The turrets can only be controled rather inacurate. My sugestion would be to have a tree view with the sub-targets can be controlled by target type, and each subtarget has a option for "line of sight" and priority. And also those setting should be saved to memory 1-10, that can be recalled by pressing alt+number while piloting. For example NPC->CV->Turrets->line of sight & priority 5 NPC->drone->priority 6 NPC->BA->Turrets->line of sight & priority 6 NPC->BA->Generator->priority 5 Player->BA->Turrets->priority 3 the subtargets should include : for CV/SV/HV thruster, guns and turrets with option line of sight on/off. for all Generator, Hyperdrive, shield dont need this option as they are rarely in the open.
Adding more layers of complexity to turrets targeting will only make the problem worse. In fact it's ridiculous that a turret could be able to distinguish between factions, predators and preys, and even specific devices. I would greatly simplify all this and make them search for heat, motion, and let the player deal with consequences of relying on turrets to decide for them. Then maybe turrets would have a much easier time firing reliably. I think it could help to have the player assign targets manually instead of leaving that to the turrets. Have a reasonably clear and comprehensive radar UI interface in which the player can assign targets for turrets, and avoid clumsy turret AI decision making.
That actually is a good argument, still in most space games turrets are able to at least decide friend/foe. I know it is some kind of game-magic there but with some kind of iff transmitter it might work. And a pilot needs the same information. i.e. in eve you know red=bad right away, so you just click the closest reds and select guns. in the kind of engagements in empyrion with dozens of drones would be bad for manual control - this would just not work out. The firing at no line of sight could be removed (fixed) all together, even making fights more interesting as you can no longer just kill the generator (i normally do that and salvage the guns) right now is looks like each turret targets as it likes without coordinating with the others, and targeting stuff they can not hit. maybe my idea would help, maybe not. But it needs a major overhaul annyways. I think having the same sub target for different target classes is stupid; sometimes i want to destroy generator for CV, but turrets for BA. lastly if a ai pilot in a game from today can control a ship and distinguish (sub)targets it only needs good sensors (and iff) to do it irl and the future...
Because the way "faction discovery" is set up even that is not "realistic" in regards of the game lore. How could turrets "know" that any "undiscovered faction" is "friend" or "foe" ? Past that point I see no problem to have the radar/ UI indicate "red" for enemies, but that's already a step ahead. When players get their first "incoming drones" alert, the "magic computer" already knows these are enemies, so they can be just red and turrets shouldn't even require player intervention to decide to shoot them down. The problem is that even with many targeting options, the turrets just get confused. This points heavily towards making the task simpler for them. Drones are easy to shoot, they have no "sub-components", and yet even then the turrets hesitate. "Line of sight" seems highly slow to process and unreliable, especially in multiplayer with "lag shot" glitches that indicate that parts of the solids displayed are not even detected by the physics engine when checking for projectile collisions and bullets pass through hulls. Add to this that some AI vessels are especially designed to "fool" turrets with dummy devices, and this cheesy tactic doesn't make a lot of sense either, because it bluntly breaks the 4th wall between the game world and the player. Just like the enemy soldiers running in a straight line towards the player shooting at them...
Well... yes and no. Now, I *do* agree that the current target listing is something that at least our own real world tech couldn't do and likely something impossible even for any reasonable future tech (e.g. faction or "predator vs. prey"). That said, even our own current AI tech can classify targets actually quite well. Well, for a reasonable choice of "good". Thus, I think we can have *some* classifications of targets even if we assume the "raw data" is based on heat, motion, and other real physical world signals. So yeah, ditch that silly "prey vs. predator" and only apply the faction choice for things that actually broadcast an IFF signal (ships and bases, mostly). But then instead of "raw choices" have classifications that are assumed to be an AI evaluation of that raw data: - Threats : In game terms this means simply turrets, guns, and gun-wielding individuals. A reasonable AI can be assumed to spot these on ships and bases. Spotting gun-wielding persons is also quite realistic, although if individuals are assumed to not have an IFF then perhaps only start targeting them *after* they have been spotted to fire at you or at your grids. The AI probably could also spot "non-gun attacks" *after* they happen. That is, a dino charging at you would be classified as threat but if it just stood there, then not. Regardless of whether it's in the current game a "prey" or "predator". In game-mechanical terms this would probably be just a simple check of the state of the NPC creature: is it attacking? - Thrusters: Easy enough from heat signature and visuals so targetable. - Generators: Again, should give off heat and radiation so legal target. - Core: If we assume it broadcasts some kind of signal, whether intentionally or as a consequence of how it works, then it's a target choice. If not, then not and you must spot it yourself. We probably should declare that it does. At least the attacking enemy ground troops seem to have no trouble locating yours. And, I think, these would actually be enough as targeting choices. Of course, being able to assign priorities for them would be nice... But the above would be enough for game purposes. Usually when you attack a POI or a larger enemy ship you only want it *disabled*. That is, destroy its weapons so it's no longer a threat and knock off its thrusters or generators so it can't escape. And then f*****g stop shooting! You're destroying *valuable* loot you dumb piece of rust covered machine brain!!
Yeah, current AI can distinguish between different breeds of cats just by the picture, it can even distinguish humans race, gender, age and possible genetic pathologies of individuals just by analyzing surveillance images so, future targeting AIs could probably distinguish different types of friends or foes If it is viable in empyrion to add countless targeting options, that is another subject altogether....
I think the current targeting options are not half-bad, its just that the turret-ai does get confused (attack that drone that shoots me, or attack that generator behind 3 blocks of combat steel, or one of the enemy turrets that shoot me versus that turret at the backside of the enemy) because it does not have enough rules to make the right desicion. so a clear priority for targets and if "behind armor" should be targeted would help those turrets be less confused. for the predator/prey thing, it could just be identifying the type of animal visually and know which of those are predators. the same for infantry: uniforms can be used to distinguish friend/foe and we need profiles to switch between. also some pre-made ones for those that don't want to tinker with it.
The current combat meta vs BA/CV/small ships being tied with "loot" is secondary to surviving the encounter, and as such if weapons are made to hit from much greater distances with much more damage, then the "loot" thing and pinpointing of specific devices becomes a problem. If the objective is to just "disable" the ship then by all means we need EMP weapons just for that purpose! Then it's not a big deal if turrets accidentally shoot a friendly because it would only disable and do no damage. Well current AI covers a very broad spectrum of things here. For example, we can't say actual surveilance satellites can detect ****** in targets, or that being able to distinguish cat species means also being able to follow any cat fleeing from a garden hose attack with the same "AI - driven device". If we look back at this recent UFO thread I started not long ago, knowing that some UFOs did disable remotely and wirelessly many ICBM inside their bunkers via re-routing programmed targets or simply taking them "offline", it shows everything is possible. But in Empyrion there is a disturbing discrepancy between the average level of technological sophistication between various devices of the same apparent "level", like the multi-tool being light years ahead of the whole propulsion or energy systems, or the crappy jetpack vs the Almighty Healing Chamber... So of course we can plead the AI could be anywhere between these extremes. That's more the actual problem. Even with the limited set of "checks" turrets now have to go through for each new target, they still waste time and get lost. Adding new checks and anything related to having some hierarchy in there is due to make things worse. If we were able to pinpoint at what time the turret targeting/ reaction time started degrading notably, we would probably see it was when some apparently unrelated feature was added that forced to make the turret AI jump through long chains of dependencies to get a simple "go".
Hey now, I'm no pirate! I just happened to start my salvaging a bit early and the guys on the ship were late in their schedule to abandon it. These things happen...
