team AI discussion

[DayGlow]

Now that I am playing the last game more and more I am seeing the warts that others dicuss. The AI handles small to medium rooms great, large ones not so well.

But at the same time, AI is always an issue with any game. I can't think of any game that I've said, 'now that's done right'. The SWAT series is the closest, but even it has it's own issues and hicups while playing.

First off one idea I have is your own team composition. I have no idea if R6:5 will have planning and multiple teams (I will always avicate it though), but I think the single teams used should be upped from 4 to 6 including the player. This way when you are working with your team you can split it into 2 element + yourself and your cover.

One of the major problems with any CQB game is how do you involve the player? The SWAT series every AI has an assigned function, the player just watches, Same with LD. What I think should happen is that there should be a training area that gives the player a tutorial on the different functions in the stack, ie number 1 through 6 and their responsibilities.

Make this position dynamic. If I tell the AI to clear a door and I'm standing where the 3 man would be, they fill out and do their job and I'm expected to do mine. This comes important if the position is either has to open/breach the door or toss in the gernade. The AI will expect the player to do it's job for a successful breach.

Any other ideas on how to make the AI smarter in the evironment and inlcude the player in the actual clearing of the room without having to do the rambo thing?

[Yen Lo]

I still think Far Cry has the best enemy AI, although alot of people like FEAR's AI. SWAT still has best freindly ai. Though none of the games ai is perfect.
You should get The Regiment, the training alone is awesome, will make you better in other games too.

[DayGlow]

I liked the awareness of teh FarCry AI, I didn't like it's contastant shoot while strafe hit you every shot behaviour. Also the difficulty in the game ramped up by making the enemies take more damage. It got to the point near the end of the game that 1mag=1kill. I got really tired of that.

[TexasRanger_562]

Maybe they should consult some SWAT experts, and see how training for CQB is actually done. Then program that into the AI.

[Woosy]

The Regiment AI works like that DG though they do it in pairs. Tell them to breah with flash, you're standing at the side of the door they will open flash it, and will take a left arc depending on what side of the door you're on and you cover their blind spot. They arn't perfect but the concept they use is fab! It keeps the player involved and includes the AI.

In SWAT4 if I stood at the side of the door like that, a team mate will say "you're in my spot" and you have to basicaly watch them do the job, which wasn't fun.

[Woosy]

Originally posted by TexasRanger_562:
Maybe they should consult some SWAT experts, and see how training for CQB is actually done. Then program that into the AI.

Special Forces experts not SWAT, the tactics are simular but not done the same. I think most american CQB tactics are pretty much the same looking at SWAT to Navy Seals there isn't any difference, though they get some leeway, other countries special forces, are very agressive and really fast on room clearing drills.

[DayGlow]

Originally posted by TexasRanger_562:
Maybe they should consult some SWAT experts, and see how training for CQB is actually done. Then program that into the AI.

Most games do have a subject matter expert, but the problem is breaking it down to something the AI can do. Much of room clearing drill is encountering setups you've never seen before and problem solving how to clear/cover while doing it.

Also you'll never get a full level of authnetisity as people will not talk about actual tactics to people outside of the 'club'. Even watch the Dallas SWAT show. They never show specific tactics or ever show them inside clearing. Part of it is safety, other is trade secrets.

[Woosy]

There is a video out DayGlow where they are showing off new and old swat Tactics, the guy who uploaded them in my youtube group has the rest of them, I may ask if he can upload the rest. It's just a question of finding them i guess, also there is a few on google.video.com aswell.
Video 1 (http://www.youtube.com/watch?v=BezXuNIe0Ew)
Video 2 (http://www.youtube.com/watch?v=x1Lpb2evvbI)

[DayGlow]

yup, I've seen those before. To go beyond that though for an AI to use is how do you define what to do when lead encounters a threat in X quadrent and engages, what does 2 through 4 do? Much of CQB is reacting off of each other and knowing what job needs to be done. If someone else is doing it, move on to the next job. Look at the first video, the wall flood, what would the team do if #1 encountered a closet along the wall. All of a sudden he has to cover that threat while everyone else adjusts to cover the rest, then someone has to move up while 1 clears the closet, etc. This flow is very hard to 'program' because it's so dynamic.

Most AI's are very good at initial entry, it's going beyond that, which I don't think AI is there yet, so how do you fix that?

I need to pick up TR to see how it works and everything as well.

[KungFu_CIA]

When talking about AI, specifically, CQB and Squad AI in general... The big thing game developers have to balance is not only how to immerse the player, but the fact most video game players are not trained CQB Operators (or soldiers) and this is where a lot of the "watch them do it for you" comes into play more than anything else in my opinion.

Close Combat: First To Fight (both PC and console) tried the approach DayGlow, suggested above about how the AI is dynamic enough to adjust to where a player stands and stacks up for a breach.

The problem is, again, most people who play video games, regardless of chronological age, are not remotely trained in any real tactics whatsoever... So, as the case in CC:FtF, they would just stand back in the number four/five position and let the AI clear the room for them without any input except to flank and cover them like most CQB games involve the player.

I think the closest thing any video gamer being "trained" in real world tactics who are not in the military or law enforcement/security services are those who play real world games like Airsoft and similar games which try and emphasize real world entry tactics and other manuevers... But this is still a small minority when compared to those who play video games in general, though.

SAS_Shield also brought up this important factor on the Council Forum in he said his clan/group tries to use real world tactics... But he realizes there are those who don't know the first thing about real world tactics and are used to playing "Rambo" style because most other games encourage this kind of play as well as most of their "experience" in tactics is what they see in TV and movies which is often far from reality as well.

I think there has to be two levels which exist in the AI in a game like R6:

One for the "casual" player -- Casual as in the most common example of the average, untrained player who doesn't know a thing about real world CQB entry and pretty much does everything for you like in SWAT 4...

...AND...

A second level similar to the AI in CC:FtF and The Regiment that allows a semi-trained player to gradually learn what their role is in a dynamic breach and how they should carry it out. As stated, a specific tutorial/training section has to be included for this as well.

Woosy also brings up a very good point in hopefully the developers will decide on which direction R6 is going to portray -- Special Forces, or SWAT -- And research (hire consultants) this area and stick with it because the two professions are very different in how they operate (and types of operations) to accomplish the same goals.

Not to keep harping on the past, but this is one of the major problems with Lockdown.

LD is more of a Ghost Recon, long range game with CQB elements thrown in... Which is the exact opposite of what R6 was envisioned as (modeled after FBI HRT teams and tactics) when Tom Clancy and Red Storm first produced the series and which focused on CQB and anti-terrorism tactics.

[DayGlow]

as much as the hardcore players hate them, how about shinny floaty things to show the causal player where to go to praticipate in the stack that can be turned off by the people that don't wan them.

ie something to indicate 'you're # 3' and when the breach happens arrows or markerss to show which way they are expected to go, then what arc they are covering?

[DreamMarine]

Very good thread! http://forums.ubi.com/images/smilies/25.gif

actually i would love to see a game which would gradually "teach" me how act as a member of a "real" hostage rescue team. this might be done by extended training maps with grading (to know how well i did) or - as DayGlow is suggesting - by certain hints in real maps (which you can turn off for more realism).

i must admit that i don't know much about programming AI. but it sounds plausible to me, that it's hard to program an AI which acts like real HRT when the player is playing rambo! ^^

but actually, i believe, that a lot of people who play R6 do this, because they like that feeling of beeing an HRT memeber (or SWAT or GSG9 or whatever).

it's all about balancing and finding a middle course. give them the feeling, that they learn to become a real "hero"... but don't ask to much of them!

OMG, i REALLY would love some kind of "close to real" HRT training in R6:5. http://forums.ubi.com/groupee_common/emoticons/icon_wink.gif

DreamMarine

[Relenquish]

None of the other Rainbow Six games reflected the pace specail forces move at. And to do so IMO would completely change the game. Do we want that?

In terms of the AI, I would love to see the game trying to teach people the basic tatics. A tactical shooter should IMO encourage people to use tactics, not just watch the AI do it.

To be honest, I dont want an AI character 1st into the room. The vast majority of the time I want to be in there 1st. I want to be doing the work. This is where I love the regiment. The ai follow you in and covers. It does its job without being spoon fed.

Maybe if more tactical shooters took this approach the following of this genre would increasingly become more capable.
What i feel important is the AI being consistant. If people dont trust the AI they wont use them. I never used the RVS AI to clear a room (Well I did once, and i learnt the leason they werent good enough).

I want to see the AI being good enough, and dynamic enough so the players have the option of using them. And training, even more detailed than the regiments. This would give the player the choice if he wanted to learn or not.

Its about time they started making tactical games that make you think.

Things like the game recognising your position in the squad, and adapting accordingly would be awesome. But when it comes to squad positions, you have to make sure the player knows his role as well.

Training should probably include a section on each position if they go for this.

So in short this kind of AI will require extensive training, and a lot of TLC from the devs. But imo if its pulled off it will be ground breaking.

[dave819831983]

Yep I agree with the original post.

I hate only being able to watch my team mates clear the room or do it myself (more or less).

The idea of taking up a position in the stack sounds great, although your right we may need floaty symbols. Especially because around confined doors (e.g. a corridor) the positions that the 1,2,3,4 etc man stands may be different to a door with loads of space around it. (play swat 4 to see this).

[Yen Lo]

Originally posted by DreamMarine:
Very good thread! http://forums.ubi.com/images/smilies/25.gif

actually i would love to see a game which would gradually "teach" me how act as a member of a "real" hostage rescue team. this might be done by extended training maps with grading (to know how well i did) or - as DayGlow is suggesting - by certain hints in real maps (which you can turn off for more realism).

i must admit that i don't know much about programming AI. but it sounds plausible to me, that it's hard to program an AI which acts like real HRT when the player is playing rambo! ^^

but actually, i believe, that a lot of people who play R6 do this, because they like that feeling of beeing an HRT memeber (or SWAT or GSG9 or whatever).

it's all about balancing and finding a middle course. give them the feeling, that they learn to become a real "hero"... but don't ask to much of them!

OMG, i REALLY would love some kind of "close to real" HRT training in R6:5. http://forums.ubi.com/groupee_common/emoticons/icon_wink.gif

DreamMarine

what you want is the game called The Regiment, itll teach exactly what and how to do it.

[DreamMarine]

Originally posted by Yen Lo:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by DreamMarine:
Very good thread! http://forums.ubi.com/images/smilies/25.gif

actually i would love to see a game which would gradually "teach" me how act as a member of a "real" hostage rescue team. this might be done by extended training maps with grading (to know how well i did) or - as DayGlow is suggesting - by certain hints in real maps (which you can turn off for more realism).

i must admit that i don't know much about programming AI. but it sounds plausible to me, that it's hard to program an AI which acts like real HRT when the player is playing rambo! ^^

but actually, i believe, that a lot of people who play R6 do this, because they like that feeling of beeing an HRT memeber (or SWAT or GSG9 or whatever).

it's all about balancing and finding a middle course. give them the feeling, that they learn to become a real "hero"... but don't ask to much of them!

No! Not so far i have heard about it. In The Regiment I am forced in continuous fast action. Though that might be actually the reality, I very much prefer the gameplay of R6.

A good mixture of sneaking, breaching, sniping and rushing into the rooms, where hostages are held would perfectly work for me. http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif

Besides, I love the overall feeling of R6 and the really great maps in the past!

OMG, i REALLY would love some kind of "close to real" HRT training in R6:5. http://forums.ubi.com/groupee_common/emoticons/icon_wink.gif

DreamMarine

what you want is the game called The Regiment, itll teach exactly what and how to do it. </div></BLOCKQUOTE>

No! Not so far I have heard about it. The regiment forces you into doing everything with high speed. Though that might be reality, I don't like that.

I prefer a good mixture of sneeking, sniping, breaching and finally rushing into a room where hostages are held. The gamesplay of R6 always felt just right. Of course, now it's time to continue and extend the original idea a bit.

[Woosy]

The thing about the Regiment with regarding to AI is that you can decide who goes in first, but who ever does you need to trust the AI to clear their side of the room. It still involves the player to watch the AI's blind spot. It's not so generic and boring and you get the feel that your AI buddy is doing the job for you, as it's a two man job and there is kills for both of you.

DreamMarine I think you can do a mix of all, what The Regiment does is it makes you breah the building loud, as talks have ended, and to be honest there is no stealth option, if there was it would lose it's challenging gameplay. That said, that doesn't mean you couldn't have a stealth mission in R6-5.

What you have to ask yourself is do I want real world tactics and if so do i comrpomise them for another combat tactic? Certain situations call for different tactics, stealth, assault, sniping duties etc... As not every map requires a sniper, not all maps will require a full assault either but stealth. But... if that stealth is compromised, you need to change your tactics up a notch to full assault, which would be hard and fast CQB to adjust to the new situation. SWAT3 does this very well. Having a variety of maps for different situations, for a range of approaches for a map would be better then compromising real tactics. That then gives the player the option how to deal with the threats, if thats assaulting stealth or the odd sniping.

What I think would make a difference, back to that TR again is having the training level kinda like theirs, you can do it on co-op and learn how to trust your Ai/Human partner in, room clearing, target acquisition, the use of real world tactics. If they do that for each thing, like a sniping partner is the spotter, stealth the use of stealth and not being caught and full assault hard and fast CQB, it would give everyone new abilities and give them training scores where they can try again and again to get a better one, it's not only fun but gives them a vast range of knowledge at the same time.

[Yen Lo]

Stealth LOL. Oh man. Still only comes down how you shoot. If you dont out shoot the competition people are gonna die, and you lose the misson. No game trains you better than TR PERIOD.
BTW NO, the game isnt perfect, yes tha AI still has its problems, and at least the Embassy map the rooms are pretty tight. Still its a heck of a challange.

[Vert22110]

As far as implimentation in team stacking goes...

I recently watched the MGS4 video, and in the beginning you see (from the FPS perspective) a squad of soldiers stack up and round a corner. The player in your view moves behind the soldier infront of him and taps his left shoulder. Apparently, this signals the go code and the two soldiers advance. It was a very short, very realistic, very COOL scene in the video (snake later owns the soldiers, hehe)...

Now, for RB6. I think something similar to this would be effective. The player should be involved in the breach, not just standing and watching.

This could be done by having the player issue a stack order. Once the team is stacked on the door, the player can move into one of several "Slots". This slot could possibly guide the player by drawing an on-HUD path for movement. The player then issues the order to strike, and he follows this path through the issued breach and clear order. The AI behind and around him would need to react realisticly to the player's actions. This means, if the player takes fire, then the AI teammates adjust their flanks, fire back and then readjust. Or should a team member die, the AI adjusts and fills in the missing slot, adjust flanks, etc.

This path line of course will be optional.

Now, this does not mean the player is locked to that path. He is guided by it, but not locked. He still has the freedom to walk out of it, or not join in the breach at all. Either way, the AI needs to react appropriately.

As you can see, AI is very important...

[Yen Lo]

[QUOTE]Originally posted by Vert22110:
As far as implimentation in team stacking goes...

I recently watched the MGS4 video, and in the beginning you see (from the FPS perspective) a squad of soldiers stack up and round a corner. The player in your view moves behind the soldier infront of him and taps his left shoulder. Apparently, this signals the go code and the two soldiers advance. It was a very short, very realistic, very COOL scene in the video (snake later owns the soldiers, hehe)...


Yes that ROCKED!!! And its in a Console game for a platform thats still a ways away. Why cant we have this in this series. That video is awesome Ima go watch again lol.

[Imparfait]

The resources are out there, you will never get fully current drills, and SOPs which are current will not be included in any AI programming, but a lot of that stuff pertains mostly to real world situations.

They could benefit from some more consulting in the general dynamic/stealth entry drills (tactics is a word used poorly on these forums btw) however what they are showed is limited, largely because it is in an official context.

There is all sorts of stuff circling around the internet, some is good, some not, some new some old.

SWAT is very different from SF CT work.

[KungFu_CIA]

Originally posted by Imparfait:
(tactics is a word used poorly on these forums btw)

I know the community will jump down my throat for saying this, but I agree.

The way the word "tactics" is used in this forum and through out the R6 Community is like this Holy grail of what other games should somehow aspire to be... When in the end, most of the "tactics" being talked about are video game, MP tactics and not real world dynamic entry, or remotely realistic military/SF/Law Enforcement tactics in general.

However, the reason I said this wasn't to start some massive flame war, or hurt people's online egos...

But to re-emphasize the points I, SAS_Shield/Relinquish, and other people have already brought up with regard to the AI and that is how most video game players are not trained SF/Military/Law Enforcement/Security Operators and are just video game players and are still using video game tactics in these games because they don't have any other frame of reference -- Whether want to admit it or not, or don't even know it subconsciously -- And this does effect how game developers program the AI in a lot of instances.

This also crosses over to the thread I just started HERE (http://forums.ubi.com/eve/forums/a/tpc/f/3801065024/m/6261043224),and another one started by DreamMarine HERE (http://forums.ubi.com/eve/forums/a/tpc/f/3801065024/m/3111093124) as these issues seem to be inter-related and also one of the key issues currently facing the R6 series as a whole.

[GSG_9_Rage]

this is my take on the whole SWAT or Special forces debate. R6 was for more CQC/SWAT style gameplay. if you wanted larger more open type of maps, then you would play GR. Those two games were very similar in the core, but the playing enviroments were very different. thats what made GS and R6 such a great pair.

[Imparfait]

As much as some might think otherwise CQB was pioneered by SF, most notably the SAS. Fast, agressive, adaptive entry designed to quickly eliminate the enemy with little to no loss of life to friendlies.

SWAT has adapted many of these techniques so as to aprehend criminals and minimise loss of life to all parties.

Rainbow is NOT a SWAT team, they are not out to arrest individuals, they are not described as a supplemental police force. They provide the means to terminate hostage situations through force, ie eliminating the bad guys at high priority when a government's local forces aren't sufficient.

A large part of the Tier One SF role is Counter Terrorism, specifically hostage rescue. This tends to focus largely on urban operations, and CQB.

Though SF aren't localised mainly in urban locations, most of the wilderness operations are undertaken by lower tiered SF units.

If Redstorm/UBI wants to truly make an accurate game (based on Tom Clancy's pre defined role for rainbow) they would profit from consulting with as many ex SFOD-D or SAS members as possible.

[Cpt.Cordalez]

Problem is, R6 does not deal with military objectives. They won't go to get secret documents, to spy out enemy positions or alike.

They are an international unit dealing with situations were local police forces can't deal with, or own SF's can't do the job due to whatever reasons.
It's not like a country would hire R6 to handle their military operations. Own SF's would do that job, but the reason why these countries still can call R6 for hostage situations is exactly because R6 is a unit, made of ex, or active SF's, that excessively train for CQB hostage rescuing stuff, or sabotaging torrorist stuff. And I speak about torroristic aggressors/offenders, not a "going to Afghanistan and hunting Taliban" style tasks or so.

Because R6 aren't local police SWAT, who work only in their hometown, R6 needs to be an SF unit, but it IS NOT a military unit, with military tasks. NO.
Sure SF's deal with hostage rescue situations, just like R6, but not otherwise (no military operations on the side of R6).

That's why you will find R6 on tasks in foreign countries, like in the destroyed Kosovo city. It is a military conflict area, but R6 isn't part of that conflict, they have just their hostage rescuing objective.
Usual SF's would do the military type tasks there, would be part of the war, R6 not. So R6 aren't the usual SF's.

Also you can find R6 doing jobs somewhere in a jungle, desert, or wherever, but not performing usual SF tasks, but their own type of work they are "designed" for.


R6 is simply its own definition. Same as SWAT and SF's share same CQB techniques, same does R6, but the tasks aren't comparable in any way.


Just remember, in the book, R6 is designed to deal with the international Terrorism.
Their working ground is beyond any international military conflicts, even if military deals with Terrorism like the current conflict with the middle east, you still need to differ between active Terroristic attacks (1), and a military/political activity against Terroristic Groups and Countries that support them (2).
R6 would be the first, military SF's would be the latter.

[DayGlow]

but in reality then they would hardly ever do work. It wouldn't make sense to fund a mutlimillion dollar unit that would never work.

Same thing happenned with the FBI HRT team, they were originally created to deal with major hostage situations in country that the millitary would/could not touch. They were never operational. They expanded their role to do more work.

I would see R6 doing the same. They wouldn't work only once every few years, I see them as a Delta equivalent for NATO. They don't do the hearts and minds work, but a lot more tactical work then just hostage rescue.

[Cpt.Cordalez]

More tactical work, yes I think. SAS style.