Non Friday 2010-04-15 Development Update :)

[constant]

Originally Posted by JG52Krupi

LOL, does that mean that soon you will be able to fly for a civilian airline

Hopefully, once ww2 is over I plan to do some commercial flying for one of those newer companies I heard about. I'm keeping my fingers crossed.

Originally Posted by Insuber

Gee I admire people keeping a flight record... or it's just an estimate?

That is actually an estimate, but I do save the records created by IL2 (DCG) for my completed campaigns where each missions time is kept. It's nice to revisit old missions from a logbook.

---

Thanks for joking over my point guys. It's coo..

---

Flight sims don't do "more" over "other" games. AI is AI, Terrain is terrain, objects are objects, etc.

Consider Oblivion, or Fallout 3. Games with weather systems. (And stars of some sort..)

It's really not a question of can it be done, it's simply will it be done. For the most part I believe companies try to avoid fine-tuning and feature adding because of the Duke Nukem Forever (r.i.p.) scenario. True perfection is a bad thing.

[TheGrunch]

Originally Posted by constant

Flight sims don't do "more" over "other" games. AI is AI, Terrain is terrain, objects are objects, etc.

Consider Oblivion, or Fallout 3. Games with weather systems. (And stars of some sort..)

I think that's rather the point - all of those games have benefited heavily from having an excess of RAM compared to more shallow games, which was what we were discussing. In any case, how do you think the physics systems of FPS games compare with contemporary flight-sims, generally? How much do you think the AI of such a game needs to consider compared to that of a flight-sim? The AI in Oblivion and Fallout 3 is an absolute joke, and the promise of Radiant AI never came about, the AI in Oblivion and Fallout from the player's point of view is exactly the same as that of Morrowind - essentially "start combat if you steal my stuff, hit me or I hate you enough and have a high enough aggression number, run for help if you hurt me too much" - with the exception that people are to be found in different locations at different times of day and that they can pick up weapons from the floor if they feel like it.
I don't really see what weather systems have to do with it, either. The weather system in Oblivion and Fallout is simple (I don't know how much modding you've done of those two games), all that happens is that each exterior cell in the overall game grid is assigned to a region, and that region has a particular percentage chance of each weather type assigned happening when the weather changes every few hours after what amounts to a dice roll that chooses the weather. It's not really a weather system as such.
As for stars, it's simply a series of domes with different textures applied with varying transparency that rotate depending upon the season/time of day/etc. No atmospheric modelling as we see in the screenshots Oleg showed of the Stuka at various times of day.
The point is, both these games needed huge amounts of RAM and powerful CPUs to run well at high settings at release, even with such simple AI and weather systems and so on. Comparing flight-sims to FPSes, even in Il-2 we require some fairly complicated modeling of aerodynamics and ballistics occurring essentially non-stop (although realism in ballistics is usually attempted in FPSes nowadays, it doesn't have to take such account of factors as relative wind), we need to have an AI that can effectively fly an aircraft without a) exceeding critical angles of attack constantly b) flying past the physical limits of the pilot (the AI are limited to 4G maneuvers, I think) and c) that understands how to effectively maneuver to a good firing solution given the capabilities of the aircraft it's flying and the position and capabilities of the opponent. Compare this to the average 2010 FPS AI and we're already talking about a more demanding AI in the 2005 4.01 incarnation of Il-2, never mind SoW. And crucially, given the focus of the next game on the Battle of Britain, Oleg and team are almost certainly aiming to have the game running reasonably on medium settings with a lot of aircraft on screen at once, after all the Battle of Britain was primarily about small groups of fighters intercepting large groups of bombers and fighters. How many FPSes do you see with more than a handful of people on-screen at once nowadays? The only FPS games that approach flight-sims in terms of difficulty of implementation would be tactical shooters (lots of fairly complicated decision-making AI and a basic physics engine), and how many of those often have lots of people on the screen? Only one that I can think of, ARMA 2, and that devours PCs even on medium settings.
So yeah, no wonder there are only a few flight-sim developers left, and thank God they have the dedication to work within such a difficult genre.

[constant]

Eheh, you make alot of really good points, if I was gonna quote you it would be the whole post You're not wrong especially when it comes to first person shooters eating computers alive these days!

However the level of difficulty for creation of the game types is still the same.

Physics is physics no matter how you roll the dice, its the implemented design that brings you flaws and limitations, not the type of game. I've used physics libraries that are powerful and fast and can be used for any game type. Aerodynamics is something else on its own, but with programming its not any different than say the basic ai: "if this occurs, do this".

Speaking of AI. FPS games do have really sad AI. Any AI is difficult. But the truth is in il2 the AI is the same as in FPS games. When I had the chance to peak at the AI code for IL2, I was amazed at how simplistic it was. I also saw code that seemed to give AI an unlimited and unreleastic "afterburner" (I believe that is exactly what they called it in the code, too)

Simplistic is not wrong, especially when it comes to programming, its always the best solution if its simple. But the AI you talk about and that actually exists are far from each other, flight sim or first person shooter.

The truth: The AI basically have pre-programmed flight manuevers and "characteristics" that are "executed" whenever a given situation exists, it is not "greater" than any fps AI, it's the same thing actually.

This is why the AI can go all wonky or do nothing at all independant of skill level, because they get caught in a situation not pre-programmed.

I was also surprised (but not completely..) to see in the code that the AI always were given the player's exact speed, location and altitude and maybe some other stuff as well, for all levels of ai skill (rookie, average, veteran, ace).

And if you watch the AI land, you can clearly see they are not actually "flying" like you and I would be flying.

FPS AI does the same thing. They get pre-programmed things to do in certain situations, but of course, and more common these days, they don't receive enough programming and therefore lack the neccessary reactions for many situations they run into.

FPS games these days are sad, focusing only on fancy shader/graphic technology, slamming the gpu with wasted effort just to put out a mediocre (or less) game that looks "pretty".

AI has not evolved much in games, no matter what type it is.

Don't get me wrong, AI is not easy, but again, the AI you talk about and the AI that actually exists are not the same.

I should wrap this up, so again, my point is the same. What it really comes down to is the game company and how well they pay their programmers.

I'm an atheist but Thank god 1c maddox is on il2 and sow

Also for an example of a non-pc killer fps game, check out Darkest Of Days was a history channel game, it uses a good physics engine, and for the most part has lots n lots of enemy on screen (and friendlies sometimes too) Of course the game isn't that great, but fighting off 100 or so enemies with a musket and a six shooter is ALWAYS fun. I run that baby on a CELERON (worst of them all!) 2.4ghz oc'd 2.9, 1.5 gig mem, geforce 850 or something like that, i forgot its model name, anyway, the same system i run il2 on and il2 still has comparable trouble with more than 8 planes. (not to mention ground units) and I've tweaked il2 as far as I can. Anyway.. time to switch class!

[TheGrunch]

Oh yes, of course, much of what the AI do is just reaction-based "if this happens, do this", but like I say, the complexity involved is quite different. In an FPS you have "point gun at player and shoot" then "oh no you're out of ammo, walk behind a wall" (if you're very lucky!), where in combat flightsims there's quite a lot more involved in producing an AI that even vaguely challenges the player.
What I think would be an interesting experiment would be to run a neural networking AI through playing a game, and then take snapshots of this AI at various stages in its development and use these snapshots to produce a rule-set for the actual game's AI implementation with each snapshot being a different difficulty level or "personality" of AI. Although I have no idea how feasible that idea would be.

[constant]

Interesting.. Don't forget even Quake3's bots had the programming to "collect armor if low on armor", "get ammo", "run away if hurt and find healing, avoid player", and I believe the bots even stuck with teammates if their teamwork level was high enough. Also capture the flag has a completely different set of rules that older games ran all the time! Stuff you don't see anymore.. The same bots could play capture the flag, team deathmatch or free-for-all. Anyway.

I have no clue as to what neural networking AI is.

I can't wait to see what they did for sow.

[TheGrunch]

It's sad isn't it, how generally AI has declined rather than improved - have a read about neural networking if you're interested. Obviously it would be necessary to give the AI a much more complex starting point that the robots which learn not to crash into things or whatever. Looking forward to seeing SoW's AI as well, although I'm not holding out that much hope since the quality of the AI stayed pretty much static for all of the Il-2 series.

[constant]

I can understand that AI not improving over the il2 series, basically each new "game" was simply a mod, just added content. Generally you wouldn't expect too much engine-revision in mods like ai-programming. They did let the AI stay in the lower-end aspect of game improvement. However its good to know Daidalos has already done some AI improvements for 4.10 .. I wanna lose ai in clouds so bad it hurts

I read the wikipage, I get a basic understanding. I might read more cuz it kinda makes me want to give it a try. I've always tried to program ai systems, never really getting much out of it, 10 years ago. maybe I've learned something since then.. The hard part is recognition of adaptative (?) changes and adapting.