The new guy: A.I. Developer #42 & Discussion on A.I.

[Dix]

Belated welcome, Gijs-Jan, and a few questions for you (apols if this was already asked - didn't see it in the thread.)

Is there any chance that we could see the aliens try to fulfill objectives, as opposed to simply fighting? For example, an alien on a terror mission will actively focus on hunting civillians, they sometimes try to swarm the chinook, on a human base attack they try to do as much damage as possible to the facilities, free prisoners, etc.?

Will aliens group together to fight, or will the ai make them act as individuals only?

Edited August 25, 2012 by Dix

[Gorzahg]

Are you going to post here when you update your blog? I'm no good at programming and math and stuff but I like reading your kind of stuff.

[Gijs-Jan]

Are you going to post here when you update your blog? I'm no good at programming and math and stuff but I like reading your kind of stuff.

I will. :-)

(...) aliens try to fulfill objectives, as opposed to simply fighting? (...) Will aliens group together to fight, or will the ai make them act as individuals only?

Its behavior Chris requested; so I do hope so :-)

As for individuality; it depends on how far I get with development. The current system in place is entirely built around simple unit based reactive behavior; I'll have to start from scratch as the paradigm is just the complete opposite.

I just uploaded some basic augmentations to the AI for pathfinding. (civilian fleeing, basic cover usage)

I missed the deadline for the weekend alpha as I spent most of the night dealing with BSOD (driver iss.); sorry guys!

Ah the great joy that is developing on Windows... *sigh* :-p

Anyway; don't expect too much, as I'm mostly now trying to figure out how to work with the old code: what to keep, and not.

But at least it should be better than the random movement ;-)

Ow and Buddhist civilian is still possible. The AI only cares about where it wants to go; not about how its going to get there.. :-p

Edited August 26, 2012 by Gijs-Jan

[Swe_Racoon]

Buddhist Civilians... That one has really snowballed...

In all seriousness though, welcome to the forums, and looking forward to see your work in action! From what's written here and on the journal, it looks very promising indeed!

[Gauddlike]

We will all know who to blame when some alien takes out our favourite trooper by doing something unexpectedly clever...

[Jean-Luc]

We will all know who to blame when some alien takes out our favourite trooper by doing something unexpectedly clever...

Like shooting off the chandelier above the X-nauts head?

[Gorlom]

Like shooting off the chandelier above the X-nauts head?

Or when they take a cardboard box, turn it upside down and goes MGS on your ass.

[Swe_Racoon]

I just saw a cardboard box moving across a big firefight, with a Ceasan with a eyepatch and all.

Glorious!

[Gijs-Jan]

We will all know who to blame when some alien takes out our favourite trooper by doing something unexpectedly clever...

I have but one personal goal in this project: Fear the Reaper..

[Max_Caine]

Ah, we're talking a smart melee-only creature who doesn't just simply make a mad dash to the sqauddies in the hope it gets to them. Flying Power Armour FTW!

[eltharion]

Personally, when I played the first X-Com I hated from the depths of my heart the crysalid. All started when my favorite colonel turned into a slimy zombies. When then acting as crysalid took out half of his former team, I had to prepare two bodyguards (heavily armed) for the most promising graduates. If you intend to make the Reaper "friendly" or even more than the old Crysalid ... it will be fun.

Good luck;)

[Moonshine Fox]

You reminded me of the article on FEAR's excellent AI.

I was thinking the exact same thing.

[Moonshine Fox]

No prob. Do you mean combat-wise?

I can't really think of examples that spring out; although I know people that designed (learning) combat A.I.'s for Neverwinter Nights.

The problem with game A.I. is that the game industry has become very, very good at creating "good enough" A.I. (An almost direct quote of Alex from GameAIDev.com). There's not much incentive to construct A.I. for the hordes of goblins that will only be on screen for a couple of seconds, or A.I. that is practically invincible. :-)

In other terms, there are games like Dwarf Fortress where aspects of "A.I." (I'm using the term extremely liberal now) really contribute to the game.

I'm going to regret this: Chris M. Parks thesis on designing emergent AI for AI:War. Hopefully, that doesn't detract too much from your work. But it's a damn interesting read, and hopefully, there's something in there you can use.

Also, WELCOME! AI Coders are something of a "hero class" to me. You guys do pure magic with numbers.

[Jean-Luc]

Finally got around to reading the journal entries. Really well written, simple to understand for someone who knows nothing about AI programming but also informative and relevant. Quite a fun read.

[Gauddlike]

I have but one personal goal in this project: Fear the Reaper..

[ATTACH=CONFIG]1401[/ATTACH]

EEEExcellent.

Ah, we're talking a smart melee-only creature who doesn't just simply make a mad dash to the sqauddies in the hope it gets to them. Flying Power Armour FTW!

Yeah not adding flying armour to Xenonauts will make the reapers more of a threat.

No chance of just hovering out of range while you point, laugh, and eventually kill them.

Now hopefully there is a higher rank with little wings to let it mimic the jump armour...

[cyllan]

I am designing everything with modability in mind. However with A.I. it's rather tough to have both "intelligent-decision" making AND complete control over what the A.I. can do. In other words: you'll need to trust the A.I. to make the right decision.

In essence; mods will allow the enemy units to change behavior (P.e. Rambo-style <-> Cover-sneaky-ninja-style) quite easily, but they probably won't be able to exactly tell them what to do.

Can't have civies becoming too smart! But maybe I'll allow the randomization to produce some that are smart enough to run to the dropship. :-)

(If they can last long enough to get there that is..)

i dont think spending time in making AI moddable is effective.....it is hard enough to do just 1 AI ........ also i dont think it should be changed depending on difficulty levels.....just increase/decrease number of enemies depending of difficulty and one good AI is more than enough.......i just hope we dont have to wait till the last hidden enemy inside a building to kill it to finish the mission......you know he could maybe "surrender" alive?

i think it would be easier /cheaper and faster to code AI this way all is needed is the number of random ufos / aliens etc...... and it woulndt impact the game at all....making an AI more defensive would be frustrating.....making it more agressive would be a mindless shootout .

however there is always 3 kinds of AI.

1...the enemy knows where you are always......cheating AI

2....line of sight and instant communication to other enemies

3...enemy cant see you , only one on one so they don't move at all .....you kill the main batch and the last one/two is a haystack chase....... a bit frustrating.

Edited September 6, 2012 by cyllan

[cyllan]

example of cheating AI

supreme commander on highest level

[Max_Caine]

Ah, cyllan, it's always a pleasure to read your posts. Now, where to begin... ah, I know.

You seem to be unaware that Goldhawk has been quite accomodating to modders, making it easy to access a lot of the surface variables. Given Goldhawk's friendliness towards modding why should GJ, if given the opportunity and the means, not produce an AI system where players can customise certain aspects of it? GJ is in a pretty golden position. With so little development done on the A.I. GJ can create the foundation for his A.I. in a way that he prefers to do so, instead of adapting someone else's mess, so he can make a nod towards modders without having to make special access available.

Just increasing/decreasing the number of enemies is a particularly lazy way of increasing the difficulty of a game, and it's one that will get pounced on when the game is reviewed. Given games can (and do) live or die by their metascore (it's more important than people would care to admit), why advocate something that will drive the metascore of this game down?

When considering A.I., "Cheaper and faster" is quite often the way to make something worse, not better. In fact, "taking shortcuts" is something at you have previously critised Goldhawk for doing. From the horses mouth. And I challenge you to look anywhere on the forums and find one thread that praises having lots of enemies and dumb A.I. You won't.

A.I. is a crucial part of this game. Poor A.I. is very visible in an environment that runs quite slowly. Please don't advocate things that encourage poor A.I.

Edited September 6, 2012 by Max_Caine
grammar and sentence structure

[Gorlom]

I'm constantly surprised by cyllans posts... You'd think I'd learn what to expect by now, but he still finds ways to surprise me.

[Gauddlike]

GJ is making just one AI really.

It is just divided into sections that can be used in different ways.

If those sections are being used anyway then making them available for modders to alter is a good thing.

i think it would be easier /cheaper and faster to code AI this way all is needed is the number of random ufos / aliens etc...... and it woulndt impact the game at all....making an AI more defensive would be frustrating.....making it more agressive would be a mindless shootout .

You assume that the only way to make the AI better is its aggression.

You could make the range of numbers used for its calculation of safe areas wider in lower difficulties meaning they would not always pick the best spot to end a turn.

On lower difficulties you may be able to adjust how precisely they calculate threats so that they won't always go for the person who is most likely to kill them next turn or who would be the easiest kill.

i just hope we dont have to wait till the last hidden enemy inside a building to kill it to finish the mission......you know he could maybe "surrender" alive?

That suggestion has been brought up a few time and I still think it is possibly the worst option available.

Much better ways exist and have been covered.

Some of them involved secondary completion objectives, AI that changed to a 'desperation' mode when solo, giving tasks to the enemies so they wouldn't just stand in a cupboard somewhere etc.

[cutterjohn]

So A.I.!

I've heard you guys were having a rough time with the subject, and after seeing a job opening I applied for the position!

A small introduction: I'm a quite passionate game developer, specializing in A.I., living in the southern part of the Netherlands.

Important answers for insight into the persona:

- Chrysalid, Chrysalid, Chrysalid

- It's called UFO: Enemy Unknown, dammit

- UFO > Terror from the Deep >>> Apocalypse

- Flying Power armor was O.P.; Real men play without.

Now onto the good part:

I want to be quite open on the A.I. development and plan to keep a journal.

The journal should be approachable by anyone (it shouldn't be technical) and the first two posts can be found here:

Xenonauts A.I. - Introduction

Xenonauts A.I. - Knowledge Systems Pt. 1

Furthermore, feel free to ask me any and all questions, apart from specifics on game content.

Also, I hope this thread turns into a discussion so you as a community can help guide me in the construction of the A.I.!

Confident he will break your unfortunate A.I. coder streak,

GJ

(Callsign Methius)

I'm late with a response, but, really, a FSM for AI?!

A neural network would seem to be better, although I suppose if you give very restricted class of responses an FSM would work, but something with a more malleable response mechanism would be better.

[caveat]

I've not seen how you're actually coding the FSM, so it might actually feature more than your run-of-the-mill FSM...

[/caveat]

[XenoMask]

You quoting that post reminded me of Apocalypse, which is my fav XCOM game. And that reminded me of how XCOM fans tend to hate it.

Then I actually started to wonder "Wait, I have always assumed that XCOM fans hate Apocalypse because its different from original(plus first sequel to original was basically identical game but reskinned and made unfairly hard so they had different expectations), but is that actually true?" So whats actual reason why XCOM fans dislike Apocalypse?

[Gijs-Jan]

I'm late with a response, but, really, a FSM for AI?!

A neural network would seem to be better, although I suppose if you give very restricted class of responses an FSM would work, but something with a more malleable response mechanism would be better.

[caveat]

I've not seen how you're actually coding the FSM, so it might actually feature more than your run-of-the-mill FSM...

[/caveat]

The whole point is to not to do a fsm as backbone for the system, that just was the system already implemented.

The only part which might remain is an extremely light version of a fsm which switches between sets of weights.

This part being so trivial that there would be little distinction between a hsfm and a nn. (except for me having to train a nn for a very trivial fsm)

So in recap; no, not a FSM. Also not a NN. Both would just result in simple reactive behavior.

[Vivisector 9999]

So whats actual reason why XCOM fans dislike Apocalypse?

I can't speak for myself, since I actually liked Apocalypse, but one common complaint were the aliens looked silly on the battlescape with their bright colors. (Granted, I have no idea why the spitters had to be bright pink of all the damn colors...) One reviewer even said he felt like he was waging war on the set of Barney And Friends, although I think that's a bit harsh.

Some of the game changes also alienated some fans. After fighting over the whole world, some felt it was a step down to be restricted to a single city, populated by future corporations that just don't have the same resonance (when you're defending them, pissing them off, etc) as the old modern world nations. Apocalypse's whole cartoonish retro-future look didn't fly with some people, either. And even though Apocalypse fans got used to the real-time mode, the turn-based mode was pretty half-assed - another thing that hardcore X-COM fanboys could take as a slap in the face.

There were a few balance issues, too - road vehicles were all but worthless, and if you didn't get certain UFO components (from UFO types that only appear in the early game), your research could get stuck in an unwinnable state.

[Gauddlike]

I thought Apoc was a great concept but flawed in the way it was executed.

The alien types were varied and had some really interesting enemies.

They looked a bit like the Muppets had launched a terrorist organisation though.

The research tree wasn't overly interesting (off the top of my head I can't think of a single thing from it).

Road vehicles were a waste of time and effort, both for the player and the developer.

The single city idea was a good one to allow greater political interaction but also felt too small.

Real time was a mistake for me, should have focused on getting the turn based right.

There were loads of things I liked as well, the androids and hybrids for example were a nice touch.

I enjoyed being able to raid hostile organisations or have them raid me.