Firefight Improvement System 2

[acoustic 82]

This is very interesting. As of now, the ai is made up of non specific terrain training. They know what they're suppose to do, but they don't have the geographic experience. Does the ai just get use a very specified spot (ie, that town), or does it gather general tactics for similar areas.

This would be groundbreaking for something persistence based like ALiVE, where this could be a map wide war going on and as the days go on, both sides improve instead of doing the same thing l. Idk if it could work together tho

[miketim 20]

That would be cool Acoustic.

Imagine a Syrian Civil War scenario in which we could simulate the "rebels" getting western support and the Syrian Army getting support from Iran, they could start using better infantry tactics or somthing.

I think once all the features of this mod work, it would be cool to see different methods of implementation, like per mission learning or somthing, like a way to have them learn stuff but only for that persistent mission or some type of stuff.

I dont know, just seems like somthing to start to look at once its reaching its feature lists completion.

[zooloo75 833]

Update: Ragdolling. Units have a chance of ragdolling when shot or hit with an explosion.

Edited June 19, 2014 by zooloo75

[diesel5187 73]

I really like were this mod is going zooloo75. But....

Remember the real fun in ARMA is to be able to play with alot of players and AI in a big mission(at least for me and the players I know). When you do that, server performance does take an impact and so does the players fps. Idk if it can even be done, but please try to keep everything as efficient as possible so that AI behavior does not "overload" the server CPU and slow everything to a crawl.

[kremator 1064]

^^ Well, duh. I doubt he will write something that is NOT light and efficient. As with all mods, if it is useful it will be ubiquitous (as with CBA)

Most FPS problems with a mission on a server are because of the mission. Some (if not most) of the stuff the zooloo75 is putting in place will be client based.

Personally speaking, I totally agree that fun comes from large numbers of AI kicking seven bells out of eachother. Putting the I into AI is the hard bit!

[zooloo75 833]

@Kremator, the mod is only used on the host/server - lifting the workload off of clients.

A large amount of AI features are expected to have a performance impact simply due to the lack of efficiency some of the scripting commands have. It is unrealistic to expect to run 100's of AI each with their own advanced thought-process and have good performance. It's going to take a host with a powerful machine to handle it all.

After all, this is a simulation, and lots of code is being put into the behavior of a unit. I already have to cut down on features because the engine simply can't handle it. I am writing my algorithms and methods with Big-O in mind.

Also as Kremator said, performance is heavily-based on the mission. If a mission is already performing poorly, then there wouldn't be much sense in throwing more code on top of it.

The mod is going to be open source and is protected under Creative Commons. If someone would like to simplify the AI's behavior in my mod to perform better for their server, then that route is available to them. I am not going to over-simplify my work because people want to run this with a 1000 AI on a weak machine. If it runs on my machine then I'm happy, and given that my hardware isn't exactly the newest and greatest, you guys should be fine.

Edited June 19, 2014 by zooloo75

[das attorney 857]

With regard to your ragdolls, did you manage to get working on DS, Locally hosted and SP?

I've got a ragdoll system working (with caveats and glitches ofc) but it is limited in the environments it will work currently. I'll be publishing the code hopefully this weekend after I fix a bug (sometimes the units die instantly or can become temporarily invincible) - I think I can fix it without too much hassle but we'll see.

I'll be welcome to share or compare approaches if it helps produce a good system for the community to use (and ultimately gently nudge BI into putting some ragdoll scripting commands in the engine).

EDIT: Just saw your thread in the completed section - be interesting to see.

EDIT2: For your reference, the method you are using is well documented here. It's worth checking out to see how they ragdolled stuff and synced up the damage on map objects:

http://killzonekid.com/arma-scripting-tutorials-scripted-charges-v2/

Edited June 20, 2014 by Das Attorney

[diesel5187 73]

From what I have observed, the main problem with AI is server optimization, on a Corei5 machine the server only used 2 cores at 60%, the other 2 cores were barely at 10%.

I was trying not to get into this, but instead say, work your magic, do your best, I really enjoyed the original version when it was working properly(before arma updates which made it weird) and I'm sure the new version is going to be great.

[taro8 806]

I love the idea. However you should watch out for spreading yourself too thin. Just keep this in mind.

Other then that, I must say that I would be more then happy with the "Skynet" feature only. That alone is just genius if it works.

Keep up the great work.

[kremator 1064]

Pah if anything ZL needs to work harder! MOAR CODING !

Just kidding mate. What you are coming up with in these forums is just genius. If you are gonna burn out, release first ;)

[spanishsurfer 58]

Keep up the good work Zoolo, once you release I'll throw it on the CTI server and see how it compares to asr_ai (which we've been using for months now).

[dronium 10]

Where i can download latest version thanks on answer?

[miketim 20]

Dronium, this is not released.

You shouldve read the comment right before you that said

once you release...

Implying that it is not released yet.

Just for the future, make sure to

never ask for updates

because it is against forum rules,

and I know that was your first post.

Also, if you do not speak english well because you live in a non english speaking country, I suggest marking your country in your profile settings,

this usually will help make people more understanding if you have trouble with english.

Your welcome.

[stoffl 13]

All this is sounding too good to be true :D

One Thing though: do you plan on making certain aspects of it modular as for example the unconsciousness for players since this is already handled by other mods such as AGM?

However I am really looking forward to it. Keep up the great work.

[SteelRain0341 10]

I loved FFIS but this has code in it that is incompatible with Linux. That's a pretty big shame because this is an awesome mod. However, Linux users cannot use this because Linux does not like any code that has upper-case letters... If you could set this up so that it just calls for lower-case codes, it will run on both Windows and Linux.

[aammooss1122 1]

Hopefully your work can complete very soon .

Because this is the best AI mod I have seen for ARMA 3 , but sadly for several times of update , the old version of your mod seems may casue some bug in the latest version of ARMA 3 .

Thank you for all your work :bounce3:

Can't wait anymore for the FFIS 2

[spanishsurfer 58]

How's this project of yours coming along?

[zooloo75 833]

Still debating if I should continue it. The chance of having realistic AI is very slim.

Performance is a killer issue as having complex simulations bogs down ingame performance too much. Would you all be okay with having the same principles of FFIS1 incorporated into FFIS2 with a slimmed down, per-mission data collection instead of per-map?

[miketim 20]

I think if you need to do per mission, do per mission.

But, see if you can make it so we can like, somehow merge files or somthing.

Outside of game, merge our per mission files to make a per map one maybe?

Just an idea.

How feasible is it to implement choices of features (ie: being able to allow certain infantry tactics or abilities, toggle them on off)

likely (assuming) outside of game, in a file or something ?

I could probably give a better opinion if I had more info to go on, clearly you are facing challenges but could you maybe elaborate on those challenges?

I am a programmer too (not for arma) so it kind of makes me want to know what is the problem under the hood :cool:

[HorbeySpector 164]

Still debating if I should continue it.

I understand that you're losing interest, probably because the other projects you're working on take up all your time. But please, don't discontinue this mod. I've put all my hope for a decent Ai in this mod. And without decent Ai, the game is near nothing. We know you've got the skills, just give it a break and maybe a(nother) genious solution will pop into your head! :)

[euly 0]

Still debating if I should continue it. The chance of having realistic AI is very slim.

Performance is a killer issue as having complex simulations bogs down ingame performance too much. Would you all be okay with having the same principles of FFIS1 incorporated into FFIS2 with a slimmed down, per-mission data collection instead of per-map?

The best course of action in this situation is to shelve the unfinished features and revise the project to FFIS1. Once you release the project, people will find problems and hopefully report back to you, which will help you identify areas of improvement and possibly inspire your ideas and methods to re-introduce the features you shelved. Right now, the options are: have a version of FFIS2 compatible with ArmA 3 that resembles FFIS1, or have no FFIS2 at all. Do what you gotta do man.

Edited July 17, 2014 by Lucidity

[sttosin 67]

What about combining your mods into one....toggle features on an off by userconfig or users delete pbo. That way you can realease one feature at a time.

Sent from my GT-P7510 using Tapatalk

[bravo409 13]

when will we be see Fire-Fight Improvement System 2 I have played ffis and the ffis 2 sounds even better so cant wait?

[fn_Quiksilver 1633]

Still debating if I should continue it. The chance of having realistic AI is very slim.

Performance is a killer issue as having complex simulations bogs down ingame performance too much. Would you all be okay with having the same principles of FFIS1 incorporated into FFIS2 with a slimmed down, per-mission data collection instead of per-map?

This is always what held me back from approaching AI coding.

When we're talking about building advanced routines for AI, this is fine with just a few, but when you get 50+ AI running around, thats a lot of eventHandlers and other loops firing, lots of knowsAbout and nearX happening.

In my opinion, you should continue but focus on building a simple, robust and performance-friendly FFIS.

I would also examine the market you are aiming at, for FFIS2, and tailor the development accordingly. You cannot hit all markets. The small 12-player COOP missions will allow you more room to flex the AI load on the CPU. Bigger coop missions with 40-64 players, will have different performance requirements.

[zooloo75 833]

This is always what held me back from approaching AI coding.

When we're talking about building advanced routines for AI, this is fine with just a few, but when you get 50+ AI running around, thats a lot of eventHandlers and other loops firing, lots of knowsAbout and nearX happening.

In my opinion, you should continue but focus on building a simple, robust and performance-friendly FFIS.

I would also examine the market you are aiming at, for FFIS2, and tailor the development accordingly. You cannot hit all markets. The small 12-player COOP missions will allow you more room to flex the AI load on the CPU. Bigger coop missions with 40-64 players, will have different performance requirements.

With that in mind, I should've kept this to heart, "Always code for the worst-case scenario."