Single Player Campaign

[Grobut]

Normally, I would take a decent playable AI as a given for any game on the Unreal engine.

However, I played RO and I have to say this proved to be the exception to the rule.

Standard UT AI does well for the kind of game UT is, or mods that play very much like it, you know, run around and shoot stuff, pick up guns, ammo and powerups, your typical Arena-shooter stuff, that's what they where coded for.

RO:Ost just reused that AI, but it's a very different kind of game, that should be played very differently, so yeah, they outright sucked at the game.

That was a budget thing really, Ost was made by about 10 guys on a microscopic budget, and the game really had no SP portion, just the practice mode, so they chose to ignore them, and spend their resources elsewhere, they really didn't touch the AI (they where just as bad in the mod, and it's probably the exact same code in use).


But they are touching them this time, since the team is much larger now (30+ people), the game has an actual budget behind it, and it will feature an SP portion, so it's a priority this time around.

Problem is though, we haven't seem them in action, not propperly, we saw a glimpse of it in one of the very first videos, and it actually looked promising enough, we saw them take cover and fire from concealment and such, a vast improvement over the headless chickens of RO:Ost for sure, but we didn't see much, it was all WIP, and they haven't really been showcased since then, they are allways set to super-easy mode in all the following videos.

So i'm not quite sure what we can expect of them, just that they will be better than they where in Ost, but by how much? I really haven't a clue..

Personally, I hope that the soldiers in those squads won't just be faceless AI. It's quite crucial to attach emotion to the AI in the SP campaign, to really make the player feel attached to his soldiers. In Jagged Alliance 2 for example, I can become very attached to my units, because some of them not only have great abilities but also have a high cost as well as a unique personality. So if my best mercenary dies or leaves because his contract ends, it sucks. I hope the AI in RO2 will be represented somewhat similarly. This is a game about one of the bloodiest battles in human history, after all.

Eh, who knows, it's possible they will try to do that via voice acting and scripted moments, or they might not, we won't know till we know

Either way, i'm not too fussed personally, it's allways nice to add some charecter depth, but then again, the R6 games never really did that (not ingame, there where some charecter bios in the selction menu, but that was it really), and it didn't bother me that it wasen't there, the combat stood on it's own without it.

 

[213]

ok, the whole emotionally attached to your squad thing is for melodramas. i think if possible, make the characters unique. so they have their own names and you can help them survive the battle and use them in the following ones...so maybe they can gain skills and become better. the stuff about heart moving speeches really belongs in a movie or a game that takes itself way too seriously.

 

[baff]

The AI from Unreal worked well enough for America's Army and Ravenshield.
Not to mention SWAT 3 and thr Regiment.

These were both significantly better than RO's AI, which were poor. So we have a good idea what they are capable of if they get enough love.
But the company hasn't earnt any positive reputation in this department yet.

I've given them the benefit of the doubt once before and next time they will have to demonstrate it in advance.

That said I am quite hopeful. I think it's a realisticaly achievable goal for them. But honestly I can't predict if they hired a decent AI guy or if they intend to spend half of their development time on SP.

So I don't have any high expectations, but I do have high hopes.
The title has an awful lot of potential.

 

[Frostedfire]

Personally, I hope that the soldiers in those squads won't just be faceless AI. It's quite crucial to attach emotion to the AI in the SP campaign, to really make the player feel attached to his soldiers. In Jagged Alliance 2 for example, I can become very attached to my units, because some of them not only have great abilities but also have a high cost as well as a unique personality. So if my best mercenary dies or leaves because his contract ends, it sucks. I hope the AI in RO2 will be represented somewhat similarly. This is a game about one of the bloodiest battles in human history, after all.

aah ja2, I'll have to admit I was a massive save-scummer in that when people died, especially when the ai did some ninja cod4pro stuff ie walking in front of all my mercs, not suffering an interrupt and sniping a guy prone on top of a building, wearing ceramic plates and bug-goo, hitting him in the face with glazer ammo and making his head asplode. How many 1's did I have to roll in a row to get this to happen

The AI from Unreal worked well enough for America's Army and Ravenshield.
Not to mention SWAT 3 and thr Regiment.

it's less that the ai was translated over, more that aa/ravenshield etc had some people build actually decent ai to play, rather than the "budget version" the dozen-or-so people at twi did when they made ro. mind you, they're lethal buggers if you turn the difficulty up to party fanatic, get one in a half-trak with you and drive around on eg hedgehog

 

[Zetsumei]

Personally, I hope that the soldiers in those squads won't just be faceless AI. It's quite crucial to attach emotion to the AI in the SP campaign, to really make the player feel attached to his soldiers. In Jagged Alliance 2 for example, I can become very attached to my units, because some of them not only have great abilities but also have a high cost as well as a unique personality. So if my best mercenary dies or leaves because his contract ends, it sucks. I hope the AI in RO2 will be represented somewhat similarly. This is a game about one of the bloodiest battles in human history, after all.

I dont want the game to go the bia way as that doesnt make me attached to players on my team.

But indeed if players in your team can gain in ranking and skill over time, and gain special abilities or special weaponry. Then you will hate losing your old guys as you'll get a newbish fresh recruit in return.

Overall that makes me most attached to my squad at least that was the case in hogs of war.

 

[Grobut]

The AI from Unreal worked well enough for America's Army and Ravenshield.
Not to mention SWAT 3 and thr Regiment.

These were both significantly better than RO's AI, which were poor. So we have a good idea what they are capable of if they get enough love.
But the company hasn't earnt any positive reputation in this department yet.

Neither of thouse games used stock UT AI, they made significant changes to their AI code to make them work well, they developed them.

RO did not, it wasen't in the budget, it used allmost stock UT AI, with just a few hooks of code in there to make them understand how to shoot the RO guns and choose a random RO class before game start, but that was pretty much it really, everything else about their combat behaviour, navigation and priorities was just stock UT code, and it didn't work well at all for the kind of game RO was.

So what we saw in RO is not indicative of what the engine can do with AI, nor is it indicative of what TWI can do with AI, sadly, what we saw was a gaping budget hole, code that never got done.

I've given them the benefit of the doubt once before and next time they will have to demonstrate it in advance.

That said I am quite hopeful. I think it's a realisticaly achievable goal for them. But honestly I can't predict if they hired a decent AI guy or if they intend to spend half of their development time on SP.

So I don't have any high expectations, but I do have high hopes.
The title has an awful lot of potential.

I'm pretty sure the HoS AI is developed by John Gibson, the president of TWI, and the man most certainly knows his way around code.
But other than that, i know as much as you do, so we'll just have to wait and see it in action.. hopefully they will make a video to show off the SP at some point.

 

[baff]

If the AI code is developed by the president of TWI, that bodes ill.
Presumably he was the president of TWI last time round and he didn't choose to "budget" for any decent AI that time.

The idea that the head of a company is an AI guy and the game didn't have any decent AI is a bloody aweful embarrassment. (I can't easily believe it's true).

I'm sorry but you don't get any kudos for the things you haven't done well.


Many other budget Unreal titles had playable AI.
The basic stock AI is every bit capable of providing decent sport. All the games I mentioned used it. What modifications do you need to make for a tac shooter AI? Run speed. That's about it.
Unless you have any special equipment you want them to use.. a heart beat sensor perhaps? A shrinker ray?
Otherwise it's just bog standard Unreal AI. Nothing that hasn't been done a hundred times before.

it's less that the ai was translated over, more that aa/ravenshield etc had some people build actually decent ai to play, rather than the "budget version" the dozen-or-so people at twi did when they made ro. mind you, they're lethal buggers if you turn the difficulty up to party fanatic, get one in a half-trak with you and drive around on eg hedgehog

The key to using Unreal AI is all in the map design. The pathways you label, the cover nodes you designate. You build the track and the AI will run on it. You don't have to mess with the AI code, just design maps to support them.


In RO, they didn't give them any random spawns or alternate pathways, didn't assign them any cover nodes etc. The code was all there in the engine still... they just didn't make use of it. Time, money, ability? I don't care why, only that they do it significantly better this time tround.


The AI in RO may not be indicative of what Unreal can do with AI, but it is highly indicative of what TWI can do with AI.
In fact it is their only track record of what they have been able to do with AI thus far.
Will there be another "gaping budget hole"?


Likewise, I've got my eager eyes wide open for more info.

I prefer not to raise my expectations up for titles so high that they can't live up to my dreams for them.

So I'm not getting all dribbly about AI that moves in formations and takes all the best covering positions. Or gun groups that set up in the best spots for guns or tank crews that bailout and fight on etc. Snipers that look for top sniping spots and relocate when they have been spotted...
You could do all that... but you can also make a pretty good fun mission without all that too. Ravenshield didn't have any advanced scripting. They just had the basic pathways drawn on the map and placed the AI with some start point randomisers.
Simple fun.
Think R6 Vegas. Simple dumb fun. It still makes for pretty challenging gameplay.



I most certainly will not be giving them any benefit of the doubt, however.
They haven't earnt it.
Quite the opposite. They have earnt a goodly amount of skepticism and mistrust from their last contribution.

I just feel it's something anyone can do on that game engine. Entirely achievable.
So here's hoping!
I surely want them to succeed, even if I'm not yet willing to bet any money on them doing so.

 

[Floyd]

Am I missing something?
Wasn't RO:Ost promoted as a multiplayer online game?
Where/when did RO:Ost say it included single player anything?

RO:Hos on the other hand.....

 

[Bluehawk]

Roost had Unreal Tournament's AI, identical to the mod's. I don't think they spent a single cent on it.

 

[baff]

Unreal Tournament was the top AI of it's day. It kicks arse.

You couldn't have bought a more expensive one if you had tried.



There isn't any need for a more advanced one than that in game of this nature in my opinion.

That said there is always room for unit formations. Convoying and all that really clever (system hungry) stuff, but I'd still enjoy it with basic UT AI.

 

[Bluehawk]

Not when you fail to tell them how to go prone, use cover or sprint.

 

[baff]

Not sure about sprinting, but they already know all the rest of that stuff by default.
You don't have to tell them.



(Although you can tell them not to if you are getting clipping errors with prone etc).


They automatically use cover whenever it is available.
But you have to designate/label in the map design what is cover or not.


You draw little Scaletrix tracks around your map. And cover nodes on all the places you think soldiers might want to shoot from behind...
And then the AI just sort of does it's thing. It can change tracks and follow it's own behviour patterns, use whatever available cover it feels like...

 

[Grobut]

Not sure about sprinting, but they already know all the rest of that stuff by default.
You don't have to tell them.



(Although you can tell them not to if you are getting clipping errors with prone etc).


They automatically use cover whenever it is available.
But you have to designate/label in the map design what is cover or not.


You draw little Scaletrix tracks around your map. And cover nodes on all the places you think soldiers might want to shoot from behind...
And then the AI just sort of does it's thing. It can change tracks and follow it's own behviour patterns, use whatever available cover it feels like...

No, UT's AI does none of thouse things, there is no prone function in that game, nor is there reloading of weapons, or lean, or a melee attack, or any of that stuff, and the AI does not know how to do any of it, it was a quite standard Arena-shooter, it's controls only one or two buttons more advanced than Quake.

And they do not take cover, ever, it is not part of their code, because UT is not a game where you take cover, it's a game where you run at 80 miles per houer, and circlestrafe a whole bunch whilst you spam your recoilless guns that never need to be reloaded (ohh yes, the bot's can't handle recoil either BTW).


You must be thinking of the AI from America's Army and such, or the SP monster AI from the origional Unreal game (you know, the one that actually had a real SP mode), but thease are nothing at all like the MP-Bot AI from the UT games, they are far more advanced, the UT stuff isen't.

All TWI had to work with was the UT MP-Bot AI, and for a game that had no SP mode at all, and the bots only existed to populate empty servers really, i can understand why they chose not to spend the significant time and money investment into makeing them good in RO.


But HoS does have SP, so now is the time for them to dazzle us on that front (ohh and for the record, Gibson may be the president of the company, but he is no stuffy buisness man in a suit who shuffles papers around all day, he is a coder, and a very talented one at that, his work is all over the weapons in RO, and in the America's Army series as well, who he has also worked for. Let's not be so quick to sell the guy short).

 

[Stephen]

(ohh and for the record, Gibson may be the president of the company, but he is no stuffy buisness man in a suit who shuffles papers around all day, he is a coder, and a very talented one at that, his work is all over the weapons in RO, and in the America's Army series as well, who he has also worked for. Let's not be so quick to sell the guy short).

That man has never worn a suit in his life.

 

[baff]

No, UT's AI does none of thouse things, there is no prone function in that game, nor is there reloading of weapons, or lean, or a melee attack, or any of that stuff, and the AI does not know how to do any of it, it was a quite standard Arena-shooter, it's controls only one or two buttons more advanced than Quake.

And they do not take cover, ever, it is not part of their code, because UT is not a game where you take cover, it's a game where you run at 80 miles per houer, and circlestrafe a whole bunch whilst you spam your recoilless guns that never need to be reloaded (ohh yes, the bot's can't handle recoil either BTW).


You must be thinking of the AI from America's Army and such, or the SP monster AI from the origional Unreal game (you know, the one that actually had a real SP mode), but thease are nothing at all like the MP-Bot AI from the UT games, they are far more advanced, the UT stuff isn't.

All TWI had to work with was the UT MP-Bot AI, and for a game that had no SP mode at all, and the bots only existed to populate empty servers really, i can understand why they chose not to spend the significant time and money investment into makeing them good in RO..

That's not all TWI had to work with at all.
By the stage RO was made Unreal had been used in over 600 games.

The knowledge base for Unreal is simply phenominal.

While the original mod for RO may well have been built on the UT engine... RO OST, was not. It was built on the Unreal 2 engine.

I have to confess that I did not play the original UT mod at all. So I'm not in a position to comment on the AI in that game and I do not base any of my judgements upon it.

The copy of RO that I bought, uses the exact same AI as America's Army and Ravenshield. (Plus all the other games I mentioned). Unreal 2.



For the record, the only time I have ever heard of Mr Gibson was when you mentioned his name. That you mentioned it in conjucture with AI and RO is good enough reason for me to pay no intrest in his career, let alone his wardrobe.
I am sorry but the whole "cult of the dev" thing, it isn't my thing.

I do know who George Broussard is if that's any consolation.


I don't feel I will really need to be dazzled by the AI in order for this to be a good game. I think really the concept art and game mechanics are far more important to me. A basic functional one would be enough for me to get a very rewarding game experience.

Understated rather than overhyped. Open to pleasant suprises. That's my preferred mental state.

 

[Grobut]

No, stop, and listen.

1) Ro, the mod, was for UT2004, which used the UE 2.5 engine. Ro:Ost, the retail game, also uses UE 2.5.

2) No, the AI you see in games like America's Army IS NOT the standard code that comes with the game engine, and it IS NOT the same AI you see in RO:Ost, it is highly modified code done by the America's Army team.
AI does not work the way you seem to think it works, it's not part of the engine, at all, it is code that exists outside the engine, and it can be, and constantly is modified and re-written by developers.
And you don't have to take my word for that, you can go and download, for free, the latest version of the Unreal-3 engine right now, and see for yourself. It's called the "UDK", and if you download the latest version, it will be a newer, and more advanced version of the game engine than RO:Ost or even America's Army 3 used.
Do that, and then try to make a bot go prone.. i promis you, you will not be able to, because no such code exists in the engine, but go ahead and try anyway.

3) It woulden't matter if there had been 80 billion titles that used the Unreal engine, because when a team modifies the AI code, it automatically becomes their copyrighted material, you can't just take it, Unreal Technology is not open-source, even Epic Games, who make the engine, do not have such rights.
And whilst there does exist a knowledge base for the engine, it's not going to write your game for you, just help you understand the basics of how to get started on something, it is not a free-for-all buffet.

 

[Floyd]

Someone correct me if I'm wrong, but I think baff also fails to realize that RO:OST never promoted, advertised or mentioned anywhere in its description or advertisement that AI, bots, or single player modes were even in the game. RO:OST was sold as an online multiplayer game. Period. The Practice Mode (which was never adverstiesd) is a 'freebie'.

So, imo to say that because the "AI" in RO:OST was done poorly means that TW has failed at an AI endeavor is baseless.

 

[baff]

Someone correct me if I'm wrong,.

Consider yourself corrected.




If it helps however, we might wish to examine this further.
Firstly you are incorrect. The practise mode is a clearly advertised feature on my RO box.
Had it not been there, I doubt I would have bought it. (I'd already been online and read they had bots available).


Lets compare this practise modes with other practise modes from a couple of it's rivals of the day.
I'd like to pick as my examples 2 online shooters of about the same era. Joint Operations and Wolfenstien Enemy territory.
JO is a AAA budget game and Wolfenstein a low budget expansion that ended up as freeware as they felt it wasn' good enough to sell.
The primary empahsis on both of those games was Adversarial Multiplay.

JO a game with a strong history of using AI, they made a few duff offline levels for practise. Duff as they were, they were more fun than RO in practise mode. But not ultimately much fun either. I must have played them less than 10 times in total.
Novalogic program their own AI and it has never been much good. Certainly no comparison to an Unreal AI. Mostly they just stood around and waited to get shot, but at least they stood around in well thought out locations. At guard posts, next to their vehicles etc.

Now lets remember Wolfenstien.
Again their aim was not to make a single player game but just to add some bots.
Their game ran not on Unreal but on the older Quake engine.
The quakebots however were writen by the same AI coder who went on to write the AI for Unreal. So though they are a bit dumber... they are comparable.
And the AI in Enemy Territory was actually pretty fun.

So while your argument holds much water with me... RO clearly was never intended to be a single player game... and the company never devoted much resources to it etc.
What they came up with was still significantly worse than their competiton even though they had superior tools to work with.

In fact it was dreadful. Plain and simple. Notably and memorably sub par.

 

[baff]

No, stop, and listen.

1) Ro, the mod, was for UT2004, which used the UE 2.5 engine. Ro:Ost, the retail game, also uses UE 2.5.

2) No, the AI you see in games like America's Army IS NOT the standard code that comes with the game engine, and it IS NOT the same AI you see in RO:Ost, it is highly modified code done by the America's Army team.
AI does not work the way you seem to think it works, it's not part of the engine, at all, it is code that exists outside the engine, and it can be, and constantly is modified and re-written by developers.
And you don't have to take my word for that, you can go and download, for free, the latest version of the Unreal-3 engine right now, and see for yourself. It's called the "UDK", and if you download the latest version, it will be a newer, and more advanced version of the game engine than RO:Ost or even America's Army 3 used.
Do that, and then try to make a bot go prone.. i promis you, you will not be able to, because no such code exists in the engine, but go ahead and try anyway.

3) It woulden't matter if there had been 80 billion titles that used the Unreal engine, because when a team modifies the AI code, it automatically becomes their copyrighted material, you can't just take it, Unreal Technology is not open-source, even Epic Games, who make the engine, do not have such rights.
And whilst there does exist a knowledge base for the engine, it's not going to write your game for you, just help you understand the basics of how to get started on something, it is not a free-for-all buffet.

Now it is time for you to stop and listen.

Unreal is the most used game engine in the world today.
It does matter how many other games have used it.

And this is why...

Because the Unreal engine is used in so many games, it has to be very flexible and offer a very wide variety of functions.
Or rather... becuase the Unreal engine has been made so feature rich and so flexible, it has been used in a wide and numerous variety of games.

Epic has full time employees adding new functions to their engine every day.
In fact lisencees who add new AI code to the Unreal engine may sell it back to Epic for inclusion in their core code.
George Broussard for example famously did this for about ten years and never even made the game he lisenced it for.

What you see Unreal AI doing in the game Unreal Tournament is by no means all the game engine is capable of doing. The game engine was designed with lisencing in mind. Unreal is one example of a game it can be used to create, not a demonstration of every single feature it has to offer.
The unreal game engine was not design to make Unreal Tornament. It was designed to make a plethora of games under lisence and Unreal Tournament.




There is a vast publicly accessable knowledge base for how to use it.
There is also a lisencee knowledge base on how to use it.

The Unreal engine is taught in schools.

This means that it is very quick to find out how to use the engine. If you don't know how to make an AI go prone or reload.... just Google it, or if you are a lisencee, ring them up and ask.


(As it happens....for TWI Unreal was "a free for all buffet". They were given it. Yay!).




Americas Army dev's were not re-inventing the wheel. Everything they did had been done before.
Their AI was not special or developed. It was standard.
I have yet to see an America's Army AI do anything at all which I have not seen many other Unreal AI's do.


Not every studio hires an AI coder. Not every studio hires an netcoder. These people are both expensive and notoriously few and far between.
Instead many (if not most) studio's simply lisence a game engine and make a mod.

Consequently, while it took 8 years to make Halflife 2, it took only 2 to make Left 4 Dead.




The AI is part of the game engine. The scripting for it is done in the map editor.


To make a Unreal AI go prone on UT, you add this header to the area of the map you wish them to do it in. "-bG0Prone".
At least you did back in 2003.
So if you want AI's to belly crawl under barbed wire, you make that barbed wire an actor and attach the go prone trigger to it.

The idea that stock Unreal AI can't even reload is highly comical to me. Of course it bloody well can. It can probably even breakdance.

 

[Agenda_Suicide]

Floyd gets +1 like.