Oolite Bulletins

Rorschachhamster wrote:

I have observed (while waiting for my train, thanks for TAF in trunk not sooo long... ), that an escort flew into a station (Kiota Solar Station, I think) after having "paniked" because it's mother flew away and it had tried to get around the station twice or so but always being stopped by the "arms".
Propably the AI set a waypoint when there was no arm because of rotation... and then the escort rammed it, when it couldn't avoid collision fast enough on the third try.

Was that under 1.76.x or 1.77?

The old (1.76.x) version of the OXP could only simulate the launch from multiple docks by taking some of the launching ships and manually moving/re-orienting them to the other side. Hence if an escort launched and ended up on the other side from its mother then that could cause some of the collision problems.

The 1.77 version has true multiple docks for launching (except the solar station, to save launching ships at the sun) so it should perhaps make things a little better. That said it may be worthwhile to consider having the trunk code try and launch mothers and escorts always from the same station dock?

Thargoid wrote:

Rorschachhamster wrote:

I have observed (while waiting for my train, thanks for TAF in trunk not sooo long... ), that an escort flew into a station (Kiota Solar Station, I think) after having "paniked" because it's mother flew away and it had tried to get around the station twice or so but always being stopped by the "arms".
Propably the AI set a waypoint when there was no arm because of rotation... and then the escort rammed it, when it couldn't avoid collision fast enough on the third try.

Was that under 1.76.x or 1.77?

The old (1.76.x) version of the OXP could only simulate the launch from multiple docks by taking some of the launching ships and manually moving/re-orienting them to the other side. Hence if an escort launched and ended up on the other side from its mother then that could cause some of the collision problems.

The 1.77 version has true multiple docks for launching (except the solar station, to save launching ships at the sun) so it should perhaps make things a little better. That said it may be worthwhile to consider having the trunk code try and launch mothers and escorts always from the same station dock?

It was 1.77.5645.
And I'm pretty sure they launched from the same side anyway (this being a solar station), just the two others got around or through, I didn't notice it then.

Eric Walch wrote:

I found the probable reason for the 'docking problem' for the tories. The docking bay is off-centre. When I examine the position, I get:

Code: Select all

PS.target.subEntities[0].position
(1, 0, 500)

And the current code sends a HOLD_POSITION when the dock of a rotating station is 0.1 meter off-centre. I think the the code should also allow one or two meters tolerance.

EDIT:
I centred the dock position in shipdata to 0,0,500 for both Tori types and from than on docking went smooth. I even added a bunch of extra traders for docking and after some time there were no waiting ships.

The slight off-centre definition was probably for cosmetic reasons, but even from close distance, I could see no artefacts with the fully centred position.

Ha! I just came by to post this picture:

There were docking proplems, too...

EDIT: Of course, this doesn't mean it's of center on x and on y , but certainly not centered on z.

Eric Walch wrote:

I centred the dock position in shipdata to 0,0,500 for both Tori types and from than on docking went smooth.

Thanks, Eric - that seems to have worked fine here at Xevera torus.

Hm, I looked at the SIRF shipdata.plist, and it has a centered dock (at 0,0,0... )
Maybe this is another problem, with the bounding box being so big that the waypoints are swallowed... but at other times it worked with this station...

Thargoid wrote:

So the problem there is not that the dock is off-centre, it's that the station main model is. Hence what you are seeing there, as the bounding box is based on the maximum distance of the model edge from the origin, which in this case is the back end of it (the opposite end from the dock).

Hence why the bounding box extends so far forward in front of the model (you could fit a second mirror-image model in there).

One of my mantras: I hate non-z-centered models. And it would be so easy to center everything before releasing it.

El Viejo wrote:

Eric Walch wrote:

I centred the dock position in shipdata to 0,0,500 for both Tori types and from than on docking went smooth.

Thanks, Eric - that seems to have worked fine here at Xevera torus.

On comparing with older code, this must be a core bug that slipped in with the new dock code. In 1.76 the misalignment of the dock was only taken in consideration for carriers (non-rotating stations). Although fixing the tori model works, it should be fixed in the code.

The question is: Is this related to the docking problem?
Because here the docking didn't work even though the dock was centered.

Rorschachhamster wrote:

Hm, I looked at the SIRF shipdata.plist, and it has a centered dock (at 0,0,0... )
Maybe this is another problem, with the bounding box being so big that the waypoints are swallowed... but at other times it worked with this station...

This certainly is an other problem than ships being put on hold for ever with the Tori stations.

In 1.74 and older, the waypoints for approach were calculated relative to the dock. That could lead to problems with stations having entrance tunnels. Since 1.75 the dock is extrapolated to the outside of the bounding box and the approach is calculated to that point. From that point on the ship goes in a straight line to the dock.
I have not examined if this has changed in 1.77, but I think it should be still happening this way.

Eric Walch wrote:

In 1.74 and older, the waypoints for approach were calculated relative to the dock. That could lead to problems with stations having entrance tunnels. Since 1.75 the dock is extrapolated to the outside of the bounding box and the approach is calculated to that point. From that point on the ship goes in a straight line to the dock.
I have not examined if this has changed in 1.77, but I think it should be still happening this way.

Should be the same as before. I tested launching and docking with the Salvage Gang from Anarchies which has a long entrance tunnel and it worked fine.

hmmm. It looked for me (You can see the police dots in the screenshot above), that they just stopped outside the inflated bounding box. I even tried to nudge one Orion ( ) away via shoving, but it returned to the exact same spot, once I flew away.

I have been lurking next to a Kiota Solar station for an hour. I have not seen any cargo floating around.

I did see one oddity and that was the station launching some shuttles. In the past all rotating stations launched shuttles, but since 1.75? only stations in an aegis launch shuttles. However the sun has also an aegis. That shuttle now has a long way to the nearest planet...

I also witnessed an Anaconda launching with 6 escorts and flying to the planet. For some reason one of the escorts ended up behind the station. The code was now constantly switching between: Formation_Form_Up and Avoid_Collision. Or visible: The escort was flying to and from the back of the station. The Anaconda already had left my scanner for a long time before the escort managed to break the circle and flew to mom. At least the collision avoidance seemed to work nicely. I only missed why it ended up behind the station. Probably it was trying to avoid a collision with one of the other escorts and fled in the wrong direction.

Eric Walch wrote:

I have been lurking next to a Kiota Solar station for an hour. I have not seen any cargo floating around.

If I remember correctly, collision detection is better when the player is in the vicinity. Or am I wrong with that?

If this is still the case, you'd better lurk somewhere in deep space, and just "watch" the vicinity of the Kiota via the console.

Commander McLane wrote:

If I remember correctly, collision detection is better when the player is in the vicinity. Or am I wrong with that?

I almost forgot. Yes, when the player is not nearby, no NPC will be able to fly through the spikes, but every ship will collide with the outer bounding box. That indeed might make a difference.

When closely watching, ships behave correctly. With the stations back facing the sun, launching sunskimmers correctly followed the way points around the station. Escorts do not follow waypoints, but head in direct line to the mother, so the last launching did find the station in between them and mother. There the collision pre-alert kicks in and sends the ship away from the station. But indeed, with the player on distance, this might trigger to late when just the bounding box is considered. That is an interesting phenomena of floating debris, that leaves the player with a mystery he can never solve by close examination.