|
Post by subunit on Apr 30, 2017 10:25:49 GMT
I think my biggest problem with CoaDE is the "shooting range" feel. It's cool to take a toy out onto the range and test it against some targets (hey, they even shoot back), but I've always had a sneaking suspicion that I'm not really playing an opposed wargame on the maneuvering layer. I decided to finally try to do some Levels.txt editing and got stopped dead right away when I tried to get an AI fleet to attack a player fleet in a nearby SOI. Here's my Levels.txt insert:
Speaker Test Speaker Alliance Iroquois Resurgence
StarterFleet LibertyOne ShipTypes Long Assault Cutter Edit
StarterFleet NipponStrike ShipTypes Gunship Large Methane Tanker Large Methane Tanker
Level Liberty or Death JulianYear 2259.1 Description Test description ShortDescription Test description TimeLimit_JulianDays 60 Location MainBody Uranus PlayerAlliance Iroquois Resurgence EnemyAlliance Nippon Prime AI Reckless IsADefendLevel true Fleets Puck 0 Iroquois Resurgence null LibertyOne Mab 0 Nippon Prime null NipponStrike Music The Edge Messages Starting Test Speaker Hey
This much works fine, just insert the level name into the Campaign list and away we go. As you can see I set this up with the "IsADefendLevel" switch to true, in case that was necessary for the AI to go aggro. The gunship has a couple of tankers, more than enough to make an intercept from Mab to Puck (closest approach ~15000 km), though the gunship shouldn't really need the extra fuel. I edited the Reckless doctrine as follows:
CapitalShipInterceptDoctrine InstantaneousDeltaVBudget_Percent 1 DeltaVBudget_Percent 1 DistanceMaximum_km 200000 OnlyEnsurePlanarity true InterceptCapitalShips true TonnageWeightFactor 1. DistanceWeightFactor 0. As far as I can tell, if the AI knows how to do trans-SOI intercepts, those settings should allow them to do so. As the thread title implies, if you set up this mission, the AI does nothing and the mission time expires. In messing around with AI doctrine settings it's not even really clear how you get the AI to stop evading player intercepts- even on Aggressive or Reckless they'll evade an intercepting player for no apparent reason other than to be coy. You can provoke an attack by completing an orbit within the AI's SOI but that's about it. As far as I can tell the player literally always has the initiative except in scenarios where they start in the same SOI as an AI and "IsADefendLevel" is set to true. If this is the case basically every scenario with enemy forces outside the player's starting SOI is a shooting gallery where the AI hangs out until the player makes a pass without completing an orbit, in which case it evades while chucking missiles/drones, or until the player obliges an aggressive AI by settling down into a parking orbit and waiting for them to figure out how to do their job. I really hope I'm wrong here and someone can point out what I'm missing, or that there's some hope of fixing this problem.
|
|
|
Post by subunit on May 3, 2017 20:24:29 GMT
I'm a little surprised by the muted reaction to this. Has no one ever seen the AI engage a player fleet in another SOI?
Maybe an analogy might help to explain why this is so weird to me: imagine that CMANO marketed itself as The Most Scientifically Accurate Naval Warfare Simulator ever, and that its mission statement was to generate a plausible model of what Near Future Naval Warfare would Really Be Like. Imagine, then, that you were interested in the possibility of a Chinese invasion of Taiwan, and you wanted to experiment with different mixed Taiwanese/American/Japanese battlegroups. How impressive would it be if you discovered that the only way to actually get a fight going in the game was to:
A) Place a Chinese battlegroup in a mainland harbour, sail into the harbour with a Taiwanese battlegroup (if the Taiwanese battlegroup doesn't come to anchor, the Chinese will not attack but will rather attempt to hide in nearby coves and inlets while launching the occasional desultory airstrike) B) Place a Chinese battlegroup unscathed in a Taiwanese harbour
Most people would think that this was not, in fact, a good representation of Near Future Naval Warfare, no matter how detailed the models of the weapons platforms themselves were. I don't think this is a bad analogy for CoaDE either, since the game represents combatants nuking each other's civilian populations- letting an enemy fleet actually establish an orbit around a militarily relevant body before attempting to intercept them is an absurdity in this case, in the same way that letting a carrier battlegroup set up shop in one of your harbours is- of course you want to intercept somewhere where they can't easily bomb your military and civilian infrastructure!
I think the lack of concern over this may arise from the fact that the community here is primarily interested in ship/component design and less interested in larger questions of strategy (one doesn't necessarily come before the other, but they have to be in conversation with one another to end up with any kind of reasonable picture of how the systems would "really" be used). This proceeds to some extent from the game design itself- if you want to play the game as a wargame rather than an engineering simulation, there's not that much to do after you've completed the campaign. IMO this is why building up a robust library of custom levels is important to broadening CoaDE's appeal- but as far as I can figure the AI design prevents anything more interesting or complicated than what can already be achieved in the sandbox.
|
|
elukka
Junior Member
Posts: 73
|
Post by elukka on May 3, 2017 22:53:20 GMT
I haven't experimented with the particulars in any detail, but I do agree on the shooting range feel. Frankly, there isn't much of a game as-is besides the puzzle-like nature of the missions. It's more like a toolbox for playing around with various scifi ideas. I bought it under that assumption, so I'm not complaining... but I would love if it evolved into more of an actual game. It would take a whole lot of AI development and probably some other things though.
|
|
|
Post by subunit on May 4, 2017 4:09:13 GMT
I haven't experimented with the particulars in any detail, but I do agree on the shooting range feel. Frankly, there isn't much of a game as-is besides the puzzle-like nature of the missions. It's more like a toolbox for playing around with various scifi ideas. I bought it under that assumption, so I'm not complaining... but I would love if it evolved into more of an actual game. It would take a whole lot of AI development and probably some other things though. What's driving me nuts about this is that CoaDE is riiiight on the edge of being so much more than just a spaceship sandbox. I mean, if the game was just the ship designer and the tactical engine, I still would have bought it and played it, and none of this would be bothering me, because I would say, you're right, it's just a toolbox with a bunch of arbitrary assumptions. But we got n-body physics and a turn based 3d maneuvering engine that should make the Kerbal devs green with envy. I had more or less the same thought the other day- "is the n-body physics engine really just there to make us solve a maneuvering puzzle before we get dumped into the tactical engine?" It just seems crazy that the AI doesn't know how to use it, to the point where I'm half-convinced I must be wrong. One way or another, I have no complaints about what I paid for- I put 90 hours into the campaign, designer, and sandbox before finally deciding to move on to do some mission editing. As fun as the ship designer is, without particular roles, particular missions, particular conflicts to be involved in, ultimately it's all just kind of contextless and drab. But by point of comparison, I have almost 900 hours in ArmA 3, most of which is mission editing time. I love scenario design, I love taking people's content and putting it in a cool context and seeing how it all works given some explicitly designed war scenario, and I really, really wanted to do that with CoaDE. And it's so close- so damn close. The AI doctrines do most of what you need them to do for this kind of thing (if they applied on a per-fleet basis they'd be perfect)- except, except... the AI can't get from here to there! If we just had some kind of scripting, or if the AI could use the automatic-orbit-matching tools that are already in the game.... *sigh*.
|
|
|
Post by shiolle on May 4, 2017 12:37:28 GMT
I think previously there was a setting in AI.txt that determined whether it would try to intercept ships outside its SOI, but I can't find it now. I also remember AI would sometime spend all its dv to fly off without hope to return to its original orbit or intercept my ships. I have no idea whether it had something to do with this setting or not. I did notice that after 1.10 AI seems to be afraid of plotting maneuvers that would put it, or its missles on hyperbolic trajectories out of its current SOI. Perhaps that's an answer to people who found themselves unable to complete missions because there was a missile fleet rapidly departing the scene they had no hope to intercept. This however leads to instances where AI would launch missiles and let them orbit indefinitely without doing anything meaningful with them.
Regarding potential, I do not disagree with you, but you have to remember that the n-body simulation we have, as good as it is, has limitations too. There is a reason why we only play around a subset of the solar system in each mission. In larger missions it is quite easy to generate a trajectory that would bring my CPU to its knees. Of course QSwitched has better data no doubt, but I suspect having multiple actors plotting trajectories in the full solar system simply wouldn't work (I know this is more than you asked for). I'm not familiar with a the exact algorithm used by AI to plot trajectories, but I think simply calculating a standard maneuver is perhaps not enough. Notice how bad AI is at plotting flybys (when you click on an orange arrow)? I find the amount of burns AI plots for intercepts a little suspicious too, but at least he does an adequate job there (unlike flybys).
I'm not saying what you want can't be done, but I'm not very clear on current AI capabilities either. In sandbox it often acts very strangely even in seemingly straight forward situations. AI is a very challenging area of programming and although I think very highly of QSwitched, I'm not sure how good an AI he could afford to create.
|
|
|
Post by bigbombr on May 4, 2017 12:48:39 GMT
... AI is a very challenging area of programming and although I think very highly of QSwitched, I'm not sure how good an AI he could afford to create. Perhaps qswitched could open the inner workings of the game's AI more, so volunteers can improve it?
|
|
|
Post by newageofpower on May 4, 2017 12:51:58 GMT
Perhaps qswitched could open the inner workings of the game's AI more, so volunteers can improve it? Perhaps Sempai~ could also grow two extra arms and an extra head, while he's at it. ;P
|
|
|
Post by shiolle on May 4, 2017 13:04:51 GMT
... AI is a very challenging area of programming and although I think very highly of QSwitched, I'm not sure how good an AI he could afford to create. Perhaps qswitched could open the inner workings of the game's AI more, so volunteers can improve it? I'm all for that, as well as having more behavior options for specific modules to help AI use them better. That said, there is a limit to what parameter tweaking can do. Ultimate solution would be to allow us to write AI behavior code and provide interfaces for that, but I don't see that happening any time soon. Even having some comments in the data files detailing what all those values mean would be extremely helpful. For example, Distant Worlds doesn't support plugin DLLs either (to my knowledge), but here is what its data files look like: 'Distant Worlds Resources - 1.9.5.0
'each line defines a single resource with comma-separated data, up to a maximum of 80 resources (0-79)
'Basic Information:
' - ID: numeric ID value of resource
' - Name: text of resource name
' - PictureRef: index value mapping to image used to display resource (\ui\resources folder)
' - BasePrice: base price of resource when buying and selling. Note that the actual price fluctuates based on galaxy supply and demand
' - Type of Resource: 0=Mineral, 1=Gas, 2=Luxury
' - SuperLuxuryBonusAmount: numeric value that indicates special development bonus for colonies with this luxury resource, range from 0 (no bonus, i.e. not a super luxury resource) to 50
' - IsFuel: Y/N value indicating whether this is fuel resource used in reactors
' - IsImportantPreWarpResource: Y/N value indicating whether this resource is important to have in home system of prewarp empires
' - ColonyGrowthResourceLevel: numeric value indicating level of resource required for growth at colonies, range from 0 (not required) to 1.0 (lots of this resource required)
' - ColonyManufacturingLevel: numeric value indicating whether resource is a manufactured resource at colonies. When greater than zero then resource is not naturally occurring but rather is manufactured at sufficiently developed colonies. Value indicates required population and development level before resource may randomly appear at a colony: value = population in billions * development level
'Distribution and Prevalance Information - Multiple entries allowed, each with 5 values as follows:
' - Type: 0=Planet/Moon, 1=Asteroid, 2=GasCloud
' - SubType: 0=Continental, 1=MarshySwamp, 2=Ocean, 3=Desert, 4=Ice, 5=Volcanic, 6=BarrenRock, 7=GasGiant, 8=FrozenGasGiant, 9=Metal (asteroid), 10=Ammonia, 11=Argon, 12=CarbonDioxide, 13=Chlorine, 14=Helium, 15=Hydrogen, 16=NitrogenOxygen, 17=Oxygen
' - Prevalence Value: chance of resource appearing at this type of planet/moon/asteroid/gas cloud. Range from 0 (0% chance) to 1.0 (100% chance). NOTE: when resource is super luxury then Prevalence Value instead indicates how many sources in average galaxy of 700 stars, i.e. 1.0 means single source in 700-star galaxy
' - Abundance Minimum: minimum abundance of resource at this type of planet/moon/asteroid/gas cloud. Range from 0 (0% abundance) to 1.0 (100% abundance). Actual abundance is random value between Minimum and Maximum
' - Abundance Maximum: maximum abundance of resource at this type of planet/moon/asteroid/gas cloud. Must be higher than Minimum value above. Range from 0 (0% abundance) to 1.0 (100% abundance). Actual abundance is random value between Minimum and Maximum
'NOTE: be sure NOT to define both gas and mineral resources at the same location, e.g. do not set Gas Giant planets to have both gas and mineral resources
0, Emeros Crystal, 0, 5.0, 0, 0, N, N, 0, 0, 0, 0, 0.102, 0.2, 0.7, 0, 5, 0.25, 0.3, 1.0,
1, Nekros Stone, 1, 5.0, 0, 0, N, Y, 0, 0, 0, 0, 0.136, 0.2, 0.7, 0, 5, 0.408, 0.3, 1.0,
2, Osalia, 2, 5.0, 0, 0, N, N, 0, 0, 0, 0, 0.102, 0.2, 0.7, 0, 3, 0.231, 0.3, 0.9, 0, 5, 0.306, 0.3, 1.0,
P.S. They also have a hundred pages long modding manual.
|
|
|
Post by bigbombr on May 4, 2017 13:09:27 GMT
It might also be a good idea if the AI was more considerate of the characteristics of the ships/fleets involved (delta-v, acceleration, weapon layout, weapon range, ...).
I think only delta-v and ammo count are currently taken into consideration.
E.g. a laserstar and a missile-barge require different tactics.
|
|
|
Post by gedzilla on May 17, 2019 5:29:16 GMT
You know, this is kinda why a play CoaDE much less now. The battles are unintresting, and rediculously easy (when ive build my custom ships). I could try to give the AI my custom ships, but it doesmt use them in the roles intended. The whole battle doesnt even give any sense of satisfaction. I could try to keep up with the current meta, cutting edge custom ships here, but tbh, i dont have that much time to devote to this. I think many of the other veteran players feel similar, as i hardly ever see them post on here, and some veteran players have gone completely and are never on. I really hope this game can get updates, or even a whole new game. I really like the idea of CoaDE
|
|
|
Post by Rocket Witch on May 23, 2019 14:01:41 GMT
I could try to give the AI my custom ships, but it doesmt use them in the roles intended. You can make an AI that will act closer to what custom ships need, but I think all fleets in the scenario will use it so you can't necessarily fight competent laserstars and lancers together.
|
|
|
Post by AtomHeartDragon on May 23, 2019 19:48:26 GMT
You know, this is kinda why a play CoaDE much less now. The battles are unintresting, and rediculously easy (when ive build my custom ships). I could try to give the AI my custom ships, but it doesmt use them in the roles intended. The whole battle doesnt even give any sense of satisfaction. I could try to keep up with the current meta, cutting edge custom ships here, but tbh, i dont have that much time to devote to this. I think many of the other veteran players feel similar, as i hardly ever see them post on here, and some veteran players have gone completely and are never on. I really hope this game can get updates, or even a whole new game. I really like the idea of CoaDE For now I recommend playing with some self imposed limitations - stock modules, reasonable reactors, sane materials, general purpose rather than mission specific ships, trying to beat campaign with very stringent mass limits or only specific weapon types, etc. The game has its weaknesses and isn't very challenging, but it will become much more fun if your battles don't resemble flattening a bunch of spear-chuckers along with their jungle with ICBMs. Ensuring common tech baseline makes for a good start. I could try to give the AI my custom ships, but it doesmt use them in the roles intended. You can make an AI that will act closer to what custom ships need, but I think all fleets in the scenario will use it so you can't necessarily fight competent laserstars and lancers together. What CDE really needs is AI layer on ship's blueprint level. Even with the same overarching tactical settings some ships in the fleet will realize them better broadsiding, some by orienting nose forward, some by beginning an aggresive joust at high thrust.
|
|