Hi Charles, Congrats for this 1st video: great job!! I’m a rocket beginner and I have a long way to go to get there... You may know my nephew Tom Hauser who has joined as well Estaca this year. Take care.
Great work, Charles! Did you think about Model Predictive Control to go further and improve your design? It could be an elegant way to deal with the varying thrust.
Thank you! I didn't look into MPC before and as I'm watching Steve Brunton's video on MPC, I think it could work and give good results. The only problem that I see is that on very short term system, like TVC at this scale, you care a lot about the initial reaction which is in a sense, kind of "guesstimated" in MPC, although it seems really good for longer term problem, like nuclear plants (I'm not sure tho). Also I don't really know how complex the structure would get as you add states (like theta, theta dot, position, position dot all in 3 axes). MPC would probably be interesting to manage more constraints though. So I think it could work, and maybe be easier than FSF, but I feel like you loose a bit of control in that case. I might be wrong though, I'll looking into it. It seems to be a good strategy for certain situations.
@@CCSTechno MPC is a lot used today for ADAS and AD systems in automotive industry and so, I'm quite confident it could be suitable for this kind of application but to be honnest, rocket science is not my preliminary knowledge area! :) Nevermind, there is an interesting video series on our website (fr.mathworks.com/videos/tech-talks/controls.html) about MPC. And don't hesitate to contact me if you need some help about Mathworks tools (www.linkedin.com/in/romainlachaux/ - I'm also a former ESTACA student ;) )
Woaw! Glad to talk with a former ESTACA student! (We can switch to French btw, idk, I don't mind!). Well, nevermind on what I said above. After talking with some people, it can work without problems for this sort of application. Maybe for pure orientation, PIDs or even FSF/LQR are easier to understand for a beginner in state control like me. But MPC is another interesting challenge, and I'll keep digging that rabbit hole with Matlab videos and other resources. From talking with a GNC engineer, it doesn't seem like MPC per say is used on bigger (orbital) rockets, just because of the additional constraints that you have. But for a project like mine, you might end up at least at the same place with MPC if not with something a bit more robust. Thank you for pointing that out! :)
Hello Charles, Very good video! Is there a chance to program in an established set of input changes from previous thrust curves? Or "chop" the averages into a pre-determined number of milli-seconds? Maybe, I don't know what I am talking about. I am nowhere near even starting with my attempt at TVC. So, if this question is dumb, please forgive me and ignore it. Just trying to get up to speed with you smart and ambitious guys. :-/
Haha, thank you very much! So on the surface, it might seem running a set of input to the servos could work to stabilize a system like a rocket. The problem is that it'd involve a *perfect* simulation of the environment (perfect inertia of the rocket, the atmosphere, the wind, the misalignment of the thrust vector control mount, where is the center of gravity of the rocket, etc. And the problem is that there is literally an infinite amount of variable that have small uncertainties, and actually for some of them, significant uncertainties. And all of them will add up overtime until you completely loose control of your system. So it would actually be a non-robust way of controlling it, and the idea behind control engineering is that you have the most robust (or at least robust enough) way of controlling your system (like a rocket). Does that explanation make sense? Btw, there is no problem at all in asking any questions! We all started at the bottom of the ladder, so I hope that it makes sense!
This was a clear explanation, thank you. One question that I have is why we are figuring out the center of mass experimentally if we can see it in the 3D model itself? (of course if we model every component and add the appropriate mass/material parameter) Wouldn't that be a more precise approach?
So, of course you can do that, but there are things like servos, servo cables, parachutes, the motor, etc, that are not homogenous (the density isn't the same everywhere in the part). So you'd have to enter the mass for each parts of the model assuming that it is homogenous, but that would already be an approximation, because the CG of a servo is not exactly at the center of the volume for example. Therefore, in my case I had the option to find the CG experimentally and I knew the uncertainty would be about the same if I did it with the 3D model. But if you're working in an environment that requires you to do it with the 3D model (because you can't balance a real rocket by hanging it with a cord), or you need tighter tolerance and you have really accurate model of your parts, then sure, go for it! (I hope it makes sense)
No problem! Let's be clear, I have close to 0 authority on any Control Theory stuff. BUT, I am planning to do a livestream early in January so that we can all, collectively try to get some Full State Control working in Matlab/Simulink. (if you're interested in that and want any update on that btw, you can follow me on Twitter @Charles_Rockets so that you know the date when I announce it)
Very great work and video! Is it possible to get the numbers of your input and output in the system identification? I'm having big trouble trying to make the transfer function blok for my servos.
@@kresoostric2688 Get simulink look at my and his simulink model, try to make a replica that works or become a patreon of mine for 5 dollar and get the simulink model for free
Hi! Unfortunately no, this is a pretty old model that I’d rather not make publicly available. That being said, I show most of the model in the video so you can piece it together and you can ask me if you have any question on a specific problem.
Bonjour ,je suis actuellement au lycée et j'aimerais savoir si l'estaca est une bonne école car elle m'intéresse beaucoup pour l'ingénierie spatiale. Sinon très bon boulot
Bonjour, Et bien personnellement je suis très satisfait de l'ESTACA. Elle a l'avantage d'avoir une bonne réputation dans le milieu aéronautique/aérospatiale (du moins de ce que j'ai pu voir). Si tu as un très bon niveau en math/physique, la voie prépa MPSI + concours pour accéder à Supaéro (qui n'a pas de prépa intégrée) ou l'ESTACA est aussi une possibilité (par contre, il faut vraiment être bon et être prêt à charbonner). Je pense que si tu aimes plus les applications que les cours théoriques, tu t'épanouiras plus en essayant de rentrer directement en prépa intégrée dans une école comme l'ESTACA.
I love rockets and now I am trying to build one i don't have a 3d printer i want to know how to make an airframe, can I use PVC as a rocket body please give me a replay and how to make and fly one plzzz
It's not gonna be easy to not have a 3d printer, especially for building the parts for the TVC mount, but it's definitely doable. You can probably use plywood or other materials to make them by hand. As for the airframe, I used regular, 75mm diameter cardboard tubes.
For Flamey, it was mainly just manually changing the gain in the simulation until I got the response I wanted it. For Epsilon, my current project, I used LQR.
To make it yourself? I wont encourage anyone to get into that unless you have a strong background. Buying them from an Estes reseller is pretty much the only solution.
Maybe. I'm not sure though because it involves intellectual property and I spent an awful amount of time on it. Maybe I should make it available for $3 or $5. I don't know if it'd be worth it. Why? Are you interested in it and do you need a specific part?
@@CCSTechno I would like to get this, because my rocket keep having problems. Even when I try simulating it in simulink(I'm new to that). This means I just spend a lot of money and time testing with motors... so if you will do this available I would be very happy:-) ＄3 or ＄5 is worth it, especially when I know it works ;-)
@@CCSTechno cool! maybe you could walk tough and/or share the code, so that we can see how you calibrate the software with the simulation? Do you use any filter for the derivative like the simulink block uses?
Still can’t believe that Flamey flown so quickly! Awesome work as always !!
This is the best Simulink Flight Sim video I know of! Sorry Joey B. ;) Very well done!
Well done broo, very niiice project , the scientific approach is fully explained !! Awesome
Hi Charles,
Congrats for this 1st video: great job!!
I’m a rocket beginner and I have a long way to go to get there...
You may know my nephew Tom
Hauser who has joined as well Estaca this year.
Take care.
Really? That's so cool! Is he in Paris or Laval? (bon on peut passer en français aussi ;) c'est comme vous voulez!)
@@CCSTechno
Il est à Toulouse.
This is awesome! Nice job with the video production!
Very well explained! This is really impressive!
This was so helpfull!! Thanks awesome work
This video is a gem 😍
Looks great!
Great work, Charles! Did you think about Model Predictive Control to go further and improve your design? It could be an elegant way to deal with the varying thrust.
Thank you! I didn't look into MPC before and as I'm watching Steve Brunton's video on MPC, I think it could work and give good results. The only problem that I see is that on very short term system, like TVC at this scale, you care a lot about the initial reaction which is in a sense, kind of "guesstimated" in MPC, although it seems really good for longer term problem, like nuclear plants (I'm not sure tho). Also I don't really know how complex the structure would get as you add states (like theta, theta dot, position, position dot all in 3 axes). MPC would probably be interesting to manage more constraints though. So I think it could work, and maybe be easier than FSF, but I feel like you loose a bit of control in that case. I might be wrong though, I'll looking into it. It seems to be a good strategy for certain situations.
@@CCSTechno MPC is a lot used today for ADAS and AD systems in automotive industry and so, I'm quite confident it could be suitable for this kind of application but to be honnest, rocket science is not my preliminary knowledge area! :) Nevermind, there is an interesting video series on our website (fr.mathworks.com/videos/tech-talks/controls.html) about MPC. And don't hesitate to contact me if you need some help about Mathworks tools (www.linkedin.com/in/romainlachaux/ - I'm also a former ESTACA student ;) )
Woaw! Glad to talk with a former ESTACA student! (We can switch to French btw, idk, I don't mind!). Well, nevermind on what I said above. After talking with some people, it can work without problems for this sort of application. Maybe for pure orientation, PIDs or even FSF/LQR are easier to understand for a beginner in state control like me. But MPC is another interesting challenge, and I'll keep digging that rabbit hole with Matlab videos and other resources. From talking with a GNC engineer, it doesn't seem like MPC per say is used on bigger (orbital) rockets, just because of the additional constraints that you have. But for a project like mine, you might end up at least at the same place with MPC if not with something a bit more robust. Thank you for pointing that out! :)
Great video 👍👍🔥
Hello Charles,
Very good video!
Is there a chance to program in an established set of input changes from previous thrust curves? Or "chop" the averages into a pre-determined number of milli-seconds?
Maybe, I don't know what I am talking about. I am nowhere near even starting with my attempt at TVC. So, if this question is dumb, please forgive me and ignore it. Just trying to get up to speed with you smart and ambitious guys. :-/
Haha, thank you very much!
So on the surface, it might seem running a set of input to the servos could work to stabilize a system like a rocket. The problem is that it'd involve a *perfect* simulation of the environment (perfect inertia of the rocket, the atmosphere, the wind, the misalignment of the thrust vector control mount, where is the center of gravity of the rocket, etc. And the problem is that there is literally an infinite amount of variable that have small uncertainties, and actually for some of them, significant uncertainties. And all of them will add up overtime until you completely loose control of your system. So it would actually be a non-robust way of controlling it, and the idea behind control engineering is that you have the most robust (or at least robust enough) way of controlling your system (like a rocket). Does that explanation make sense?
Btw, there is no problem at all in asking any questions! We all started at the bottom of the ladder, so I hope that it makes sense!
Hello, could you make the simulink model available? Is there any way to obtain it?
This was a clear explanation, thank you. One question that I have is why we are figuring out the center of mass experimentally if we can see it in the 3D model itself? (of course if we model every component and add the appropriate mass/material parameter) Wouldn't that be a more precise approach?
So, of course you can do that, but there are things like servos, servo cables, parachutes, the motor, etc, that are not homogenous (the density isn't the same everywhere in the part). So you'd have to enter the mass for each parts of the model assuming that it is homogenous, but that would already be an approximation, because the CG of a servo is not exactly at the center of the volume for example. Therefore, in my case I had the option to find the CG experimentally and I knew the uncertainty would be about the same if I did it with the 3D model. But if you're working in an environment that requires you to do it with the 3D model (because you can't balance a real rocket by hanging it with a cord), or you need tighter tolerance and you have really accurate model of your parts, then sure, go for it! (I hope it makes sense)
@@CCSTechno Yes, this makes sense, thanks. Will you also explain the full state control in the future? That would be awesome :D
No problem! Let's be clear, I have close to 0 authority on any Control Theory stuff. BUT, I am planning to do a livestream early in January so that we can all, collectively try to get some Full State Control working in Matlab/Simulink. (if you're interested in that and want any update on that btw, you can follow me on Twitter @Charles_Rockets so that you know the date when I announce it)
could u share the simulink file in the description or a photo of the entire block? please
Very great work and video! Is it possible to get the numbers of your input and output in the system identification? I'm having big trouble trying to make the transfer function blok for my servos.
Any luck finding input and output numbers in the system identification. Thanks !!
@@kresoostric2688 look at some of my simulink videos
But hey man, you don't even have to implement that! You only need to model Inertia and dead time!
@@hanbergsspacemission4184 Thank you. And how do I do that...?
@@kresoostric2688 Get simulink look at my and his simulink model, try to make a replica that works or become a patreon of mine for 5 dollar and get the simulink model for free
I am sorry, but how could i get that model? is that any other option to get it?
Hi! Unfortunately no, this is a pretty old model that I’d rather not make publicly available. That being said, I show most of the model in the video so you can piece it together and you can ask me if you have any question on a specific problem.
Bonjour ,je suis actuellement au lycée et j'aimerais savoir si l'estaca est une bonne école car elle m'intéresse beaucoup pour l'ingénierie spatiale. Sinon très bon boulot
Bonjour,
Et bien personnellement je suis très satisfait de l'ESTACA. Elle a l'avantage d'avoir une bonne réputation dans le milieu aéronautique/aérospatiale (du moins de ce que j'ai pu voir). Si tu as un très bon niveau en math/physique, la voie prépa MPSI + concours pour accéder à Supaéro (qui n'a pas de prépa intégrée) ou l'ESTACA est aussi une possibilité (par contre, il faut vraiment être bon et être prêt à charbonner). Je pense que si tu aimes plus les applications que les cours théoriques, tu t'épanouiras plus en essayant de rentrer directement en prépa intégrée dans une école comme l'ESTACA.
@@CCSTechno merci pour cette réponse très complète, je vais donc charbonner pour rentrer à l'estaca
Hello Charles are you going to make that model available for download?
I love rockets and now I am trying to build one i don't have a 3d printer i want to know how to make an airframe, can I use PVC as a rocket body please give me a replay and how to make and fly one plzzz
It's not gonna be easy to not have a 3d printer, especially for building the parts for the TVC mount, but it's definitely doable. You can probably use plywood or other materials to make them by hand. As for the airframe, I used regular, 75mm diameter cardboard tubes.
@@CCSTechno Thank you so much if I have any doubt about my rocket journey I'll ask for help
How to tune the gates in this model?
For Flamey, it was mainly just manually changing the gain in the simulation until I got the response I wanted it. For Epsilon, my current project, I used LQR.
what type of propulsion did you use for your rocket model?
An Estes F15 motor (so same as a mini SRB).
If I want to do by myselft a propellant, what materials do I have to buy?
To make it yourself? I wont encourage anyone to get into that unless you have a strong background. Buying them from an Estes reseller is pretty much the only solution.
Норм:)
Hello Charles are you going to make that model available for download?
Maybe. I'm not sure though because it involves intellectual property and I spent an awful amount of time on it. Maybe I should make it available for $3 or $5. I don't know if it'd be worth it. Why? Are you interested in it and do you need a specific part?
@@CCSTechno I would like to get this, because my rocket keep having problems. Even when I try simulating it in simulink(I'm new to that). This means I just spend a lot of money and time testing with motors... so if you will do this available I would be very happy:-) ＄3 or ＄5 is worth it, especially when I know it works ;-)
I’m also interested 🙋♂️🙋♂️
Hey! Sorry for getting back to you so lately. I'll have some time at the end of next week to think about it and see how we can work this out.
@@CCSTechno cool! maybe you could walk tough and/or share the code, so that we can see how you calibrate the software with the simulation? Do you use any filter for the derivative like the simulink block uses?