Jump to content

Huge problem: passive aliens inside ufo's


Recommended Posts

I'm playing my first "official" playthrough now and raiding Corvettes ATM. While I´m mostly having great time, I feel completely unchallenged inside Corvette (and scouts too) for two main reasons:

  • Aliens are super passive and just hunker down. They won't come at me, but just sit in one position and wait for reaction fire possibility (which they won't get when my high TU, high reaction modifier shotgunners rush them) or just move from one cover location to another doing nothing to kill Xenonauts invading their ufo.
  • Aliens can't open doors if Xenonauts are adjacent to it on the other side. I can take the Corvette entrance hall without any risks as long as I can get my troops to block all the doors. (easy with high TU units). Doors might open from time to time, but LoS won't update (only black shround is visible through door), aliens do nothing and the door closes.

Assaulting ufos should be terrifying experience, but with the currect AI in place, I feel almost completey unthreathened when inside. Aliens should act more agressive when they know Xenonauts are inside. Only officers (and their possible guards later on) should fall back.

This is my experience till Corvettes and I presume it's the same with the latter ufos too. Is anything being done to this anymore or are we left with incompetent UFO defensive AI when the game reaches completition? It would be a huge letdown as raiding ufo's is one of the main things in this game.

Related to this problem: people that have experience modding AI, can something be done to this if Goldhawk can't have time to tune this anymore? If so, I'd be very interested on having a mod that makes aliens defending ufo's more active and agressive.

Edited by Skitso
Link to comment
Share on other sites

I'm actually looking at the problem. Most of the problem can be fixed with the weights in ai props.xml.

Did this behavior pop up in the latest builds, because I haven't seen it mentioned that much before it. Also, if you have any saves, please post them.

Link to comment
Share on other sites

I posted a bug report about this a few days ago:

http://www.goldhawkinteractive.com/forums/showthread.php/10607-v1-05-GC-Passive-Command-Room-Aliens

I've not got any saves, but I've posted a video on that thread and have another one too I can upload if that's helpful.

I can also possibly replicate at least one issue with a good level of frequency, so I might be able to produce a save game if its useful.

EDIT: As a side issue, I've been watching how the aliens behave while the UFO isn't under assault and the aliens inside the UFO tend to wander backwards and forwards a lot - like they do in the video in fact - while their waiting. While it makes sense to do this in terms of moving to new positions throughout the UFO, mostly it just involves them moving from one tile to the next and back. This wastes their TUs and reduces even the chance that they will take reaction shots.

EDIT 2: Foe what it's worth, as well, I'm not sure it has anything to do with the pathing AI, I've been playing with it a lot over the weekend and nothing has stopped this behaviour, at least for command AI aliens which seem to be the major problem. I've also tried lowering minimum shot chance to 0.05 which seems to have no effect. So it implies it's a problem with the scripts rather than the variables.

Edited by kabill
Link to comment
Share on other sites

On the old Corvette layout, with the big command / powercore room in the back at least, I get similar behaviour. Aliens move around the room but seem extremely unwilling to do anything. On the layout with the 3-way doors, aliens do sometimes pop into the entryway corridor.

Link to comment
Share on other sites

I'm actually looking at the problem. Most of the problem can be fixed with the weights in ai props.xml.

Did this behavior pop up in the latest builds, because I haven't seen it mentioned that much before it. Also, if you have any saves, please post them.

Anecdotal evidence but;

On my play-through of 1.02 I am experiencing plenty of aggressive AI inside the UFO.

Examples;

1. Casean is hemmed into pilots room (in corvette). He has two guys to his front, both with shields though they are behind cover. On my turn I open up door behind him and spray him with rifle (miss and don't kill). Close the door and run back a few steps. On the AI turn, instead of shooting at shield guys (and risking reaction fire), he turns round. Opens up door, walks through and kills my rifleman.

2. Lining up behind the three doors in the corvette. Sebillian officer opens up door, launches a grenade right into them and kills two out-right. On the same turn, Sebillian opens up side door and shoots at another and kills him. Interestingly though the Seb from the side ignored a guy in the same room and shot at someone through the side door (but in the corridor).

Link to comment
Share on other sites

It's possibly related to minimum hit chance?

I'm getting inconsistent results testing. Much of the time, it seems more or less fine and when I leave soldiers out in the open, they get shot at as expected. But put my guys behind cover and sometimes (though not always!) the AI can't cope and just wanders around. Other times, however, it decides it wants to advance to attack me. It's strange.

Link to comment
Share on other sites

Trouble is with those is that when I've used them the aliens have tended to stack at the door because they're trying to move towards your soldiers even though they're also being held in place by the command/objective room tiles. So it might work but it might be a bit of a blunt object.

In any case, the fact that the tests produce varying results is perhaps the most problematic thing. As I've outlined in the bug report, *sometimes* the aliens work exactly how they're supposed to. But under similar circumstances on other maps, they tend to hang back and only after a while switch to 'kill mode' and come and attack you.

As such, it implies that the issue is at heart in the scripts running the AI. It could well be simply the random variation in AI moves in play, but if it is it's causing problems because sometimes it results in utterly docile enemies. And, worse, they don't just sit their and ready reaction fire but move around, losing any advantage they might have kept by not attacking in the first place.

Link to comment
Share on other sites

I count five words... ;)

He's suckered you in with that one. Only the randomly capitalised ones apply.

It's possibly related to minimum hit chance?

I'm getting inconsistent results testing. Much of the time, it seems more or less fine and when I leave soldiers out in the open, they get shot at as expected. But put my guys behind cover and sometimes (though not always!) the AI can't cope and just wanders around. Other times, however, it decides it wants to advance to attack me. It's strange.

Without having played around with it, I've been under the impression that the min hit chance is theoretically the problem. Try with it set to 0. Here's why. Lots of cover and props in the interior of a UFO. Even if a unit is not directly behind cover, there's a good chance that there'll be at least one piece of cover in any LOS. Throw in the obscure LOS system - picture one of those scenes where one thinks "well, he can shoot me, but not me him" - and switch it around. By setting it to 0, technically the ET would shoot if a naut was in the LOS, which then might put an end to the shuffling.

However, I suspect the shuffling has also got a lot to do with certain points in the UFO, as the shuffling follows a distinct pattern - it's always the same tiles they head for.

Link to comment
Share on other sites

@Mikhail: I did actually try that (well, I set it to 0.01, anyway) and on that test the aliens fired merrily. But when I set it back to check with the same alien type (Caesans) they were firing fine anyway (for clarity, this was on an entirely new game as I think the aiprops data is saved with ground combat and possibly even the UFO/crash site).

What's especially ironic is that Androns seem to be the worst culprit for this kind of behaviour, and when I was checking their stats they actually seem to have the highest accuracy (the higher ranking ones have 65, while Caesans seem to cap at 55). so if anything, Androns should be the least likely to not fire out of all of the common races, if minimum fire chance is indeed the cause.

I really want it to be the minimum fire chance, of course, because that's a relatively easy fix. But my testing doesn't seem 100% in accordance with this (though this might be my testing at fault!).

As for shuffling: yes, you're right. But the fact that they're shuffling at all is the problem. It seems like the AI is coded to always make use of x% of its TUs when it's not shooting at things, which is a massive waste for aliens inside the UFO. Especially when they're using those TUs simply to wander backwards and forwards across the same two tiles!

@Skitso: I'm really not sure. The AI is definately capable of opening the exterior UFO doors when there's soldiers blocking all of the tiles on the other side, as I've seen that happen a lot recently. I can't actually tell whether this is the AI opening the door, however, or whether it's a bug with the door opening automatically, or whether it's an intended feature to make the door open occasionally (I say this because it seems related to the no-LoS-after-the-door-is-opened issue).

I'm not getting it at all with interior doors, however. I'm not sure why there's a distinction: it might be because the AI will sometimes 'camp' on the inside of the exterior door, and therefore is in a position to open it, while it never does that with the interior doors so it never comes up. It might equally have something to do with 1 vs. multi-tile wide doors. The impression I get from comments and videos I've seen is that the new Corvette layout with the cargo bay and double doors is quite deadly and active, while you're own account of the original layout (which has only single doors) is not, so there's some possible evidence for door size mattering. On the other hand, in my new modded corvette layout (which is actually quite similar to the alternative vanilla one), aliens seem to have a fair attraction to the centre room, so it might be something else entirely. I don't know.

Link to comment
Share on other sites

@Mikhail: I did actually try that (well, I set it to 0.01, anyway) and on that test the aliens fired merrily. But when I set it back to check with the same alien type (Caesans) they were firing fine anyway (for clarity, this was on an entirely new game as I think the aiprops data is saved with ground combat and possibly even the UFO/crash site).

What's especially ironic is that Androns seem to be the worst culprit for this kind of behaviour, and when I was checking their stats they actually seem to have the highest accuracy (the higher ranking ones have 65, while Caesans seem to cap at 55). so if anything, Androns should be the least likely to not fire out of all of the common races, if minimum fire chance is indeed the cause.

I really want it to be the minimum fire chance, of course, because that's a relatively easy fix. But my testing doesn't seem 100% in accordance with this (though this might be my testing at fault!).

As for shuffling: yes, you're right. But the fact that they're shuffling at all is the problem. It seems like the AI is coded to always make use of x% of its TUs when it's not shooting at things, which is a massive waste for aliens inside the UFO. Especially when they're using those TUs simply to wander backwards and forwards across the same two tiles!

A flat 0 would remove doubt, either way. Even at 0.01, there's a chance that with all the cover props inside a UFO, and the funky LOS instances where one side can see the other but the other has not hit %, that a shot will still not qualify to be taken. 0 should remove all that as variables and allow it to be seen if min hit chance is the problem - then the ET's will fire if they have a unit in sight, regardless of LOS and cover.

The shuffling, bless 'em, is best illustrated on your video, where the two sebs, as if on a merry-go-round, head for the same tiles each time as soon as the other is vacated - which suggests the weighting for the preference tiles in the UFO could be over-riding everything else.

Link to comment
Share on other sites

A flat 0 would remove doubt, either way. Even at 0.01, there's a chance that with all the cover props inside a UFO, and the funky LOS instances where one side can see the other but the other has not hit %, that a shot will still not qualify to be taken. 0 should remove all that as variables and allow it to be seen if min hit chance is the problem - then the ET's will fire if they have a unit in sight, regardless of LOS and cover.

You're right. I set it at 0.01 as I didn't want the AI to try and shoot through walls. But it would probably be a better test to eliminate any doubt.

The shuffling, bless 'em, is best illustrated on your video, where the two sebs, as if on a merry-go-round, head for the same tiles each time as soon as the other is vacated - which suggests the weighting for the preference tiles in the UFO could be over-riding everything else.

In theory, though, it shouldn't matter as *all* the UFO tiles have the relevant command/objective room markers so they should all have equal weighting in that regard. I've also seen it happen with a command room alien outside the UFO which exhibited the same behaviour (and what was especially odd in that instance is that my soldiers weren't in any kind of cover at all. Although it was pistol armed and might have been slightly beyond maximum range, which might explain that discrepancy).

(On the subject of the video: what gets me in particular is the bit right at the end, where the seb moves across the room, then back, *then* decides to take a step forwards and actually shoot. It's stuff like that which implies to me that there's something more than just the pathing variables responsible).

Link to comment
Share on other sites

I count five words... ;)

D:

I lose all the coolness

Kabill tgere is something that your not considering, you can try give a gun to a reaper and set all his AI parameters like a ranged combatant..... he will become like a reaper, some part of the aliens behavior is hardcoded, the AI props only that to stablish a priority of weights for desicions.

Still, I prefer a alien that open the door of his car and shot me.

Link to comment
Share on other sites

@TD: I don't disagree with you. This is why I was pointing to the AI scripts as a possible source of the problem. There's definitely something else going on beyond the variables which are accessible in aiprops.

Yep Im only complementing master kabill ^^

Maybe and maybe not... Why its possible that I manage to set a very flexible AI in XNT Into Darkness?

I believe that in the mod Accuracy and TU dynamics its heavily changed, giving more freedom to the AI. I ask multiple times for developers to review the parameters in that mod but they believe that is only fan made work.

Link to comment
Share on other sites

It might equally have something to do with 1 vs. multi-tile wide doors. The impression I get from comments and videos I've seen is that the new Corvette layout with the cargo bay and double doors is quite deadly and active, while you're own account of the original layout (which has only single doors) is not, so there's some possible evidence for door size mattering. On the other hand, in my new modded corvette layout (which is actually quite similar to the alternative vanilla one), aliens seem to have a fair attraction to the centre room, so it might be something else entirely. I don't know.

Hmm, I've seen aliens open the two tiles wide corvette middle interior door many times when my units are not adjacent to it, but never ever when both tiles are occupied. Not sure about the two 1 tile wide side doors though. I'm pretty sure the doors are okay, but aliens just can't open them when the adjacent tile on the other side is occupied. This in addition with passive ai and shuffling between few tiles causes surprising situations to never occur.

Edited by Skitso
Link to comment
Share on other sites

Fair enough; t'was just a thought. It's definitely possible with the exterior doors, though.

(Hmm, random thought: If there's something specific about the external doors, the *only* thing that I can think of which differentiates them from regular doors is that they have the "UFO door" variable checked in their spectre. I'm not confident this will make a difference, but I wonder whether it's worth setting all the interior UFO doors with this variable checked, just in case).

EDIT: I guess the other thing that differentiates them is that there's no UFO objective tiles on the other side of them. Might be worth removing the those tiles from the front-facing sides of the doors as well, in case that changes anything?

Link to comment
Share on other sites

I've had couple of instances where I've stormed in a building with low TU's only to discover an alien there. I've retreated back out and closed the door. Next turn if I open the door again, the alien is always camping on the other side, adjacent tile to the door. I've tested it numerous times and even if I wait for several turns, alien will not open the door but just camp on the other side.

Link to comment
Share on other sites

Ok, two hypotheses:

1) Aliens just can't open doors without actually moving through them. The fact that they can open the external UFO doors when there's troops stacked there is therefore either a bug (though actually a welcome one), or a feature related to that specific kind of door (which brings me back to the point above about what makes them special).

2) The alien in those instances had relatively passive/non-aggressive AI parameters, therefore didn't try to open the door. (There's a fair chance they were passive aliens, since they're the ones most likely to be inside a building.)

Given that it's hard to test the latter of those possibilities (as you can't necessarily tell the difference between a passive AI and an AI that's unable to open doors) it probably makes sense to test the former. There's two ways I can think of doing this:

1) Remove the 'unique' elements of external UFO doors and see if the door-opening still happens; or

2) Apply those features to other doors.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...