My experience with OBSP autopilots always run me in the same direction: If you can calculate a variable exactly instead of relying on control algorithms... DO SO!

If you take killrot for example... you should be able to calculate with a very high precision how long each thruster needs to fire at full power to bring the angular velocity down to 0. With things like rigid body physics, this works VERY well.

You still have to get rid of the jerkiness at close to 0 though.


This kind of system works well even when you can only come up with a rough estimate for the problem - where it's impossible to calculate because you simply don't know some variables or because math won't allow it.

For my atmospheric autopilots, I calculate when a plane should return the aileron from full deflection to no deflection so that when the aileron gets there exactly at the moment when the plane hits the target roll and no angular velocity.

This doesn't work exactly. Just too many variables that I can't predict. Some because of limitations of Orbiter, some because I treat them as constants when they're not,...

If you look closely, you'll see that the air control surfaces shake, but the shaking is small enough so that the plane itself doesn't shake.



When you're working on something as specific as your autopilot, you can go ahead and choose a particular control algorithm and tweak it. But the moment you apply it to a different craft, it'll misbehave, so setting desired angular velocity and angular acceleration can come back and bite you in the ass...


For practice, you can try writing a "fast horizon level" autopilot. The idea would be to fire the RCS at full power and let the angular velocity increase. Calculate when the RCS should be reversed in order for the lander to have 0 angular velocity just as it reaches horizon level...

I can generate them from altitude curves, but it would take time. Nevertheless, I'll do research and see what I can find regarding actual data.

I don't have them but I could do them as meshes perhaps.
The only problem is having tons of ongoing projects at the same time :-)

Anyway, for nice terrain check this Apollo 11 landing game: http://www.planetinaction.com/moonlander/#
That's about 10cm terrain precision or less..

Looks cool. If you need textures or panels (2D / 3D) I have some than might be used. Can also help with real terrain ;-) for the Apollo landing sites.
A LK lander version would be great.

Late last year. I've dabbled with 2d games (look for my big hit 'Move the red square around the green window'), and a little bit with the Panda3D engine in python, but for the most part I've stuck to CLI programs.

Looks nice....just a question, will it have accuracte collision detection?

Wow, impressive. You have made huge progress, congratulations!

Nice!

Oh nice job man, you're getting wonderful at this stuff. I can see a bright future ahead of you. Good job, George.

Sam

Your welcome george7378.

Same here, the progress you've made compared to me is just stunning.

Wow!

Nice stuff!

If you need a beta-tester just PM.

Could you post a link to this tutorial? I've been looking for something like this, but I've been completely unable to find any good one.