,dA,dB,dC,dD,dx0] = idssdata(ss)

A =

   -0.2101   -3.0985   -0.8638
    2.5007   -4.9519  -11.9558
    0.1966    0.4627    8.4746

B =

    0.0056
   -0.0048
   -0.0562

C =

   31.8116   -0.8537    0.0245

D =

     0

K =

    1.3349
  -32.7004
   40.9704

x0 =

   -0.0165
   -0.0724
    0.0023

dA =

   1.0e+14 *

    1.2124    0.1556    0.1185
    2.6893    1.5376    1.1212
    2.1587    1.5017    1.2671

dB =

   1.0e+11 *

    0.2318
    4.1463
    5.7636

dC =

   1.0e+14 *

    1.0754    0.0817    0.0737

dD =

     0

dx0 =

   1.0e+14 *

    0.0877
    1.6986
    2.8014

I want to simulate the system. It's SISO (velocity -> position) but because it's third order I have to duplicate the velocity three times, one for each system state variable.

U = velocity * ones(1, 3)

Here is what I am calling lsim with:

>> size(U)

ans =

        2415           3

>> size(time_sec_)

ans =

        2415           1

I attempt to simulate the system:

lsim(ss, U, time_sec_)

Matlab returns the following error:

Error using DynamicSystem/lsim

When simulating the response to a specific input signal, the input

data U must be a matrix with as many rows as samples in the time

vector T, and as many columns as input channels.

This error message doesn't make sense because:

• U has as many rows as time samples in time_sec_ vector
• U has as many columns as rows in "B" matrix in the model (third order model, B has 3 rows)
• time_sec_ vector has one row for each time sample
• Time samples are absolute values increasing in time, in seconds

Answer 1

[Paul]

Hi Valeriy,

Because the B matrix has one column, the input to the system is a scalar. Hence the U input to lsim should have one column and as many rows as time samples. Each entry in U is the scalar input to the the system at the corresponding time sample in time_sec_

Also, I suggest not using ss as a variable name. Doing so overrides the same-named functions in a couple of different toolboxes.

which ss -all

/MATLAB/toolbox/control/ctrlmodels/@ss/ss.m                 % ss constructor
/MATLAB/toolbox/control/ctrlmodels/@DynamicSystem/ss.m      % DynamicSystem method
/MATLAB/toolbox/ident/ident/@idParametric/ss.m              % idParametric method
/MATLAB/toolbox/mpc/mpc/@mpc/ss.m                           % mpc method
/MATLAB/toolbox/shared/controllib/engine/@StaticModel/ss.m  % StaticModel method

[Valeriy]

This is the first thing I tried but it didn't work. I finally realized that the first error message I got was complaining about the time vector, because it was the original time samples which were not completely evenly spaced.

It finally succeeded when I generated the time samples synthetically, and used that as the time input to lsim:

linspace(7, 64, 29000)