Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[stormbringer951]

Is it possible to make the AI more aware of burst weapons it currently has firing and weight attempting to hold them on target a bit more heavily? AI ships tend to waste the damage of phase lances, needlers, etc, by firing them when they are at high relative velocity or turning relative to their target, so half the burst misses or (very noticeable with phase lances) sweeps across the target rather than holding it against the same point on the armour of the enemy ship to punch through armour in the case of anti-armour weapons.

Easiest to reproduce this with those weapons mounted on small fast ships (Tempest and Medusa come to mind), especially when there are multiple enemies around and the ship will likely start juggling targets.

I understand that interfering with ships' maneuvering logic might result in more suboptimal behaviour generally, especially on larger ships, but thought I'd flag this issue for attention because it's a bit annoying and very noticeable on smaller maneuverable ships.

[Thaago]

I've seen this happen many times as well.

[SafariJohn]

Weapons with charge-up times could also use the help.

[Morrokain]

I've brought this up in the past. Here is a quote from Alex regarding it:

Unfortunately, this isn't something I can easily fix. The reason I was asking about the TL is if you were seeing something that the TL didn't suffer from, then that might indicate a potential avenue for a fix. But generally speaking, the AI isn't aware enough to keep facing at the same target while firing a burst, that is, positional tactical considerations drive its facing, and there's no mechanism to weigh "but I'm also firing something" into this. One could easily see a situation where the AI sticks to a target for too long and dies due to chain-firing several burst-type weapons, for example, so it's potentially fraught.

So the issue here as I understand it is that fixing this issue potentially leads to other undesirable AI behavior. Mainly along the lines of stubbornly staying on target even if situational awareness considerations (like a surround) would indicate that this is a very bad idea.

So it might be a judgement call that the AI cannot reasonably make.

I 100% agree that it's painful to see though. Subjectively, I'm not sure which situation I would dislike more. It's one thing when an AI ship gets surrounded because I gave it an eliminate order and it's another that it defaults to it because of its loadout.

Hmm, I wonder if maybe there could be improvements to "sticking to target" burst fire AI specifically under an aggressive officer or eliminate command?? That at least wouldn't feel bad if the ship were lost as a consequence - or rather it would feel bad but not be the AI's fault if that makes sense. They player has some amount of control over the overall behavior.

Idk just thinking as I type. Not really a fully fleshed out idea or anything.

[Alex]

One thing that's semi-related but may help overall: there's also a case right now where turreted burst weapons will miss a lot of the burst when first approaching the target, due to firing too early - before they're quite aimed at it. That should be much improved in the next release.

I also think that some other AI improvements (namely, deciding which enemy to pivot around at a longer range than before) result in less target switching, so that may help indirectly. For example, I just ran a quick simulation of a Tempest with 2 Phase Lances fighting off 2x Hound 2x Cerberus 2x Lasher, and I think I saw a lance slide off-target once over the whole thing, and that was for the very tail end of its shot. Running a test with a Medusa with 2x Light Needler, yeah, it definitely happens more, but again, it wasn't dominant.

[SafariJohn]

Besides AI stuff, a big contributing factor to burst and charge-up weapons missing is the rotation slowdown while firing. Reducing or removing that would change a lot of weapon balance, but at least it could feasibly be done.

[intrinsic_parity]

I feel like some of this is related to the behavior where the AI tends to turn to the side unnecessarily a lot even in a 1v1 (I think trying to distribute armor damage or something). I've definitely seen that behavior when testing loadouts. I vaguely remember hearing that that was being addressed at some point.

[Wyvern]

Yeah, the rotation slow-down while firing is bad for burst (or charge-up) weapon turrets - wouldn't help with hardpoints, but some better is better than none better.

Might be worth considering adding that as a secondary effect to the Advanced Turret Gyros hullmod? Have it reduce the turn-rate-while-firing penalty in addition to just increasing the turn rate?

[Alex]

It's a rotation bonus while not firing, not a penalty while firing, and weapons are explicitly balanced around the "value while firing". That's that value in the weapon_data spreadsheet, etc.

But: they also get a significant rotation bonus for *keeping a firing weapon facing in the same direction*, which is more what you're talking about. So e.g. if a turreted HIL is firing and the ship turns under it faster than the HIL can turn to keep on target, the HIL will get a turn rate bonus for staying on target, it's just not enough to counter a very high turn rate. But that doesn't matter at all forward-facing hardpoint weapons.

I vaguely remember hearing that that was being addressed at some point.

Yeah, somewhere between 90-100% of that - when it happens without a reason, that is - should be taken care of.

[SafariJohn]

But: they also get a significant rotation bonus for *keeping a firing weapon facing in the same direction*, which is more what you're talking about. So e.g. if a turreted HIL is firing and the ship turns under it faster than the HIL can turn to keep on target, the HIL will get a turn rate bonus for staying on target

Maybe Auxiliary Thrusters and whatever skill buffs ship rotation should also buff that hidden bonus?

[Morrokain]

Honestly what if the bonus was just doubled or even tripled without the need for a hullmod? Would that help the problem/not look weird visually? Especially if target switching likely won't be as much of a concern, ship rotation still keeping burst weapons on target would be the last piece of the issue right?

[Wyvern]

Eh, at this point my attitude is "wait and see what it's like in the next version" - we can revisit this topic if it's still a notable problem then.

Or we can suggest specific simulator battles where this is reliably an issue and see what Alex thinks of them; that would work too.

(I mostly don't have this problem... but that's largely due to avoiding setting AI-controlled ships up in ways where it would be a problem.)

[Kriby]

I see this occur with a ship that can comfortably keep its enemy locked in its forward arc, using swivelling turrets:

Equip a Medusa with phase lances and tactical lasers in the small energy slots, give it an integrated targeting unit (and advanced turret gyros by all means). Pit this medusa against the leftmost hammerhead and leftmost condor in the simulation.

When the Medusa engages the hammerhead, the condor will zone out the Medusa, but send missiles intermittently. The hammerhead largely cannot bring its weapons to bear so it will be zoned out by the Medusa. When there are nearby incoming missiles, the phase lances will spin off target right away; when not, the phase lances will stay on target. Note that the missiles are typically out of range when this occurs, when missiles are in range the phase lances will shoot them down if able.

[Alex]

Hmm - none of the weapons involved would target missiles, and I'm not seeing the behavior described in this situation, so I'm guessing something's not quite right.

[Kriby]

Hmm - none of the weapons involved would target missiles, and I'm not seeing the behavior described in this situation, so I'm guessing something's not quite right.

On revisiting the test I can see how it doesn't always give a ton of valid situations for what I wanted to show. Here's a video that I recorded showing what I mean. Timestamps 1:40 showcases it really well, 4:00 again but less egregiously.

Edit: I realize I didn't crank the CR to 100% in the video, but redoing the test with 100% CR still shows the same behavior. The video is recorded in unmodded 9.1a.