This is a reply of mine from a thread in gameplay. Actually.. WOW. I think that would be a fantastic idea. Everyone is saying how Comming, you don't have much control over the situation. That the comm cannot change the situation no matter how good or bad. He just goes with whatever team he has and makes the best of it, hoping its a good one. HOW ABOUT.. We can make AI infantry and tanks. Just like an RTS. The players would be HEROES. They have the upgrades and all, and the AI soldiers are just dumb unskilled soldiers who are cannon fodder. BUT this would give so much improvement to both the commander as well as the players. The comm could make a big difference. The players could take control over these squads and tell them to help build or follow him. I know that making AI form into squads is crazy. But you can script a follow me command for each bot you look at, or the comm can assign bots to a player, and you can have a "move here" command. thats the only two commands you need for bots really. I think this would greatly enhance the game. It would make the environment seem more alive as there are more people there, and there are easy kills for new players to get and feel like they are winning. You want to play a game that you feel like your winning. So both new and experienced players can benefit from this and this would widen our playerbase a lot. It would appeal to a lot more people and see how bots are dumb, unupgraded, and are not as good as the HEROES of NF or BRENODI it makes everyone happy!
Really, and who's going to make the AI? Who's going to host all these bots on their servers? Who's going to incorporate all this? We all have brilliant ideas on how to improve this game, and the devs probably do better than any of us. There's just not enough resources to do this.
Dude, bots would be brilliant. Fill out server, players can actually meet enemies, and skill-adjust by admins to meet any kind of stacks and whatever. Who is gonna implement this? The devs paying attention to the Suggestion forum. In other words nobody. This is just as good as any other suggestion.
I give the devs a deadline of 20 years. That is JUST enough. enough but barely. Didn't want to give them too much time but still need to keep them on their toes ;-) they can look at iron grip mod if they need a general idea how it works.
Essentially, the only way I see AI being implemented is via a third-party plugin. The two best candidates are Omni-bot and RCBot2. RCBot2 is working for TF2 (i.e. OB engine), but I haven't been able to get it to load correctly with Empires. RCBot2's source code is available so, in theory, an Empires specific version probably could be compiled, but I don't know C++ well enough (yet!) to do it. Personally, I think bots/AI would be one of the most helpful things for Empires. Playing on a server with less than 20 people sucks and bots could be used for everything from populating servers, teaching newbs, different gamestyles, etc. NS1 does this to great effect.
Bots in EMP would be pretty sweet, but because of the complexity of the game, all they could do is engage enemies in infantry combat probably. Not to mention AI might not be too high on the TO DO: list.
Unless you can find a willing coder, forget it. AI is hard enough in normal FPS, let alone FPS/RTS construction/tank combat.
I'd like AI that just move with attack orders and move orders. I would imagine a working implementation would end up looking like Brothers In Arms.
Well, there are already some bots out for other source games. They really don't need to do anything other then move and shoot, considering we could just leave building to the real people. It would actually be fun for the comm to push around some mediocre cannon fodder, while the real players all do their own thing. My #1 concern would be server lag though. All the servers we have right now work fine at max players, up until a tank or two shows up. At max tanks, turrets, players and buildings the game becomes a powerpoint slideshow, somehow I think a handful of bots would
Everyone seems to fail to understand how buildings affect things. A bot .nav files, what is used on a map by map basis to determine where the bot goes and what it does, is based on map geometry. Empires has pretty dynamic map layouts due to buildings. Geometry is all over the place, being relocated. To constantly update the nav data would be immense on server load. An alternative would be difficult.
Same thing happens in NS1 and yet they still found a way to do it. Bots in Empires are doable, but probably difficult.
make a unique building for both teams NF will have some kind of animal cage for their genetic mutants which releases antlions and antlion guard AI BE will have some kind of robot factory for their 'nanobot terminator robots' which produces combine soldiers and combine elite AI (they seem robotic enough) do NF and BE have separate reasearch trees? if they dont,you can research 'improved mutants' in bio tree for antlion guards and research 'improved terminators' in the mech tree for combine elites
Maybe we shoud not talk about bots but about creeps like in DoTA. They would act like walking LVL1 MG turrets.
Really? I'd have said AI for infantry only, that way you can make more useless crappy infantry to throw at the tanks and their incompetence doesn't cost the team much. I'd also suggest making them not use tickets, in fact they probably shouldn't be bot controlled players, they should be a specific AI controlled unit type, maybe little remote controlled gun bots like they have in Iraq now, just a gun mounted on treads basically, more like a mobile turret than an infantryman. You could give them their own weapons and maybe the ability to self destruct at tanks to damage them. I don't think AI will ever make a good infantry or tanker but it would make a good 'fairly stupid robot' which could be assigned the boring tasks like blowing up refineries and even building new ones if you give them a limited build ability. I am greatly in favour of the new FSR units being implemented immediately. You know there are other methods of navigation other than nav meshes? A simple 'aim at the objective, drive forwards, if you hit a wall, turn 90 degrees and go for 50 metres, then turn towards the objective again' would probably work fine for avoiding buildings. You could then mix it in with a very basic navmesh, like one node in every open area and one in each chokepoint so the robots can plot their path over long distance, just tell them to pick a random point in the area specified by the node and head for it, that stops them bunching up. You could also maybe mix in a 'keep away from walls' function to stop it getting too stuck, and also an entity which denotes a dead end on the map and contains a marker which anything stuck in the dead end zone can head for to get out of the dead end. A little foresight on the part of the mapper mixed with some code support to handle buildings being in the way would probably work fine. Hell if nothing else you could just use a navmesh and have it blank out all nodes within a certain area of a building, you can activate and deactivate navmesh nodes already in basic HL2 using map entities so doing it with code would not, presumably, be impossible, it just means when a building is placed all nodes are deactivated, both client and server should be able to do this so you don't need to sync anything other than the actual building being placed which is already done. The only major issue with performance would be syncing the AI, and that as far as I know is usually easier than syncing players because AI has a predictable set of responses, you set two identical AIs off on the same path and they'll both react identically to any issues that come up, players have dozens of inputs coming in every couple of seconds in a firefight, AI only has one every time it changes direction and every time it changes target or starts/stops shooting.
Like I said, it's doable, but theres 2 sides to it. Either it's difficult, or it rapes the server because it has to rape it's CPU updating the nav data every second.
You don't update anything every second unless it changes every second in an unpredictable fashion. Unless you're constantly dropping and recycling large buildings the navmesh wouldn't have to change every second, it would change once when a building is placed, and once when a building is destroyed, and I can't really imagine why that change would need to be transmitted across the network because you know exactly where the building is already, and you also know exactly how the building would affect the navmesh in that position, so clients can update their navmeshes individually, if they have to update them at all, because you could also completely ignore the client navmesh and simply use the server navmesh to calculate the trajectories of the AI units and then update them when they change, which would not be very often. Minimal server lag, and only as much effort as is needed to write any sort of moving AI. Being small units the AI don't need much in the way of code to handle being a large object, they can probably be treated as having zero size for the purposes of navigation. It's less complicated than having a player connected, and it would make the game playable with much fewer players than it currently is.
