Question AI

[RisingFury]

Alright! time to settle down and wait...
Oh, do you guys want some sort of Kinetic Energy Weapon for OBSP? I'm working on a simple demonstrator right now with SC3...

I'm assuming your orbital launch thing. Is it supposed to be self guided or just launched very accurately?

 

[Grover]

@risingfury

i meant:
how do the autopilots KNOW how to do the plane alignment/base alignment? it takes an MFD code to generate those solutions, which would need to be put into (or "bolted" onto) the autopilot coding, making the autopilot a very chunky piece of code, which older computers would struggle to digest in realtime

 

[MaverickSawyer]

I'm trying to get it to the point where it will run in Orbiter. It's simply to test the idea. I don't intend to try and hit anything in particular yet, so it's going to have to be launched very accurately. However, I am planning on putting translation thrusters on it to allow for fine tuning of the trajectory, and possibly moving fins fore and aft for atmospheric flight. I'm still thrashing out the .cfg file...

 

[RisingFury]

@risingfury

i meant:
how do the autopilots KNOW how to do the plane alignment/base alignment? it takes an MFD code to generate those solutions, which would need to be put into (or "bolted" onto) the autopilot coding, making the autopilot a very chunky piece of code, which older computers would struggle to digest in realtime

Indeed - the autopilots need to calculate everything themselves
The most challenging part of making autopilots isn't defining the behavior, but writing all the math to do the various calculations.

A lot of CPU power in MFDs goes towards drawing the solution to the MFD. Autopilots don't have any of that. Once they calculate what they need, they can go on to execute the maneuver.

The most challenging calculation in OBSP right now is the function that figures out where the bomb launched from a plane is going to land. It's about two orders of magnitude more draining than anything else OBSP has. The problem of the trajectory is such that you can't solve the differential equations that define the trajectory and any simplification you make turns out not to be accurate enough.

A game called Jane's USAF had the same problem - they used a simplified solution, so at high speeds, you had to release the bomb a bit early and at low speeds you had to release it late. A human can compensate with the gut feel, but a computer doesn't have that.

I'm trying to get it to the point where it will run in Orbiter. It's simply to test the idea. I don't intend to try and hit anything in particular yet, so it's going to have to be launched very accurately. However, I am planning on putting translation thrusters on it to allow for fine tuning of the trajectory, and possibly moving fins fore and aft for atmospheric flight. I'm still thrashing out the .cfg file...

You won't hit anything without active guidance. Even the slightest tumbling will spoil your pre-calculated trajectory and cause a miss. I had a similar problem with making the bombing autopilot work.

 

[MaverickSawyer]

Still, it's a start. Eventually, I want it to "explode" upon impact with appropriate energy, but that's not a priority right now.

 

[RisingFury]

Still, it's a start. Eventually, I want it to "explode" upon impact with appropriate energy, but that's not a priority right now.

OBSP already has a system in place. Currently the DumbBomb module used for all the bombs has a constant detonation energy that you set through a .cfg file, but I could program it to take into account its kinetic energy.

The KEW would need an autopilot though, to guide it to its target.

 

[MaverickSawyer]

I know. I'm still trying to get the damned thing to run.

 

[Grover]

and if itsw gonna be going through the atmosphere at high speed, it needs to take into account aerodynamics of equal detail to the orbiter aerodynamics engine

 

[MaverickSawyer]

Yes, which is why I am planning on having both aerdynamic and thruster controls.

 

[jangofett287]

OK, here it is:

Comments: leave them on the Youtube page.
Be sure to check out my other vid, its far more polished!
Any problems, take two and call me in the morning.
Now I've gotta go to bed before I get my parents beathin-- EEPgottagobye!

 

[RisingFury]

Well, you have an old version indeed. It wouldn't happen to be from last year, would it? I can see some really old bugs in there.

One in particular stands out - the following XR2's fly very high into the sky before returning down. There was a bug in the control function that for some reason didn't figure out it's supposed to pitch up or down. The bug is there since one of the first version of the function.

Also, the vessels oscillate left and right. Recently I put in oscillation dampening that prevents that.

@jangofett287: Please post the scenario where you started from, so that we can have the same thing as a reference to compare the new version once I'm ready to do the vid

 

[jangofett287]

Here ya go, one scenario, and one lua script. (the commented out parts are from my without collision and damage test I'm yet to run) Enjoy!

 

[RisingFury]

Ok, the first thing I did was spread the vessels in the scenario wider apart. I won't be able to do a video with XR2's because looking at five of them brings my framerate down to 15 FPS. Even so, the autopilot performed just like normal.

Takeoff:

No complications there. I hit full throttle until I was up, then reduced throttle to half for the remainder of the flight.

Shortly after takeoff:

Squadron gains formations without ans oscillations:

I pitch up gently, squadron follows with no oscillations:

I pitch down sharply, there are some oscillations, but the dampening corrects them within two oscillations:

I roll 90° and pitch up as hard as I can. The autopilots don't copy my controls exactly (I wanted them to handle like humans can, when surprised):

Quickly after I turn, they follow and they start to regain formation:

They regained formation shortly after I took that image.

Very sharp movements can still induce oscillations and out of control flying like in the vid jangofett287 posted, but it's much more rare now. I think I can also improve the oscillation dampening if there's a need.

 

[MaverickSawyer]

:sweet:
Love it. When's the beta release?

 

[RisingFury]

:sweet:
Love it. When's the beta release?

When we feel like it.

We're not gonna rush the release. There's a huge thread on OBSP and a lot of requests have been made. While we won't be able to do everything for the first release, we are committed to listening to what the community wants and are using the thread to pick up some ideas.