Micro, AI, and the third resource: Initiative

There are two confirmed resources in PA, stuff to use, and motive force to make it useful, commonly called "metal" and "energy". However, there is no resource for "what to do" which I think is an oversight.

There is a common problem in RTS games, that of increasing complexity getting out of hand, putting more and more stress on the player's mind until they can no longer keep up. The best player has the greatest ability to micro-manage their units, infrastructure, and resources to match the situation. This is as it should be. I do not propose to do away with this, but I do suggest making it a strategic choice.

Why do engineers just stand there when a building across the base needs repairing? Why do my units loiter at the factory when the perimeter is under attack? The AI exists to easily employ these units (since theoretically the AI could be running my entire army) but instead it does nothing, waiting for me (the player) to initiate every action. What is lacking is an in-game method to build, employ, channel, and combat the existing unit AI. We should be able to harness the game's AI to our ends. The AI should be available to help us, as well as a foe to fight against.

We need a third resource, intelligence, or perhaps initiative, or intrinsic-personality. I'm going to call it IN for the rest of this post. When you start off, the commander produces a small amount of IN, as well as metal and energy. This IN will be enough to automatically start building a base (if you want to surrender control of the commander to the AI) and control a handful of units. If your IN runs out, units will start to stand around idly (like they do in a normal RTS) and require micromanagement. Resource buildings and infrastructure can be built to produce (and distribute?) IN, just like the other resources. An excellent player will have little need for this resource, being able to micro-manage all of their units without the help of the AI. However, even the best player may find a bit of AI assistance helpful. In any case, building IN structures will consume resources, and add another strategic decision to the game-space.

With sufficient IN resources, the your army could run itself as if the AI were running your units. You can sit back and watch your bases unfold and armies conquest with no direction. Of course, if your foe is spending those resources on more guns and micro-managing instead, you'll get your butt kicked every time, but every player has their limits, even the best. Being able to hand over the operation of an asteroid, planet, or army to the AI should improve any player's game, and deciding what gets AI controlled must still be a play choice.

Note that I have stayed as far from implementation details as possible. The UE team knows their work, and how to implement ideas. Can we get on-board with this general concept though? Am I just crazy, or is this putting the (already planned) extensive AI system at the player's disposal and finally banishing the specter of limitless micromanagement?

There is a reason that nobody does this. It's not fun to have an AI that only helps you some of the time. I'm all for having an AI that does obvious things, but why would you cripple it?

Yet another thread on this topic?! Have you notified Guinness yet?

I was hoping to have the built-in AI notify them for me.

I'm all for a third resource, but i'd rather have a variety of rare metal than a vague concept act like a resource..

given a universe populated with asteroids there should be plenty of opportunity for resource locations and metal types strikes my mind as the most realistic and easily introduced third resource, you still have to gather it, it wont just spontaneously generate itself like some sort of command point and it adds just that little bit of extra realism and strategic management to the game

In a future so advanced as PA, one that nanolathes manufacture everything in, it strikes me as unlikely that they wouldn't be fissioning and fusioning their elements to get the atoms they need. Another type of metal is redundant.

Interesting. Another RTS from way back had a limited form of this, Earth 2150. You built a building and told it to take control of a small, usually fairly mundane, but still important part of the game micro-management.
The example that sticks out in my head (because it was all I used) was automatic defense-building weapon upgrades.

Not all of us are 200 apm players, and adding a degree of late game automation would take some of the intense micro out of the last game, where we should be blowing **** up.

People really have no understanding of what they’re asking with all their automation crap.

Firstly, situations are very specific. The player thinks “gee, why won’t my units do X in situation Y?”. The player thinks this is an easy decision for the AI to make. IT’S NOT. situations can be very dynamic. Even in the examples used in the OP- perhaps the damaged building being repaired will crash your economy? Perhaps the units milling around in the base are better sitting there while you build reinforcements, rather than wandering off to their death against a superior force?

Look at any RTS AI you care to name. what do they all have in common? THEY SUCK. I do NOT want these god-awful AIs taking over the automation of my base.

Secondly, I guarantee you that these AIs will be unpredictable. Which isn’t good. You need to be able to have faith in them to be doing things the same way, everytime. Otherwise its just a mess, “No Mr Engineer I wanted THIS building repaired, not THAT one”; “Christ, why are you attacking THAT enemy force way away over there when THIS enemy force is much more dangerous!?” etc etc etc.

All these simple decisions the human brain makes in a split-second are actually very difficult to design an AI for. You don’t appreciate what you’re asking.

for the same reason that in the universe you describe mass fabricators aren't used exclusively

I agree up to a point, I see what you mean as wanting something like the Hearts of Iron III AI where you could select how much control you wanted to help deal with the massive areas you fought on but I would prefer for this game where you had to manage the solar system you were in by hand. The AI could help in Galactic war assuming it's all real time and each battle dosn't represent and instance of a turn, if you have fifty solar systems realtime there is not a snowballs chance in hell you can manage that against an invading army by the AI on multiple fronts. If I could designate AI to fight for me there until I can take care of it personally I would be happy. I don't want it to be able to fight my battles but at least be adept enough to hold for awhil while I'm getting there or busy fighting.

That may be. However, I guarantee that having my units do something marginally useful will be better than having them do nothing (which is what they normally do). Having the factories build something is better than resources piling up and doing nothing (which is what I inevetably end up building).
So:
1: If you're way smarter, quicker, and broadly aware than the AI, great. Good for you. Most of the time, I'm not, and I would like the AI on my side, watching my back. I think I should have to pay for it in-game though, which is why I suggest making it an in-game resource.
2: I have a decent amount of faith that the UE guys are going to pull this one out of the hat. What if they come up with a radical fantastic AI? Even better, no?

PKC:

I'm not suggesting huge alterations. Telling your engineering units to automatically repair a building within 'X' distance. Exploring unknown section of the map. Keeping your economy in the black by shutting off metal makers for a few moments. Firing base defenses at non-visual targets (this one is straight out of TA...). Things that are important, but generally tedious.


Honestly though, check some of the options from Earth 2150 out. For a game released in 2000 it still does things that no one else does in an RTS.

Agreed. Earth 2150 had some really good elements. Unfortunately, it didn't implement them all in a well balanced and enjoyable package. Like Ground Control, it succeeded brilliantly at a few things.
I'm pretty sure the Planetary Annihilation team is trying to make a well rounded enjoyable game. If they can use some of these revolutionary elements, great. If not, at least it will still be fun to play!

For the ideas in the OP i agree.

For general automation I agree with everything except the part written in red (I would put it less rudely though). I think PKC captures basically why I think most AI suggestions on this forum are not usable.

Firstly discuss general automation instead of coming up specific ideas. We are already at a point where a lot of people seem to dismiss automation due to specific bad ideas. The creators of those ideas also need to know what is likely to be accepted by many people. Instead of discussing you could just find the previous threads because I would probably be reiterating myself.

Here is the short version of what I said: Players have to know the effect of every command they issue.
It's not a finished definition because many words are debatable. Express interest in discussion if you want more. If not then continue to come up with ideas without thinking about what automation means.

??? What did Total annihilation and Supreme commander have in common that we all love.


Some comparisons to "common RTS's"

• Large number of units that are useful in different situations vs. ~10 units that have only upgrades that are rock-paper-scissors
  Real terrain with real weapon physics (ie ballistic, line of site, time to target) vs. calculated weapon fire and hits with after the fact graphics.
  Massive Maps vs. closterfobic ones with 1-2 elevations and no alternate terrain.
  Massive and diverse armies vs. small squads of 5-30 max - with some rock-paper-scissors coverage.
  Air-sea-land units! plus transports of various kind vs. mostly land with a few air and one transport.
  Useful Super units vs. one big unit that if you get to it the game went too long anyway
  Expansive/extendable economy vs. limited to map resources.
  Units that did some thing on there own vs. completely dumb units
  Patrols/waypoints/queuing/repair vs. maybe one of these?
  Setting facing/rally points/and disposition of units and buildings vs. again pick one or two
  being able to build real bases! vs a cluster of building with a few defense towers.

Some additional items

• Ratio of economy size and vulnerably (ie the bigger and more powerful the more vulnerable)
  GOOD AI (the stock AI was good but the 3rd party AI was amazing - not just hard but it played smart)
  Automated transports/repair/refueling etc

We liked Total annihilation and Supreme commander because they where different from the normal RTS's out there. NOT over simplified to rock-paper-scissors and fast mouse work.

RTS's like MMOs have gotten stale but people are all like "you can't change it or it is not the same?!?" People forget that both MMOs and RTSs have been around since computers barely had graphics! The old arguments do not apply anymore and the limitations of 30 years ago should not dictate today's games.

The argument that "well AI sucked before" and so it must now does not hold water with multi core/threaded systems with dedicated graphics cards nowadays. With good programmers the AI can be VERY good (as demonstrated in SupCom MODs).

Also I want to play a game where I can get down and dirty and micro but also do not want to have to tell each unit not to stand there and die because I happened to be else where. Why do people think we have to micro manage everything? because we HAD to because of older computer tech. Now (as also seen to some extent in SupCom) That units can act "smartly" but still follow orders.

BTW this initiative resource already is being used but the user does not see it the Pathing and AI gets dumber are system resources get over utilized.

Do people want to play a Supreme commander with units that require your strategies but can operate on a small scale on there own? or just one more hive mind game where you need to tell each mindless unit to take a dump before they explode?? :lol:

I'm sorry but if you think all it took was multicore cpus to enable decent AI in games, you're gonna be mightily dissapointed.

By the way, we've had them for years, I haven't seen any improvement in game AI.

Even the very best AI I have seen, whilst mildly entertaining to learn the game against, is nothing compared to a human opponent.

I’m not sure it’s a matter of processing power (sorian or someone qualified could comment) though. I think it’s more that it’s really difficult to teach strategy to an AI for a real-time game.

Also, whilst the modded supcom AIs were a huge improvement, they were still absolutely terrible.

You’re one of the people that doesn’t quite understand what you’re asking. Units should do EXACTLY what they’re told. Predictably. ALL THE TIME. nothing more. In your example, what should your units do? Rush off to engage the enemy? So you’re constantly having to make sure you’re not being baited into a fight you can’t win? Maybe you want them to run away? Where to exactly?

incorrect. Warcraft unit AI for instance would have your units run at the enemy when damaged. It was beyond frustrating. Units in supcom do exactly what they’re told.

No. because they can’t. well, if its something that you can toggle off in the menu I don’t care. I will enjoy smashing noobs that can’t manage their army.

I guess you COMPLETELY missed the point. @zordon

In game there WILL be a Skirmish AI with multiple varriations. But I am not talking about that. (but is an example how those AI's can be good with less Strategic work)

This has to do with basic unit/group functionality while your "eye" is some where else actually doing something important rather then being forced to microing a unit to do something insanity basic.
What you and other seem to miss? it that some people would like to see some basic autonomy (yes with limitations) so WE CAN use our human brains and use more complicated strategies?!?! Most games force humans to play DOWN to the computers level, buring time micro managing. The computer can do that faster and more efficiently then ANY human could ever hope to do. Where they historically fail is the "big picture". This Strategic level requires good programming and lots of resources. So, sans that, bog the Human down microing or trying to manage a large force and they slow down and start making mistakes, how most RTS AI's fight. How about let the mundain stuff to the computer/units where it belongs and use the brain to do higher level work?? that combo would allow larger more epic battles that microing would be impossable. (also if I wanted to play a small totaly micro and fast human Vs human battles I would not be funing this game I would be playing Starcraft - but I want bigger)

Also a better level of AI not only possible but has been done before. The problem has ALWAYS been the extra time of development and limited system resources. It is much easier to give players micro managing since it requires less development. Less Development/QA = less $$ spent win-win..
But LOOK at the SupCom AI MODs! some great work there even with a limited framework shows what can be done with time and effort.
I am suggesting that this game in the light of the advances the other incarnations showed during there time, the Devs here may bring this up a notch because they can.

PS. I know the development that goes into this, I have experience in it

@jctorta
Try being specific because there are a lot of bad ways to implement AI. Play Zero-K for examples of what you want.

@PKC
You point out many examples of poorly implemented unit AI but that doesn't argue against the concept.

Do you know exactly what you mean when you say this? The statement may not be as exclusive as you believe. For example if I set my units to Roam then I have effectively ordered them to chase enemies.

For an example of AI which is not bad take automatic kiting. This is a command which is identical to A-move except that units also move away from opponents which are too close to them. This command has been used for years but 1000s of players, it is useful. Do you have any problem with it?

I was referring specifically to suggestions made by other posters, and provided real-world examples of these suggestions being used in other games (and being horrible).

As I have mentioned here and elsewhere, I am not necessarily opposed to player-input commands (such as roam) or toggle options if you don’t wish to use them.

If you’re saying “identical”, by which you mean you press say, K-right click on a piece of the map, which is where units would kite to, then I could see that being fine. However an AI should never ever “think” where it should be kiting to, because it will be unpredictable, and often get it wrong. Additionally, kiting means moving to AND fro, so does a unit/blob move back to where it was when given the kiting order? How long does it wait before doing so? If it has no intel (say, radar is destroyed), does it just walk straight back to the original point when given the kiting order and get annihilated by longer-range units with intel?

I honestly don’t think players appreciate the kind of questions that need to be answered when designing decent unit AI. Definitely they’re not raising the right questions in the multiple threads about it. They just take, at face value, that the right decision (that THEY want) will be made automatically.