|
Think you are smart at math? (pg. 2)
|
View this Thread in Original format
| Krypton |
| quote: | Originally posted by Pokit
You mean ax^2 + bx + c? :toothless |
With a twist because he wants to plot on a 3-D plane...kind of just made it up lol:p |
|
|
| Joss Weatherby |
Haha, well its actually fairly more complicated, but it can be assumed to be a flat plane.
I am trying to script a simulation of counter-battery artillery radar for a game.
I already have the tracking and data gathering part. Its just getting it all plotted out.
Granted I could fake it and just watch for when the shells first enter the game world, but this seemed more fun. |
|
|
| Rose |
This thread makes my head hurt.
Yeah, that's all I was going to post in here. Go on. |
|
|
| winston |
| I believe you are focused on CB intelligence? If so, are you dealing with triangulation and on what levels? I sense it must be different since this Is a platform to model CB artillery radar functions. |
|
|
| tachobg |
so you're trying to determine the approximate landing position (and possibly the whole trajectory) from some samples of the objects position? here's how i see it: the trajectory is subject to real world physics, and physical parameters. what you measure are samples of position that depend on those parameters. what would be nice to do is to get back the parameters that generated the given data. Then, because these parameters determine the physics of the scenario, you can use them to determine the entire trajectory.
In this case, the parameters could be something like {initial position, initial velocity, some constant of proportionality for the air resistance force}. This comes out to 7 parameters. What you can do in rough terms is to calculate the parameters that most likely generated the data. If your data is noisy, you probably want to use some probabilistic inference, and calculate the probability of params given data measured, and take the params that maximize this probability. If everything is deterministic, you should be able to set up some equations and in principle get an exact solution for the parameters. This might need some numerical solver though. Good luck and have fun :) |
|
|
| ziptnf |
| quote: | Originally posted by winston
and this...
http://jbrwww.che.wisc.edu/tech-rep...mcc-2001-01.pdf
I could be wrong, but you're dealing with 'disturbances in velocity/disturbances in acceleration', The link above would give you an idea of a similar algorithm. I never said I was right or smart. |
This is an excellent example. It's very specific to positioning in free space. It gives you a state vector, input vector, output vector, and 3 matricies of positions, assuming matrices A B and C represent the x,y,z directions. |
|
|
| winston |
| quote: | Originally posted by tachobg
so you're trying to determine the approximate landing position (and possibly the whole trajectory) from some samples of the objects position? here's how i see it: the trajectory is subject to real world physics, and physical parameters. what you measure are samples of position that depend on those parameters. what would be nice to do is to get back the parameters that generated the given data. Then, because these parameters determine the physics of the scenario, you can use them to determine the entire trajectory.
In this case, the parameters could be something like {initial position, initial velocity, some constant of proportionality for the air resistance force}. This comes out to 7 parameters. What you can do in rough terms is to calculate the parameters that most likely generated the data. If your data is noisy, you probably want to use some probabilistic inference, and calculate the probability of params given data measured, and take the params that maximize this probability. If everything is deterministic, you should be able to set up some equations and in principle get an exact solution for the parameters. This might need some numerical solver though. Good luck and have fun :) |
this is my wingman, right here... |
|
|
| Joss Weatherby |
Thanks tachobg.
I am going to have to look at how it models air resistance in the game. From what I remember it is fairly simple.
I can also probably get initial velocity by "guessing" the round type and the probable launcher and from that using what its assumed muzzle velocity is.
As far as the sample positions go, its actually samples inside a known cube that has positive and negative coordinates and I am tracking its actualy x,y,z position in it, not a relative position to the radar. That should make it a bit more simple.
I am thinking in reality all I need to do is determine its muzzle velocity or initial velocity and its angle and then apply those to it to determine its range.
This site actually has a bunch of algorithms and examples that seem to be helpful.
http://hyperphysics.phy-astr.gsu.edu/Hbase/traj.html
I also am wondering if I can see the code they use to project aiming points for artillery, because their aiming mechanism shows the approximate fall lines for varying destination altitudes and also the flight time for those fall lines. |
|
|
| Omega_M |
| you might wana look at kalman filter and how you can apply it to your problem. Kalman filter is used to predict present / future position of an object in the presence of noise. You will find a lot of literature, and perhaps code for implementation. I don't know if this will solve your problem, or how easy it is. It probably is not. |
|
|
| Joss Weatherby |
I looked at some of their code and its totally beyond me.
They have lots of code to generate ballistics tables and they have some pre-generated ones for use with the artillery pieces in the game already...
s confusing. :p |
|
|
| Acton |
| I'm unsure about programming and stuff, but ALL that is required for calculations like that are the simple equations of motion with a drag coefficient associated with them.....and that's it. It's not rocket science..... well, not all of it anyway ;). |
|
|
| ziptnf |
| quote: | Originally posted by ********
|
Holy cow, when did you get unbanned? |
|
|
|
|
