The AI wishlist

[D-Scythe]

Just interested in what everyone thinks, as this issue comes up a lot. The question is, in everyone's own opinion, what would be your ideal AI, either as a wingman or an enemy? What would they do, and when?

 

:?:

[EricJ]

I think some friendly AI needs to be improved, such as heading out 100 kays to make an attack (not real so, but if you want to boogie, mit wingman, it's impossible) and having to repeatedly call them in. On the other hand though, the AI is much more dependable in LOMAC than it ever was in Flanker. This creates the bias that wingnuts are good targets for the enemy AI, and treated as such, or an additional bomb truck. But I feel more safer when I got one.

[GGTharos]

Already fixed.

[D-Scythe]

Well, ED can do some other stuff with the wingman, to provide much better mutual support so we don't see them as just bombtrucks or expensive decoys. For example, expand the functionalities of the "Cover me" command (like if the order is given, and they detect an enemy fighter flight where you do not, they engage each plane with active radar missiles as you engage you're targets), or if you're flight outnumbers the enemy, that the extra elements drag left/right accordingly.

 

Also, a reduction in reaction range for fighters assigned the role of CAP or Fighter Sweep would be awesome, as currently fighters will typically make a beeline to targets over 150km away. And there should be a difference in behaviour between flights that are tasked with fighter sweep and those tasked with CAP: CAPping jets should do anything they can to defend a piece of airspace, while jets on fighter sweeps should opt to lob their missiles at long-range and be more inclined to avoid the WVR fight unless their advantage is definite. Such a distinction would make the game much more immersive, as aircraft like the F-14 and MiG-25/31 would be awesome for fighter sweeps while others like the F-15 and Su-27 be assigned CAP duties.

[Caretaker]

I see three major aspects for AI behaviour:

 

1) controllabilty

2) efficiency

3) realism

 

Let's start with 1) - when designing missions, it's important for me to have the AI do what I want them to do. How to react to threats from both air and ground, which targets they attack or ignore, and when to abandon a mission. Ideally a complex trigger system is in place that enables me to react to any kind of event that happens - enemy fighters scramble somewhere, a SAM site locks onto the plane, a wingman gets damaged - whatever, the more, the better. Jane's F/A-18 is a good example of this. Such triggers have to be planned from the start for a sim, but are not too complicated to implement - often it's rather simple binary logic.

 

Controllability is also an issue for AI wingmen of course - there has to be a reasonable amount of commands that can be given, and the AI should be able to execute them as ordered.

 

Now for 2): this is a more complicated area - to be efficient, AI has to be able to fly sensible maneuvers within a 6DOF environment, manage target priorisation and make "judgements" that humans do more intuitively, so that they are hard to put into rules. Choosing the right weapon, at the right distance and aspect to an enemy plane, together with the fitting attack profile, is as much an art as a technique:) Also, rfficiency relates to such basic (but not trivial) things like evading mountains.

 

Point 3) means that the AI behaviour has to stay within believable limits - both for the physical bounds that human players have to live with, SA (like losing tracks of targets in clouds or behind the plane, or simply when there are too many enemies around) and also human behaviour like panic or target fixation.

 

Those aspects do of course influence each other. With a complex trigger system, it's possible to have planes react realistically to a situation, even if it may need a lot of scripting for that. This can also produce the diversity in behaviour that makes AI less predictable and thus, more interesting.

 

 

For Lock On we won't see a complete redesign of the AI part to allow for all this. I think controllability is the biggest problem with the AI, along with some other issues. It's still possible to get it to a good level as long as work on that area continues. And in some regards, Lock On's AI is also quite good already - I'm not totally happy with it, but I've also seen worse. A much neglected topic in many sims unfortunately.

[Guest ThomasDWeiss]

Hear! Hear!

[GGTharos]

AI behaviour can be easily made 'better' with application of certain AI concepts (which I won't go into here) that would enable maximum performance for minimum CPU cycles - I don't know what choices ED has made in their AI design nor why, but then not everyone is an AI expert -nor- is it an easy subject so AFAIK they're forgiven for the time being ;)

 

For some AI (SAMs for example) the AI 'code' is pretty simple as all you need to do is prioritize threats (the sorting algo for those shouldn't be hard to implement, doesn't need to be complex either as you're only prioritizing on things like size of return, vector and speed. You could get more complicated and get the SAMs to assist friendly fighters or protect areas of interest ie. ground assets which won't really cause CPU-bound problems but the algorithmic solutions become a lot more complex)

 

So far I'm happy with what I hear they've accomplished in 1.1, it will at least allow us to contorl missions better.

[MikeRocker]

One really annoying AI behavior I often come across is the way AI wingmen change formation. When I am going head to head with some bogies I would tell my wingbecile to go line abreast and widen the formation, or tell him explicitly to bracket right. He then goes and does a big break maneouver where he ends up MILES behind me instead of just sliding outward gradually. Also I had the impression that on receiving the bracket command, AI will revert to the altitude entered in the flightplan. I can tell you that its quite annoying when you enter BVR at 20k and your 'best friend' suddenly dives for the ground just when he's supposed to be supporting you..

 

My pet peeves...

[D-Scythe]

There are some other things they can do with the SAM and ground AI logic now too, as right now it is far too two-dimensional:

 

• i) For one, they can try to implement some real life SAM tactics, like ambush (where a SAM waits until a targeted flight is well within no-escape range before shooting) or buddy launching (where if two identical SAMs are hooked up close together either site can use the other's radar) or simply turning off and on the radar periodically

 

• ii) In terms of AA combat, lower generation fighters like the MiG-23 should seek to stay low or behind a moutain, and thus undetected when on CAP duty. Resulting engagements would then start off from very close range as the enemy is not only detected at the last possible second but also missile range is reduced, thus nullifying the technological disadvantage of these early fighters. This could make for some spectacular air combat engagements over the Caucasus mountains

 

• iii) In BVR combat, the "hit the burners and make a beeline to target" gets boring fast and is unrealistic. Most fighter pilots don't just make a head-long charge at the enemy. Again, ED can do a lot with AI fighters providing mutual support for each other, attacking and defending in pairs rather than singly as they do now. In a 4 vs. 2 battle, the two extra planes on one side should support their lead rather than fly stupidly forward; likewise, any outnumbered F-15C or MiG-29S should let loose with their active radar missiles at all targets within range.

 

• iv) Target priority should be a bit better. F-15s or Su-27s assigned Escort or Fighter Sweep should engage hostile fighters first, then move on to engage strike aircraft. Fighters on CAP should be more inclined to engage strikers, even if it means lobbing a few missiles at any escorts first.

 

On the subject of targeting, I find it hard to believe that modern battle tanks like the M1 and the T-80 need 10 shots just to kill a single vehicle from 2 km range. That's ridiculous. Typically, such tanks should open fire at 4 km, and accuracy should almost be tripled from what it is currently to reflect a realistic level of gunnery from the tank drivers. If the M1 gunners in Iraq were as bad as they are simulated in Lock On, whole cities would be flattened due to errant shooting.

[GGTharos]

i) THat would require IADS simulation for some of them, if not all (SA-8 for example and SA-15 can use optical tracking, but they have to knw the plane's coming first - S300 is far too long ranged to work on a visual detection basis so it would have to be linked to an EWR)

 

ii) No. You simply assign'em to targets where shorter range of their systems is ok, and they're backed up by SAMs. THat's how it works. You won't always have mountains available, and it also might cause you to miss those fighter-bombers zipping by in an adjacent valley ...

 

iii) agreed

 

iv) Agreed.

 

On the subject of tanks, assume active evasive maneuvers and whatever countermeasures vehicles woudl have available such as smoke. The sim doesn't model them so worse gunnery is a relatively good approximation against them.

[D-Scythe]

THat would require IADS simulation for some of them, if not all (SA-8 for example and SA-15 can use optical tracking, but they have to knw the plane's coming first - S300 is far too long ranged to work on a visual detection basis so it would have to be linked to an EWR)

 

Not entirely. Currently, fighters on Intercept, CAP, GAI or fightersweep already act differently with a system like an AWACs than they would without AWACs - a MiG-29 with AWACs will engage targets 150 km away, while without it it would only engage if it detects a target. So I can't see how it could be that much different with SAMs - with EWR (or AWACs), advanced tactics are used, and without, it reverts to being stupid :D

 

No. You simply assign'em to targets where shorter range of their systems is ok, and they're backed up by SAMs. THat's how it works. You won't always have mountains available, and it also might cause you to miss those fighter-bombers zipping by in an adjacent valley ...

 

It's called an ambush CAP. And staying low is a great tactic against an opponent with a longer reach. For example, let's say you're in a MiG-29S fighting F-14s, and their AIM-54s have 80% more range than your R-77s. At 35 k ft, when your R-77s can kill things at 30 nm, the AIM-54s will out-range you by an extra 24 nm. However, if at low alt, you're R-77s have a range of 10 nm, then that difference shrinks to only 8 nm. It's not nearly as simple as that, but the point is that for all missiles they glide a lot longer than their motors burn, but at low altitudes 'glide' is considerably reduced due to higher drag. If I was a MiG-23 or F-16A pilot on CAP without any SAM support and my RWR picks up two MIG-31s inbound, I'd drop to low altitude to take away the *very* long-range missile shot and lure them into a knife fight.

 

In any case, Falcon 4 had it :P

 

On the subject of tanks, assume active evasive maneuvers and whatever countermeasures vehicles woudl have available such as smoke. The sim doesn't model them so worse gunnery is a relatively good approximation against them.

 

Take the velocity of a tank gun, measured in kilometres per second. Take a 40 ton tank, whose max speed is measured in centimetres per second (Assuming that it is travelling at max speed - in all likelihood, it would be zero). What kind of evasive manuevers can a tank possibly do to avoid such a projectile? And smoke is not that effective - in most cases, it reveals your actual position to the enemy if anything. And in tank on tank combat, staying hidden is the ultimate defense, even better than depleted uranium armour, which won't help you much if you got T-80s shooting at you from the front and Vikhr missiles streaking in from behind.

[GGTharos]

If your RWR picked up MiG-31's you'd be about to die (AFAIK they can lock you up and fire well beyond your RWR's capability to detect them. AFAIK)

 

I've never heard of anyone doing a CAP like that, it's insane. The F-14 can jump up to 40k and launch its phoenixes which will loft to 160k feet, which will then drop RIGHT DOWN on those MiG's - the geometry mostly prevents the MiG's from doing anything useful like hiding in mountains, and if they do, they're also liekly out of contact with their EWR and unaware that those 14's are just repositioning for better detection geometry. At 20nm 40k feet there's almost no way to hide! (Sure, you could do this and that, but technically speaking those MiG29's are toast)

 

As for SAMs, again they could wait for their quarry to be in visual range and then just slave the trackig radar to the optical sight (they spot with mk1 then train the optical on it) and get a lock that way right away, then fire. Very nasty.

 

Oh by the way, if smoke 'isn't that effective' then why's it being used? :P

 

Smoke is -highly- effective. It'll screw up the LR/D, cause you to lose sight, and in general multiply your problems given the slightest deviation of your target from its original course, and you have zero guarantee of a first-shot kill against an enemy such as a T-80.

[Starlight]

LockOn's AI is quite poor compared to other aspects of the game and compared to the AI of other sims, even if they're much older (F4). And with the AI it has today I think it would be hardly able to manage any dynamic campaign too, if this was to be added.

 

The first problem is in mission planning/execution. The mission planning is very far from being precise and flexible. You can just assign waypoints with simple actions to perform, but, even if they were accurate, you're not sure that the AI will perform what it's supposed to do.

I'm talking about simple things... today I started a mission with 4 flights of four F-16s each (16 A/C together), starting from shelters. To have all of them in the air they took 10 minutes, and when the last aircraft was taking off, the first one was really close to its target. How could this be possible? Why the leader of each flight does not wait on the runway or loiter in the air for its wingmen?

 

Then, the attack phase was awful, with the F-16s slowing down to nearly minimum speed to drop their bombs.... mmm... the mission editor doesn't let you choose which attack profile to use, but this is nonsense anyway!

 

Even the RTB was a problem, because at the first try all aircraft dispersed in different airports, instead of returning to the one they took off from. Still wondering why, since it was fully operative... The second time I tried the same mission all the aircraft returned to their home airfield, but when landing they had troubles in loitering and in the final leg, so some diverted to other airfields (even if they didn't hit bingo fuel).

 

To sum up, LockOn has serious problems with the mission planning phase, both because it's limited (it doesn't allow formations, ROE (rules of engagement), attack profiles, weapon delivery type and things like that) and because the execution of the orders is really bad, and it's really hard to have the AI units do what you really want them to do.

 

LockOn has great graphics and great gameplay but it really lacks the sense of "real thing" given by Falcon 4. The devs of LockOn really cared a lot to convey realism through graphics, realistic weapons and FMs. But without a good AI which uses them, the single player part of the game is much like an "airborne shooter", not a realistic sim.

[D-Scythe]

Oh by the way, if smoke 'isn't that effective' then why's it being used?

 

Smoke is -highly- effective. It'll screw up the LR/D, cause you to lose sight, and in general multiply your problems given the slightest deviation of your target from its original course, and you have zero guarantee of a first-shot kill against an enemy such as a T-80.

 

If smoke is that effective, why wouldn't every tank be equipped with them? The way you're making them out to be, they may as well be as important to tanks as chaff is to fighters.

 

Smoke gives away your position. And in ground combat, position (and the ability to change position) is everything. Sure, it'll screw up LR/D, but the enemy now knows *exactly* where you are. So unless you intend to flee the battle afterwards, using smoke is almost suicidal, because I can guarantee that once you're cover is blown, at least four other tanks/anti-tank teams would be bearing in on your position looking for the kill.

 

Smoke may be effective against fast CAS jets, cause the next moment they're gone, but against other tanks and gunship helicopters? Or concealed anti-tank infantry? It's suicidal - you may as well tag a big "Kill me" sign on your tank. Moreover, how are YOU going to see past you're own smoke? ;) It puts you on the defensive and keeps you there.

 

It's analogous to jamming, sortof, in that it instantly gives away your positon (and you're exact position if your enemy has RHAW). A lot of fighters are equipped with them, or can be equipped with them, and use them, but they don't leave it on. Except that jammers are continuous, where smoke grenades are not, and fighters can move a lot faster to hide themselves again if necessary, while tanks again cannot.

 

The F-14 can jump up to 40k and launch its phoenixes which will loft to 160k feet, which will then drop RIGHT DOWN on those MiG's - the geometry mostly prevents the MiG's from doing anything useful like hiding in mountains

 

I'll guarantee that an F-14 at 40 000 ft will not be able to detect, much less engage, enemy bandits hiding in ground clutter beyond AMRAAM range. Thus, the Phoenix then becomes basically a $1 million AIM-7 with active radar. And, I've always understood from what I gathered from actual fighter pilots that it is not a good idea to stay more than 20 000ft above your target, or else certain advantages can be exploited by the enemy against you ;)

[GGTharos]

Every tank IS equipped with smoke launchers! It's STANDARD EQUIPMENT!

 

It's also USED against tanks! That's the standard tactic when you're udner attack in a tank, or APC - back up, POP SMOKE!

 

Man!

[D-Scythe]

I'm not saying it isn't, but that it is a last ditch move to do when you want to retreat.

 

Think about it - how would a tank even know it's being targeted in the first place? And if it did, how would it see past its own smoke? And what happens when its own smoke clears? Does it pop another one, or engage? Having just been shrouded in smoke, how does it know who to engage? You'd need precious seconds to reacquire a target, while the enemy already knows EXACTLY where you are once you're smoke fades.

 

You do the math.

 

Just look at it this way - FACs use smoke to designate targets for a reason ;)

[D-Scythe]

Here an excerpt from an old tank sim, M1TP:

 

Turret Smoke Grenade Launchers: The smoke grenades are commonly used to defeat incoming missiles. If you fire a salvo when a missile launches, then switch to thermal sights, you can continue firing while the missile frequently misses. As the missile gets closer, smoke has less and less effect on its accuracy.

 

However, missiles have very little chance of penetrating your front armor. About the safest thing you can do is just face toward them, close the hatches and pray. Don't bother wasting smoke GLs on a missile that's close to hitting you. Save it for escapes (see below).

 

Engine Exhaust Smoke Generators: The VEESS exhaust smoke generator lays a continual smoke screen behind your vehicle. Obviously, this is useful when retreating...

 

Escapes: Inevitably, in some battles you'll drive a tank into a bad spot. To escape quickly, aim your turret at the biggest concentration of enemies and fire a smoke grenade salvo. Next turn on the exhaust smoke screen. The smoke GLs protect you while you turn around, then the exhaust smoke covers your rear as you retreat.

 

The article is here: http://oldsite.simhq.com/simhq3/sims/turret/m1tp2/m1tp1man/tactics.html...courtesy of SimHQ.

 

You can tell that smoke is not used to defeat SABOT rounds ;)

[Brit_Radar_Dude]

Many currently fielded ATGM's require the operator to keep his sight on target for the duration of the engagement. Tactics a tank crew might use to cause a miss are:- (1) to pop smoke so he can no longer see the target and (2) open fire at him with a main round snapshot or MG fire and maybe make him duck! Latest Chinese tanks (it is claimed) may have a laser system to damage aiming system optics or even the eyes of the ATGM operator.

 

Some ATGM's may have a thermal sight which may still see the tank through the smoke, which is why more modern smoke grenades utilise anti-thermal imaging techniques - hot/burning material (plastic fragments, graphite etc.) mixed with the smoke.

Even against sabot fire from another tank, popping smoke may be a worthwhile tactic. If it was you in the target tank, wouldn't you pop smoke just in case it increases the probability of a miss by a few percent?

 

 

FACs use smoke to designate targets for a reason

Indeed they do, they use smoke to enhance the visibility of a spot on the ground. A tank would use smoke to obscure the visibility of a spot on he ground. Smoke is a wonderfully versatile tool, useful in attack or defense - why not go buy some today! :wink:

[D-Scythe]

Even against sabot fire from another tank, popping smoke may be a worthwhile tactic. If it was you in the target tank, wouldn't you pop smoke just in case it increases the probability of a miss by a few percent?

 

Maybe someone knowledgeable can chime in and settle this issue once and for all :?:

 

Anyway, back on topic. I really hope some of the more powerful SAMs like the S300PMU/SA-N-6, PAC-2 and the SM-2MR incorporates some type of loft profile to better engage low altitude targets at range. Apparently, all these missiles got some sort of effectiveness increase in V1.1, but I hope lofting is one of them.

 

As for ships, the major thing is to vastly increase the range and aggressiveness of SLCM attacks, so that any target on the map could be targeted and struck. Furthermore, ships like the Ticonderoga and Moscow guided missile cruisers should each be able to engage a lot more aircraft/missile targets simultaneously (especially the Tico, who can only engage half the number of targets as the Moscow).

[GGTharos]

Lofting is what they got (although I suppose lofting might be the wrong name in this case)

[Drakkhen]

Reasonable AI taxiing speed

 

The current AIs don't allow efficient/realist GAI use. For example, when you attack an airfield, the closest planes to the runway should head for it as soon as you're suspected by radars and not at grand'ma's pace if possible...

 

They're so slow you'd better make them start takeoff procedure as soon as you do, you still can reach objective before they switch gear-up. And it's almost impossible to miss them if you strafe the taxiways once air-defenses are down... :lol:

 

They don't allow the player to make a grouped take-off without waiting for minutes on runway either. Ah... almost forgot... when you're waiting for them, you have to do it "behind" taxiway to runway point otherwise, they wait for you to take off before "entering" runway, even when they're your own wingmen.

 

Possibility to set a pilot as wingman not only leader. Human can be just as good as AIs do. :wink:

 

Most other requests I had in mind have already been discussed.

[D-Scythe]

Lofting is what they got (although I suppose lofting might be the wrong name in this case)

 

So they've got "loft" in V1.1? Cause in V1.02, the missiles typically just make a beeline straight for the target, spending WAY too much time at low altitude.

 

Besides, real SAMs do "loft," flying to a higher altitude before pitching down for the kill against low altitude targets.

[GGTharos]

Offical line:

 

'SAM Tajectories optimized'