Limiters for AI maneuvering

[Dragon1-1]

It would be great if we could limit, in ME, the things AI will do to maneuver. There's very little control over AI flying currently, and that makes it hard for mission makers sometimes. Introducing the following limits, to be set by the mission maker, would help with that.

Pitch angle
Climb/descent rate
Roll rate
Turn rate
Roll angle
Airspeed (separate upper and lower bound)
Altitude (lower and upper bound)
G load

Right now, we have no way of influencing those parameters. I propose that it should be possible to set strict limits on the parameters listed above. That is, if you turn on the "don't go below 10kft" mode, then it won't go below 10kft, period, even to save its own life. It'll act like there's terrain there. Of course, good judgment from mission creator would be required, so that the AI can actually fly those parameters, and also to disable them for dogfights (except the hard deck, that'd be very useful for realistic NTTR hops) unless the mission author wants a turkey shoot.

Note that those would be limiters, basically, the rest of the AI would remain the same, and if the AI doesn't want to, say, turn as hard as possible, then it won't, even if the limit has been set. The point is to be able to hard clamp AI performance if the mission maker wishes. It doesn't even have to be "aware" that this is in effect, of course, this could prevent it from fighting or completing certain maneuvers properly, but it'd be on the mission maker to ensure this is not a problem.

If the current FM doesn't allow implementing this, please consider this feature for GFM. Anything that makes wrangling the AI easier will be very, very much appreciated by both campaign creators and players alike.

[irisono]

With all due respect for your suggestions to make this combat simulator more realistic, but none of this would be necessary if the aircraft AI had been properly developed. AI-FM cheating is the main cause of the whole misery and this should be corrected as soon as possible. In dogfight missions (e.g. with the Mig-21, Mig-15 or F-5) many users select AI experience level exclusively to "rocky" and set conditions such as "restrict jettison", "restrict afterburner" just to achieve a somewhat plausible simulation. It doesn't surprise me that with such absurd aircraft AI, serious DCS users are trying to find workarounds.

[Exorcet]

7 hours ago, irisono said:

With all due respect for your suggestions to make this combat simulator more realistic, but none of this would be necessary if the aircraft AI had been properly developed.

It very much would be needed even in that case. Smart AI still needs context. No matter how good the AI is, it won't know the difference between a training mission and a emergency interception unless the player tells the AI what is what.

Floors are needed for common training scenarios, and would also be very helpful in setting up the AI for different types of missions. Other limitations may be desired for specific maneuvers like tankers, transports, or fighters doing a show of force.

Better AI would mean less tinkering would be required, but you'll never be able to get rid of all of it.

[Gunfreak]

Since we don't have a me109 that is useful for anything we have in DCS. I have often dreamed a toggle to make the 109K4 get throttled to earlier modles of 109.

Flying a spitfire in late 1942 fighting agiast late 1944 109s just feels off.

So to add a toggle that throttles the max speed of the 109k to 109g6 or 109g speeds would be welcomed.

Can't do anything about that 30mm nose. But at least you wouldn't be fighting a plane that flies 70km/h faster then the plane you actually should be fighting in late 42.

[Dragon1-1]

8 hours ago, irisono said:

With all due respect for your suggestions to make this combat simulator more realistic, but none of this would be necessary if the aircraft AI had been properly developed.

It would be necessary even then. For instance, consider an instructor on a training mission. He can say "I will roll my aircraft slowly so you can practice staying in formation with me, then slowly increase speed with each turn". AI will never be able to come up with that on its own, but the limiters I propose would make setting such a situation up a breeze. You could have an aircraft that, in-story, has a failure that forces it to limit its maneuvering, but it's an AI-only aircraft without detailed failure simulation. Again, easy with artificial stops. The big one is altitude: the exercise says hard deck is 10kft, go below that, you crashed, and you have to RTB to get yelled at for busting the deck. AI won't come up with that on its own, deck in the actual combat is deck. 

There are many reasons for restricting certain performance parameters. For a human, you just say to the pilot that he better not exceed certain parameters. For an AI, you need a trigger to enable a restriction for a given phase of flight. Even a competent AI won't obsolete this ability.

[irisono]

vor 40 Minuten schrieb Dragon1-1:

It would be necessary even then. For instance, consider an instructor on a training mission...

I understand some of the points made but your reasoning is very specific and beneficial for high-end mission creators. The UFO behavior of some ED modules affects the entire SP community, i.e. the majority of DCS users.

[Dragon1-1]

It's beneficial for everyone who deals with AI. Limiters can be a clumsy, but possibly an easier to implement way "de-UFOing" the current AI, but that's not their primary purpose. Mission creators of all sorts would find them useful, not only high end ones. For instance, if AI is crashing into the ground, set hard deck at some reasonable value to maybe prevent this. It might still dip below the hard deck because it's DCS AI, but that would not be an instant crash, but instead it'd try to get back over the deck, overriding all other behaviors. Also, mission makers could use this to make AI follow published limits and procedures when they won't do this on their own.

ED is working on GFM, which will hopefully help with other issues. I'm not suggesting these are a fix-all solution. They're not. However, limiters would remain immensely useful for anyone working with triggers, who wants the AI to behave in a specific way.

[Mike_Romeo]

On 12/15/2023 at 12:29 AM, Dragon1-1 said:

Right now, we have no way of influencing those parameters.

Its actually possible with some scripts. Some where used to make this formation flight trainer: https://www.digitalcombatsimulator.com/en/files/3327920/

[irisono]

vor 13 Stunden schrieb Dragon1-1:

It's beneficial for everyone who deals with AI. Limiters can be a clumsy, but possibly an easier to implement way "de-UFOing" the current AI, but that's not their primary purpose. Mission creators of all sorts would find them useful, not only high end ones. For instance, if AI is crashing into the ground, set hard deck at some reasonable value to maybe prevent this. It might still dip below the hard deck because it's DCS AI, but that would not be an instant crash, but instead it'd try to get back over the deck, overriding all other behaviors. Also, mission makers could use this to make AI follow published limits and procedures when they won't do this on their own.

ED is working on GFM, which will hopefully help with other issues. I'm not suggesting these are a fix-all solution. They're not. However, limiters would remain immensely useful for anyone working with triggers, who wants the AI to behave in a specific way.

Although it is and remains a workaround for "de-UFOing", your justification sounds convincing. Sometimes it's better to have a workaround than none at all. Thank you.

[Sérvalpilot]

I absolutely agree that there should be limiters.

I found out that the "limit afterburners" in waypoint actions for AI does NOT stop them from using afterburners in combat.

I found one way to reduce their maneuverability. Make them carry really heavy things that they won't use in a dogfight (like bombs), then set the option to restrict jettisoning. That slows them down a bit.

[twistking]

On 12/15/2023 at 12:29 AM, Dragon1-1 said:

It would be great if we could limit, in ME, the things AI will do to maneuver. There's very little control over AI flying currently, and that makes it hard for mission makers sometimes. Introducing the following limits, to be set by the mission maker, would help with that.

Pitch angle
Climb/descent rate
Roll rate
Turn rate
Roll angle
Airspeed (separate upper and lower bound)
Altitude (lower and upper bound)
G load

Good idea. I think i would add sth. like "% of max engine power". even with the upcoming GFM and better AI this would be cool to balance difficulty of engagements and simulate exercise scenarios or simulate other (lower powered) variants of an aircraft.

Another aspect to think about is, how that many options could clutter the UI. Maybe it would be good to have a limited set of options as standard (max g-load, altitude bracket, max engine power) that offer decent customization for many scenarios and have the other limiters as advanced options (i do see the usefulness of limiting pitch, roll angle, roll and turn rate), but i think for 99% of mission design AI scripting it would be sufficient to give speed/altitude brackets and limits to max power and max G load).

[Dragon1-1]

US wise, I actually envision them as trigger or waypoint flags, not something you set like afterburner limits and such. I don't think they need a separate UI at all, just adding to the options already present. 

[SickSidewinder9]

What we also need is a Merge task for waypoints.  As it is now, an AI adversary won't really do a merge, they will immediately start maneuvering or even firing weapons.  A task that limits that behavior until a certain waypoint or condition is met (distance decreases to a very small number than starts increasing again?).

[Gunfreak]

11 hours ago, SickSidewinder9 said:

What we also need is a Merge task for waypoints.  As it is now, an AI adversary won't really do a merge, they will immediately start maneuvering or even firing weapons.  A task that limits that behavior until a certain waypoint or condition is met (distance decreases to a very small number than starts increasing again?).

This can be done with a weapons hold order and then a weapons free order on a waypoint. You have to check the time.

 

Set enemy at 450kt. Set original waypoint as weapons hold. Make a second waypoint and add weapons free check the time the enemy will reach waypoint 2.

Now make your flight and match speed and time so you'll reach your waypoint 2 as the AI reaches it.

And you'll do a merge.

 

Like this

Or 

This 

The hard part is getting the correct speed in fight. So you don't arrive too early or to late for the merge.

[Jself]

I came here to post this exact suggestion.  We have limit Afterburner in the ME, it would be awesome to have other options, turn rate limit, climb dive limit, etc.  It would make escort type missions much more realistic

[ShuRugal]

On 12/15/2023 at 3:15 PM, irisono said:

I understand some of the points made but your reasoning is very specific and beneficial for high-end mission creators. The UFO behavior of some ED modules affects the entire SP community, i.e. the majority of DCS users.

This comment makes it sound like you think ED cannot do both.  The people who work on flight modeling and the people who work on AI logic are probably not the same people, this work can be done in parallel.

[SickSidewinder9]

On 12/23/2023 at 5:38 AM, Gunfreak said:

This can be done with a weapons hold order and then a weapons free order on a waypoint. You have to check the time.

 

Set enemy at 450kt. Set original waypoint as weapons hold. Make a second waypoint and add weapons free check the time the enemy will reach waypoint 2.

Now make your flight and match speed and time so you'll reach your waypoint 2 as the AI reaches it.

And you'll do a merge.

 

Like this

Or 

This 

The hard part is getting the correct speed in fight. So you don't arrive too early or to late for the merge.

That's very precise, but ok.  Was thinking of something like Sedlo's BFM trainer where a bad guy can be spawned in front of you.  Doesn't really work that way.  They need some kind of merge behavior not based on anything so precise.