NLMPC state function parameters

I am currently developing a NLMPC controller for my system and the plant model described in the MPC uses a large array of parameters. Since I would like to adapt these parameters through my simulation I do not want to define them directly in my stateFcn.

According to the matlab documentation each one of the parameters should be define din the handle of the stateFcn, e.g.:

function stateFcCT(x, u, p1, p2, p3, p4, ..., p37)

These parameters are fed through a predefined bus which is also a little bit troublesome to define. There is no example using more than one parameter and since I am using an EKF I am not sure how to feed all these parameters into the EKF block....

Is there an easier way to define these parameters for my state functions?

I have tried using an external function but this seems to slow the simulations too heavily.

varargin also doesnt work in this case...

Answer 1

[Emmanouil Tzorakoleftherakis]

Hi,

Given that it is not possible to package all variables into a struct and feed them into the MPC controller that way (it's not supported), one alternative would be to create a vector with all the desired values and pass it as a single variable to the state function. You can then index into the vector to access individual parameters as needed

[Eduardo]

I have seen an example of NLMPC in the newer versions of Matlab (2021b upwards) that uses more than one parameter for the state function (Example)

This, however, is not working in my case. I get the following error while the model is compiling:

The line mentioned corresponds to the first use of the additional parameters in the state function:

ts = p(3);  

In the base workspace "p" is defined as a column vector just as in the example [1;2;3;4;.....] and then fed into the bus creator element with a constant block.

Any idea why this is happening?

[Eduardo]

I have tried using an IC block and also a signal specification block to get the dimension of the output but to not avail...

[Eduardo]

I just found the source of error.

I am also using a Kalman filter to estimate the states from measurements and since it uses the state function the parameters should also go into the block.

Currently I commented out the filter just to see if the NLMPC runs and it does. Now the question would be how to input the array into the EKF so that it can be compiled