Lafayette out here producing banger after banger - also this is so niche but I have a whole rant about how there is no true "optimal control" approach like LQR, since making a system like this LTI removes all the hard stuff like the nonlinearities you mention, so you're tuning for an entirely fake system. Anyway this stuff is so impressive!
With optimal materials, a decent infrared camera and 4 axis gimbal system as well as object tracking, lead and trajectory computers and better optimised guidance fins, this man could turn a project into a project that would get him in trouble (please don't do this they arent afraid to come by on a Thursday morning)
BPSspace: I am scared of ITAR so I´m gonna only show half of how the rocket works Lafayette Systems: Lmao here is how the code in matlab works, have fun. Man i love armchair engineers who know what they are doing
A while ago, i wanted to learn how all of these things work behind the scenes, in greater detail for the thesis i chose back in university, it was like a final project. Nobody ever said a useful thing from everywhere i could find online. I had to search seemingly unrelated stuff and slowly glean tidbits of physics from nasa website and things like this. The guy you mention at first would not help in the slightest, only saying "Buy my thing for 200$". Right... a dirt eater poor like myself would become a US citizen and buy his trash, sure. I've managed to learn and discover everything by myself, reinventing the wheel... didn't even know about these simulink subsystems... did everything by hand from scratch, slowly and painfully. I watch here, for a dose of catharsis, seeing how there's still decent people that offer knowledge for free.
Having such deep domain expertise on so many engineering disciplines is WILD! I am beyond impressed, but simultaneously my head hurts just thinking about the mountain of hours that rocket has devoured. On the plus side, your resume can now just be a picture of that bad boy with a link to your channel.
As a mechanical engineering student that has mainly been focused on mechanical design so far, this project blows me away with what you’ve been able to achieve here. Projects like these fascinate me because they encompass all the things I’m currently bad at as a student but want to get better at in the future. I’m starting systems and controls this semester but my fluid dynamics and programming skills are very rusty and I have little experience with electronics. So, I would be interested if you had any advice for people such as myself on how to tackle projects similar to yours.
The best advice I have is to tackle learning new things one at a time. There’s lot of free resources out there that can help with aerodynamics and controls. They are kind of dark magic even to the people with PhDs but learning a practical amount is surprisingly accessible!
This is a really kickass intro to not just control theory but approaching problems from a systems level. If I was a freshman engineering student, a video on scoping out avionics/hardware and one for creating a nice top-down cad model for manufacturing would have saved me 2 years of figuring out how to actually engineer something. Amazing work!
Man I feel bad watching your videos, at the same time feel really inspired. I can't believe how much a better engineer you are. Please do more deep dive into all the sections you have done before
OpenRocket is much more accurate than you would think. The problem is you have to put in ALL the parameters and have it reflect your design. Most people using OpenRocket do not even know just how many parameters there are. Did you know you can input the approximate surface roughness of the paint on your rocket? You need specialized measurement tools to get that value but if you can, OpenRocket can be really good. Where it struggles are atmospheric modeling which is pretty simplistic and doesn’t allow for measured high altitude winds and it struggles with compressible flow dynamics. But if you’re flying a subsonic rocket to less than 10,000ft you can get within less than a percent error almost every time.
Hey great video! If you are looking for topics, there is one black hole in all this that I have never filled. That is the physics and math of the aero forces and moments and how they get modeled in all this. A body and a fin will have lift and drag, and a moment. Where does one get those parameters? And then, how are they combined to model the flight dynamics in the software such that a controller can do something to control the rocket intelligently? I think that is worth at least one video if not two or three. First video might be how to find, get calculate derive the aero coefficients of the rocket parts. Next video could be how to take those aero coefficients and combine them to get net rocket linear and rotational forces. And a third video could be how to take those forces and integrate them into final net rocket accelerations, velocities, positions and resultant path. Thanks for considering!
Just found you. Ive built a few Estes and just bought some wanting to progress from Beginner to Level 5. Lol your like level 9000! Now ive got open rocket and open motor not sure what to do with it yet, but clear i need it. Love your videos keep it up!!
This channel is an absolute gem dude. Keep of the great work, these videos are such good quality for a newer channel; I'd imagine if you keep up with this level of quality and consistency you'll gain quite the following in no time. I appreciate the look at things I'm interested in tinkering with... Unfortunately, a certain four-letter agency has made it incredibly difficult to find any information on such things. I'm more interested in the simulation and control theory side of it all anyway; but ITAR will be damned if I can find information on any of it...Llet alone get a comprehensive look at the implementation and experience of someone else's attempts. Love the content man.
I think the hardest part for me has been figuring out how those control inputs then affect the output, so please make a video about that! I've tried running CFD at various fin deflection angles but I didn't find it to match my experimental flights, and I'm curious how you do it!
Have you ever thought about using a learning based solution? I'm working on quadcopters and we have created a learned dynamics model, where we augment a physics based dynamics model with machine learning, obviously this requires some real world flight data, but it can be super sample efficient with the right approach. With our own project we have an infinitely better dynamics model than we ever could create by traditional means, and we strongly believe this is the only feasible way to simulate the aerodynamic effects to this extend. You could also use a MLP trained with something like PPO in sim for the actual flight controller, instead of relying on more traditional control systems.
That's very interesting! In what ways is the model better? In the sense that it takes into account tiny things that traditional rigid-body modeling + some aero forces doesn't?
@@dashs2597 you can model the rigid body stuff really well, the problem comes down to the aerodynamics, for example with a quadcopter you have 4 propellers creating all sorts of turbulence around the body, and the worst part is the airflow generated by each propeller interacts with the other 3 as well, it's impossible to simulate this with CFD. This is actually the trick, to use a well put together rigid body simulation and augment it by estimating the external forces acting on the body given the current state. I'm not that familiar with rockets, but obviously simulating the aerodynamics is a huge issue there as well, although significantly less compared to quadcopters. Not sure as to whether it is worthwhile using a learned dynamics model, because it's complicated, computationally slower than using approximation for aerodynamics, and it requires diverse flight data. 🤷♂
@@boldizsarszabo I think this learned approach is generally done as well, but in a more manual fashion for rocketry projects. It's usually the case that a autopilot tuning is done after a flight test through engineering intuition and experience. I do think that the reinforcement learning benefits you'll see for rockets are probably going to be less than quadcopters, just simply because there is less to learn. Generally subsonic to supersonic rocket aerodynamics is pretty well understood and captured by empirical and data based aerodynamic tables (see DATCOM), and the flow around said vehicle is generally pretty linear since static airframe body contact to air dominates. The nonlinearities are usually a result of the smaller fins that actuate, but if the flow is subsonic and the fins are placed rearward (meaning fins are downstream of a majority of the rocket body), you can get away with linear assumptions for aero. Of course this breaks down a bit for supersonic flow, but still, at low angles of attack you can get away with a lot of linear assumptions. Compare that to very large non-linearities in the flow for a quadcopter, due to the large rotating propellers that heavily determine the flow structure around a quadcopter body.
This is really cool! Love your explanations about the software stack and variables used in your rocket I wonder if, given enough data of your launches, a neural network can be trained to make the rocket steer towards a moving point in space That would be equally interesting to see
I have a question more for the previous video, but i will ask here since its the latest one. You got some amazing results in waypoint mission, but, i wonder if you verified your Inertial positioning system against GPS or some other reference frame to make sure that the rocket really "knows where it is"? ;) Impressed by your work
This is a superb project and presentation is top as well! Thank you! btw: have you ever experimented with rollerons like they have on sidewinders to stabilise flight? They would autocorrect trajectory during flight, might mess up the landing phase though. (:
Lafayette Systems: correct calculation and accurate simulation are very important. Mistakes are not allowed. Me: 😮 🚀 he has a rocket and it does whizz-whizz
Nick...question....Do you think that your approach to model rocketry design... aka OR / RockSim + Matlab + ground simulation + electronics... will ever take off in the model rocketry arena? In other words, will the average rocketeer (that wants more than just the "how high" and "how fast" to seem like an expert) ever catch on to what you're doing, as I'm sure that you have followers that have already...aka...me and the others.
For the average rocketeer, the last two phases probably won’t be very common. They really aren’t super necessary without a control system unless you’re doing them for fun/experience. Some of the commercial flight computers have some ground test options available, but without a control system they don’t need to be and aren’t very sophisticated.
You not going to show it to the viewers but you want their views to earn? 😊 quite cool quick manipulation. If you don’t want to show then don’t talk about it at all this tell viewers that although you know the best way out but intentionally want them to be ignorance of the truth. Will you feed your child stone as bread when you know you have bread
![Chlöe - Shake (Feat. Jeremih) [Official Visualizer]](http://i.ytimg.com/vi/aDs_1ufpfv8/mqdefault.jpg)
Lafayette out here producing banger after banger - also this is so niche but I have a whole rant about how there is no true "optimal control" approach like LQR, since making a system like this LTI removes all the hard stuff like the nonlinearities you mention, so you're tuning for an entirely fake system. Anyway this stuff is so impressive!
Non linear model predictive control enters the chat
@BPSspace
Lol, I thought this was you looking at the notif. 😅
Wow
thank you for the tutorial, they'll never see it coming
- « This guidance test system seems accurate enough for my purposes »
- Mmh what purposes ??
they will never see it coming
@@alert2 """"they"""" o.o
With optimal materials, a decent infrared camera and 4 axis gimbal system as well as object tracking, lead and trajectory computers and better optimised guidance fins, this man could turn a project into a project that would get him in trouble (please don't do this they arent afraid to come by on a Thursday morning)
You mean thermal camera?
BPSspace: I am scared of ITAR so I´m gonna only show half of how the rocket works
Lafayette Systems: Lmao here is how the code in matlab works, have fun.
Man i love armchair engineers who know what they are doing
A while ago, i wanted to learn how all of these things work behind the scenes, in greater detail for the thesis i chose back in university, it was like a final project.
Nobody ever said a useful thing from everywhere i could find online. I had to search seemingly unrelated stuff and slowly glean tidbits of physics from nasa website and things like this.
The guy you mention at first would not help in the slightest, only saying "Buy my thing for 200$". Right... a dirt eater poor like myself would become a US citizen and buy his trash, sure.
I've managed to learn and discover everything by myself, reinventing the wheel... didn't even know about these simulink subsystems... did everything by hand from scratch, slowly and painfully.
I watch here, for a dose of catharsis, seeing how there's still decent people that offer knowledge for free.
same here. my thesis is about 6 DOF two stage rocket
Lafayette chose to not share any code or CAD in any video yet, as far as I know
It's a simplified model. They did nothing bold.
@@vikramgogoi3621 +1
Having such deep domain expertise on so many engineering disciplines is WILD! I am beyond impressed, but simultaneously my head hurts just thinking about the mountain of hours that rocket has devoured. On the plus side, your resume can now just be a picture of that bad boy with a link to your channel.
As a mechanical engineering student that has mainly been focused on mechanical design so far, this project blows me away with what you’ve been able to achieve here. Projects like these fascinate me because they encompass all the things I’m currently bad at as a student but want to get better at in the future. I’m starting systems and controls this semester but my fluid dynamics and programming skills are very rusty and I have little experience with electronics. So, I would be interested if you had any advice for people such as myself on how to tackle projects similar to yours.
The best advice I have is to tackle learning new things one at a time. There’s lot of free resources out there that can help with aerodynamics and controls. They are kind of dark magic even to the people with PhDs but learning a practical amount is surprisingly accessible!
@@LafayetteSystems Thanks for responding!
This is a really kickass intro to not just control theory but approaching problems from a systems level. If I was a freshman engineering student, a video on scoping out avionics/hardware and one for creating a nice top-down cad model for manufacturing would have saved me 2 years of figuring out how to actually engineer something. Amazing work!
Man I feel bad watching your videos, at the same time feel really inspired. I can't believe how much a better engineer you are. Please do more deep dive into all the sections you have done before
OpenRocket is much more accurate than you would think. The problem is you have to put in ALL the parameters and have it reflect your design. Most people using OpenRocket do not even know just how many parameters there are. Did you know you can input the approximate surface roughness of the paint on your rocket? You need specialized measurement tools to get that value but if you can, OpenRocket can be really good. Where it struggles are atmospheric modeling which is pretty simplistic and doesn’t allow for measured high altitude winds and it struggles with compressible flow dynamics. But if you’re flying a subsonic rocket to less than 10,000ft you can get within less than a percent error almost every time.
Hey great video! If you are looking for topics, there is one black hole in all this that I have never filled. That is the physics and math of the aero forces and moments and how they get modeled in all this. A body and a fin will have lift and drag, and a moment. Where does one get those parameters? And then, how are they combined to model the flight dynamics in the software such that a controller can do something to control the rocket intelligently? I think that is worth at least one video if not two or three. First video might be how to find, get calculate derive the aero coefficients of the rocket parts. Next video could be how to take those aero coefficients and combine them to get net rocket linear and rotational forces. And a third video could be how to take those forces and integrate them into final net rocket accelerations, velocities, positions and resultant path. Thanks for considering!
Just found you. Ive built a few Estes and just bought some wanting to progress from Beginner to Level 5. Lol your like level 9000! Now ive got open rocket and open motor not sure what to do with it yet, but clear i need it. Love your videos keep it up!!
I just love this man!!!
Really informative video! Thanks for sharing.
Let's go new video I'm so excited to watch this thank you you make good videos
This channel is an absolute gem dude. Keep of the great work, these videos are such good quality for a newer channel; I'd imagine if you keep up with this level of quality and consistency you'll gain quite the following in no time. I appreciate the look at things I'm interested in tinkering with... Unfortunately, a certain four-letter agency has made it incredibly difficult to find any information on such things. I'm more interested in the simulation and control theory side of it all anyway; but ITAR will be damned if I can find information on any of it...Llet alone get a comprehensive look at the implementation and experience of someone else's attempts. Love the content man.
I think the hardest part for me has been figuring out how those control inputs then affect the output, so please make a video about that! I've tried running CFD at various fin deflection angles but I didn't find it to match my experimental flights, and I'm curious how you do it!
7:01 "turned out to be useful for" or "turned out to have the potential to be useful for"?
Thank God i found your video man, i just got into Rocketry because i had a stupid idea😭🙏🏾
cant wait for the next video. please make more detail videos, specially about how you use Simulink.
we got senku ishigami irl before gta 6 🔥🔥
Have you ever thought about using a learning based solution? I'm working on quadcopters and we have created a learned dynamics model, where we augment a physics based dynamics model with machine learning, obviously this requires some real world flight data, but it can be super sample efficient with the right approach. With our own project we have an infinitely better dynamics model than we ever could create by traditional means, and we strongly believe this is the only feasible way to simulate the aerodynamic effects to this extend.
You could also use a MLP trained with something like PPO in sim for the actual flight controller, instead of relying on more traditional control systems.
That's very interesting! In what ways is the model better? In the sense that it takes into account tiny things that traditional rigid-body modeling + some aero forces doesn't?
@@dashs2597 you can model the rigid body stuff really well, the problem comes down to the aerodynamics, for example with a quadcopter you have 4 propellers creating all sorts of turbulence around the body, and the worst part is the airflow generated by each propeller interacts with the other 3 as well, it's impossible to simulate this with CFD. This is actually the trick, to use a well put together rigid body simulation and augment it by estimating the external forces acting on the body given the current state.
I'm not that familiar with rockets, but obviously simulating the aerodynamics is a huge issue there as well, although significantly less compared to quadcopters. Not sure as to whether it is worthwhile using a learned dynamics model, because it's complicated, computationally slower than using approximation for aerodynamics, and it requires diverse flight data. 🤷♂
@@boldizsarszabo
I think this learned approach is generally done as well, but in a more manual fashion for rocketry projects. It's usually the case that a autopilot tuning is done after a flight test through engineering intuition and experience.
I do think that the reinforcement learning benefits you'll see for rockets are probably going to be less than quadcopters, just simply because there is less to learn. Generally subsonic to supersonic rocket aerodynamics is pretty well understood and captured by empirical and data based aerodynamic tables (see DATCOM), and the flow around said vehicle is generally pretty linear since static airframe body contact to air dominates.
The nonlinearities are usually a result of the smaller fins that actuate, but if the flow is subsonic and the fins are placed rearward (meaning fins are downstream of a majority of the rocket body), you can get away with linear assumptions for aero. Of course this breaks down a bit for supersonic flow, but still, at low angles of attack you can get away with a lot of linear assumptions.
Compare that to very large non-linearities in the flow for a quadcopter, due to the large rotating propellers that heavily determine the flow structure around a quadcopter body.
You are preparing us for our WW3 air defense system. Amazing job and thank you for your fantastic explanation!
Great video. I was hoping for proper HITL test with software running on the ground with real time simulation data
This is really cool! Love your explanations about the software stack and variables used in your rocket
I wonder if, given enough data of your launches, a neural network can be trained to make the rocket steer towards a moving point in space
That would be equally interesting to see
Woohoo!
Amazing video as always!
I have a question more for the previous video, but i will ask here since its the latest one. You got some amazing results in waypoint mission, but, i wonder if you verified your Inertial positioning system against GPS or some other reference frame to make sure that the rocket really "knows where it is"? ;)
Impressed by your work
This is a superb project and presentation is top as well! Thank you!
btw: have you ever experimented with rollerons like they have on sidewinders to stabilise flight? They would autocorrect trajectory during flight, might mess up the landing phase though. (:
Great content. thanks man
Why dont you use Simulink's code generator to convert your model into c++?
Watching this before interviewing for a GNC engineer internship later this week
Good luck with your interview!
@@LafayetteSystems Thanks!
Great!🚀
Very systematic. I love it.
Very inspiring, my student rocketry team plans to attempt something similar this year, would love to connect to ask for some advice
Lafayette Systems: correct calculation and accurate simulation are very important. Mistakes are not allowed.
Me: 😮 🚀 he has a rocket and it does whizz-whizz
Super cool keep up the good work
I like to think of a shuttlecock when thinking about COM/COP
Nick...question....Do you think that your approach to model rocketry design... aka OR / RockSim + Matlab + ground simulation + electronics... will ever take off in the model rocketry arena? In other words, will the average rocketeer (that wants more than just the "how high" and "how fast" to seem like an expert) ever catch on to what you're doing, as I'm sure that you have followers that have already...aka...me and the others.
For the average rocketeer, the last two phases probably won’t be very common. They really aren’t super necessary without a control system unless you’re doing them for fun/experience. Some of the commercial flight computers have some ground test options available, but without a control system they don’t need to be and aren’t very sophisticated.
16:49 "hatchet" aka switchblade 😅
WOW NICE ROCKET 🚀 AND WERE IS THE CAT
Sehr gut
getting sponsored by aura
+10000 aura
i love it , thank you !
Dudeeeeee great video
Rollerons! We need rollerons!
Insane sponsor segway
So... how much of a payload can that mi... model rocket with wings and sea... camera carry? Asking for a friend.
This dude keeps forgetting to set his application videos for RTX and LM as non listed videos...
I argue it is not unprofessional
Hey man awesome new video. Thank you for answering my email i know it must have been uncomortable to answer. Youre the man!
mükemmel
I want to make rockets but sadly it is banned in my country 😢
yolo
same
Hi
Hi! How r u making money? I mean whats the business model here?
lol this is just a hobby!
@@LafayetteSystems is there any chance that this could be turned into profitable business?
It’s possible but I’m totally ok just doing it as a hobby for now
@@LafayetteSystems hmm ok
Bro made a Patriot missile in his backyard
Bro is making a cruise missile 😂
FBI open up
IS this guy a fed? I feel like hes trying to get certain people excited in this so they can get arrested preemptively.
2:10 What the F indonesian sim Card😅
You not going to show it to the viewers but you want their views to earn? 😊 quite cool quick manipulation. If you don’t want to show then don’t talk about it at all this tell viewers that although you know the best way out but intentionally want them to be ignorance of the truth. Will you feed your child stone as bread when you know you have bread