0

Cross-posted from signal processing stack exchange:

Consider the following approach for de-rating proportional feedback gains to account for a measurement delay by controlling a future-projected response:

enter image description here

In practice, I've found this works pretty well for the simple case of controlling the velocity of a DC motor whose position is measured by a rotary encoder. However, position control is troublesome, because typically we don't have the same phase lag in the position and the velocity measurements due to the need to compute velocity via some form of finite differencing (which has some inherent lag).

Given optimal PD gains for such a position-velocity system (e.g. as computed via LQR or pole-placement), is there comparably simple way to "de-rate" the gains in the above manner, but reflecting this difference? A naive attempt to divide the future-projection into "chunks" quickly devolved into calculus/algebra soup, and I admit such things are not my forte.

Any advice, including fundamentally different approaches to the problem, are welcome!

user3716267
  • 1,421
  • This question is similar to this one. Do any of the answers to it answer some of your questions as well? – Kwin van der Veen Jun 25 '25 at 01:27
  • @KwinvanderVeen Unfortunately, not really. I'm particularly interested in the case where our effective measurement delay is different for our velocity and position data. – user3716267 Jun 25 '25 at 01:36
  • Can you clarify what you mean exactly with "de-rating"? Are you referring to compensating for time delay? The model that your attachment shows is continues time, but for derivative calculation you mention finite differencing, which implies a discrete time. So what is the exact problem that you are considering? – Kwin van der Veen Jun 25 '25 at 03:07
  • @KwinvanderVeen By "de-rating," I mean shrinking the gains so that you do not get a wildly unstable system. The above approach (i've edited to show the discrete version) just shrinks the LQR optimal gain by a factor of exp((A-BK)^tau). – user3716267 Jun 25 '25 at 13:40

0 Answers0