Cheating AI

[aimmaverick]

I cant believe that nobody doesnt care about cheating AI in air combat. The enemy always detects you first, can burn through yours ECM sooner and have overwhelming SA. If I go with my F-15 against F-16 or Mig-29 for instance they can always detect me first and burn thru my ECM despite the fact that Eagles radar is vastly superior. And this happens against all aircraft exept F-5!!!! How much realism is here if AI is so strong? Where is each aircraft unique radar moddeling?

And strangely it doest matter if you go against F-18 or -16. It appers there is no difference at all. In that game I am feeling beeing in the Tiger and not Eagle.

Will this be finnaly fixed in 1.2 or are we have to be satisfied with current model?

It doesnt make sense to have any serious campaign if you are from the start in worse situation over the enemy.

[GGTharos]

Actually, HOJ lock by Russian aircraft will give you a lock warning now ... did you miss this in the readme?

 

Yes, you're right, AI has uncanny awareness, and it won't be changed soon: It's just not a simple thing to do. All AI have the 'same' radar and other sensor logic, AFAIK, and the only thing that changes it probably the range.

[S77th-GOYA]

Yes, they cheat, but they're morons. So, it kinda evens out. I seriously doubt it will get any better in 1.2.

[S77th-konkussion]

Yep. Very predictable. 2 A10's could take on an AI flanker and win- he's too stupid to engage them both.

[Gazehound]

Very little point terrain masking when the AI knows exactly where you are and locks you as you pop up :(

My advice - fight people until they fix it :D

[504MrWolf]

Very little point terrain masking when the AI knows exactly where you are and locks you as you pop up :(

My advice - fight people until they fix it :D

 

Are you saying he should go out and kick some ones head in ?? :p

[D-Scythe]

This would fix a lot of the problems associated with the AI in A/A combat: http://forum.lockon.ru/showthread.php?t=11915

 

Unfortunately, I don't think any of it would be implemented.

[tflash]

The biggest problem is that the AI's ueber-SA turns every fight into a head-on engagement. The AI always turns your way.

 

This makes practicing realistic scenario's rather difficult.

[bSr.LCsta]

SA or lack of it is a human quality. If you think about it, all you are is some code and the Ai is reacting based on the perameters it was written to.

You can never "hide" from Ai, only know its perameters and engage/avoid/react accordingly.

 

p.s. they are very frustrating though!

[GGTharos]

No, SA is not just a human quality.

 

Those same parameters that you mentioned constitute SA, and you'd be 'all knowing' if you were being fed the same data ;)

[Gazehound]

Are you saying he should go out and kick some ones head in ?? :p

LOL Yeah, he should also get drunk and puke on a girl's lap. :icon_jook

[bSr.LCsta]

hmmm...I guess. I don't think of it as situational awareness when talking about a program's code, compared to a persons ability to deduct what information is present (not necessarily fed to them) and make a decision.

I think the human element is a big part if not the biggest part when describing SA. ;)

 

Originally an aviation term used in human factors to describe awareness of tactical situations during aerial warfare. It has now been adopted throughout aviation, and increasingly in transportation and other dynamic complex systems requiring human control.

 

Situation awareness (SA) is important for effective decision making and performance in many domains, including aviation, nuclear power, chemical processing, automobiles, air traffic control, medical and health systems, teleoperations, trains, space operations, maintenance, and advanced manufacturing systems. In these complex and dynamic environments, human decision making is highly dependent on SA — a constantly evolving picture of the state of the environment. Situation awareness can be described broadly as a person’s state of knowledge or mental model of the situation around him or her.

 

A general, widely applicable definition describes SA as ”the perception of the elements in the environment within a volume of time and space, the comprehension of their meaning and the projection of their status in the near future” (Endsley, 1988). SA involves perceiving critical factors in the environment (Level 1 SA), understanding what those factors mean, particularly when integrated together in relation to the decision maker’s goals (Level 2), and at the highest level, an understanding of what will happen with the system in the near future (Level 3). These higher levels of SA allow people to function in a timely and effective manner.

http://www.reference.com/browse/wiki/Situation_awareness

[Yellonet]

LOL Yeah, he should also get drunk and puke on a girl's lap. :icon_jook

It would probably happen in the opposite order though.

[GGTharos]

hmmm...I guess. I don't think of it as situational awareness when talking about a program's code, compared to a persons ability to deduct what information is present (not necessarily fed to them) and make a decision.

I think the human element is a big part if not the biggest part when describing SA. ;)

 

 

http://www.reference.com/browse/wiki/Situation_awareness

 

What you quoted describes what the machine is doing (or trying to do) as well.

 

The difference is that a person can lose SA more easily due to factors that the machine doesn't experience, ie. panic, etc.

 

Also, you're right, a machine can't 'predict' or 'suspect' as such, but this is of no consequence in simple situations where it's 'looking around' ... since it doesn't need to predict or suspect in that case.

Either way, you could incorporate prediction/intuition, but it would be computationally expensive.

 

If you're talking about the intuitive part of SA, then I agree with you.

[bSr.LCsta]

Either way, you could incorporate prediction/intuition, but it would be computationally expensive.

 

Yes the "if", "then" coding possibilties are virtually limitless....where to draw the line?

[GGTharos]

You don't use if-then's for that sort of code.

 

With all due respect, read an intro to AI book ;)

 

If-then's constitute a strict and unbending ruleset.

 

AI will use things like fuzzy logic (let's call this a 'relaxed' if-then), neural networks, support-vector-machine classifiers, and all sorts of other things that most people don't really undestand.

 

While you'll certainly find 'if-then's' in that code, the if-thens no longer dictate the decision - the mechanisms are quite different.

 

It's when you try to make up for things with if-thens that AI becomes poor. Also, poor training of your neural nets/SVMs and whatever other method you wish to use will cause issues.

 

Not having enough/correct inputs/outputs in the AI code will cause issues. (Eg. you cannot avoid over-g if you don't have a g-input, and you can't turn left if you don't have a turn left output - you can take these farther than such simple actions, of course)

 

It's a pretty complex subject.

[bSr.LCsta]

always gotta get the last word in, huh?

I was making a general statement regarding coding (as I am one of the morons that don't understand the technical side), and was in agreement that tons could be done, but at what cost.

[GGTharos]

I'm sorry, I misinterpreted your post.

 

Yep, if we only have 2-3 boxes in a cluster to run the AI... ;)