The second of my research threads, please choose an option for the default behaviour. Note that it would all be configurable in the server settings: a) AI suck. No AI. b) AI Commander by default (if no player chosen), immediately yields to any player who enters the comm vehicle. No other AI. c) AI Commander with infantry AI who fill up the server limit but disconnect when a player joins. d) AI Commander with small number of infantry AI (say, 4) but disconnect when a player joins. Note that the infantry AI are nothing special. They'll shoot enemies, follow orders, take cover and build structures. This thread doesn't mean AI are coming to Empires. It it purely theoretical.
Jokes on you! It is and is being featured in the new tutorial, along with a way for mappers to now make survival maps. Disclaimer: It's not like infantry bots will be roaming around. Some of this stuff will be featured in the tutorial, other things wont.
Hmmm, I was actually thinking of integrating them into gameplay differently; that you could purchase AI bots as a commander; the first person to invite those to a squad gets the bot following that squad leader. That means the possibility that it will be advantageous for each player to pick their own squad and fill it up with 4 bots, and then try to keep them alive. (because, when a bot dies, they wouldn't respawn; you purchase each life of a bot) The idea of an AI commander is interesting; please tell me more about what it would do; I think placing buildings at all usefully would be a bridge too far in terms of programming it; but maybe it could just place refineries and research at random? Or perhaps we could build some positions into each map where the AI comm would place buildings when conditions are met. Edit: So, I'd love to discuss all these possibilities - I think a poll is a bit premature, and when I read the choices they all seemed along the same lines anyway.
Uhh, I think what was talked about in the AI commander was it would get research going but be on a delay so it isn't better than a human, would not give targets, and if you "request commander attention" it would try to place every type of building (Ref first in the area) and then place every other building possible but have a limit of how many of each it can place. IE 2 VF's allowed so it will place one at main near itself and one wherever someone requests to prevent building spam and trolling. Research setup to go down a certain meta maybe? Also, I like your bot idea. Would make games seem even more massive. However, the problem is that bot's can't navigate around placed objects atm but they can navigate everything else per commander order.
My current commander AI model is as follows: - Find closest resource point, order engineers there, build a refinery. - If resources > radar cost, build a radar. - Begin research (preset path, one after the other if resources allow it). - If resources > vehicle factory cost, build a vehicle factory.
Honestly an AI commander would fix so much shit. Just do like-> if unit close to ref and unit is engi and we have 100 res drop ref. Same with radar etc like you said in the post. And for research we can just let the people vote for research paths. You barely need commanders anymore if that gets done properly
I wonder how ai works in other rts games. Modifying that to accept player prompts might not be too hard, I dunno. Voting for research reminds me of science and industry(hl1 mod), what a cool game. Holy shit, it's still being updated.
mostly on grids (some on navmeshes) without buildings you can get into (makes pathfinding a whole lot easier). for decision making you use behavior trees.
i know how its done (and why it sucks) in wc3. first you have inbuilt behaviors like miners, builders, wheter units tend to flee or fight to death and whatnot. you cannot (easily) change those behaviors. then you have a build order much like "build peon, build peon, build peon, build altar, build barracks, build peon, build hero, build grunt, build grunt, build grunt, build warmill, build grunt, rearch upgraded weapons, build grunt, ..." the game figures out building placement by itself, its done by proximity to either your spawnpoint - a prepalced entity - or the resource the building acts as depot for (townhall close to goldmine, warmill close to trees) and lastly there is a thing called attack groups where you can set a composition and minium and maxium value of units to be satisfied for attacking. theres a bit more to all that, but in general thats how it works. the reason why it sucks is that, if at any point players figure out anything of this they can just explot it by lets say removing a building thats very high up in the build order (high up = high priority) even tho it might not be of much use anymore (like for example many grunts in late game might not be the best idea, so a barracks would mabye be less important than a wyvern building thingy) or liek for example if they have tons of wood but you have x woodcutter peons the ai will desperately try to build peons even if the 75(?) gold would be put to better use. translated to empires research should be rather easy and straight forward, you dont need to build units and it doesnt cost money anymore, so as long as you have a radar you just follow a set of "build orders", either randomly chosen or set on a per map basis (or maybe even both to allow for even more variety). its pretty much how comms research anyway. maybe, after you followed the main path, you could introduce a counter system, but this is a bit tricky because how would the AI know what the enemy has (ofc the server knows it right away, but thatd be an unfair advantage over human players who have to guess) the only complicated thing will be buildings (wc3 uses a grid and its rather easy to check wheter or not there is free, buildable space on a uniform grid, just translate the coordinates and iteratate over the area to check no grid cell is unbuildable. ultimatively you could see the world as grid in empires aswell, the only issue is that the cell size is about 1 inch and theres thousands if not millions of cells under the footprint of a rax) and especially forward bases (in wc3 this was relatively easy since an expansion location did directly correspond to a goldmine, its not always directly related to refinaries in empires and not all refinaries make for a sane expansion location) but duke (i think it was duke) proposed a rather smart way to figure out buildable space in another thread making use of something similar to how bsp is used for collision detection if i remember correctly, so the only question remaining is how to decide where the ai will be building up - as forward rax are crucial on most maps. and ofc implementation details and im really curios how the pathing problem will be solved as for bots in a classic mode empires game you cannot just use the inbuilt precalculated navmesh based pathfinding solution. im guessing something like a navmesh with some sort of nodes that allow to connect map-pathing with building-pathing, much like you do ladders in such enviroments. but idk if theres a workaround, i think youll still need to remesh at least the areas where the building footprint intersects the underlying navmesh geometry. i really wonder if its worth the work. but i guess thats more an issue of the one who wants to do it. oh and i want to add, im not very fond of AI controlled units. they are aimbots (no shit huh). i feel cheated when a bot kills me because a RNG decided the bot headshots me. and even tho some superubermlgmylittlepony pro will come and tell me that hes better than bots, you are not, they decide in the very frame you enter their range and could fire at the exact coordinate your head is with 32bit precision (unless its 64bit floats ofc). if you can beat bots its because theyre artificially dumbed down.
Can we get and option E? No to AI commander, but AI to assist commander? Can't get a newbie to build that ref? Send in a bot to do a newbie's job and maybe train the newbie by example at the same time.
There's nothing wrong with artificially dumbing down AI. Add some lag to reaction times, add some randomness to aiming (go from rough angle to precise over time), etc. to mimic human players. I do take your point about feeling cheated when shot by one though.
I would definitely like AI bots to train our new players. For example, new players can set up a local bot server(not open to other players), with bots loaded into the game. Kinda like those "V.S. AI" games in League of Legends and World of Warship. But I am neutral about adding bots into actual gameplay. It's not exactly that. Bots still can't bypass "weapon spread" and "weapon cycle time". Sure, they can put their crosshair dead straight at my head but in the end they still have to deal with unpredictable weapon spread. Weapon cycle time also means they don't kill me in 0.001 second. Back when I was a semi-pro CS player, I fought against a bot with 0ms reaction time and the fight wasn't one-sided. There's even more tactics in Empires for good players to outplay "0 ms reaction time bot". With all that being said, I am neutral about throwing bots into the actual gameplay. (AKA bots in NUBS)