Hello:
I have a question about the definition of adaptive control, since I´m researching about making a model-free adaptive control system. I will appreciate your help.
The definition I found says that an adaptive control modifies its parameters or structure in order to achieve a performance index. Reading about the subject in different sources I noticed that when they refer to an adaptive controller, it always has a model of the plant and implies an adaptation law which is usually obtained by taking the model and manipulating expressions.
My deduction is that when these sources refers to adaptive control, is about a kind of this instead. Am I right? I will really appreciate your support on this.
Thanks.
Pablo.
The wheel circumference of the bogie wheels changes as the locomotive travels. There is no correction related to circumferencial speed of the bogie wheels. That is why slip/slide control is necessary and each wheel must be separately driven and controlled.
Bernard Schaffler.
People do many things without adaptive control.
However, at some point, we may understand that the performance may not be what we were looking for. Either the plant parameters were not exactly what we thought, or they might have changed from what they were in the beginning. Moreover, the object of control could be time-varying, nonlinear, etc.
This is what led people to think about “adapting” the control parameter to the actual situation.
A first idea was to only change the main control loop gain. This was called the MIT rule and it worked nicely and was even implemented on a plane that flew nicely…until it crashed. Comes out that working with time-varying control parameters is not the same as fixed parameters.
This made people reluctant to the idea of adaptive control.
Then, some names, such as Prof. Kumpati (Bob) Narendra from Yale, contributed a lot to the theory of stability and also came up with their own idea of adaptive control. This appears in literature as Model Reference Adaptive Control (MRAC) and this is what you usually see in literature.
It requires the plant to behave like some ideal model of the same order and indeed, it tries to modify the parameters of the big plant.
As it happens, by 1980, after some engineering experience, I came to RPI (Rensselaer) for a PhD. The “problem” was that the Professor’s grant was for Large Flexible Structures (LFS). They were not only large plants but also MIMO, so there was not much we could do with the classical MRAC.
They started using another idea, which was called Command Generator Tracking (CGT). You don’t try to make all state variables of the big plant follow all state variables of some big model. Instead, you define some low-order model that only defines the trajectory that you would like your plant output to track. They had some publications, yet the idea was negatively received, so they were almost on their way out. However, the grant forced my advisor to let me “try to do something.”
As I was coming from the industry, I liked this idea of a simplified controller with varying parameters, and I even “felt” that it should work. The plant can be of order 20 or 50 and yet the model and the controller can be of order 5.
What can I say? I ended up with a low-order adaptive controller that managed to control the pretty big LFS. It was interesting that people at JPL liked the idea and even dared to implement it on their large flexible antenna.
After playing with it, I decided to call our approach “Simple Adaptive Control (SAC).”
I won’t make this message a dissertation, but if you write Simple Adaptive Control on Google, you can see quite a bit, of publications from all around the world.
Also, please always feel free to come back with any questions.
- Badges
- Science method