How to solve the discontinuity of MPC controller output？

In simulink, I used the MPC toolbox to obtain the control value calculated by MPC with a sampling period of 0.05s.

In the following calculations, I need to derive the control variable. However, 0.05s is much smaller than the simulation step size, which leads to an infinite value for the derivative.

I tried to use low-pass filtering, but it still can't solve the problem well.

The MPC algorithm should calculate the control variables in the control horizon together, and then output them one by one. Is it possible to perform interpolation processing before the output of the MPC controller? Or is there any better way to solve this difficulty?

Answer 1

[Jingchao Li]

When I use the discrete derivation module, the problem is solve

Answer 2

[Fangjun Jiang]

Use a rate transition block or zero-order holder block to deal with different sample rate

[Jingchao Li]

I tried to use the rate transition block, but when the MPC output is not connected to other modules, the output becomes very smooth, but once connected to the controlled object, it cannot work.

[Fangjun Jiang]

Then you might have a closed-loop oscillation (stability) issue. Taking derivative is prone to this type of issue. low-pass filtering should help and maybe you should play with it more.

[Jingchao Li]

In the process of using the low-pass filter, I also encountered the same problem:

I added a low pass filter after the output of MPC.When it is not connected to other modules, the output becomes very smooth, but once connected to the controlled object, it cannot work.The system has no feedback, and the MPC output to the low-pass filter is exactly the same.

[Jingchao Li]

Thank you, when I use the discrete derivation module, the problem is solved