Project Universal Autopilots alpha 100324

[Sideshow]

Just did a bit more testing, using the Air-Dg from runway to orbit scenario I got into orbit fine (200km circular orbit). I then cleared all the sequences and tried to use the attitude sequence to hold either a prograde or retrograde attitude.

The retrograde did not work at all, as in the DG oscillated wildly without ever pointing retrograde.

The prograde sequence did not remove all the rotations necessary to maintain a prograde attitude, the nose was briefly pointed heads up prograde but then drifted away. Since the autopilot dissengages as soon as it is prograde it does not correct the drift.

Currently using the Alpha version 100310.

 

[Artlav]

After a manual orbit sync, I tried an autopilot docking to the first shuttle.
...
The autopilot wouldn't engage, but I imagine that is because the ShuttleA does not have an IDS to lock onto.

Not quite, it targets the port, and that should have worked.
Were the two within limits (<50m/s dv, <10Km distance)?

The retrograde did not work at all, as in the DG oscillated wildly without ever pointing retrograde.

Weird, to be fixed.

The prograde sequence did not remove all the rotations necessary to maintain a prograde attitude, the nose was briefly pointed heads up prograde but then drifted away. Since the autopilot dissengages as soon as it is prograde it does not correct the drift.

More or less correct - attitude autopilots are intended to hand the attitude on to the next step in the sequence, so they neither persist nor extremely accurate.

 

[mjanicki]

Not quite, it targets the port, and that should have worked.
Were the two within limits (<50m/s dv, <10Km distance)?
Weird, to be fixed.

Limits? There are limits?

I had the vessels well within 2 km of each other, but I didn't think to check my closing velocity. I was definitely closing too fast for the autopilot. I reran the scenario after manually reducing CVEL and the autopilot worked just fine close in. I then backed up to an earlier quicksave and let the autopilot take over from 9 km out. Again, flawless docking.

While running these scenarios, there were two things I did notice with the docking autopilot, though.

1) A quirk with the UAP blocks that get written to a quicksave scenario. I began the docking autopilot, then stopped the autopilot and did a quicksave. The UAP block in the scenario began wih a ">" as if I hadn't stopped the sequence.

2) CVEL when the vessels mated was 0.25 m/s. I typically aim for 0.01 m/s with manual docking, so it felt a little fast to me. That might just be my overly cautious nature though, so don't worry too much.

Other than that, everything looked great.

-- Mike

---------- Post added 03-14-10 at 12:26 ---------- Previous post was 03-13-10 at 16:53 ----------

It just keeps getting better, Artlav. I managed to setup the UAP to launch a Chapman Probe aboard an Atlas-V 422 from LC-41 at Canaveral. I used a standard short coast to geosynchronous transfer orbit ascent profile as outlined in the Atlas launch system mission planner's guide. I didn't hit all the numbers that Lockheed directs, but it's not a bad launch. I busted allowable max dynamic pressure by about 6 kPa, but the release of the Centaur stage happened within 1.5 km of what the planner's guide says. Not a bad launch overall. The Centaur still had 2.3% of its fuel on board, so there's some room to play with.

Here's the UAP launch I used. It's not overly pretty, but it gets you where you're going.

BEGIN_uap
  $AtlasV:lift_off(engine=main,mode=1,heading=90.000000,altitude=2000,stage=0),
  lift_off(engine=main,mode=2,heading=90,pitch_tgt=35,pitch_duration=110,off_duration=5,stage=0),
  trans_orbit(engine=main,kind=1,apoapsis=35786000,ta=0,mode=0),
  tools(type=0,key=102),
  tools(type=1,wait=10),
  tools(type=0,key=j).
END

There was only one problem. UAP wouldn't allow me to enter the "f" key for a tools keypress. I tried entering the decimal value for the ASCII code instead, as you can see, but it didn't seem to work.

The wait and the keypress "j" still worked, so I imagine it's just a matter of allowing the "f" key. That key is used to release a fairing in some Spacecraft2/3 based vessels like the Atlas V and Centaur. The "j" key worked, which jettisoned the Chapman probe, but since the fairing hadn't been released, it wasn't possible to try to program the next phase of the launch.

Oh, and when I saved the scenario after programming the UAP, a ">" was written at the beginning of the block again, instead of "$".

Overall, I contine to be impressed by this great bit of programming.

-- Mike

 

[Artlav]

Alpha 100314:
orbides.1gb.ru/orbf/uap-alpha-100314.zip

Changes:
-Added approach autopilot
-Added sync orbits autopilot
-Added target to runway_off
-Added no time accel limits button (Shift+A)
-Usability fixes in MFD
-Fixed engines kept on in lift-off AP
-Fixed vessel misnaming in scenario causing CTD
-Fixed CTD on empty sequence input selection
-Fixed and tweaked defaults in some APs

Added autopilots:
Sync orbits autopilot
Performs orbit phasing to put you and your target at the same point in space-time.
Only works if your orbit is below the target orbit
- engine specifies the engine to use
- target specifies the target vessel to align into a plane of
- max_dv_to_use is the maximum delta-v to use in planning the maneuvre
- tgt_distance is the distance to target that is considered as success

Approach autopilot
Moves your vessel withing a given distance from the target
- engine specifies the engine to use
- target specifies the target vessel to align into a plane of
- max_velocity is the maximum velocity in approach
- tgt_distance is the distance to target that is considered as success

Scenarios:
-Mission-Atlantis - Gets Atlantis from the pad to ISS docking. Horrible sight so far.
-Mission-DG from runway to ISS - Launches a DG from KSC runway to ISS docking port.
-Mission-Lunar station ascent - Launches a DG from Brighton Beach pad to Luna-OB1. Collision detection wanted badly.
-Mission-XR2 from runway to ISS - XR2 flies to ISS

Plans:
V 1. Sync orbits autopilot, and thus - a full ground to docking autopilot
2. Land airless from orbit autopilot - full pad-docking-pad mission autopilot
3. Land runway autopilot and re-entry autopilot - full runway-docking-runway autopilot, and beta/gold release.


The interesting feature of this alpha - the first reached milestone - is the ability to do a full mission runway to docking.
That is what to be tested, unusual places, unusual vessels.
Sync orbits autopilot is particularly interested in being tested, being run by something quite arcane.

A quirk with the UAP blocks that get written to a quicksave scenario. I began the docking autopilot, then stopped the autopilot and did a quicksave. The UAP block in the scenario began wih a ">" as if I hadn't stopped the sequence.

Something is weird here, can't trace it yet. To be fixed.

CVEL when the vessels mated was 0.25 m/s. I typically aim for 0.01 m/s with manual docking, so it felt a little fast to me.

A machine is better than human in precise manoeuvres, so sit back and enjoy
Add a velocity limit variable?

There was only one problem. UAP wouldn't allow me to enter the "f" key for a tools keypress.

Fixed the key.

 

[Pablo49]

I tried going from lift off to approach with a module on an Ariane 6. Everything worked perfectly, and very neatly. Though a few things to comment on:

Is it possible to enter a ctrl/shift/alt + key combination in tools? Like Ctrl+D for undocking, etc.
It would be nice if Approach had a time until approach and distance, like Sync_orbits does.
Possibly add simple engine burns, like burn main/retro for n seconds etc. Would be useful for auto-deorbiting tugs or moving away from a station before doing another task etc.

Also, there was a UAP Planner MFD. Is that just something you're working on but isn't functional yet? Or does it do something as of now? Didn't look like it was working.

 

[Sideshow]

Just tried the Mission-from runway to ISS scenario. When I initially start the scenario the autopilot will cut out after a few seconds and the engines will stop. There does not appear to be any pattern to the cutoff since I have a different velocity each time, but consistently don't get to V1.

If I then stop the sequence and then restart, the autopilot functions flawlessly resulting in a docking with the ISS.

This is quite an amazing addon Artlav, thank you for your efforts :thumbup:

Using Alpha 100314.

---edit---
Just tried it again after having dinner and restarting my computer. This time there are no problems. Can only assume it was something odd on my end.

 

[mjanicki]

Using UAP-alpha-100314

I ran all the scenarios I'd been using to test earlier alpha builds, and everything still looked great. One difference I did notice was with the launch autopilot in hold heading mode. Using an Atlas V, the autopilot did a lot of "hunting" causing the vessel to roll back and forth quite a bit. The 100311 build did not do that so I imagine something changed. Once the second part of the sequence kicked in, the "hunting" stopped. The Atlas V was the only vessel I tried which displayed that behavior though, so it might have something to do with just that vessel.

I saw that you allowed for the "f" key in the keypress tool, but I can't enter that key directly in the MFD. If I enter "f" for the keycode in the scenario file, it works but the MFD shows numeric keycode of "33" for "f". I can't figure out why 33 for f.

Everything else in my old scenarios worked as advertised.

Now for the fun part. I created a scenario to test out as many things as I could think of and settled on this:

- Launch ShuttleA 01 from Olympus into a 450 km x 450 km orbit.
- 75 seconds later, launch ShuttleA 02 from Olympus chasing the previous shuttle with target launch mode to a 250 km x 250 km orbit.
- ShuttleA 02 completes ascent and waits for plane alignment.
- ShuttleA 01 completes ascent, turns prograde, and opens its docking port cover.
- ShuttleA 02 aligns plane with ShuttleA 01.
- ShuttleA 02 syncs orbit with ShuttleA 01.
- ShuttleA 02 approaches ShuttleA 01.
- ShuttleA 02 opens its docking port cover.
- ShuttleA 02 docks with ShuttleA 01.

BEGIN_uap
  >SH-01:lift_off(engine=hover,mode=0,target=,heading=70.000000,altitude=200.000000,pitch_tgt=0.000000,pitch_duration=1.000000,off_duration=0.000000,stage=0),
  tools(type=0,key=g),
  lift_off(engine=main,mode=2,heading=70,pitch_tgt=45,pitch_duration=15,off_duration=0,stage=0),
  trans_orbit(engine=main,kind=0,apoapsis=450000,periapsis=450000,mode=0),
  attitude(engine=main,mode=0),
  attitude(engine=main,mode=2),
  tools(type=0,key=k). 
  >SH-02:tools(type=1,wait=75),
  lift_off(engine=hover,mode=0,target=SH-01,altitude=200,stage=0),
  tools(type=0,key=g),
  lift_off(engine=main,mode=2,target=SH-01,pitch_tgt=45,pitch_duration=15,off_duration=0,stage=0),
  trans_orbit(engine=main,kind=0,target=SH-01,apoapsis=250000,periapsis=250000,mode=0),
  attitude(engine=main,mode=0),
  attitude(engine=main,mode=2),
  align(engine=main,target=SH-01,rinc_delta=0.10),
  attitude(engine=main,mode=2),
  sync_orbit(engine=main,target=SH-01,max_dv_to_use=400,tgt_distance=50000,stage=0),
  attitude(engine=main,mode=2),
  approach(engine=main,target=SH-01,max_velocity=100,tgt_distance=5000),
  attitude(engine=main,mode=2),
  tools(type=0,key=k),
  dock(target=SH-01,port=0,with_port=0). 
END

That was the plan, and I have to say after running it 3 times that this is just fantastic.

Now for the details:

- Targeted launch and ascent was much more accurate than in the last build. Average rinc for this scenario was 0.56 (compared with 1.8 in the previous build).

- Align planes was doing odd things on a first attempt. Although rinc was only 0.56, the align planes autopilot failed to improve that much on a first pass. A second pass corrected it perfectly though. I also noticed that after the first failed align plane pass that the vessel was spinning a bit. I did a quicksave and found this "THLEVEL 6:0.000 7:0.000 10:0.000 11:0.000 14:0.018 15:0.018" in the scenario quicksave. Somehow the autopilot was leaving some thrust being applied. At first I thought I might have hit a key, but this behavior showed up on all the runs of this scenario. This made the second pass of align planes difficult, but the autopilot managed. Following the killrot I had sequenced after align planes, those thrust levels were gone and that was the last I saw of vessel spinning.

- Sync orbit worked well. For my first attempts I tried for a 10 km success value and it just wouldn't happen with nonspherical gravity. Bumping that up to 50 km as success got a sync orbit in the first try. It looked like the autopilot would probably get no closer than 30-35 km, but approach and dock will work from 50 km, so that should be good enough.

- Approach worked well. Two things of note with the approach autopilot:
1) The approach autopilot seemed to only use full throttle on the mains, so that once it got close to the desired velocity it would alternate between full throttle and no throttle rapidly. All the other autopilots seem to throttle back the mains once precision is needed, but this one didn't.
2) With a max approach velocity of 100 m/s I found that 5 km was a good approach distance because it takes the autopilot some time to align for the deceleration burn. Anything closer than 5 km at that speed and the vessel would overrun the target before the autopilot could slow the approach.

- Opening docking ports with key=k worked fine and the docking phase completed successfully.

Once I had the scenario set up and programmed the autopilots, I literally just sat back and watched events unfold from chase launch to docking. Very impressive.

-- Mike

Oh, and what's the UAP Flight Planner?

 

[IronRain]

can someone help me?

I just can't get the docking autopilot working.
i'm far within the limits (I think)..

do I have to create a IDS for both vessels that I want to dock with each other?

 

[Artlav]

Is it possible to enter a ctrl/shift/alt + key combination in tools? Like Ctrl+D for undocking, etc.

Could be possible, but unlikely, since none are possible according to the API.
For the undocking - that will be an option in tools AP, besides, system keys like that are not pressable at all.

It would be nice if Approach had a time until approach and distance, like Sync_orbits does.

Not quite possible, since this time is nearly impossible to predict.

Possibly add simple engine burns, like burn main/retro for n seconds etc. Would be useful for auto-deorbiting tugs or moving away from a station before doing another task etc.

Some open orientation+duration burn tool? Could be useful. Deorbits are to be handled by a landing module, but get-away-from-station job is empty.

When I initially start the scenario the autopilot will cut out after a few seconds and the engines will stop. There does not appear to be any pattern to the cutoff since I have a different velocity each time, but consistently don't get to V1.

If I then stop the sequence and then restart, the autopilot functions flawlessly
...
Just tried it again after having dinner and restarting my computer. This time there are no problems.

Have to assume it was something odd on your end, unless steadily reproducible.

If I enter "f" for the keycode in the scenario file, it works but the MFD shows numeric keycode of "33" for "f". I can't figure out why 33 for f.

Hopefully, fixed. Orbiter uses it's own key mapping, so i have to translate them.

Now for the fun part. I created a scenario to test out as many things as I could think of and settled on this:

Nice! I'll put that in the demos, if you don't mind.

- Align planes was doing odd things on a first attempt.
...

The thrust remnants are now fixed.
The sync not happening is an artefact of the way thing work - it tries to "land" the vessel onto the target plane the fastest way, and if the difference is too small, it might not succeed. I didn't do the return-"ascent from underplane" mode, so it waits for the next node.

- Sync orbit worked well. For my first attempts I tried for a 10 km success value and it just wouldn't happen with nonspherical gravity.
...
It looked like the autopilot would probably get no closer than 30-35 km

More or less so, it should approximate along simple elements, adding non-spherical gravity increases the complexity by few orders of magnitude, and it's already the most complex autopilot in the entire UAP with a big lead.

can someone help me?

I just can't get the docking autopilot working.
i'm far within the limits (I think)..

do I have to create a IDS for both vessels that I want to dock with each other?

What exactly fails?
What have you tried?

Vessel should be slower than 50 m/s and closer than 10 km.
No IDSs are needed, "port" defines port on target, "with_port" defines port on vessel.
Is the target port free?
Do demo scenarios work?

Also, there was a UAP Planner MFD. Is that just something you're working on but isn't functional yet? Or does it do something as of now? Didn't look like it was working.

Oh, and what's the UAP Flight Planner?

Greetings.
Where do you want to end up today?
<- 2. Docked to ...
<- 3. Landed at ...
<- 4. In orbit of ...

# 2

What do you want to get docked to?
<- 2. => 

# ISS

Computing trajectories...
Done. 
How much DV to use?
<- 2. L9.1-AS0.4-AD0.3 (9.8 km/s, ~43643 sec)
<- 3. L9.8-AS0.8-AD0.9 (11.5 km/s, ~24870 sec)

# 2

Sequence set.
Press GO to start flying.

Tell it what you want, it will tell you what can be done.
To be functional by release-time, currently is experimental and is an accidental inclusion.

 

[Pablo49]

So, I tried a similar scenario to the one I tested with yesterday (Ariane 6), but with launching South-East. I got this:

That's with setting a target in lift_off and tans_orbit. I tried it again, but this time setting the appropriate heading manually. It lifted off fine, when it got to trans_orbit (Which still had a target set) the rocket immediately turned and the orbit ended up like the one pictured above. It seems the autopilot can't tell ascending from descending.

Then after all of that, with the opposite orbits, I ended up with a relative inclination of 57 degrees (Which is within the limits of the autopilot, correct?) I tried it out to see if it could correct such a far off orbit. It had trouble orienting and after the second burn the RCS got hung up which made the rocket bank out of control. EDIT: Apparently this is already fixed, nevermind.

 

[IronRain]

thanks for the respons!
I tried a demo scenario and I noticed a difrence between the demo settings and my settings.
in the demo, the value of ''with_port''(in the docking autopilot) was 0
my value was 1.

so that went wrong..

thanks again!

 

[mjanicki]

Hopefully, fixed. Orbiter uses it's own key mapping, so i have to translate them.

Artlav, do you ever sleep? You're fixing and adding features faster than we can test them out. Thanks.

Nice! I'll put that in the demos, if you don't mind.

Mind? I'd be honored. I have a full scenario file if you want me to send it your way. It'll save you the trouble of having to position vessels.

The thrust remnants are now fixed.
The sync not happening is an artefact of the way thing work - it tries to "land" the vessel onto the target plane the fastest way, and if the difference is too small, it might not succeed. I didn't do the return-"ascent from underplane" mode, so it waits for the next node.

Sounds good. Again, great job tracking down and fixing these issues.

-- Mike

 

[Artlav]

So, I tried a similar scenario to the one I tested with yesterday (Ariane 6), but with launching South-East.
...
It seems the autopilot can't tell ascending from descending.
...
I tried it out to see if it could correct such a far off orbit. It had trouble orienting and after the second burn the RCS got hung up which made the rocket bank out of control.

Interesting. The heading is to be fixed, the align planes is already fixed.

Artlav, do you ever sleep? You're fixing and adding features faster than we can test them out. Thanks.

Being on the other side of the planet i work while you sleep and vice-versa. It's more fun than work anyway.

I have a full scenario file if you want me to send it your way. It'll save you the trouble of having to position vessels.

Two SH-A's on Olympus? Long since positioned.

 

[Sideshow]

I've been playing around with the tools autopilot, in particular the key press one. It does not appear to accept key presses such as "[" or "]" or any other of orbiters default autopilots.

Also it does not appear to distinguish between the numpad numbers and numbers on the top of the keyboard, interpreting them all as the numbers at the top of the keyboard.

 

[Pablo49]

I was running through sync orbits a few times. Once it dropped my Periapsis down to around 130 -140km. Another time it took my Apoapsis to around 1.5M, which ended up working, but was highly fuel inefficient. I ended up running out of fuel during the following approach. Both of those orbits were only one time, so may have just been random. (This was with a Progress M1, for the record.)

Ctrl+A didn't seem to work during Sync Orbits either, but that's probably a good thing with that pilot.

I also had an idea. Would it be possible to have UAP trigger another craft's autopilot to start? Like a rocket launches, get's into orbit, then jettisons a capsule or another craft, and then the rocket's AP would tell the capsule's AP to start. You could already have the Capsule's AP all loaded up for aligning and rendezvous or whatever.

 

[Artlav]

Also it does not appear to distinguish between the numpad numbers and numbers on the top of the keyboard, interpreting them all as the numbers at the top of the keyboard.

Correct, it is designed to send a key to a vessel (gears, fairing, staging, nose cone), not to Orbiter (undock, fire engine, change view), so it won't press special keys.
Others are being added.

I was running through sync orbits a few times. Once it dropped my Periapsis down to around 130 -140km. Another time it took my Apoapsis to around 1.5M, which ended up working, but was highly fuel inefficient.

It should find an optimal solution distance-wise, within given DV limit, is this condition met? Although, optimization definitely needs tuning and taking DV into consideration - if there is a solution for 30Km and 10 m/s and one for 29 Km and 200 m/s, it will take the second now.

Ctrl+A didn't seem to work during Sync Orbits either

Ctrl+A does something?

I also had an idea. Would it be possible to have UAP trigger another craft's autopilot to start?

Nice idea, added.

 

[Pablo49]

It should find an optimal solution distance-wise, within given DV limit, is this condition met? Although, optimization definitely needs tuning and taking DV into consideration - if there is a solution for 30Km and 10 m/s and one for 29 Km and 200 m/s, it will take the second now.

It did get it within the DV limit, which was at the default amount. I didn't expect it to try to use all of the DV available, since the orbits weren't too far away. I assumed it would balance out DV and distance. But you already know about that, so nevermind.

Ctrl+A does something?

Whoops, typo. I meant Shift+A, the shortcut for removing the cap on time acceleration.

 

[Sideshow]

Correct, it is designed to send a key to a vessel (gears, fairing, staging, nose cone), not to Orbiter (undock, fire engine, change view), so it won't press special keys.
Others are being added.

Understood, thanks.

I have a strange request. Would it be possible to have an autopilot that did the following? In a specified amount of time it would pitch the vessel to a specific angle. In that same amount of time it would throttle two different engine groups linearly a specific percentage. So for example,

Time = 30 seconds
Pitch angle = 80 degrees
Hover thrust initial = 100%
Hover thrust final = 0%
Main thrust initial =0%
Main thrust final = 80%

Why would I want something like this? Well I have been playing around with the UMMU 2 Shuttle A and launching it from earth. The above autopilot would allow the vessel to climb to about 500 meters, pitch up to a suitable launch attitude, while at the same time transition the thrust from the hover engines to the main engines. Performing such a manoeuvre manually is very challenging (as in I’m not quite skilled enough :lol but it would look fairly impressive. I’m sure that such an autopilot may have other applications too.

I realise this may be a rather ambitious request but maybe it is something you would consider.

 

[Artlav]

I have a strange request.
...
I realise this may be a rather ambitious request but maybe it is something you would consider.

Actually, that's about the first thing i done when designing UAP, and the code is still around.
Added.

 

[Sideshow]

Actually, that's about the first thing i done when designing UAP, and the code is still around.
Added.

That's great news, thanks!