Wingman AI Discussion

Here I am again, trying to open a little discussion. This time about wingman AI, comms, mission profile and mission planning. Please feel free to add comments and suggestions.

First I would like to say that all I have is reading knowledge, no first hand experiance. Most of the following subject comes from the excellent SimHQ Air Combat Library. So please excuse me if some of my suggestions are unrealistic. I tried to suggest items I think that are realistic, as far as I have read about them.

In the past time, large parts of the flightsim community judged realsim by amount of buttons and radar modes. If a sim had to be realistic, it had to have as much radar modes as possible. With this dogmatic view on hardcore sims, many aspects that determine the overal realistic representation of an airplane and it's operations get overlooked. Aspects that not only have a much bigger contribution to overall realism of the product, but that also can greatly enhance the gameplay and fun of a flightsim ( something that is often considered as counter realistic by the current dogma ). One of this aspects I would like to go a bit deeper into now are wingmen and teamwork. Something that is much more important in real life areal warfare than your available radar modes.

In the history of flightsimulations wingmen have never reached the importance they enjoy in real life. Understandable, as programming a human like AI is ( at least yet ) impossible. In many sims wingmen were not more than armed decoys, just some other jets that follow yourselfe, distract the enemy and shooting them down occasionly. In the flightsims with real good AI, wingmen emerged to something like player controlled weapon systems. The player could use his wingman as tools to do certain tasks. Of course this is a good thing for flightsims and many sim would profit if it would feature such wingmen ( *hint* ). But it is still a incredible big simplification of reality, something that weights much more than absence of radar mode xyz. A fully new approach to the flightsim genre would be to leave the known concept of wingman behind and emerge to the concept of tactical partners and teamwork. A sim that would put its design scope on how the player acts together with his AI flight as a team. Btw, it is interesting to note that the tactical shooters have started to make that approach to gameplay since quite some time ( Rainbow Six etc. ), altough of course on a much simpler level than flightsims. Of course this is a somehow utopic vision, because AI will never ( in the near future ) reach the perfection of the human. But a humanlike AI is not the point of the subject, it is more about a change in the design approach. To forget the way we have flightsim wingman in mind and trying to work in the direction of the tactical partners concept.

So then, lets try to make some suggestions to reach that goal. I will concentrate on the A-10 here, relying heavily on the "Hog Basics: RAF Bentwaters Tactics Guide, 1982" articel by SimHQ's Andy Bush.

-Mutual support. This is perhaps the most important aspect in flying as a team. To cover each other and complete each others situation awareness by voice. The wingman ( basic formation is always a 2-ship, everthing else is just a combination of 2-ships ) should be much more vocal about the envoirment. He should call out radar contacts ( with all infos ) and visual contacts. And not only the usual 1-2-3 o'clock, but also height and range ( = Tally bandit 1 o'clock level, 3miles ). Also he should call out threats to the player like AAA sites or SAM launches, if he is in a position to actualy see them. The wingman should try to comlete the players SA.

Now remember the tactical parters. The player should also try to support his wingman with informations. Because your wingman isn't perfect either and doesn't see everything 360° all the time. He scans the skys like the player and has incomplete SA. To communicate fast and efficient to your partner, a interface would have to be designed for this. I envision something like a hotkey, combined with the direction of your view. So lets say you slew you view to the 4 o'clock position and hit the "Tally Bandit" button, then your caracter makes a "bandit at 4 o'clock level" call and your wingmans SA improves. And there would be a "threat" call, using the same methode. Lets say your team does a shooter/cover attack the player makes the covering part ( never been done before ). The player would cover his buddy and call out SAM launches against his wingman.

At the moment I can think of 3 hotkeys, for simplificity sake:

*KEY_VISUAL_AIR ( I see an enemy plane )

*KEY_VISUAL_GROUND ( I see an enemy ground unit )

*KEY_VISUAL_THREAT ( I see a missle, cannon fire aiming at you )

This is only a first brainstorm and much tought has to be spend on that, but you see where I want to go.

-Shooter/Shooter tactic. Both members attack the target simultaniously and coordinated.

-Shooter/Cover tactics. Something essential for CAS aircraft but with current sims impossible. A shooter/cover attack is when one member flyes the attack, while the other covers his partner and calls out threats and if the situation permits surpresses them ( surpression; -> ground AI ). As we accept our wingman as partner, the player can fly the role of the shooter or coverer, as flightleader he has the choice.

-Decoy tactics. One member of the flight ( or one element of a 4-ship ) flyes a decoy attack at the ground target ( per example a long range strafe ), while the other one attacks with surprise from a different direction ( element of surprise; -> ground AI ). Player can fly as attacker or decoy.

This advanced tactics open the question of controllability for the player. How is everything controlled in the heat of battle. Also in this regard, lets search the answer in real life procedures. We are going to formulate a attack plan to our partner, but first we need time to actualy plan it and then communicate it to our wingman. So lets introduce holding patterns:

Circle: http://www.simhq.com/_air/images/air_092a_6.gif

Racetrack: http://www.simhq.com/_air/images/air_092a_7.gif

We order our wingman one of the holding patterns and get time to study the map, work out a plan and communicate it to our partner ( something that is more fuzzy in flightsims than in IRL due to the need for comms menus ). The attack plan would include all the informations the AI needs to fullfill it's role as shooter or cover man. These include:

-target position and type

-initial point ( one of several pre briefed or FAC assigned positions, here is the start of the attack run )

-attack formation

-tactic and role ( shooter/shooter, shooter/cover etc., who is who )

-weapon type ( this also influences weapon delivery profile )

-Nr. of attacks

-egress ( egress formation, split or crossturn etc. )

-egress point ( here you team up again if you get lost )

This tells your wingman the basics of the attack plan. It is basicaly a "five point brief" used in real life, perhaps with some additional infos. I could imagine to use a basic menu structure for it. You select to send a "five point brief", then you run trough menu pages with the different options for each item. Sounds time consuming and it is. But that is why they set up a holding pattern in real life too to do this stuff. At the end you send it, but nothing happnes yet. You only told your wingman the plan, so he will do the right things at the right time.

I could also imagine a different planning tool that is more map based. Imagine the Il-2 style map that is in the corner of your screen, covering only parts of your view. That is the planning tool on wich you work out your plans using the map itselfe.

You can also create a battleplan in the flight planing phase that is the default attack profile that is used if no special commands are given while on the mission.

Ok, that is enough for this evening. As you see, I didn't make a simple feature request. This is more a request for a fundamental design scope that is determing the face of the sim. Also please note that I concentrated on the A-10, altough it is unknown if the A-10 is featured in ED's next sim. But many of my proposals also work for other planes. And my main intend anyway is to show the possibilitys of a different scope of the sim, independed of the flyable plane.

Last I would like to point out SimHQ excellent library, where you can read in depth about the tactical parters concept that is very prominent in Andy Bush's articles. You can find it here:

http://www.simhq.com/_air/acc_library.html

And in particular the genious 'Hog Basics: RAF Bentwaters Tactics Guide' article written by Andy Bush with special regard to Lock On. There you can find out more about all I wrote above in great detail. This article was a real eye opener for me how Hogs would have operated in a 1980s scenario. It is a long read but well worth and I can only recommend to read it ( also to the Devs. ). You can find it here:

http://www.simhq.com/_air/air_092a.html

Alright, now my fingers are falling off, so I will stop :)

Hmmm, next time I am going to write less...

I agree that LockOn, like most sims, needs an AI improvement. And one of the parts where AI is worst, is the "artificial" wingman.

I also agree that LockOn lacks many useful commands, but I think the problem has many more sides which must be explored.

First, LockOn lacks not only commands to execute tactics, but also lacks tactics. Wingmen are able to track and shoot targets and to fly on your wing, or onto a specified path. That's all. They do not apply any tactic, instead they attack in a chaotic way, sometimes in a dumb way (just think at the bomb-runs, performed at slow speed, with brakes deployed). So the point is that ok, it will be good to have new wingman commands, but first you must implement the tactics they have to follow ;)

Second point. There are some tactics that are simple, and are common knowledge for all pilots. Others are a bit trickier and require something else. This "something else" usually comes in two ways: pre-mission planning and training. Tactics planning in the briefing allow to setup attack routes, IP for attack runs, attack profiles and other things like that. Training instead makes pilots act as a team, and allow the creation of new and particular tactics, often tailored to the particular scenario/situation one is gonna face.

And under these points of view LockOn, like most of the other sims, doesn't offer anything. The training phase is a chimera, because would require a tough AI, but accurate mission planning is doable with the technology available today, and so it would be nice to see it added in some future.

So, my point of view is that a few new wingman commands in LockOn would be good, but the uselessness of AI wingmen in LockOn will not be solved just adding commands, but improving AI, implementing some tactics, and allowing a more accurate mission planning, which keeps track of every aircraft in a flight/section.

(I think this could be possible because LockOn is a simulation in a "tactical" scale, i.e. with small numbers of aircraft fighting in a limited scenario. In a dynamic scenario with dozens of aircraft in the air this would be tougher)

First of all, kudos for making a serious investigation into actual CAS formations and tactics! Very interesting links.

IMHO, this is an issue with many aspects - AI behaviour, the user interface to control that behaviour, the ability to model SA and peripheral vision through a monitor, how to extend this model to other non-A-10 aircraft, how to extend this model to multiplayer human wingmen, etc.

Starting with AI, the list of things to fix in Lock On is long:

http://206.116.70.212:8811/board/index.php?a=topic&t=32

Someone once said "the problem is too complex - like every other sim, we are always going to have bad AI." There's some truth to it, but I think it's more precise to say we will never have "good" AI, rather than that we will always have "bad" AI. I think if the AI wingman behavior is simple enough, then at least it can be avoided to make it "bad".

-SK

Great thread, and top work MBot and Caretaker.

Anyway, the wingmen issue (and also the CAS link!) has already been pointed out in this fourm. I think folks at ED know that it's a issue that must be improved.

http://forum.lockon.ru/viewtopic.php?t=2230

http://forum.lockon.ru/viewtopic.php?p=35750#35750

http://forum.lockon.ru/viewtopic.php?p=31672#31672

The point is that wingmen abilities must be improved, like it should be improved overall AI in LockOn. And for the wingmen, it's not just a matter of CAS. Also air combat can have it's own tactics/commands, and also the takeoff/landing/flying/formation procedures can have better commands (and better wingmen)

All we can have now is a "significantly improved AI" in Flaming Cliffs, that's what's they say....

Let's hope that 1.2 will bring a new dimension to LockOn since putting the KA-50 (which shares most of the avionics with su-25t) shouldn't be the hardest effort! (and I don't know how many people in the jet sim community would buy an add-on which features just a new flyable copter!)

bump.

Any dev reading this? I doubt...lets hope im wrong.

SwingKid and Starlight are right, this is a subject with many aspects that are linked closly together. AI, procedures, tactics, communications, flight planing, mission editor. Somehow this elements together build the core of how planes operate in flightsims. And realistic operations of airplanes is in my opinion what realy makes the realsim part of flightsims ( oposed to the popular avionics criterium ). It is difficult to discuss this "core", as it has so many facettes, but I think that AI is a good approach to it. You have to start somewhere...

I am looking forward to the AI improvements of 1.1 and 1.2, but honestly I don't expect a revolution. I expect small changes that make the current game more playable, but all in all I think the level on wich operations of our flyable planes are presented remain the same. Fine tuning is the motto.

Expecialy if you consider how many roles the current flyable planes fullfill and almost non of this have the accompaigning procedures/tactics/comms/AI ( what I called the "core" above ). And additionaly to the wide field soon the field of helicopter operations will be added, wich is extremly complex on its own and could easly fill an independed sim. This broad scope makes depth in operations impossible.

My hopes are on EDs next generations sim. There they have the possibility to clearly define the scope of the sime from the start of the design work and I realy hope that portraying the operations of the choosen flyable plane will be an major goal. To be honest, I don't think we will see the depth I described above, but I would love to see it and that is why I am writing here.

One thing that is clear to me is, that EDs next sim will have to feature only one flyable plane, if we want to see its operations in great detail. And I think that even this single plane should be one with only limited mission types ( e.g. no multirole plane ). Every mission type needs its own tactics and comms. And with multiple planes or even just a single multirole fighter there are just too much mission types. The alternative would be less detail for every single profile ( whats IMO the biggest problem of the current LOMAC ). So my advise is, narrow down the scope. Make an A-10 or Harrier sim with a FAC-CAS scope, a Tornado sim that concentates on strike, a helicopter sim, or an F-14 sim around fleet air defense ( insert favorite plane here ).

To come back to the suggestions part. Here is an SimHQ articel on 2on1 fighter tactics, that realy nicely shows the Engaged Fighter/Support Fighter ( shooter/cover) approach in A-A combat. The tactical partners concept is realy a underlaying philosophy in aircombat, independent of A-A or A-G.

http://www.simhq.com/_air/air_027a.html

I agree in the narrowing of the airplane choices but just not to one plane.

In terms of what's going to be on the post-LOMAC sim for ED, I think we're going to see a Red plane and a Blue plane.

For the extra headaches modelling 2 planes provides, it taps a market for the CIS that ED seems to like.

And as multi-role is more common in western planes, the more likely it is that the multi-role is what it is going to be.(F16/F18)

There also the online aspect - The ED preference has to be Air to Air - that's what the majority of fans expect - take a look at IL2 as well.

The marketing wants drives the AI needs.

It's a shame they're not able to do a full AFM Russian fighter as one of the paying expansions to LOMAC. The work in that could be done and polished long before they did work on the Western follow-up.