search

Rajesh-Siraskar / Reinforcement-Learning-for-Control-of-Valves

This project uses DDPG for "optimal" control of non-linear valves. Uses MATLAB and Simulink
67 stars 11 forks source link

Request file #1

Open TangJichuan opened 3 years ago

TangJichuan commented 3 years ago

Hello, I am very excited to see how you use matlab/simulink to apply reinforcement learning in the control system. Can the pre-trained model Grade_IV_ver.3_600.mat in your code provide me with in-depth research verification? Grateful

Rajesh-Siraskar commented 3 years ago

Hello Mr. Tang Jichuan,

Thank you for your appreciation.

Your question is: ".... provide me with in-depth research verification?"

So the model was verified as working with many experiments carried out on a Simulink non-linear valve. The description and observations made from these experiments are explained in this paper: https://arxiv.org/abs/2012.14668

Please see section 6: Experiments, Results and Discussion, on page 10.

I hope this helps.

With kind regards, Rajesh

On Mon, Mar 1, 2021, 6:25 PM TangJichuan notifications@github.com wrote:

Hello, I am very excited to see how you use matlab/simulink to apply reinforcement learning in the control system. Can the pre-trained model Grade_IV_ver.3_600.mat in your code provide me with in-depth research verification? Grateful

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/ASKXGU3M3X4N5GMADQNADU3TBOFCJANCNFSM4YMPRKGQ .

TangJichuan commented 3 years ago

Hello Mr.Rajesh Thank you for your apply.I will carefully review your suggestion. By the way,When training the reinforcement learning agent, I noticed that you are using the same constant reference signal as in the matlab official website. How about other complex signals for agent or controller training such as sine waves ?

Thank you very much

Jichuan Tang

------------------ 原始邮件 ------------------ 发件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves" <notifications@github.com>; 发送时间: 2021年3月2日(星期二) 中午1:38 收件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves"<Reinforcement-Learning-for-Control-of-Valves@noreply.github.com>; 抄送: "喜欢啃苹果"<1152249351@qq.com>;"Author"<author@noreply.github.com>; 主题: Re: [Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves] Request file (#1)

Hello Mr. Rajesh,

Thank you for your appreciation.

Your question is: ".... provide me with in-depth research verification?"

So the model was verified as working with many experiments carried out on a Simulink non-linear valve. The description and observations made from these experiments are explained in this paper: https://arxiv.org/abs/2012.14668

Please see section 6: Experiments, Results and Discussion, on page 10.

I hope this helps.

With kind regards, Rajesh

On Mon, Mar 1, 2021, 6:25 PM TangJichuan <notifications@github.com> wrote:

> Hello, I am very excited to see how you use matlab/simulink to apply > reinforcement learning in the control system. Can the pre-trained model > Grade_IV_ver.3_600.mat in your code provide me with in-depth research > verification? Grateful > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <https://github.com/Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves/issues/1&gt;, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ASKXGU3M3X4N5GMADQNADU3TBOFCJANCNFSM4YMPRKGQ&gt; > . >

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

添加到短语集

没有此单词集：英语 -> 中文（简体）...

创建新的单词集...

拷贝

Rajesh-Siraskar commented 3 years ago

Hello Jichuan,

It is not really a constant reference training signal. By moving up and down randomly, it results in training the agent for following a complex signal composed of ANY combination of "straight-lines".

But, yes, you are right that we can try sine-waves as well. This will generalize it for OTHER types of signals.

Jichuan - are you a researcher (working or student)? I am glad you are interested in RL for optimal control. It is an interesting subject.

I request that you cite my work if you use somewhere please.

Thank you, Rajesh

TangJichuan commented 3 years ago

Thank you for your reply. I am a PhD student in China. As early as last year, I did something similar to your work, but the effect was not satisfactory to me. I hope that in the future there will be more exchanges with you in the areas of reinforcement learning/machine learning/optimization control. Your research is of great value. I noticed that the paper is already in the review stage and I will cite it in the future.

------------------ 原始邮件 ------------------ 发件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves" <notifications@github.com>; 发送时间: 2021年3月2日(星期二) 下午2:15 收件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves"<Reinforcement-Learning-for-Control-of-Valves@noreply.github.com>; 抄送: "喜欢啃苹果"<1152249351@qq.com>;"Author"<author@noreply.github.com>; 主题: Re: [Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves] Request file (#1)

Hello Jichuan,

It is not really a constant reference training signal. By moving up and down randomly, it results in training the agent for following a complex signal composed of ANY combination of "straight-lines".

But, yes, you are right that we can try sine-waves as well. This will generalize it for OTHER types of signals.

Jichuan - are you a researcher (working or student)? I am glad you are interested in RL for optimal control. It is an interesting subject.

I request that you cite my work if you use somewhere please.

Thank you, Rajesh

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

Rajesh-Siraskar commented 3 years ago

Thank you Jichuan, I wish you luck in your PhD.

I too plan to apply and take up PhD maybe by next year.

YES - it will be great to be connected and learn jointly.

Thank you, Rajesh

On Tue, Mar 2, 2021 at 11:53 AM Jichuan Tang notifications@github.com wrote:

Thank you for your reply. I am a PhD student in China. As early as last year, I did something similar to your work, but the effect was not satisfactory to me. I hope that in the future there will be more exchanges with you in the areas of reinforcement learning/machine learning/optimization control. Your research is of great value. I noticed that the paper is already in the review stage and I will cite it in the future.

------------------ 原始邮件 ------------------ 发件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves" < notifications@github.com>; 发送时间: 2021年3月2日(星期二) 下午2:15 收件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves"< Reinforcement-Learning-for-Control-of-Valves@noreply.github.com>; 抄送: "喜欢啃苹果"<1152249351@qq.com>;"Author"<author@noreply.github.com>;

主题: Re: [Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves] Request file (#1)

Hello Jichuan,

It is not really a constant reference training signal. By moving up and down randomly, it results in training the agent for following a complex signal composed of ANY combination of "straight-lines".

But, yes, you are right that we can try sine-waves as well. This will generalize it for OTHER types of signals.

Jichuan - are you a researcher (working or student)? I am glad you are interested in RL for optimal control. It is an interesting subject.

I request that you cite my work if you use somewhere please.

Thank you, Rajesh

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves/issues/1#issuecomment-788649013, or unsubscribe https://github.com/notifications/unsubscribe-auth/ASKXGU6TQVPDMFNEP3V7OJLTBR76RANCNFSM4YMPRKGQ .

TangJichuan commented 3 years ago

Thank you, and sincerely wish you！

------------------ 原始邮件 ------------------ 发件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves" <notifications@github.com>; 发送时间: 2021年3月2日(星期二) 下午2:28 收件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves"<Reinforcement-Learning-for-Control-of-Valves@noreply.github.com>; 抄送: "喜欢啃苹果"<1152249351@qq.com>;"Author"<author@noreply.github.com>; 主题: Re: [Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves] Request file (#1)

Thank you Jichuan, I wish you luck in your PhD.

I too plan to apply and take up PhD maybe by next year.

YES - it will be great to be connected and learn jointly.

Thank you, Rajesh

On Tue, Mar 2, 2021 at 11:53 AM Jichuan Tang <notifications@github.com> wrote:

> Thank you for your reply. I am a PhD student in China. As early as last > year, I did something similar to your work, but the effect was not > satisfactory to me. I hope that in the future there will be more exchanges > with you in the areas of reinforcement learning/machine > learning/optimization control. Your research is of great value. I noticed > that the paper is already in the review stage and I will cite it in the > future. > > > > > > ------------------&nbsp;原始邮件&nbsp;------------------ > 发件人: "Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves" < > notifications@github.com&gt;; > 发送时间:&nbsp;2021年3月2日(星期二) 下午2:15 > 收件人:&nbsp;"Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves"< > Reinforcement-Learning-for-Control-of-Valves@noreply.github.com&gt;; > 抄送:&nbsp;"喜欢啃苹果"<1152249351@qq.com&gt;;"Author"<author@noreply.github.com&gt;; > > 主题:&nbsp;Re: > [Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves] Request file > (#1) > > > > > > > Hello Jichuan, > > It is not really a constant reference training signal. By moving up and > down randomly, it results in training the agent for following a complex > signal composed of ANY combination of "straight-lines". > > But, yes, you are right that we can try sine-waves as well. This will > generalize it for OTHER types of signals. > > Jichuan - are you a researcher (working or student)? I am glad you are > interested in RL for optimal control. It is an interesting subject. > > I request that you cite my work if you use somewhere please. > > Thank you, Rajesh > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub, or unsubscribe. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <https://github.com/Rajesh-Siraskar/Reinforcement-Learning-for-Control-of-Valves/issues/1#issuecomment-788649013&gt;, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ASKXGU6TQVPDMFNEP3V7OJLTBR76RANCNFSM4YMPRKGQ&gt; > . >

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.