AI is on top of my list
 

Now that the look and sounds of Cliffs are just about there the next fix on the top of my list is AI. I'm a SP guy and was chasing a Hurricane last night and it was shaking, flopping around and it just takes you right out of the game. With IL2 1946 the AI might have been dumbed down and unrealistic in many ways but at least the AI flying was fluid and looked real like someone coud be flying the plane. This just looks silly. Hope AI is near the top of the developers list as well.

PS Love the game:grin:

I read somewhere that AI definately is on top of their list too. Not sure where though. :-P

I've tried finding it by screening through some of the recent dev stickies and didn't find it. But: could a mod please clean up the sticky mess? It's seriously 2km long. A ton of posts can be de-stickied I believe. Only the most important stuff should be a sticky. Like the bug reporting thread for the beta patch.

It is something that I hope for too...I play offline myself and the AI needs work.
But, I'm sure their AI man has not being twiddling his thumbs whilst the sound and graphics have been reworked.

Not top but 2nd.

My current wishlist is:
1-communication
2-AI
3-dynamic campaign

-1 online is where its at :P ;)

Only joking, offline flying is useful training but on the top of my list is realistic FM's :cool:.

The right priorities...:cool:

+1

Give the sim an AI of a 2011 level. Hire Buddye from A2A BoBII team for success if you could not develop proper AI in all these years.

would really love to see improved ai, plus what krupi mentioned.......realistic flight models!!!

Honestly, FMs seem fine to me. Also, unless we have some pilots on board who have flown WW2 fighters, how would anyone begin to have a feel for accuracy. In terms of AI behavior, anyone can see it's WAY, WAY off.

LMAO okay I kind of agree but we know the aircraft speeds and they are off by more than a few mph and the spitfire mk2 has an anti-stall system that would impress the engineers of 8000 AD.

i would love to see the AI that Wings of Victory II has :-P

+1
i want as accurate models as possible....even to my disadvantage.
btw,...didnt know about the spit2 and the anti stall system;)
is there a difference between the mkII and the mkI when its about flight characteristics, except performance?

Mk II (Type 329)


Spitfire Mk IIa P7666 of 41 Squadron. P7666 "Observer Corps" was flown by Squadron Leader Donald Finlay; Finlay shot down two Bf 109s in P7666 in November 1940.
In the summer of 1939 an early Mk I K9788 was fitted with a new version of the Merlin, the XII. With the success of the trial it was decided to use this version of the Merlin in the Mk II which, it was decided, would be the first version to be produced exclusively by the huge new Lord Nuffield shadow factory at Castle Bromwich.[69]
Chief among the changes was the upgraded 1,175 horsepower (876 kW) Merlin XII engine. This engine included a Coffman engine starter, instead of the electric system of earlier and some later versions of the Merlin, and it required a small "teardrop" blister on the forward starboard cowling..[69] The Merlin XII was cooled by a 70% to 30% water glycol mix, rather than pure glycol used for earlier Merlin versions.[70]
In early 1940 Spitfire Is of 54 and 66 Squadrons were fitted with Rotol manufactured wide-bladed propellers of 10 ft 9 in (3.27 m) diameter, which were recognisable by a bigger, more rounded spinner: the decision was made that the new propeller would also be used exclusively by the Mk II. This engine/propeller combination increased top speed over the late Mk I by about 6-7 mph below 17,000 feet (5,200 m), and improved climb rate.[71] Due to all of the weight increases maximum speed performance was still lower than that of early Mk Is, but combat capability was far better.[33] The Mk II was produced in IIA eight-gun and IIB cannon armed versions. Deliveries were very rapid, and they quickly replaced all remaining Mk Is in service, which were then sent to Operational Training Units. The RAF had re-equipped with the new version by April 1941.[33] The Rotol propeller units were later supplemented by de Havilland constant-speed units similar to those fitted to Mk Is.
A small number of Mk IIs were converted to "Long Range" Spitfires in early 1941. These could be recognised by the fixed 40 gal (182 l) fuel tank which was fitted under the port wing. With a full tank manoeuvrability was reduced, maximum speed was 26 mph (42 km/h) lower and the climb rate and service ceiling were also reduced. Several squadrons used this version to provide long-range bomber escort.[72] Once the Mk II was taken out of front line service, 50 of them were converted for air-sea rescue work, at first under the designation Mk IIC (type 375) but later referred to as the A.S.R Mk II. The Merlin XII was replaced by the Mark XX, a "rescue pack" was fitted in the flare chute and smoke marker bombs were carried under the port wing. [73]
A total of 921 Mk IIs were built, all by Castle Bromwich.[68] A small number of Mk IIs were converted to Mk Vs.[nb 7

Stolen from wiki ;)

Yeah, these are my three wishlist toppers too mate. :)

AI is top priority for me also, as a single player pilot...

AI is one of, if not the most important part of Offline play.

All your combat and flight members are AI. The AI affects everything.

Sadly, AI has been ignored by Oleg for years. The Online game with humans against humans makes development much simpler. Programming for AI takes alot of work, lots of behavioral probabilities stuff has to be programmed. It can be done, because people most often do predictable things, especially when they have been trained a certain way as combat pilots are. You don't try to gain altitude when the guy on your six has alot of E, you don't turn against a better turning opponent,etc.

The AI in Battle of Britain II WOV is the most exceptional AI performance I've ever experienced. The game hasn't been very popular, because it is strictly OFFLINE game. The BOB II WOV campaign has been done well, in fact, the game is very top notch. Yet, it becomes boring just doing air combat, with not much else.

COD will become passe' soon enough unless other maps, aircraft aren't applied soon.

Well, I agree that it would be nice with better AI but how about a few secs of love by someone briefly acquainted with usability design on the MP GUI would not hurt either ;)

Anyone else remember the improved AI that were introduced with one of the patches for the original Il-2 series, and that were then taken away again in the next patch soon after that? Oleg himself stated when the patch came out that the new AI were a glimpse of the ones being developed for the new upcoming game (that would eventually become Cliffs Of Dover).

As I remember it, those temproary AI were a vast improvement over the ones we had at the time, but were quickly replaced with something not quite as good in order to keep them as a feature of the new upcoming game. From memory, they were better in some ways than what we finally did end up with, not prone to the unbelievable routines we had back then or that we have now. Might just be rose-coloured glasses I'm remembering them through, but they were realistic to the point where it was hard to tell sometimes if they were AI or human pilots.

I recall making a very thorough research of AI nonsense and such several years ago.

The AI is ignored, not just with Il2 or COD. It is very standard practice, because it is much easier to build another aircraft, map or object. People will be happy with the new stuff.

It is hard to care about the Offline player when that isn't your interest. So the developers just make the graphics stuff rather than work with through the very difficult process of programming probabilistic functions and algorithms for a competent AI performance.

The benchmark for AI performance is Battle of Britain II Wings of Victory. The improvements have been coming steady to their users for several years. All the COD development team needs to do is get a copy of the BOB WOV and study the AI, also the BOB WOV AI config files are openly shared.

We will see what comes of the AI, but I think it will probably require some of that passion associated with the term..."Keep their feet in the fire, they'll get the message". LOL

I'm sure online is a barrel of fun, but offliners shouldn't be ignored. I like playing offline because I can play at my own pace and do my own thing without having to worry about someone letting me down or me letting someone down.

All very true and what you say echoes my thoughts exactly.

Hey, all I want really is AI that doesn't do what it's doing now...really immersion-killing stuff like ignoring enemy planes even when in full view, not reacting at all to being shot at, and the crazy rates of rolls these boys are pulling off.
In short, give me old Il-2 AI, but maybe a little bit more advanced. I'm sure a talented team like MG can pull that off, surely?

I'm content to wait and see what happens...there is still a lot of things that need their attention still and I'm confident that in time we'll see AI that is improved. (I'll just bang the drum from time to time to remind them. :))

I wish there was a way to set passive/aggressiveness of enemies in custom missions, I want enemy bombers to fight back!

I agree with nearmiss but ....

Why is it so difficult to make AI planes that follow the same rules of physics and gunnery that human-controlled planes do in flight sims?

I also am very interested in AI. As a volunteer BDG (BOB Development Group) member, I have worked on the BOBII AI over 6 years trying to improve it and make it more human like.

In BOBII the AI and the player use the same FM and code so they are equal with only two exceptions (1) the AI do not black out or white out and (2) the AI can see through clouds (the AI do have blind spots as they can not see through the solid A/C parts [wings, rear, nose, etc] or see into the sun). Customers do sometimes complain that the AI can pull more G's and does not black out but in my opinion the advantage is small against a human pilot.

In BobII the AI can stall, spin, crash, and do stupid things. The AI ability to fly effectively is controlled with "skill level" which the player can select in Instant Action Missions and skill level is assign or user controlled in the Dyamic Campaign. The higher skill level AI are more effective in Air Combat as they have the ability to fly more maneuvers and they fly each maneuver more effectively and efficiently.

Why are the AI not more human like?

In one word the answer is "cost". The AI can not just be designed and implemented. It must be tested, re-worked, and re-tested over and over. The quality of the testers giving feedback must be exceptional and the amount of time and manpower is very large and it can never end as you are never done. The AI is not just designing to a spec. The AI is more of an art. It is cheaper to put you game money into things that can be designed, scheduled, and implemented (Multi-Player, Landscape, cockpits, and more eye candy). Also eye candy sells well but customers seem to always complain about the AI.

Been reading all the interesting comments and thought I would jump back in.

I don't know if I'm in the minority or majority. I get a little window of time to play and fire up Cliffs for about 15 minutes and then I hear my wife or kids call and I'm gone. Don't think I've got the time to really get in to MP. Maybe I'm wrong. Also seem to rememeber seeing a poll that showed the vast majority of players are SP, not MP. Maybe I'm wrong about that too. I would rathther go back to IL2-1946 than the current Cliffs in terms of AI. Maybe that would be easier for the developers to implement than fixing current AI. I'm for whatever is the quickest / easist fix but would like a fix of some sort. Right now planes just sit there like on auto pilot or go nuts like stunt pilot monkeys on crack.

+1

you guys seem to forget a simple truth: the IL2 series numbers are based on singleplayer, not multiplayer.

the game sold tens of thousands of units, yet a very small fraction of those could be meet online playing the MP part. a thousand? maybe 2? for sure under 10% of game's buyers.

stop the nonsense that IL2 strength is in MP. The longevity yes, it is based and rests on MP, but not the game's sales number, which is the thing that makes MG tick..

And the thing is, even now, half a year later, the SP part of IL2 CoD is completly broken. One can't play anything else excepting simple dogfight missions because of this.

The game's stability, and then its performance, were the logical first needed fixes, as you can hardly play a game which is a crash fest, or crawls at 5 FPS on 2011 supercomputers. As those were fixed, the next pririty ones should be, in exactly this order:

1) Radio Comm - affecting both game modes, it completly porks SP and gravely affects with AI MP (until much more MP players will join, with the actual online numbers you can't really play anything else except dogfights if you're not adding complementary AI)
2) AI - same as above, out-of-this-world AI maneuvers, performances and behavior are really killing both game modes
3) dynamic weather
4) dynamic campaign

Right on.

You did a great job with the BoB2 WOV AI Buddye. They could do with some of your expertise on CloD.

As many in this thread have said, the AI is a mess. I haven't fired up CloD much in the last month or so because the awful AI has always been the biggest immersion killer for me since release. The eye candy is all very nice but IMO it's worthless without some pretty serious tweaking under the hood on the AI and AI flight models as many of them are laughable at the moment.

Here's hoping that getting it fixed is also near the top of the devs list.

Didn't Illya say in one of his updates that he had recently hired someone or was going to hire someone with sole purpose of reworking the AI in CoD?

Buddye has done a fabulous job with the BOB II WOV. I mean fabulous AI performance after hundreds and hundreds of hours working with the AI. Then going back and forth with the community of users for feedback.

Alas, the problem with most gamer type developers their ego gets in the way. Then we get stuff like MSFT CFS3 from Hatfield that just ruined a great sim with his so-called improvements. I knew it would be junk when I started it the first time, the pilots were hanging around the garbage cans.

It would great if Luthier consulted with Buddye, he couldn't do better. I say that because NO ONE has a better AI performance package for WW2 air combat sims. I think I own every WW2 air combat sim produced for the past 12 years, so I haven't missed much.

Glad you dropped in Buddye, I'm glad to toot a horn for you.

Yes, adonys is making sense here. MP is essential but should be built on a strong SP platform. Anyway, AI is required for MP as well.

+1

Some very fine posts have been made in this thread, thanks for that. I recommend anyone who's new skimming through - it'll be worth it.

That said I want to point one other thing out:
The thing is that AI also affect all online combat if it's not pure dogfighting. A lot of the dynamic campaign efforts and co-op missions depend on it. So I would say it not only affects offline players but almost certainly the majority of all online players as well.

True. It's the same guy that will redo the radio commands menu. According to that post by Luthier, he will do the radio commands first and then move on to the AI.

By the way there are some new placeholder commands in the latest beta patch (they don't work yet, they are just there in the menus) that are giving me much hope for what will come.

I saw commands to specific crew members in bombers, for example asking your co-pilot to report speed or altitude, have him operate flaps or landing gear, order the gunners to hold fire, etc.

Good stuff is coming and yes, it will be useful both for SP and MP ;)

I remember several years back building missions. I was stopped dead in my tracks so often by poor AI, and lack of comms ability to have any power of your player flights...except go home.

Pokryshkin developed tactics to beat the Germans in G2s and 190s, by using his flight groups at tiered altitude levels. The p40s were at lower altitudes to take on the german bombers. The P39s would drop in from tiered altitude levels when the german fighters went for the p40s. Pokryshkin kept the E advantage with flights at several levels higher and as the battle advanced they would drop down incrementally. The p39s with those big cannons in the nose made short work of the german fighters.

I could not keep the Russian fighters at the tiered levels. The moment the germans attacked the p40s all the p39s from all levels, except player would drop into the fur ball. The P39s were no match for the G2s or 190s without E advantage.

So, yes communications with power for managing flight groups is very necessary. I'm just thinking in terms of this one example.

You can tactically fly Online with comms, but Offline you are subject to whatever a stupid (ultra-stupid) AI aircraft will do.

I realize good ai performance makes a difference online, but there was never much flak from online players about AI performance as I recall.

Actually what I meant was that with this new completely open design AI planes could be used much more constantly than in the old IL2.

Actually there could be a constant presence of certain flights on almost all servers. Basically what I want to say is that the new multiplayer could involve a lot more AI than it did in the past. Huge dynamic campaigns or constant scenarios that have a realistic amount of AI planes, ships and ground vehicles present at any time.

Good example regarding AI though. :) I wonder if somethink alike would / could be implemented.

As I understand it. WHen COD is working at finished levels 128x128 should be realistic number of aircraft in air combat at one time = 256 planes.

That would be a whopping big order to put together that many human players at one time.

AI is the only possibility for large groups of planes in the air at one time.

So... I would naturally think the devs have good understanding that competent AI performance would be required.

+1

Once they have the game optimized, revamping the AI is necessary.
I play a lot offline, because I don't have enough free time to look for a populated server and pair up with wingies. Just 10 minutes to start the game, and try to shoot some bad guys. Fighting against either planes flying bee lines or crazy rolling AI planes is not very funny and also it teach you a lot of bad habbits ;).

256 online players mixed with AI ... it would be an amazing experience but, today, it seem unlikely to me. That will put our system above their capabilities. If I try to join a sever with 256 players and Ai planes buzzing arround, my PC will laught at me.
My system is not a "Top specs" one, but when I play online, everytime an AI flight is spawned or a player joins the server, i have heavy stutters.
My guess is that this situation will improve with the following patches.

Maybe in 2013, we can experience massive online raids over London, who knows. (Fingers crossed)

Well, I know I'm certainly not alone in appreciating the efforts of development teams and testers in trying to make AI more advanced and believeable. As you say, it is something of an art form and sadly something that seems to be a dying art form at that.
However, to me and I'm sure many others, great AI is every bit as important as other aspects of a game, and it's something that should never be ignored or skimmed over by developers.

In the new system you cannot strictly draw a line between SP and MP and the AI, because the AI is getting more a part of MP, compared to 1946 (without mods).

Therefore I would love to see better AI, but what is better AI worth, if the netcode cannot support more than 15 AI planes in a coop mission, otherwise the planes warp, lag or are not drawn at all (see ghost/phantom planes)? Seeing improvement of the AI as long term project, I would first focus on the obvious problems and bugs:

FM (performance of the existing planes before introducing new ones f.e. service ceiling :rolleyes: ) -> SP&MP
Network (get rid of the ghost dots and warp-fest of AI) -> MP
LOD (dots visibilty) -> SP&MP
AI -> SP&MP

I have been reading this thread over the last couple of days at work, along with some other threads about the new beta patch and thinking to myself, it seems like it might be time to have another crack at this (after having shelved it for a couple months).

Tonight I got to have a crack at the new beta patch for the first time and while I can agree there is considerable improvement in over all graphics and sound I still couldn't find the hook I was hoping for (maybe too much forum hype, who knows). 109 sounds although, pretty much hit the bull's-eye! Nice work fellas!

smink1701's original post sums it up for me almost perfectly as I see it. The AI with CoD has always seemed absolutely mad to me and one of the great put offs, I mean absolutely spastic manoeuvres that make a chase an absolute joke. 109's with a roll rate that would shame the best 190 pilot, not only in passive flight, but also in steep dives to the point that I thought they were going in without hope only to find them pull up at the last minute and return to the fight LMFAO!!

I also read that some forum members believe that development focus does/should favour online play, well, from what I have read over the years to date, statistically off line play makes up the greater number of paying punters. I'm a 100% offline player as is everyone else I know that plays PC games. AI performance is not a trivial issue with gaming development, just ask the developers how important it is.

This is certainly a very significant patch for CoD, but after all it was either that or shut the doors and call in the cleaners right?

Well, for me it's off the shelf and now sitting on the desktop waiting for the next patch.

I have faith now at least it will get there eventually.

Cheers, CrazySchmidt. :)

+1


...and if world were perfect, to hire the talented guy who made -BOB II WOV- AI, no need to reinvent the wheel.

Said it before and I'll say it again, the AI in BoBII WoV is the only AI I have competed against that made me actually think I was against a real player.

At the point I ended up in a Rolling Scissors against a 109 I was really hooked!

If you want to experience true Flight Simulation AI, load up BoBII WoV and then compare it with the AI in Clod.

BoBII is like playing chess against a Grandmaster while Clod is like playing against your mate who thinks he can use the Pawn's to jump over pieces and remove them from the board!

Sign up Buddeye and pay him a salary...he'll be worth every penny!

BTW, I fly mainly offline at the moment as CloD doesn't really interest me that much online as I don't have the time to go through learning CEM, so offline is fine for now.

Eventually I will learn CEM but for now I just want to have a little fun and compete against quality AI, and at the moment....we don't have any!

Cheers, MP

The problem with realistic and Human like AI is that it is so very costly.

The developing of realistic Flight Sim AI can “not” really be coded to a spec (like graphics, landscape, textures, hardware interfaces, multi player, canned missions, etc.) where the cost can be accurately estimated and bounded. Developing AI is a process (develop, test, refine, test, refine, etc). You are trying to simulate a human whose options and thinking are almost endless.

AI Robots can be built today if their functions are limited (repeatable task like the auto robots) but the building of human robots would require the refinement and testing process with endless delays and the cost gets too high or out of control. It is almost impossible to "bound" human thinking (say a good pilot, as individuals solve problems and invent solutions in an almost infinite manner).

Combat Flight Sim AI has the same problem. The development cost to make the AI more human and thinking is just too high. The game developers are force to get the AI working and call a stop to the refinement process. They can not continue the process too long because of cost.

In BOBII, we work for free. We get endless ideas from our players and testers which we can use for the BOBII AI refinement process.. The implementation and testing is still very difficult requiring much refinement and some really good testers who can offer refinements. We have been refining the AI in BOBII using this refinement process for 5 years. We could go on forever with the only limit being “new” ideas and experienced manpower.

The answer IMHO is simple, the Flight Sim game industry, the developer, can not afford the money for an effective and complete human like AI system (a commercial game developer really can not afford a BOBII like refinement process). Only the military or NASA can afford to develop smart, effective, thinking, and human like AI and even they have had limited success, as yet..

BINGO! I remember over ten years ago, playing European Air War and thinking "The A.I is ok in this sim but in ten years it's gonna be like freaking Skynet type of A.I!". Here we are, multiple CPU cores later and the A.I is dumber than ever. Why? As the poster above mentioned, money! A.I programming is so costly that even the biggest of the biggest game studios has to cut corners to get their software out the doors. You see games like Call of Duty where the A.I has the IQ of a golf ball but it doesn't matter since 90% of all the action is scripted. When I think more about it, I'm amazed MG managed to get the A.I they have now out the door considering their -probably- scarce resources. IMO, let the community experts get their hands on the code give it a gosh because they are willing and willing to do it for FREE!

Below is a horrible example of how scripting is almost replacing A.I:

http://www.youtube.com/watch?v=RULv6HbgEjY

buddye

It seems to me from what you say, if a developer like COD wanted a competent AI performance then the best course of action would be to open the source in such fashion as to allow coding for the AI.

AI performance improvements would come from competent programming and a community involvement over time.

There could be benefits to developer, because it would take the "hot potato" of AI developement from the developer and hand it off to the community.

Naturally a reasonable AI performance package would have to be coded as a starter package. The AI would thereafter be done over time with 3rd party community developers and feedback from the community.

I could live with that, and actually I'd be thrilled with the prospects that we could eventually have an outstanding AI performance package.

That's exactly what we should ask for: MG to open AI over-coding in an additional dll.

Maybe then the buddye and his mates from WoV BoB 2 could help with it, or anyone else with spare time and interest.

Maybe the community should compile a list of AI attributes or features that the developers or modders could use as a sort of checklist.

Something separated into two general categories of Friendly and Enemy AI. Then further separated into three categories of, Essential, Desired, and Ideal AI qualities or abilities.

In the Essential category you could have not just positive pre-requisites, but negative ones too, things along the lines of what the Friendly or Enemy AI must never do or fail at.

In the Desired category you could have more advanced features that should be possible given the current technology, but which aren't game-breakers like in the Essential category.

In the Ideal category you could have wishful thinking, 'it would be good if we could one day have such and such a feature' type ideas.

Just putting it out there. Might be more constructive and proactive than just waiting to see what the developers or modders do then reacting to that after it happens. Create some guidelines so the developers know what the community considers to be fundamentally necessary or essential, what they consider to be worth trying to achieve beyond that, and what they don't really expect to be done but would hold as the highest achievement or aim.

I am not sure if a commercial game developer would feel comfortable releasing his product (his source code). His source is his future with respect to new products which he plans to use as a baseline for future products. A flight sim developer must sell a string of products base on his engine to make any real money these days (you do not develop a new engine for only one product).

The COD manager might be willing to assign a programmer to work with a committed team of AI testers who would test, offer suggestions, analysis, and refinements. All parties would need to understand that the project would be long term (multi-year) with updates maybe every 6 months. A new priviate forum could be used to coordinate the work (the BOBII approach).

I do not want to bore you guys with BOBII but for those interested in AI (and I understand that some players are only into MP), I will offer a document. I offer the document to show an overview of BOBII AI's the scope and complexity. Please skip it if not interested and am sorry for boring anyone.

Overview of the BOBII AI Performance and Design

The redesign of the AI Maneuver Selection

The redesign of the AI maneuver Selection Criteria was driven by the need to become more deterministic and less random in selecting AI maneuvers. I felt the need to move in this direction to improve the AI offence and defense so the AI selection software had more control. Of course there is a big danger in this strategy. We do not want BOBII to become repeatable. This will become a fine line to follow in the future and I will need both our testers and customers help and feedback.

Here is a summary of the new selection criteria design:

1. Altitude (how much altitude and rate of change)
2. Speed (how much speed, and rate of change)
3. Position of A/C to each other (none to tail, tail to tail, nose to beam, tail to beam or left, right, front, back)

To implement the new selection criteria I built a 3 X 6 matrix (a truth table) and many new programs. For each A/C (the unfriendly and the AI or the player), I designed programs to look at Altitude, Speed, and position and to first try and select the best maneuver option (aggressive or defense) based on AI skill level.

Altitude and speed can also be thought of as “energy” as altitude can be turned into speed.

The new design uses the A/C’s current position but I am thinking about implementing the A/C’s “lead” position (his future position which would be a small delta in the future).

Anyway, this new design is a long term work BOBII AI strategy (work in progress) that I can work on as I have new ideas for the AI from our testers and customers. It should position the BOBII AI design for future improvements by using more deterministic rather than random selection criteria. The goal is very simple "to improve the AI performance" and keep BOBII the best off line AI.

My gut feel based on my experience and testing is that the AI is now stronger at selecting the best/correct maneuver, avoiding bad positions (like low altitude), better at avoiding low energy, and in general a bit stronger fighter and defender. Of course, the Player will always “win” with experience/practice but if we give the player a better fight (scrap), I personally consider that real progress.

BTW, I also gave the Terminator AI a bit of boost in performance and tune-up so you might want to try it out. With the boost in Terminator performance, he was getting “cocky” so I implemented a spinout feature. The Terminator flies so close to the edge now that he has a tendency to “spinout”. This is when you can get him.

I also implemented a new AI feature called “Flying Factor (FF)”. This is the knowledge of the AI pilot to fly a given maneuver (experience) and how well the AI pilot will actually fly the given maneuver (skill). The FF is based on the Skill Level (customer selected in Instant Action Missions and software assigned in the Campaign). The Terminator AI is assigned a Skill Level of Hero2 (the highest in the game) so that is where he gets his boost in performance (edge).

AI Maneuver Selection is KEY

First, the BOBII AI has a special case for selection called “Evasive maneuver selection”. The Evasive maneuver is required when an AI is shot at (either a hit or near miss). The AI will select a defensive maneuver based on the criteria/data for both Player and enemy AI (speed, altitude, and the position of the targeting AI and the AI being targeted with respect to each other).

The first key decision to be made is to select either an aggressive or defensive maneuver. This is a complex decision based on the available information on both the Player and the Enemy AI or the friendly AI and the enemy AI. The data considered for both Player and enemy is speed, altitude, and the position of the targeting AI and the AI being targeted with respect to each other.

After selecting either an aggressive or defensive, then a random approach is used to select a category (choose good, choose bad, or choose “neither” good nor bad maneuver).

Maneuvers are then divided into three parts Climb, Horizontal, and dive for each of our categories (choose good, choose bad, and choose neither good nor bad maneuver).

The individual maneuver selection is then based on speed, altitude, and position of both the player and the enemy AI.

In BOBII we have over 80 complex maneuvers for selection (both the aggressive and defensive maneuvers) for the fighters (Spit,Hurri,109.110) and over 50 simple maneuvers for selection for the JU87 and Defiant. The following are BOBII’s complex maneuvers (each maneuver is a significant program for completing the assigned maneuver):

MANOEUVRE_SELECT
MANOEUVRE_LOOKROUND
MANOEUVRE_WELDEDWINGMAN
MANOEUVRE_BAILOUT
MANOEUVRE_TOPCOVER
MANOEUVRE_FORCETOPCOVER
MANOEUVRE_LINEABREAST
MANOEUVRE_PINCER
MANOEUVRE_MULTIWAVE
MANOEUVRE_DIVEANDZOOM
MANOEUVRE_INSIDELOOP
MANOEUVRE_LAGPURSUIT
MANOEUVRE_SPLITMANOEUVRE
MANOEUVRE_HEADON
MANOEUVRE_LINEASTERN
MANOEUVRE_VICATTACK
MANOEUVRE_BARRELROLLATTACK
MANOEUVRE_SCISSORS
MANOEUVRE_MILDSCISSORS
MANOEUVRE_TURNINGFIGHT
MANOEUVRE_SPLITS
MANOEUVRE_ZOOMANDDROP
MANOEUVRE_STRAIGHTANDLEVEL
MANOEUVRE_SPINOUT
MANOEUVRE_DIVEFORHOME
MANOEUVRE_GOHOME
MANOEUVRE_MAKEFORFRIENDLY
MANOEUVRE_MOVEAWAY
MANOEUVRE_QUICKROLL
MANOEUVRE_IMMELMANNTURN
MANOEUVRE_IMMELMANN
MANOEUVRE_STAYWITHPREY
MANOEUVRE_CLIMBFORHOME
MANOEUVRE_STRAIGHTDIVE
MANOEUVRE_OUTSIDELOOP
MANOEUVRE_SHOOTTOFRIGHTEN
MANOEUVRE_SHOOTTOOEARLY
MANOEUVRE_GAINHEIGHT
MANOEUVRE_LAGROLL
MANOEUVRE_EXTENSION
MANOEUVRE_DIVINGROLL
MANOEUVRE_REVERSETURN
MANOEUVRE_SELFASBAIT
MANOEUVRE_JINK
MANOEUVRE_BREAKTURN
MANOEUVRE_LAZYTURN
MANOEUVRE_BREAKLOW
MANOEUVRE_BREAKHIGH
MANOEUVRE_BREAK90
MANOEUVRE_BREAK180
MANOEUVRE_HIGBARRELROLL
MANOEUVRE_PANICTURN
MANOEUVRE_TURNANDRUN
MANOEUVRE_LOWALT
MANOEUVRE_ZOOM
MANOEUVRE_INTERCEPTHIGH
MANOEUVRE_INTERCEPTLOW
MANOEUVRE_GAINSPEED
MANOEUVRE_HEADONATTACK
MANOEUVRE_LUFBERRY
MANOEUVRE_STEEPDIVE
MANOEUVRE_UPANDUNDER
MANOEUVRE_STALLTURN
MANOEUVRE_SLASHINGATTACK
MANOEUVRE_CLIMBTURN
MANOEUVRE_ROLL360
MANOEUVRE_STRAFFE
MANOEUVRE_TRANSITION
MANOEUVRE_PEELOFF
MANOEUVRE_SNAPSHOT
MANOEUVRE_STAYONTAIL
MANOEUVRE_FLYTHROUGHCLOUD
MANOEUVRE_REGROUP
MANOEUVRE_DISENGAGED
MANOEUVRE_SPINRECOVERY
MANOEUVRE_COLLIDE
MANOEUVRE_LAST
MANOEUVRE_ALIGNMENT
MANOEUVRE_DROPANDZOOM
MANOEUVRE_COLLISIONAVOIDANCE
MANOEUVRE_TURNTOHDGANDPITCH
MANOEUVRE_SCREWYOUGUYSIMGOINGHOME

The AI Performance Criteria/Dependencies

Skill Level

The AI performance is dependent on AI Skill Level (which is customer selected in Instant Action and SW assigned in the campaign). BOBII AI do make mistakes (spin, crash, dumb maneuvers, bad judgment, shoot late , shoot bad, etc) which is after all very human.

The skill level of the AI is key in making decisions on about everything with respect to AI performance like (1) how well the AI fly, shoot, shoot fast, slow, accuracy, or not shoot, (2) how well the AI fly, what maneuvers are selected, and how well the AI will fly the chosen maneuver.

Random Numbers (Luck of the draw)

The BOBII AI A/C is also dependent on luck (specifically on random numbers). Random number decisions are coded through out the AI code. BOBII’s random approach keeps BOBII from doing the same thing each time. Even something as simple as the direction to start a maneuver (left or right), I will use a random number to decide (why hard code something when you can use a random number).


For example, most BOBII vertical maneuvers use a random number to assign a length of time to for a specific vertical maneuver (Like Zoom). The AI pilot will sometimes cut off early, or late, or somewhere in the middle. If early the maneuver may carry too much speed and if late the AI may slow down so much that control is lost (very human).

The bottom line is that the customer will always see a somewhat different maneuver (very good, good, not so good, and loss of control) both because of the random implementation and the different physic’s parameters (speed, roll, heading, pitch, and overall energy) going into each maneuver.

The very real downside of random numbers is it is very hard to test (not repeatable) and the processing power used.

Conditions

The conditions for each maneuver are always different (energy, speed, altitude, skill, damage, and enemy position). This also changes how the maneuver is performed. A damaged AI will not fly as well as an undamaged AI.

I wasn't saying about them giving the code away, but just provide an AI SDK or a script mechanism similar with the one for missions in order to overwrite AI functions.

I'm going to sticky this thread for a few days. Give it some time for response from other members that are very interested in this topic.

Interesting insights Buddye, thanks for that.

It looks to me, as someone who has no idea of how these things really work, that you have a set of potential manouvres that are triggered by where the planes are in relation to each other and what their individual state is (damage, speed, skill level etc.) With variations in how well the manouvres are carried out also based on the planes individual state, but with some randomness thrown in as well.

In laymans terms, that would mean the more of those manouvre routines you have, and the more often the game re-calculates which routine is required under the changing circumstances, the better.

What a nightmare lol :grin:

And by that I mean, where do you set the point at which the game still remains playable and all your computing power doesn't get chewed up by calculating what the AI should do?

It's always got to be a compromise.

But even so, it's hard to accept that some of the behaviour we see in Cliffs Of Dover is the unavoidable result of making compromises to work within those fundamental computational limits.

It could be done better, right? There are ways to make the AI conform to at least a basic set of player expectations aren't there? And if those expectations were conveyed to the developers they wouldn't be too difficult to meet, would they? Just asking for your opinion based on your understanding of what's involved and what you've seen of Cliffs Of Dover in it's current state. I know you've already said the AI could be worked on endlessly and that the developers couldn't afford to do that, just wondering how far you think they might be able to get.

+1

Definately my top 3 as well. Better FPS would be #4! ;)

Exactly. MP is never going to be great without good AI in the mix. particularly for me, since I only like co-op with friends.

GREAT idea :cool: thanks!

I think you understand the Overview of BOBII AI performance and design.

Actually, the processor requirements for the AI are very low as compaired to GFX, weather , water, and landscape which run every 40 MS or faster. Remember, it takes seconds to complete a maneuver (range of 3 to 15 seconds on average but some maneuvers take longer (some of the aggressive maneuvers). As an example, a break low maneuver takes 7 seconds. Seconds are easy for a modern PC.

The COD developers are good. They can fix anything they want. Yes, they can make significant improvements to the COD AI. Right now I think it is a question of priority as they have a very long work list with higher priorities. My hope and I think some of you agree is that the Developers will agree and give higher priority to the AI performance.

How difficult might it be to remove the flaws which make it obvious that the plane is flown by AI? E.g. the dithering you see when they can't decide which way to break.

It seems to me that there are two levels of AI improvement for CoD: the first would be to remove these aspects which really break immersion because you instantly realise that the plane is AI-flown. The second would be to turn an AI plane into a tactically strong opponent.

I hope we get some dev involvement in this thread in particular. I think CoD is heading in the right direction, especially after the last patch, but there seems to be a lot of community expertise that could be tapped. Thanks for sharing the document, buddye.

Well said, CrazySchmidt
 

Thanks 'CrazyScmidt' I could not have said it better myself. I am just puting the finishing touches to a Ju88 'offline' campaign for the 'bomber boys' ( which will be posted shortly).
I have spent hours and hours historically researching this campaign and thats what 'offline' campaigns are all about - fully historical or semi-historical campaigns that really put you in the 'drivers seat' re-creating what our forefathers (on both sides) actually did! For me personally, the immersion is far better than an 'Online Campaign', although I do enjoy battling with my sons on 'Online campaigns' occasionally.

This flight sim has put the air war during WWII firmly on the map (around the world). Ilya and his team will eventually fix the AI situation (I really can't understand why they haven't followed the programming in IL-2 Sturmovik 1946 which got it fairly right! Hello Ilya?)

This is what Oleg really created in IL-2 Sturmovik, an awareness about what really happened in the air war during WWII. He will always be remembered for this (I hope you are still out there Oleg?)

Finally, the more and more I immerse myself in creating my 'Offline' campaigns in COD, the more and more I really like what the team has tryed to create - eventually (with the help of lots of very intelligent members input) this flight sim will be the leading flight sim. Ilya will then develop a 'Korean War' flight sim then onto 'Vietnam', my personal era.

DFLion
P.S. I was very sad to hear about the crash of the Yak42 with a Russian Ice Hockey team aboard. I was very glad to hear that the Russian President will be looking very closely at Russian Aviation following this crash. The 'Ruski's' have some of the best aircraft designers in the world, they just have to get their act together with Civil Aviation in Russia!

DFLion

+1 Great thread.

I think improved AI would well increase the playability, appeal and longevity of the game (and sales!)

I'm sure there is a hell of a lot of purely SP action going on that doesn't ever make the headlines in forums.

Does someone have a recent vid of the fast roll rate? I used to see this a lot in the quick missions but have lately been playing custom SP missions, mostly against veteran 109s and 110s, and the roll rate is fine.

For those who haven't seen this video I did, here is an example of BoBII in action.

Now the initial attack from behind on the bombers I'd never get away with in IL2, but if you watch the rest you will get an idea of how good the AI is when under attack.

Also if you watch the titles, you'll see Gun camera footage I made from the game and in it I am chasing a damaged 109 low on the deck. Just as it goes out of view I get into position and when I get onto him he is stalling and goes into the ground.

Now that is impressive AI to make it try a move and then stall...brilliant.

Here's the video:

http://www.youtube.com/watch?v=8lPmkkBFT8M

Cheers, MP

I agree that the AI should be held at one of the highest priorities now. Its importance for single-player goes without saying, however I feel that it would benefit multiplayer as well.

Not sure if eveyone is aware, but you can actually spawn with AI wingmen in multiplayer. This is a key feature in regards to helping recreate the sheer number of aircraft that featured in the Battle of Britain - it is impractical to expect to get the same amount of humans to join the one server. ;)

At the moment its not worth taking AI wingmen because they just can't fly properly, on the ground or in the air. Coupled with the fact you just don't have any real control over them and it just turns into a frustrating experience. For the amount of work they put into enabling AI to dynamically spawn with players regardless of the missiontype, it just seems wasted when they just didn't come through with the code for the actual AI itself.

I don't know what the full story is. Maybe they lost the team member who was working on the AI and someone else had to fill in his place. I would hope that now however that they have enough team members working on it to actually get it working within a reasonable timeframe. A significant chunk of enjoyment that someone can get out of this game is lost because of the broken AI.

I so do agree with these points above.. i only play offline and because of that COD is unplayable an just collecting dust atm. for me. I travel a lot and have mostly bad internet so online is out anyways. sometimes i used to let the AI fly for me and i just give orders to pass the time (which entertained me for many hours in 1946) ,when i'm in a hotel but even that is not possible in COD :(

Hope they get it done right because i long for some flightaction again (now i can play only SH5 when not at home :wink:)

ooh and throw in some medals that you can earn :)

This my comment is not directly related to the topic, but BoB II: Wings of Victory seems to be even more unique WW2 flying combat simulator where is the opportunity to be at the same time hundreds of AI planes without the game's FPS dramatically fall, visual displays, even now, sometimes pretty good, although quite an old sim is already the case.

- Excellent video.

The Battle of Britain II WOV AI performance should be the benchmark for any recently released WW2 air combat simulation. That is, if it is still true that as improvements are made as they are continuously improved thereon.

Like Microprocessors, hard disk drives, computer graphics... don't think I need to say more.

Buddye and the BOB WOV community, set a benchmark that is achieveable with AI performance. You can virtually fly the BOB II WOV, and experience the AI. It's not pie in the sky and the community made that happen. There should be no going back to old AI standards or non-standards. The kind of AI performance in the BOB II WOV should be the starting point for current developers at minimum.

Buddye and the BOB WOV community will be glad to share what they know and how to achieve the kind of AI performance they have achieved. You can't ask for more.

Same here now that MP is getting decent numbers!

Hi all,

Many thanks for this! :)

BTW, I remember how we tweaked the AI for F4 RPG 10+ years ago using similar techniques...


Leo "Apollo11"

"What method do testers and players use when testing and refining the BOBII AI maneuvers"? (we have over 80 complex maneuvers for selection [both the aggressive and defensive maneuvers]) for the fighters (Spit,Hurri,109.110) and over 50 simple maneuvers for selection for the JU87 and Defiant).


Sorry, I have just one more boring technical post to try and communicate how the BOBII players and testers worked with a single programmer on the AI for the last 6 years (version 2.03- 2.11). Each and every maneuver was reviewed, tested, evaluated, and refined. This was a committed team approach involving a large amount of work. Is the COD community ready to commit to a large long term task?

We use Move Code Labels in real time as a tool to show which maneuver has been selected.

The best way to learn more about the BOBII AI is to turn on Move Code Labels and watch the AI. A good way to watch the BOBII AI is to use Auto Pilot (control+A) so you can just sit back and watch the move codes change.

BOBII’s label options are the following:

1. Move Code Labels - Used to test BOBII AI maneuvers
2. Short Labels - Short text
3. Full Labels - Full text
4. Single Character Labels - Player selected single character
5. No Labels - For realistic game play

Here is a Pic example and text description for Movecode labels:

1. Move Code --Examples like AUTO_FOLLOWWP (following WP like bombers), AUTO_COMBAT (fighting move code but depends on individual maneuver code), AUTO_LANDING, etc.
2. Maneuver code -- Examples like MANEUVER_DISENGAGE (stop fighting, reasons are many), MANEUVER_TURNINGFIGHT (aggressive maneuver), MANEUVER_BREAKHIGH (defensive maneuver), etc.
3. Range to player (meters)
4.Skill level (Novice, Veteran, Ace, Hero and each subdivided, example Ace1, Ace2, Ace3)
5. Alt (feet)
6. Speed (MPH)

http://www.shockwaveproductions.com/.../movecodes.jpg

Getting community involved in an improved AI wouldn't be difficult. Look at all the community that jumped into improvements for the IL2 Sturmovik. This would not be difficult at all. There are a myriad of enthusiasts of the IL2 series that would be very interested to help develop the AI performance.

At the least I would think the people that are doing IL2 mods would be interested to pursue such a project. The IL2 still has good possibilities from all we have experienced in the past year. The AI would be more difficult, because there are just sooooo many flight models represented in IL2. Possibly, if some type of performance spreadsheet was created for each of the various aircraft it might not be as difficult as I perceive. The data from the spreadsheet could be plugged into the Ai performance coding possibly without too much work.

I remember the 1% flight model spreadsheets made by 3rd party developers for the MSFT CFS2. All to get players on consistence basis of performance, yet the AI never were improved by MSFT to take it to the next level.

The COD devs may not have the option for opening the code for quite awhile, unless they can provide a window for coding as per an SDK or something.

Thanks for the extra information Buddye.

It's yet to be seen how many members of the Cliffs Of Dover community would be interested and committed enough to put in the work required to get the AI up to the highest level. Even amongst the large amount of work done by the modders and hackers of the old IL-2 series, advancing the AI wasn't something that figured prominently IIRC, though maybe that was because the old series already had better AI than what we seem to have now in Cliffs Of Dover.

From what I can gather though, there's no way it's going to happen anyway without the co-operation of the game's developers.

It seems to me there are only two (legal) ways the AI in Cliffs Of Dover will reach the highest level of refinement possible.

One is if the developers hire some hard-working and talented individual who has access to a large crew of play-testers, and get them to work full-time for years on nothing but the Cliffs Of Dover AI.

The other is if they separate the AI source-code out and allow it to be modified at it's most base level by the modding community.

And looking at it that way, I think I know, in the long run, which way would have a greater chance of success.

I'd put my money on the group that can work on it whenever they want, however they want, just because it interests them. The group that can determine for itself, as actual end-users of the game, what exactly needs to be worked on. The group that may even have amongst it's members people as talented and capable as anyone the developers may be able to hire.

Of course, opening up a part of their game like that would be a slightly unconventional approach for a developer to take. But surely the question should at least be examined, is following convention in this instance doing what we have to do to give our sim the best chance of reaching it's fullest potential?

Ideally, what I'd like to see is the AI-modding community working in conjunction with the individual in charge of the AI development at Maddox Games. To have the modders and the developers working on the same page, with a common aim of improving the AI to an unprecedented level. What could be better for the game than to have the enthusiasts doing what only they can do, in line with the efforts of the official developers themselves?

Definitely not top of my list.

Before we worry about the AI we need to be able to fly properly ourselves and we can't with the current FMs.

1. FMs
2. AI Comms
3. Dynamic Campaigns

And for some there is this...

Doesn't matter how far you go with anything, you will always come up short... for someone.

Actually Klem, the AI is an extended labor that could take years. The FM could be fixed within a few days or weeks a most.

What you finally get for FM, may not satisfy you either. Afterall, there are still complainers about IL2 flight models, whose complaints have echoed on forums like this one for years.

I'll take Rotol Prop Spits that perform close to their RL performance versus 109s that do the same. :)

If it gets so close that it comes down to hair-splitting I'll live with it.

Maybe you will get what you want, hopefully not at the expense of an improved AI performance.

Well I understand its a 50/50 call.

I fly exclusively on-line so I'm more interested in human players and our FMs.

For the off-line players the AI is a big issue.

Actually, good FM is a must for the AI, if the AI is subject to the same laws of physics as the player (which is a "must" in MHO).

It would be almost pointless to have well designed, tested and refined AI maneuvers/performance and have poor FM controlling the AI trying to fly the maneuvers.

The Ai as it is is just downright cheating. Tonight I was just messing around and shot the end of a Hurricanes wing off. Instead of losing maneuverability he did a few barrel rolls as quick as a Red Bull air race and lost me before I even had a chance to react.

Thats what i thought. don't they go hand in hand? You cant tell AI to behave a certain way in a FM that's incorrect or doesn't have the correct parameters?

The interesting ingredient missing in discussions of FM. So many people have their own ideas of what is a competent FM. Oleg had his idea and he was rebuffed many times by the community. Responding to a community of people, many of whom are pilots can have it's problems.

I mentioned the 1% flight models created by third parties for the MSFT CFS2. The spreadsheets that were created were a very good tool to plug the numbers and stats, then come up with what was called a 1% flight model. The spreadsheets were developed using actual flight model specifications from the aircraft manufacturer, which seemed adequate. Yet, the missing ingredient to all performance specs... Air combat required a new set of rules. Test pilots didn't take the aircraft to untried and unproven thresholds for failure. These failure points were determined by experience, not by test pilots pushing out to reasonable expectations of failure. So... there is that.

Those spreadsheets are still floating around, and probably attainable in many places.

I won't think to argue about their validity. I'll just say they put everyone on the same playing field.

I agree with nearmiss, the fist and most important issue in FM is to find and agree on the best A/C specs and then stick to the agreed to A/C specs unless a better more accurate spec is found and agreed to use.

Then the FM software uses the the A/C spec data. If you find a FM issue you can not assume it is the A/C spec data. The objective is to fly accurate FM (based mostly on pilot reports, test reports, and old problem reports) using many A/C specs with one integrated FM software package.

This is a very tough debugging and analysis problem and often results in player A/C flight opinions that almost all player have with certain A/C (roll rate, speed at specifixc altitudes, dive speed, climb speed, etc. etc.).

The hard working FM engineer must stick by his baseline A/C spec data and try and understand why the A/C is not performing based on old pilot reports and other old A/C testing/problem reports as many "expert Players" complain and give other spec referrences. The job of a FM engineer is never ending with alot of negitive feedback.

The AI is clearly dependent on the A/C FM with respect to AI quality and performance.

AI is top issue now I agree.

Some for good reason.. most are not.. my sig says it all

I have been reading and tonight just reloaded BOB WoVII to my PC and updated to latest patch. I already have Cliffs of Dover but not beta patch, just latest Steam patched version. I don't think that makes a difference to AI?

I played a 4 vs 4 dogfight with both sims in quick mission builder/instant action, where I was in Spitfires vs 109s. Here is my notes I wrote while pausing and playing, just the AI I was considering:

Cliffs of Dover: Merge is good, both sides trying to swing around ea. other / no one opened fire during merge? / wtf 2 109s go away from fight / one 109 shot down by RAF AI / I have not been targeted yet after two minutes / help shoot second 109 down / look at map, 2 other 109s running away, so chase / catch up and fire, they do not react, shoot one down, no reaction from last 109 / boring so leave him to fly away. no one shot at me in whole dogfight.

Battle of Britain II Wings of Victory: Merge, 109s open fire with MGs at long range, one Spitfire damaged / I have two 109s on my tail / call for help on radio / Spitfire comes to help, 109s break off my tail / I get onto tail of 109 but he dives away / he extends and turns back at me! / I get hit with cannon / I spin and recover at sea level / I climb back up to fight (rudder damage I think) / I fire at passing 109 and he breaks off / I follow and fire and he makes evading turn / I cannot maneouvre good and break off, so does he.

This was only one fight but it is as I remember. You cannot compare these two games for single player AI. BOBII rocks, CoD sux. Very sad for such a beautiful sim.

It's good you did a recall of a personal combat event. No one really connects with competent AI performance until they really experience it.

Keep it up and you'll get very absorbed in the air combat experience. Actually, it is a very excellent training tool as well for Online air combat

You will be a pro at it, if you stay with it. When you go online you'll not get your head handed to you near as often.

I haven't run the BOB II for awhile myself. I, like many others was waiting for a multi-player. It's not that offline play is that bad, it just gets old flying aircombat hour after hour with no human interaction. Guess it's OK for reclusive personalities, but I'm not one of those types. LOL

Anyway, my interest has been tweaked so I may reload and run it as well. Last time I fired that puppy up I was using XP and Vista. Now I'm on Win 7 64bit. Guess there may be some startup issues, but like always I'll know to stay with it until I get it. LOL

I recall there was a more competent AI performace created by "Certificate" when the IL2 mods were a hot commodity. He made a few changes, and made a world of difference in ai. Actually as I best recall the AI were more aggressive for friendly flights. the enemy ai were about the same

AI yes! Defiants aren't suppose to do continual barrel-rolls .. the gunner would be sick all over his turret! :grin:

Pity we don't have scripts that could be changed. eg Tactics of a defiant is going to be a million miles different to any single-seater fighter!

He111.

The strategy of the BOBII AI has been a major topic of player discussion. Do the AI fight, how long, when do they break off and then what?

In BOBII we have evolved to a different strategy for our Instant Action Missions and our two Campaigns (Single Player and Commander Campaign).

In Instant Action (IA), we have assumed that the player is interested in a good scrap (a longer dog fight where he can practice his combat pilot skills). In the Campaign, the AI Squad will break off sooner to prevent unrealistic Squad losses. In both the Campaign and Instant Action Missions, the AI will break off when the Squad losses reach a certain level or the Squad is outnumbered by a certain number.

A individual AI will disengage when he is out of ammo, at bingo fuel, or has damage that prevents effective fighting. When disengaged, the AI will head for home after diving away but the AI will take evasive action if shot at (either a bullet hit or a bullet near miss).

Interesting points Buddye, a good reminder that developing the AI is about more than deciding how the individual planes dogfight each other. Anyone who's ever tried in vain to get through an IL-2 series campaign with minimal squad losses will understand how crucial the development of the strategic aspect of the AI is.

On a related note, I'm guessing that similar to how the individual AI are dependent upon the flight models of their planes, the strategic or tactical aspect of the AI as a group would be dependent on the game having an effective way for the player to command their AI team-mates.

Is working on the command system a part of developing the AI, or is that as separate as the flight modelling is? Seems to me there might be a bit of a grey area or overlap there in terms of what the AI are coded to do, as opposed to what the player can tell them to do. I imagine there must be some kind of heirarchy or order of importance that would have to be determined. For example, did refining the BOBII AI require fixing things like the tendency we see in the IL-2 series for several AI to form a conga-line behind a single enemy AI, often following it to the exclusion of all other considerations? Would be good if stuff like that could be written out of the picture once and for all.

BOBII uses Radio Commands to permit the Player to get information from the AI and command the AI. Some Radio Cammands are indpendent and some must work with the AI. The problem with many AI attacking a single enemy was solved with a option permitting the player to select the max number of AI that are permitted to engage the same enemy.

Here is a list of the BOBII Radio Cammands taken from the User manual:

Radio Keys Fighters Fighter Bombers

Group Info
R-1-1 Report in Report in
R-1-2 Report position Report position
R-1-3 Fuel check Fuel check
R-1-4 Report status Report status
R-1-5 Tight left turn now Tight left turn now
R-1-6 Tight right turn now Tight right turn now

Pre-Combat
R-2-1 Test guns Test guns
R-2-2 Patrol position Patrol position
R-2-3 E/A vector E/A vector
R-2-4 Airborne Airborne
R-2-5 Give freedom Give freedom
R-2-6 Take command Take command

Combat
R-3-1 Bandits! Bandits!
R-3-2 Sighting request Sighting request
R-3-3 Break! Break!
R-3-4 Covering? Covering?
R-3-5 Attack my target (Help!) Attack my target (Help!)

Post-Combat
R-4-1 Regroup Regroup
R-4-2 Send all home Send all home

Tower
R-5-1 Mayday! Mayday!
R-5-2 Home tower Home tower
R-5-3 Nearest tower Nearest tower
R-5-4 Surface wind Surface wind
R-5-5 Wind at altitude Wind at altitude
R-5-6 Land at home Land at home
R-5-7 Land at nearest Land at nearest

Ground Attack
R-6-1 N/A Begin your Bomb run
R-6-2 N/A Strafe the target
R-6-3 N/A Leave area

Content deleted.

Some good AI example for Cliffs of Dover, so it is not all bad. I am trying the campaign mission where in Spitfire you have to intercept incoming raid flying from Manston. Here there is also a Hurricane squadron, already on their way, as your Spitfire squadron takes off. In BOB for real, the Hurricanes generally concentrated on the bombers while the Spits took on the fighters.

In this mission that is exactly what happens, even though the Spits are close to the bombers as they are climbing up, they leave the bombers for the Hurricanes to attack, and they go after the 109s escort. The Hurricanes do manage to bring down some Heinkels and the Spits do OK against the 109s but not great I will say.

Anyway, some good things to say about this element of the AI!

i think that behaviour is not much about the AI itself, but more about how the mission was created in the FMB.if they selected the bfs as a target for the spits, then its no wonder that they will ignore the bombers.
that is more a strength of the FMB instead of the AI itself i think.
but what i recognized in a similar mission today, created by myself, was that a group of bf's, who escorted He's over the channel to dover, flew s lines 500 meters above the He's, to stay in contact with them, although they flew faster.i set both groups at the same altitude, but as soon the bf's spotted the bombers, they climbed up to 1k and began to bank right and left all the time, to keep the high six flying 400kph, while the He's flew at 500m with 300kph.i flew as the wingman of the leader, and was pretty surprised by this behaviour.

Actually in FMB ,you can add scripts for aircraft groups .. does anyone know how that works? :confused:

.

I suspect the script box in FMB is just what is needed, except the FMB box is placed within the code for working with the mission in progress.

I would say it pokes the code in a specific place. There might be some possibility to code some AI in this box, but I doubt it would allow for much.

Luthier did mention SDKs are coming, this might be the best way for coding AI. We will have one for maps, aircraft and AI SDK would be great. The AI performance coding will take a long time and require alot of beta testers among users to get things to make sense.

WOW the prospects of a great AI are just exciting to think about.

Agreed wholeheartedly.


And yet I wonder what exactly happened in the AI development division of Maddox games between say 2006 and 2011 on a typical work day ...


I mean, they had an extremely rich resource of what sim players want and what they like/ don't like about all aspects of the product. More feedback and community interaction than most other flightsims. Themselves worked on AI for IL2 1.0 and later refined it so the development team had long time experience. It was not their first flight sim.

And then look at what we got in the spring of 2011!?

Has anyone created one-0n-one dog fight Missions, so that the enemy AI can be tested at different skill levels and the AI at the different skill levels can be compared in a simple example?

Sorry , I do not use the FMB, I can not change the skill level in the quick missions, and I do not even know the AI's coded skill levels.

What did you have in mind? I could fix those up pretty quick...

AI if I remember is same as old IL2, Rookie, Average, Veteran and Ace...

PM if you want them making up and tell me your email address. Let me know which aircraft you want to fly and which enemy AI aircraft you want to fly against. Merge or dead 6? Altitude etc...

Farber, if you do make some test missions up, you could put them in a zip file and attach them to a post on the forums here, in case others want to use them as well. Zip files can be as big as 14MB per post.