Lock-On's limitations and the "Bubble" system

[coldcrew]

if you bought a sim why you so worried with AI? Lockon excels as a sim but fails as a game. It fails as a game because the AI is weak, so maybe you wanted a game instead of a sim? AI isn't what makes it a sim, MSFS and xplane make that quite obvious.

[GGTharos]

Hawg11, that's a neat idea, but there are problems with it...since the Mig isn't theer, how do you produce a radar blip? How does the AWACS know about it? And so on. A bubble system is no easy thing, and it isn't going to be done through LUA.

 

D-Scythe, you hit the nail on teh head. Threat priority sorting and team tactics are absent in LOMAC AI. A better algo for classifying threats and sorting out which to attack is what's needed - but I don't think we'll see this happen.

 

Anyone notice that LOMAC's AI doesn't understand what an F-Pole is, either?

[DayGlow]

if you bought a sim why you so worried with AI? Lockon excels as a sim but fails as a game. It fails as a game because the AI is weak, so maybe you wanted a game instead of a sim? AI isn't what makes it a sim, MSFS and xplane make that quite obvious.

 

says who? At simulates a battlefield experience. Part of that is correctly acting planes and ground units. I fail how the combat side of a combat sim doesn't add to the overall product.

[GGTharos]

Well, to use Falcon as an example, a radar blip won't show from anything outside the bubble.

 

However, you do hear comms from AI planes outside the bubble. This could be achieved with the method I'm speaking of quite easily.

 

Notice that most radio traffic in Falcon (except wingman commands), happens within the "campaign" rather than the "game." To prove this, load up Falcon and start a campaign and set time compression to max. You will hear packages talking before you even choose a mission to fly. This is because those packages are talking within the "campaign," not the "game" itself.

 

Basically, with the method I'm speaking of, you could have all sorts of voice traffic going back and forth through events happening within the "campaign" rather than the "game" with VERY little overhead on sytem resources.

 

As you can see this method enables the possibilites of having truly amazing voice comms from flight packages, airport towers and ground contol, artillery units, FAC, etc.,

 

And yes, I stand by what I said. It could all be done with LUA, a simple c++ campaign engine, and access to portions of the multiplayer code.

 

 

Let's *not* use Falcon 4 as an example then, since its system is merely adequate and not particularely good - it works, but it only works 'so much'. There are probably ebtter ways to do things instead of preventing the player from interacting with the full range of his platform's capabilities - of course, on the other hand it also so happens that the F_16's radar range ain't all that, either ... but in any case it's a bad example (it was good when it was conceived) and a bad resolution (because detection is posisble at over 100nm for some targets). Writing a game, and in particular a sim, isn't about 'let's do it the way the other guy did it'. You want to do it *better* ... why limit yourself to solutions that applies to computing capabilities of 10 years ago?

 

Your method really enables nothing but someone patching code on top of code which is definitely NOT a great solution - you're running your campaign and comms in AN INTERPRETED LANGUAGE on the same computer as you're running everything else, and expecting it to pretty much run AI, too (you don't think you're proposing that? THink again!) it is in fact a pretty terrible solution as you need to pass data twice to do things with them. This is only valid if it's in the game already - anything you do otherwise needs to be measured. Worse, how it is going to sync in MP? You get even more MP overhead now because of this.

[Mako]

Yeah, you're right a bubble system would allow for more players online as well as much more units involved.

 

How does a bubble work in multiplay? If another human group is fighting AI (or other humans) outside of my bubble, how is it handled?

 

Is my question irrelevant to this discussion? Does the bubble system only apply to offline play?

[DayGlow]

each player has their own bubble I imagine. If you attack a target outside another player's bubble, all your machine would do is update the other boxes of what you attacked/destroyed, etc and it would adjust it's statistical computations for that.

 

I think F4 works with the server running the campaign engine and sending and recieving info to the networked computers of what was attacked and or destroyed. Some people run F4 SP with a server box to offload the campaign computations. No idea how much FPS at the FLOT they save doing so.

[Prophet_169th]

Sounds like a great idea Hawg.

 

The only problem, is that it doesnt appear that LOMAC can add units after the mission starts.

 

I have tried setting units to spawn later than when you enter the mission, and they never show up.

[GGTharos]

It's apparent you don't understand what I'm trying to explain to you.

 

I thought it would have been clear that I used 30 nm as an example only. Of course, in Lo-mac you would want to have the bubble spread a little further than the player aircraft would be able to detect.

 

Also, what I proprose is not "code on top of code," it's modular design. The advantage of a modular design system is that the campaign COULD run on a separate computer. I know I've got two outdated computers in my closet that would go to good use running this type of system.

 

For more reading about modular design see xplane. The code is made so that an insructor station, extra monitors, guages, etc,. can be run on a separate computer. Modular design improves perfomance, it doesn't hinder it.

 

The game always runs the AI; the campaign only runs the campaign. The campaign never runs the AI. Please re-read orignal post for clarification.

 

Bottom line is you said with with the current engine it's imposssible to have a DC. This is wrong and I've shown you why.

 

Clear as mud?

 

Cheers.

 

 

I didn't say anything about a DC - I said what the issue is regarding things while playing a mission. As for a DC, the AI is too insensitive to what's going on in 'the world' to have anything but the most rudimentary DC (Which would be better than the 'nothing' we have right now) ... however I'd prefer some solid triggers and AI behaviors to be programmed WELL before anything like this is implemented. To attempt a DC without those extra features is asking for dissapointment. Really, you're not adding that much ebcause you can't really interact with things while in the mission - it may be a bit more interesting, but it'll be sterile while you're flying.

[coldcrew]

They don't have to unlock anything. ED could add spawning ground units after a certain time. AI planes can already do this in multiplayer but not ground units. Without access to the code there are a lot of limitations on implementing certain features. It's too easy for non developers to say how trivial it is to add stuff.

[coldcrew]

There is a difference between "it can't be done" and "why it wasn't done". Developers make decisions to add/remove or not implement features all the time. AI planes can spawn at different times in the missions, AI ground units cannot. Only ED knows why this is so. As for the source code, forget it about.

[coldcrew]

you're welcome. sometimes the clueless need to be clued. Why ED *needs* to release portions of the multiplayer code to the community still remains a mystery

[D-Scythe]

Anyway, so about that pesky AI...

[coldcrew]

I know what you are talking about. Suggesting that a bubble system could be made with lua scripts or plugin code showed you don't understand how F4's bubble system is implemented.

[Stormin]

Just about anything can be done in development provided the proper amount of time and money are allocated to the effort. Even then, little things like practicality, opportunity cost, and return on investment have a tendency to influence the decisions.

[Dmut]

"bubble system" in F4 is not a good way to develop in novaday sim. They implement it only because of limited CPU resources at that time. It simplify overall mission calculations but making missions tracks and massive multiplay weird. Thanks to ED LO didn't used bubble system from the start. IMHO, implemention that system now would be a step back for LO.

[SUBS17]

Oh comme on. F4 AI are you kidding ???? Where did you see in real life an A/G mission being ended with 4 kills for a plane heavy loaded with A/G ordonance ???

 

Air defense system is much beter in Lockon than it is in F4. F4 AI isnt a real AI. It will always react same no matter the situation wheras AI from Lockon wouldn't.

 

If you call an immersion the fact that the AI plane should rush on air defenses then I suggest you to study a litle bit better ROE for every army in the world.

In Lockon the AI trys to preserve its "life" and in "F4" the AI is a terminator.

 

Can you telle me what makes you think that the AI in F4 is much better ???

 

 

For Note :

 

I have a copy of F4 patched at SP4 and I'm playing Balkans.

 

 

The ai in Allied forces is way beyond anything available at the moment if even half the statements from the developers are true. The new ai invloves pilots taking into account survivability when on the ingress. if the target is too heavily defended then the ai aircraft will just head home. And thats just one comment on the new generation F4 ai. If you don't believe me then goto the lead pursuit site and read ramping up the ai part 2. Absolutely amazing :cool:

[ALDEGA]

"bubble system" in F4 is not a good way to develop in novaday sim. They implement it only because of limited CPU resources at that time. It simplify overall mission calculations but making missions tracks and massive multiplay weird. Thanks to ED LO didn't used bubble system from the start. IMHO, implemention that system now would be a step back for LO.

Track recording would indeed not give you a "regular" track with all objects at "100% simulation resolution" (perhaps objects will simply disappear when they were not in the bubble when the track was recorded).

 

I think people will gladly accept a less complete track-recording feature if it means that the game itself can become "more".

 

In order to simulate a large number of AI (tied with a DC or not) something in the concept will have to be changed.

 

The challenges appear to be right now, for a future ED project, or a project afer that future project ( ;) ):

- increase number of AI

- increase realism of AI

- implement DC

- maintain good performance

 

If the current "all units at 100% simulation resolution" concept stays then it will be very hard to achieve good performance. Don't forget that in future projects there will be new/enhanced features that will also take more CPU cycles, besides the AI simulation.

[Orao]

You all seem to forget something. All of you want an AI which feels and act as a human. However I don't know how many of you tryed to program the AI, for those who did they know that during programing process of AI a "tree" should be elaborated.

 

I'll expalin what do I mean.

You have an AI. An AI encouters another AI or a player. Based on that situation the AI should try every possible awnser test it in simulation and take the awnser which gives it the greatest probability of success. Now imagine the number of possibilities for for initial situation and imagine the nomer of awnsers. If you consider that you have lets say 4 initial situation then the number of possibilites to test would be something like 4 power "n" where the "n" is how deep you want the AI to go in the awnsers. The greater the n is the more likely the AI will behave as a human.

This is just for one AI. Now add more AIs and add a physic engine plus graphic engine.

 

NO PC in the world will be able to handle that. Once again the AI from F4 is using scripted actions. The awnser will always be the same no matter what.

 

However there was a game on the market made by guys from Matemathical Institutes called "Conflict Zone". They developed the "self-learning" AI. The AI was learning from human player and was generating its own code which was stored withing game directory. After 50 to 100 hours of play it was almost impossible to beat it. That game was a real time strategy type.

Such approach could be an awnser for what we all want.

Those students wanted to develop a Direct AI. Some sort of AI sub-routines just like Directx for gfx but for some reason game industry didn't want that.

 

Check the link for this game and give it a try (it's only 6€ now)

 

http://conflictzone.ubi.com/ind_01.htm

 

The name of company who developed it is MASA.

[SUBS17]

Hopefully the Lockon successor will be able to handle more objects and much greater depth to the ai. Theres also alot of other things that can be created by accumulating information on ai behaviour in past flight sims to produce better ai responses for each circumstance. Ultimately Human players are the ultimate answer as every reponse is orginal and not scripted. BTW not all scripted behaviour is bad, neither is scripted missions if they are done the right way.

[Starlight]

"bubble system" in F4 is not a good way to develop in novaday sim. They implement it only because of limited CPU resources at that time. It simplify overall mission calculations but making missions tracks and massive multiplay weird. Thanks to ED LO didn't used bubble system from the start. IMHO, implemention that system now would be a step back for LO.

 

mmmm... you may be right but your example was not that good :)

Falcon 4 has a DC, LockOn has not.

Falcon 4 can manage a large number of flights/units, LockOn can't.

It seems that the F4 approach seems at least better than the one used in Lomac.

 

Today's CPU are maybe 10 times more powerful than the ones we got when F4 came out. But when you're talking about AI, trees and in general about algorithms, a non-optimal approach can lead you to performance which are much more than 10 times worse (sometimes we can increase the computational cost by orders of magnitude, or even with exponential growth rates).

 

I'd also like to remind you that F4 was SLOW on those machines, even on the suggested high-end ones (P2 450Mhz, from F4 manual). Given that a modern sim would have to manage a lot more details (at least about graphics, even if helped by GPUs), I don't see a reason why the bubble model (maybe with some improvements) could be labelled as obsolete due to the new CPUs. Even today we can't afford to waste CPU cycles by using a non-optimized approach.

 

And, just IMHO, tomorrows' sims shouldn't be published without a good AI and a DC. They would just be games, not sims, like those for consolles.

[Orao]

May I remind you that F4 when was released had multiple memory leaks and still had it even after 1.08 US patch. Performance hit had nothing to do with the AI.

Besides F4 and Lockon Algo for computing the AI have nothing in common. If you admit that then you will also admit that the buble system simply can't be implemented. Lockon is computing the AI on the theater level while the F4 is computing the AI on player's level.

 

As for last sentence I agree with you. Just remeber that ED had planed to build the DC for Lockon. Why they didn't do it only them can awnser eventhough we have the begining of the awnser.

 

However ED has prooved us that they are listening to the community and I beleive that with the experience gained over Lockon developement they'll give us an ulitimate warrior in simulation world with their next title.

[Dmut]

mmmm... you may be right but your example was not that good :)

Falcon 4 has a DC, LockOn has not.

DC and bubble system is not connected in anyway. It is possible to implement DC with current units calculation system (but implementation of DC is not worth a time, regarding to LO brand)

Falcon 4 can manage a large number of flights/units, LockOn can't.

I would call it "half-managed" in case of F4's bubble system ;) in opposite of LO's true units management.

It seems that the F4 approach seems at least better than the one used in Lomac.

not so "seems" to me, sorry ;)

Today's CPU are maybe 10 times more powerful than the ones we got when F4 came out. But when you're talking about AI, trees and in general about algorithms, a non-optimal approach can lead you to performance which are much more than 10 times worse (sometimes we can increase the computational cost by orders of magnitude, or even with exponential growth rates).

F4 was slow on typical hardware in that time of release, about to similar to LO on current typical PC in FPS case. but F4 used "bubble" and LO - doesn't, so who has made a "true" optimal SW? :)

I'd also like to remind you that F4 was SLOW on those machines, even on the suggested high-end ones (P2 450Mhz, from F4 manual).

good, at least we agree on this statement :)

Given that a modern sim would have to manage a lot more details (at least about graphics, even if helped by GPUs), I don't see a reason why the bubble model (maybe with some improvements) could be labelled as obsolete due to the new CPUs. Even today we can't afford to waste CPU cycles by using a non-optimized approach.

Why do you think that CPU cycles must be spend to increase a number of unit instead of computing every unit's small features like phisics, AI, tactics?

And, just IMHO, tomorrows' sims shouldn't be published without a good AI and a DC. They would just be games, not sims, like those for consolles.

I agree on this one too. DC will be a truly nice addition to LO's successor.

 

ps: I must mention, that I really respect F4. It was and it IS a great sim, with some advancement over other sims. I didn't played much but I pass a traners lessons of SP3.

[Orao]

Lockon do same thing only the AI of Lockon is much more advanced that's all so it demands more cycles to compute.

[GGTharos]

But that's where you are wrong, sir. To prove this, load up a campaign in F4 and and set time compression to max. You will begin hearing packages dropping on target, RTB, sunrise, etc., all before you ever even choose a mission to fly.

 

At this point the 3D world has not been loaded; the campaign is running completely independently of the 3D environment.

 

What happens is, the "campaign" runs in 2D tables. Once the "campaign" decides that the player aircraft would be able to see certain units, it loads those units in the 3D environment.

 

The campaign and the bubble system are completely intertwined in F4 (one was created so that the other could work properly); they could not be separated.

 

 

No, that's pretty much where -you- are wrong. The two are very separate. They may be complimentary, but hey are two different things created for two DIFFERENT reasons. They just happen to cooperate, which is good - what they are not, is perfect. The bubble system as it exists in F4 at this point in time is antiquated and needs major tweaking at the least to take advantage of what can be done with computers today.

[Dmut]

But that's where you are wrong, sir. To prove this, load up a campaign in F4 and and set time compression to max. You will begin hearing packages dropping on target, RTB, sunrise, etc., all before you ever even choose a mission to fly.

he-he... you mean - if visualization phase is no running then we call it "bubble" ? :)

than you will see LO's "bubble" system in dedicated server - screen is black but planes are flying and packages are exploding :D

seriously, I don't call "before-a-fly-time" as evidence of using "bubble" system; it can be easily modelled in LO - just start you mission and turn monitor off. From programmer point of view - everything what F4 calculate in menu time can be calculated in LO without visialization, which itself takes a lot of CPU time.