AIs behaviour/improvement

[maxqubit 1]

In fact I guess more lifelike AI is not that much magic as you might think. BIS is already half-way there and I´m sure they will be walking this way as close to the end as they can go.

You want something and BIS has to program it for you:)

Don't get me wrong, i love OFP:E and BIS did a stellar job, but they just cannot pull off the impossible.

Imho THE strong point of OFP:E is the balance in thing (of which AI is a part). The balance like stars, wind, reasonable AI, insects, grass, clouds, etc make for a believable world. And that is the direction for Game2, a believable world in which you are a lone soldier.

Now i they want to use this extra CPU power they have to upgrade and extend everything a very little bit to maintain balance. So, snow falls, melts, footsteps stay, dead bodies decay, rats will eat it, birds will fly up if you stumble thru the woods, etc. Because all things interact with each other you will see little specific improvements but you will notice it is 'more immersive'. That's why ArmA looks like OFP but i'm sure it will feel much more immersive when you and i will actually play the damned game:)

For me it is the immersion of OFP:E which is super compared to other games. BIS are masters of balance. A bit better AI in a dead wireframe world would not be my choice.

[maxqubit 1]

It's not that hard to make the soldier move behind the nearest rock in a certain angle so that the rock is in the line of sight of the attacker and then lean out behind of cover to see where the attacker is going - you could say that there are too many rocks in OFP for the ai to not destroy your pc with all the calculations, well, limit the range at which the AI conciders cover, lets say 50-100 meters, if anything is further then that they try to surpress you or surrender.

First, sorry for the bit harsh post.

Second, if it was 'not that hard' wouldn't you think BIS would implement it. Obviously there are issues with this.

I admire the OFP gamers for there love for the OFP world but to say thing like 'not that hard' without programming it yourself, i find too easy. (I always think 'do it yourself better than';)

[NeMeSiS 11]

Don't get me wrong, i love OFP:E and BIS did a stellar job, but they just cannot pull off the impossible.

But VME (I thought..) came very close to what we are asking for, and thats NOT AI WHO REACT LIKE REAL PERSONS, just overall improvement in taking cover(and leaning) and formations  

But it had a big impact on the performance on the game, because the "normal ai scripts" aswell as the VME scripts were running at the sime time.

EDIT: Im just hoping for some small improvements, or a way to turn the standard AI off, but i dont think it will be in

[Balschoiw 0]

Quote[/b] ]You want something and BIS has to program it for you:)

It´s a technical idea and suggestion to BIS.

If there were no suggestions we would still be playing pong.

Quote[/b] ]Imho THE strong point of OFP:E is the balance in thing (of which AI is a part). The balance like stars, wind, reasonable AI, insects, grass, clouds, etc make for a believable world. And that is the direction for Game2, a believable world in which you are a lone soldier.

We play OFP for 5 years now. Of course you get used to features like you mentioned above. That does not mean they are bad, you just get used to it.

This specific thread is about AI improvements, nothing else. We are not debating the envoronmental features here. There are other threads for that.

Of course a game has to feature more than just AI but as it is a vital part for longterm motivation and a vital part for overall game-rating and playability it has to be focussed upon more than other features.

Quote[/b] ]A bit better AI in a dead wireframe world would not be my choice.

Noone sayy that this is what we´d like to see. Processors and gfx cards have updated a lot since CWC was released. Why not take benefit of that not only for eyecandy but for ingame realism and better AI handling ?

Noone plans to take away any features you´ve mentioned.

Games are evolving. OFP is evolving to ArAs and finally Game 2. The process of evolution includes all aspects, but playability and a more sophisticated and real-life-like AI should be on top of the list for anyone who wants to create another addicting game like OFP.

Eyecandy and environment are fast forgotten if the engine is not able to portray good AI behaviour. Remember Doom 3 ? What a blast in gfx and effects, still it rots in my shelve for years now.

Ok back on AI now pls

[lwlooz 0]

Personally I agree with the people who want the single AI or the squad AI improved. I think any higher level of organization can wait for Game2. Besides,I dont think "The community will fix it" will work with Single AI behaviour in an engagement,thats pretty much something for the core AI .What you happily can do tho is affect the larger scale of things (Guard-Waypoint,Scripts -> I may repeat we need a way to find out if the AI got into a battle or had a contact).Also I would happily trade in birds,butterflies for a little more cpu for Mr.AI,because in the end that what in comes down to.OFP's strength is the attempt at trying to simulate the larger scale of things.You are just one little guy in a big fucking war.Not like in all the other games where the whole concept is based around focusing solely on the player(with all the "neat" little details that come with that)

[Heatseeker 0]

If the system requirements rise too much we will have alot of sad people crying that they cant run the game .

I like to use alot of a.i. groups in maps and manipulate their movement/behaviour based on events, its good fun.

A.i. that runs behind a rock sounds simple but its not because nothing is ever simple in this context.

Should a group of 12 units take cover behind the only small rock available in the area? (how stupid would that be?)

Should they hide there forever?

When should they consider it safe to move out?

When should they decide if they should stay behind cover or engage the enemy?

People forget that Armed Assault is just not CoD .

I'd rather have the option to edit it myself and leave the a.i. has is, if under fire=take cover, its something that was present in OPFR (hide) but didnt work, maybe BIS will make it work in Armed Assault .

[maxqubit 1]

Personally I agree with the people who want the single AI or the squad AI improved.

I think nobody will disagree on that one (question is at what costs)

Also I would happily trade in birds, butterflies for a little more cpu for Mr.AI, because in the end that what in comes down to.OFP's strength is the attempt at trying to simulate the larger scale of things.

But i can't agree with this. I prefer immersion hence I would prefer bird-AI which fly up (and give away your position) when you stumble thru the woods over a slightly better soldier AI in a lifeless environment. (While this isn't a soldier AI point it IS a general AI point so imho not really OT)

But anyway, i will go easy on expressing my opinions despite someone who said

Quote[/b] ]If there were no suggestions we would still be playing pong.

;)

Out. (for me)

[Heatseeker 0]

Out of all the OPF veterans posting here nobody considered the TAKE COVER comand already present in OPF, the one that switches the team member action/status icons to hide, it doesnt really work, the a.i. will try but they wont succed and a.i. groups dont use it.

What does it tell us? BIS considered it long before we even know about their existence but were unable to successfully implement it at the time.

So instead of asking them to implement new features into the a.i. maybe we should just remind them of it .

[Guest Ti0n3r]

They removed the take cover command (and many other commands) in Elite, guees they'll do the same for ArmA

[lwlooz 0]

People forget that Armed Assault is just not CoD .

Let there not be a misunderstanding.There aint any other game I know of(FPS anyways) BIS could look to and take note.They rather should base it like all things on military history,facts and manuals.If I can buy silly books and find links or play wargames in order to get a slight grasp on military procedures,BIS can do too.Even better,they have connections to real military type of persons for which they develop training simulations. So,I guess they know how AI has to react, its probably just a matter of them not being able to bring this up to a sim-like level because of VBS or they think gamers just missed fancy graphics in OFP

[GBee 0]

They removed the take cover command (and many other commands) in Elite, guees they'll do the same for ArmA

IMHO they removed those commands in Elite because the menus would have been too long/complicated for a controller interface. I doubt they would remove them from the PC version, especially when the existing OFP community has become used to them.

[DEAD_RABBIT 0]

What has always bugged me with OFP AI, is their incapability to understand the situation and act accordingly. They have so little situational awareness and that's also the hardest to program. AI lack it on both the individual level and on the group level.

It requires a self-made list of a large amount of applicable tactics (tactics that can be applied dynamically) that are being used when certain conditions are met.

If I would program an AI with more situational awareness, but maintain the dynamic aspect as well, so I can apply the same system onto higher levels of AI (individualtactics/group tactics/platoon tactics), I would do it this way:

DYNAMIC SYSTEM LOGIC:

-1) Condition: CONTACT?

An AI observes the area. When the AI 'sees' an object that it cannot identify, condition is set true.

-2) Action: (RE)GATHERING INFORMATION FOR TACTICAL USE

The AI observes the object and creates an array of values about its terrain, weaponry, physical condition and the information of the contact of course.

-3) Action: SELECT TACTIC OR CREATE CUSTOM TACTIC

With this array, a syntax can then select a tactic from a large list of tactics. This is actually very simple in mind. Each tactic in the list has conditions to which the array values have to coincide. These conditions can have for example a wide or short period the corresponding array value has to meet. You can also combine these array values for complex conditions.

A different way can also be for the syntax to produce a stepping stone plan based on the values it has been given. I prefer this approach to some degree, because it enables you to add an immense diversity to the AI. Or you could do both.

-4) Action: EXECUTE TACTIC

With the chosen or custom created tactic, the AI is now going to do all the things it has to do according to the tactic. What happens, is that with each step of the way, the step contains simple 'to-do' things, like "move to bush at 'position' without being detected" results in an action being performed by the AI in question that conforms to the conditions of the action. If the AI completes the action without failing one condition, it proceeds with the next action.

However, if the AI fails because for example, it was detected by a patrol, the program has to update the array of values and rewrite or reselect the tactic based on this new information of the new (unexpected) situation. The fact that the AI has to know that he (or she) is detected also plays a role in being stealthy. You can't be stealthy if you don't know if the enemy knows where you are. Therefor, it is an intricate part of stealthy behavior, besides lying on the ground and hiding in bushes. This same philosophy should be applied on all actions. That they have to be complete, in the sense that with each action, there is a guideline for the AI to follow.

-5) Condition: 'LOOSING' is true : goto STEP 2

Even when all the 'to-do' things have been completed, the AI has not yet completed the 'objective' (in this case, staying alive by being stealthy). The LOOSING condition is a separately determined by a condition of the last action from the tactic.

-6) Condition: 'VICTORY' is true : goto STEP 1

Similar to STEP 5, except that it is activated for the opposite reason.

[berghoff 11]

They removed the take cover command (and many other commands) in Elite, guees they'll do the same for ArmA

I don't think they copy everything :P from OFP:E as it was designed for a controller and scrolling thru too many commands just takes too much time in battle.

[crashdome 3]

If BIS would just expose the appropriate values for scriptors and add-on makers, we could easily improve the AI to atleast an acceptable level.

In fact, we can do it now but the nature of scripting and doing it for hundreds of units causes intense CPU usage.

Functions provide a bit of relief and a simple library of basic commands to set values such as skill, target, and knowsAbout dynamically would do alot IMO. The AI appears to be something of a state-machine (although I cannot verify this obviously) and simply allowing us to change states and override certain features would be incredible.

For example, the DisableAI command has great potential if it was a little bit more flexible... could I create my own add-on and turn off Targetting AI for my unit (leaving movement.. etc) and run my own functions instead to handle targets?

Something like that would be incredibly flexible to add-on makers because we could actually make units that are specialized actually be... "specialized" and not merely different looking. Stealthy recon units that don't fire so aggressively and run away when facing larger groups... Machine Gunners that targetting multiple units within small increments of time (not just laying 100s of rounds into one guy)... etc...

We can do it now.. it's just heavily scripted because there are numerous workarounds to what would normally be simple commands (like the ability to fire in a specific direction without needing a target, etc..).

Conclusion of this post is, if BIS would just supply some basic and much needed commands that eliminate a majority of the workarounds we use, we would be golden. However, because BIS never reports any new features in terms of scripting we have no way of knowing this could already be available.

[bn880 5]

Christ, these forums are sounding more and more like the Fanboi Falcon 4 Allied Force ones on SimHQ.

A person can't even make valid and important suggestions/criticism without putting up with a load of BS coming backtheir way.

Current AI is totally inept for fast ... anything. Only thing AI is good at is some manovering and good marksmanship. Face it. Try going up against 12 humans versus 12 AI one day, you will see. Let alone in a city complex.

[crashdome 3]

A person can't even make valid and important suggestions/criticism without putting up with a load of BS coming back their way.

Uhhhhhhh....... am I missing something? Who are you directing this towards and about?

[EDIT] btw..I'm just a little lost between the two threads covering this same topic... not trying to suggest anything by this question.

[benreeper 0]

That's what I would like to know also Crash.

Max you have my sentiments exactly. People will accept less from games like COD, BF2 et al., (whose graphics aren't even better than and the gameplay is worse ) but demand perfection from BIS and when you point that out they call you a "fanboy". I guess I;m a fanboy then. I mean COD's AI is heavily scripted, BIA's AI has narrow paths to follow and there is usually only a max of 16 of them in the game at once and BF2 has no AI AT ALL!

--Ben

[Friedchiken 0]

Well, I guess I can summarize my views about AI with this:

I hope the AI can mimic real combat situations as well as it can.

Basically it's up to BIS to decide which algorithms and strategies to use in approaching this goal.

While you can critique and art style or data of research, I guess it is hard to make very useful suggestions for programming something difficult like AI. You could suggest a certain way or method of doing things, but in reality maybe they already tried that approach and are having difficulties adjusting to the funny choices an impersonal program will make.

In the end, I would just encourage more play testing to see what important AI issues need to be fixed. It's sorta about improving the things that are great, fixing the things that are broken, and ignoring the mediocre parts until they are seen as broken.

[sniper pilot 36]

idk i hope the AI will be modable somehow...

[Bouben 3]

Guys AI will be different from demo what we saw! This is our hope!

[Zendjir 0]

Quote[/b] ]Guys AI will be different from demo what we saw! This is our hope!

Yep, let's hope for the best.

[lendrom 0]

Well, according to hardrock our hope can be now backed up by Rastavovich's (a BIS employe) statement:

Quote[/b] ]The AI wasn’t that impressing in the demo, but Rastavovich assured us that this will change for the final product.

One of the most important news from the hardrock's article IMHO.

[mr.peanut 1]

A couple of comments:

1) The AI need not act realistically as much as they need to appear to act realistically. There is a difference.

2) The cost benefit of improving the AI is small. A lot of money and time can be spent researching it without a guarantee of improvement. Graphics tend to impress the plebes more than AI, as long the gameplay is decent.

3) Computationally, the cheapest way to deal with individual AI behaviour is statistically. For example you can calculate every line of sight for a single AI with high precision to see if it spots an enemy (expensive), or you can calculate a probability that an AI will spot an enemy based on various factors such as distance, visibility, terrain, direction AI facing (cheap). For each factor for which you want a single AI to respond, there is a significant increase required in computation.

4) I'm not sure that BIS can improve the ArmedAss AI much beyond what has already been done in the OFP community by scripting, without seriously affecting the minimum system requirements.

[deanosbeano 0]

A person can't even make valid and important suggestions/criticism without putting up with a load of BS coming backtheir way.

Lets debate and discuss arma not discuss what these 4 or 5 guys think we should discuss.

[Friedchiken 0]

You know, you could just ignore such threats and publish your opinions.

What can anyone do? Stab you over the internet?

I'm sure that since your posts can't be casually deleted, someone at BIS will read them.

Personally I'm under the camp that says beta nitpicking is a useless endevour, but I encourage you to post away. If anything, it saves me time from not needing to post.