AI plans and a more responsive AI

[MikeyD]

A couple years ago, just for fun one afternoon, I took an old, very simple all armor ME scenario that had the enemy racing down a big map and I redid the AI orders. I substantially slowed down the battle, subdivided the force, added precisely timed movements, with bounding overwatch and intervals where the stationary subgroup is allowed to sit and observe the battlefield. The formerly 'easy win' shooting gallery scenario turned into The Punisher. 

It is possible to make the AI appear as though it knows what its doing. Or at the very least you can make a scenario where the player thinks 'This is not what I expected the AI to do.' You can also make a scenario where the player thinks 'Same old stupid frickin' AI'.

[mjkerner]

Freyberg, you will be releasing that beautifully ruined map, no? Pretty please!

[Freyberg]

4 hours ago, mjkerner said:

Freyberg, you will be releasing that beautifully ruined map, no? Pretty please!

Definitely!

It's in the zip file I linked to above (of maps in progress - haven't done much since then, too tired from work). Here's the link again

https://www.dropbox.com/s/auilzpybhm0w206/FI map pack in prog May20.zip?dl=1

The map is called 'Frey Tiny 07 [RU] - City, ruins - Northwards ATTK.btt', but all the maps with the code [RU] are slices of the same master map - that one is basically the very centre of the map.

 

 

[mjkerner]

Thanks Freyberg. I missed it.

[Freyberg]

1 hour ago, mjkerner said:

Thanks Freyberg. I missed it.

I'm really glad you're enjoying them

[BornGinger]

I wish there was an option to withdraw troops a bit to deal with the situation they're suddenly facing if they would get attacked by a stronger enemy force. But that would probably require an if ... then ... kind of AI coding. The way it works now the attacked AI troops will either cower at the spot or maybe even run back a bit to cower further back. But it isn't possible to make the AI troops withdraw to another position and prepare for an attack or code it to try to avoid the situation by going another way.

The option we have is reverse. But as far as I understand that option only works with vehicles and could be used together with a trigger line and would order the vehicles to reverse if the enemy crosses a certain line. But that option doesn't give room for a surprise attack.

The option with if... then... coding could maybe be a little help in trying to make the AI act like someone in a H2H battle.

I also think we need more than 16 AI groups. We need at least 24 groups to be able to split companies into smaller groups on large and huge maps.

Edited June 3, 2020 by BornGinger

[weta_nz]

Thanks for the maps @Freyberg I've only played 1 map so far but it looked great and was a fun battle

[Sgt.Squarehead]

3 hours ago, BornGinger said:

I also think we need more than 16 AI groups. We need at least 24 groups to be able to split companies into smaller groups on large and huge maps.

Depending on the length of a scenario, it is possible to re-use AI slots (facilitated by either an Exit Zone, or The Grim Reaper).....This can require some experimentation and quite a bit of testing, but it's a handy trick.

[Glubokii Boy]

5 hours ago, BornGinger said:

I wish there was an option to withdraw troops a bit to deal with the situation they're suddenly facing if they would get attacked by a stronger enemy force. But that would probably require an if ... then ... kind of AI coding. The way it works now the attacked AI troops will either cower at the spot or maybe even run back a bit to cower further back. But it isn't possible to make the AI troops withdraw to another position and prepare for an attack or code it to try to avoid the situation by going another way.

In a prepared AI defence...

This is actually quite possible already if i'm not misstaking...

Use a terrain objective infront of the AI possition covering the max-advance line of the players troops...as desired...

When any player troops moves past (on to) that line (terrain objective)...use the AI withdraw command...it works perfectly well for infantry also...

The troops will withdraw in bounding fashion and even use smoke to cover their withdraw...

Covering groups will have their guns facing the advancing enemy...

 

 

[Glubokii Boy]

If the AI is forced back while it is not in a prepared defensive posdition  (expected by the designer)...things no doubt becomes trickier...

One solution for the future might be to be able to 'lock' a terrainobjective to a certain AI group...the terrainobjective moves forward stuck to the AI group...sort of a circular cover arc...extending a set number of actionsquares around the AI groups as it moves...set by the designer.

This would make what you desire possible even if the AI is forced back unexpectedly...That is...

If any planer troops move into that arc...new orders could be given to the AI group...including a withdraw order..

But this is something for the future maybe...

 

 

 

[Glubokii Boy]

2 hours ago, Sgt.Squarehead said:

Depending on the length of a scenario, it is possible to re-use AI slots (facilitated by either an Exit Zone, or The Grim Reaper).....This can require some experimentation and quite a bit of testing, but it's a handy trick.

These are decent tricks but they are no substitute for more AI groups...

More testing and tweaking is not what the scenariodesigners need...

They need LESS tweaking and testing 

[Freyberg]

3 hours ago, RepsolCBR said:

More testing and tweaking is not what the scenariodesigners need...

They need LESS tweaking and testing 

This is my position, too. What I would personally love to see is the task of the designer being made simpler.

My impression is that the AI already has a lot of power that is not currently being fully exploited. Some of the ideas above are excellent, such as 'if..then' routines, but I don't really want to micromanage AI plans.

[MikeyD]

An easy solution to the problem of too few AI groups is to make smaller scenarios.

Okay, that was a joke. I had filled all the AI slots in the last three scenarios I made, myself - and I'm not known for making particularly big scenarios. Still, there's a 'sweet spot' where force size, AI capabilities and map dimensions all fit to produce a scenario that works. If you insist on testing the boundaries of scenario design you will butt up against the outer fringe of AI capabilities. There's nowhere near enough AI slots to manage two battalions on a map but two battalions is an awfully unwieldy force size to manage at the best of times.

[RockinHarry]

3 hours ago, RepsolCBR said:

In a prepared AI defence...

This is actually quite possible already if i'm not misstaking...

Use a terrain objective infront of the AI possition covering the max-advance line of the players troops...as desired...

When any player troops moves past (on to) that line (terrain objective)...use the AI withdraw command...it works perfectly well for infantry also...

The troops will withdraw in bounding fashion and even use smoke to cover their withdraw...

Covering groups will have their guns facing the advancing enemy...

 

 

thought about that one too, but I don´t like a wimpy player scout team make a good defensive position get abandoned that easily. So again, some particular "If" trigger would be needed to make it work in more reasonable ways. But there´s other good uses for "retreat", either offensively due to deliberate smoke usage, or to bait a player into a trap or something.

But maybe a terrain trigger works in reverse. In case an AIP unit gets to retreat through normal enemy fire & morale effects, one can possibly place a friendly terrain trigger to the rear of that unit (formation/group). In case the retreating (or evading) unit hits the friendly map zone trigger, then either that single unit or all of its formation can be forced to "retreat" to a position as selected by the mission designer. Needs lots of testing, the more if considering the TacAI´s irrational retreat & evade behaviors. But I believe there´s some possibilities.

Since I very much dislike map zones to be used as "victory objectives"(occupy), they´d be all free for triggering AIP movements or combat order changes (i.e from hide to active etc).  Think there´s lots of yet unexplored possibilities with all the trigger options, both for defensive and offensive AIP plays.

[Glubokii Boy]

10 minutes ago, MikeyD said:

An easy solution to the problem of too few AI groups is to make smaller scenarios.

Joke or no joke...that's actually true...

But it kind of limits the options...

 

[Glubokii Boy]

14 minutes ago, MikeyD said:

 There's nowhere near enough AI slots to manage two battalions on a map but two battalions is an awfully unwieldy force size to manage at the best of times.

I agree...

Personally though...one of my favorit kind of scenario i like to play...or would like to play is...

Defencive, reinforced company kind of scenarios vs the AI...

There are not a whole lot of them though...

The problem with this is...it kind of requires the AI force to be of atleast battalion...if not reinforced battalion size to have a decent chans for success...

Simply having more AI groups will not make the AI brilliant...i know...but it  will most  certanly help...

 

[Glubokii Boy]

24 minutes ago, RockinHarry said:

thought about that one too, but I don´t like a wimpy player scout team make a good defensive position get abandoned that easily. So again, some particular "If" trigger would be needed to make it work in more reasonable ways. But there´s other good uses for "retreat", either offensively due to deliberate smoke usage, or to bait a player into a trap or something.

But maybe a terrain trigger works in reverse. In case an AIP unit gets to retreat through normal enemy fire & morale effects, one can possibly place a friendly terrain trigger to the rear of that unit (formation/group). In case the retreating (or evading) unit hits the friendly map zone trigger, then either that single unit or all of its formation can be forced to "retreat" to a position as selected by the mission designer. Needs lots of testing, the more if considering the TacAI´s irrational retreat & evade behaviors. But I believe there´s some possibilities.

Since I very much dislike map zones to be used as "victory objectives"(occupy), they´d be all free for triggering AIP movements or combat order changes (i.e from hide to active etc).  Think there´s lots of yet unexplored possibilities with all the trigger options, both for defensive and offensive AIP plays.

Intresting idea...

And I agree...all possibilties are probably not explored yet  !

[RockinHarry]

11 minutes ago, MikeyD said:

An easy solution to the problem of too few AI groups is to make smaller scenarios.

Okay, that was a joke. I had filled all the AI slots in the last three scenarios I made, myself - and I'm not known for making particularly big scenarios. Still, there's a 'sweet spot' where force size, AI capabilities and map dimensions all fit to produce a scenario that works. If you insist on testing the boundaries of scenario design you will butt up against the outer fringe of AI capabilities. There's nowhere near enough AI slots to manage two battalions on a map but two battalions is an awfully unwieldy force size to manage at the best of times.

in case of a mostly immobile defending AIP the amount of AI groups is mostly sufficient. But for an attacking AI one can run out of groups fairly quickly. Anything bigger than Plt sized AI groups doesn´t work, seriously. So counting 16 AI groups x Plt size  leads to 2-3 Coy size AI force at max, also considering that some groups are needed to deal with HQ and support units. 

 

[Sgt.Squarehead]

23 hours ago, RepsolCBR said:

These are decent tricks but they are no substitute for more AI groups...

More testing and tweaking is not what the scenariodesigners need...

They need LESS tweaking and testing 

I don't disagree, but in the absence of an upgrade, it's worth trying to get the most from what we already have IMHO.

[Glubokii Boy]

10 minutes ago, Sgt.Squarehead said:

I don't disagree, but in the absence of an upgrade, it's worth trying to get the most from what we already have IMHO.

Very true...

[BornGinger]

As this thread is about the AI, I take the opportunity to bring up something about the AI plans for quick battles.

If we allow the AI to decide what enemy troops we are meeting in a quick battle we sometimes get funny results. Of this reason some, or maybe even many, of us pick the troops for the enemy and while we do that we might decide to give one or two tanks or other armored vehicles to one or two platoons. After we have made our preparations the game engine is preparing the quick battle with the map of our choice and we don't know what troops gets allocated to which AI group. So there is the possibility that some infantry men in those platoons probably start the game by riding those armored vehicles as they have been allocated to the same AI group by the game engine.

In connection to this I was thinking wether it's possible to somehow control which group(s) that will have infantry only while doing the AI plan for a quick battle. When we make AI plans there is the option to order a group to dismount passengers. 

 

If I make plans to a group and keep the No Dismount order throughout the plan for a quick battle, is the engine taking that as a confirmation that there shouldn't be any vehicles in that group or is it going to put vehicles into that group and have the infantry riding those vehicles until they are forced to jump off them when being attacked?

 

I have only made a few quick battles and gave the order Passengers Dismount to every group just to make sure possible tankriders do get off the tanks in case the AI would put infantry and tanks or other armored vehicles in the same group.

[BornGinger]

On 6/3/2020 at 7:59 PM, RockinHarry said:

Since I very much dislike map zones to be used as "victory objectives"(occupy)

How would you like those objectives to be made? Occupy objectives are sometimes necessary to control strategic and important positions.

[benpark]

The game engine will randomly assign the AI Groups, as it must be adaptable to all available forces. In order for the QB system to allow for all possibilities, it must use some generic solutions- such as the assigning of units randomly among the AI slots. Programming it with this in mind is tougher than scenarios, but pretty decent if done more with a "big picture" view than down in the weeds of a single tank.

There isn't currently a toggle for foot/wheel/track in the AI plans, and adding one now would impact the existing QB maps and plans with the current system. I've thought about possible solutions to this- dedicating 3 or so slots to being vehicular, but that's getting into substantial engine work with possible unintended consequences. So "dismount" will only impact the mounted units, and be ignored by those units that it doesn't apply to. Better to have it work this way than not for either.

[Bulletpoint]

2 hours ago, benpark said:

There isn't currently a toggle for foot/wheel/track in the AI plans, and adding one now would impact the existing QB maps and plans with the current system.

I think it wouldn't really impact older maps. If you used a new engine with, say, 3 categories to load up an old map without these categories, it could just lump all units into the default group 0: general mixed units.

The old maps would then continue to work in the same way as when they were designed, but new maps could take advantage of the new feature.

[RockinHarry]

On 6/10/2020 at 3:42 PM, BornGinger said:

How would you like those objectives to be made? Occupy objectives are sometimes necessary to control strategic and important positions.

"Control" and victory objective zones aren´t quite the same IMO. I personally prefer a more general mission goal and choose control objectives by myself, not from mission designer. At least not in (too much) details. I.e mission goal "capture bridge", there can be multiple map areas that actually "control" the bridge site. Wouldn´t make any sense to place a victory objective zone (of whatever size) right on the bridge. So I´d like figuring out myself on how to control the bridge, without getting hands tied with going for given victory zones. Main goal in this example would be to deny the opponent any map spots that he can exert observed direct or indirect fire onto the bridge and its approaches. Mission scoring here can be made by placing a general bridgehead area victory zone (but hidden from player) and assigning high point values to certain enemy units. That could be any heavy weapons and FO that exert mentioned control on the bridge and surrounding. These high point enemy units then remain unmentioned in briefing and the player just finds out about them at end game scoring. While not all easy for beginners I find it more interesting and challenging for my own game play and mission design. But there´s countless possibilities on how to set up a mission and a matter of taste at last.