Dear Amardeep,

For resolving your problem i suggest you to see links and ayyached file in topic.

-Adaptive fuzzy sliding mode control of uncertain nonlinear systems

www.scielo.br/scielo.php?script=sci_arttext&pid=S0103...

-Double Adaptive Fuzzy Sliding Mode Control Scheme for Electro ...

https://scialert.net/fulltextmobile/?doi=itj.2012.620.625

-Double Adaptive Fuzzy Sliding Mode Control Scheme for Electro ...

https://scialert.net/fulltextmobile/?doi=itj.2012.620.625

- Adaptive fuzzy sliding mode controller for wheel slip control in ...

https://www.researchgate.net/.../305521124_Adaptive_fuzzy_slidin...

-An adaptive Fuzzy sliding mode controller for reactive power ...

ieeexplore.ieee.org/document/7848639

Best regards

Hi Amardeep,

Adaptive control methods are commonly used to estimate the unknown nonlinearity or process in a dynamical model.

What does your adaptive control do?

Can you plot a comparison between the actual signal and the estimated signal?

Hi Yew!

I am trying to control the attitude of my airship using AFSMC wherein adaptive fuzzy is being used to approximate the unknown part or the uncertain part of the dynamics. It is able to control the system when the desired trajectory is either a fixed point i.e stabilization problem or if the reference trajectory is slowly varying. Can you tell me if there is way to improve the response of AFSMC in the case where the desired reference signal is fast varying by fast varying I mean y_d=.6*sin(t) and by slow I mean y_d=.6*sin(.01t). Here are the plot of output theta and psi. Notice how the output behaves differently in 3 different cases.

Hi Deep,

You seem to have overlooked my queries. See Post #2. Also, for the convenience of the readers, please attach the image in JPG format.

Do you have any idea of how you might troubleshoot the control system?

Could you show the results for the ideal case (non-adaptive)? It is important to find out the performance of the non-adaptive controller for tracking fast-varying reference signal.

Hi :

My suggestions are as follows:

(1) Do you must use the "adaptive" fuzzy sliding control for your UAV? If not, fuzzy logic control is more appropriate.

(2) If you must use an "adaptive" fuzzy sliding-mode control, an adaptive compensation of the coefficient(s) of upper bound of uncertainties is more suitable for the implementation of UAV.

(3) The motor dynamics of UAV should be considered to deal with high-frequency behavior.

(4) A (An adaptive) finite-time tracking control can be considered.

The attached paper is your reference.

CL

You can use a type-2 fuzzy logic system written in C languaje

I'm unsure if Amardeep is willing to switch to Type-2 Fuzzy System at this moment in time because his research progress is already more than halfway through.

Looking for alternative solution is probably not his intention for posting this question in the first place. He is looking for advices and tips to tweak his adaptive control algorithm.

Dear Amar,

I suggest you to look how many fuzzy rules you have taken. In fuzzy as universal function approximator, it is important what initial values you have taken in adaptive rules. You should focus on these two issues. I hope you may get some better result.

Dear Mishra ,

As far as know, generally adaptive laws are extremely sensitive to design parameters. I personally have faced lots of similar problems in adaptive control design. So, as an trial and error procedure, I suggest to go with exaggeratedly high parameters, in adaptive laws, to see if the designed controller is able to operate properly or not. if yes, then you may decrease the value of parameters. this is the case especially for the part of the law which is going to guarantee the stability. for example, say p_hat_dot=f(x)-mu*p_hat, where, p_hat is estimated parameter, f(x) is a part of law which cancels the known part of dynamic, and -mu*p_hat is the part of adaptive law which guarantees the stability. the parameter mu, based on my experience, plays a vital role in closed-loop performance. so, I suggest to check different value for mu, starting from very big one.

I hope it is helpful.

regards,

Hi Yew!

Yes, since I am almost done with my results, I merely wanted to tweak the performance of my adaptive fuzzy sliding mode control and thats why I asked this question. Changing the entire control strategy at this juncture will not be fruitful.

Hi Amardeep,

Can you perform the parameter convergence analysis for the 3 cases and post them here? Parameter drift is suspected.

1. Step reference

2. Fast time-varying reference

3. Slow time-varying reference

Hi yew! I'm do it by tomorrow and post the results. Thanks in advance

Hi Yew, here are the plots that you asked for-

1) Case 1: i.e parameter values theta1 and theta2 for fixed point stabilization. As we can see the learning is perfect and they converge to a value.

2) Case 2: i.e parameter values theta1 and theta2 for a fast varying reference signal- i.e .4*sin(.1*t). Here in this case as it is evident that parameters fail to converge at any value and are unboundedly increasing.

3) Case 3: i.e parameter value theta1 and theta2 for slow varying reference signal- .4*sin(.01*t). Here in this case, both the parameters settle down to some value just like the first case.

So, yes as you suspected, the parameters are drifting in the second case.

Good. The adaptation dynamics is probably excited by the oscillations from the fast-varying reference input. You should modify the adaptation law to avoid the parameter drift. There are several modification techniques you can take to solve the problem of parameter drift. What was your original adaptation technique?

Hi yew!

First of I'd like to clarify that two parameters theta1 and theta2 are for two control u1 and u2 in my system.

I'm using lyapunov based synthesis, I.e parameter update law is

Theta_dot=gamma*s*h(s). Where s is sliding variable and h(s) is fuzzy basis function arising out of fuzzy system with only s as input to that fuzzy system.

I don't know which technique is suitable for your case. You may try:

Dear Mishra ,

You have not say what is the type of fuzzy logic system you have used ? I think that the zero order takagi sugeno can improuve the quality of your performences, abous the sliding mode you can see the fast terminal slididng mode control this approach can be usefull for you