ไม่สามารถเล่นวิดีโอนี้
ขออภัยในความไม่สะดวก
Machine Learning Control: Tuning a PID Controller with Genetic Algorithms
ฝัง
- เผยแพร่เมื่อ 9 มิ.ย. 2018
- This lecture shows how to use genetic algorithms to tune the parameters of a PID controller. Tuning a PID controller with genetic algorithms is not generally recommended, but is used to demonstrate evolutionary control algorithms.
Machine Learning Control
T. Duriez, S. L. Brunton, and B. R. Noack
www.springer.c...
Closed-Loop Turbulence Control: Progress and Challenges
S. L. Brunton and B. R. Noack
appliedmechanic...
Code: faculty.washington.edu/sbrunton/DataDrivenControl.zip
www.eigensteve...
This video was produced at the University of Washington
Great work.
I would also like to point out that it can be a good idea to also plot the control law output as function of time, and play with different values of Q and R to see their effect on the outcome.
this was a great video steve, appreciate it a ton!
very clear illustration! huge respect!
Holy shit, that was crazy, i never heard of some self tuning PID like this! Amazing!!
Love your work, but I have a question that other than using GA what are the other easier algorithms that is recommended for PID tunings ? I’m doing a project on this and is really scratching my head off 😵 please help 🙏🏻 thank you very much in advance
What a great example Steve, really well explained, thanks.
I know that you discouraged the use of this for real PID but i was thinking, maybe i could actually use it to tune some PI controllers when the system is non linear and matlab fails to identify the plant using diferent approachs.
I mean if everything fails and other classical methods are too time consuming, just let the pc runing overnight to check if this can tune it. Or maybe you have a better approach to this.
If you can afford to run it overnight, I don't see why not to give it a try. For this type of optimization, there are a ton of approaches, including Monte Carlo and simplex methods, which might also work and may be pretty efficient.
@@Eigensteve How exactly use simplex method for this kind of task?
How can we introduce upper and lower bounds on tuning parameters on this ga (genetic algorithm). Kindly help me.
Excellent work
Excellent work...
How do you convert your gains values into genetic components?
At 6:18 - "High pass filter." Shouldn't it be a low pass filter as the 1/(1+0.001s) term is attenuating high frequencies?
Hello sir, I was working cardiac pacemaker, Can this techniques can be used over there since pacemakers are planted externally , could this technique is practically feasible w.r.t to computations
Excellent informative lecture, sir is it possible to imply the idea to higher-order state space model?
pure beauty..thanks
Thanks
Plz can I use the same things in the code with a discrete PID controller ?
Excellent dear. I wanted to know how can we tune 2 control inputs ( for example u1 and u2) in order to minimize a cost function the Dynamic system of which is Nonlinear. Sadly, we can not derive the Closed loop transfer function. Can You guide me with that please?. I want to do this with Genetic Algorithm.
sir ho to apply genetic algorithm for a simulink model for PID controller implemented in a boost converter
Thanks for the extensive explanation and I have an inquiry in this regard.
Can we use this algorithm in the state space (in terms of q and q_dot)?
you have mentioned that the transfer function is a function of (s) in the frequency domain. Then what happened if we have a model based controller and we want to tune its parameters??
Have anyone try it with MIMO( multiple input multiple output) system?
Thank you a lot...professor!
@8:05, line 11 should be u=lsim(CTRLtf,1-y,t). am I right?
yeah, I also have doubt of it
No. The closed loop system is simulated on line 8 by giving it a step input (yd =1), and the output y is recorded. In order to get the controller output (u), you simulate the controller transfer function K directly with its own input "yd- y" which is (1-y). However, you could also simulate CTRLtf with a STEP input using "step(CTRLtf,t)" and get the same results.
7:29 shouldnt line 10 be CTRLtf = (K*G)/(1+K*G) ?
Thank you. 👍
How to define the bounds of PID?
Thanks man!
hey, im new to MATLAB so i just wanna try this code that you write... but after i run this code, it's error and it said that error in using save, error in myfun, error in gaoutput and many more. can you explain what happen please. thanks
Sir can you provide a cost function that penalises on the maximum value of state or input?
In general, I don't see why not. Genetic algorithms can handle all kinds of strange, non-convex and non-smooth cost functions. Although this might make the optimization a bit more difficult. Model predictive control is really good for these types of hard constraints though.
@@Eigensteve Sir apart from the constraint used here can we include an another constraint on u simultaneously? If yes then how?
Hi steve. I just want to know that, is there any boards that will support the database with the PID controller? If exists, could you please help me in knowing the boards. Thank you steve for ur very useful videos.❤❤😍🧡
Excellent job! 👏 Is the script available to download anywhere?
Thanks in advance👍
Thanks! All code available at databookuw.com
@@Eigensteve Lot of thanks!👏👍👋
Where should I get this code.i have to do project
Is it the same for fractional order pid ??
DEAR did you find Any anserwers or not yet
Sir superawsome vedio....can u please tell me also the MOGA multiobjeective genetic algorithm in MATLAB ?
can you tell me where can i get the code
is the code still available?
can i have the matlab code of this
Certainly. All code is available at databookuw.com under the CODE.zip link
👍👍
he didnt show the control signal and there is no noise added, not very good video
also if you only have 3 parameters you can just loop through them with 3 loops, it would give similar results