WW AICover

[GShock 10]

Maybe a workaround?

Force leader to a new order for a "scatter formation" upon establishing LOS?

I've noticed that "Take Cover" and "Move" actually detach soldiers from a squad (Take Cover --> Hide). IMO a detached soldier has less restraints and should find cover a lot easier.

My concern is that the soldier must put the cover between him and the gunshot sound / LOS.

Order to "Watch this direction" is also there... the engine appears to already have what's needed to make the AI stay alive. I'm sure you'll find a way, WW. :)

[froggyluv 2136]

Agree about the 'need to hold formation' basically breaks down the very essence of a firefight -that of staying alive. IIRC, the AI in Arma2 right around the time of the ACR release -was pretty well perfect on this. I remember getting in deep woods firefights as leader of AI squad, which once the bullets started flying, the AI held their allotted tree as cover very well EVEN if I was moving around. They weren't easily pulled by the leash of the squad leader's movement.

[Windwalking 18]

I can make the AI hold their cover as long as they have an enemy in sight, and the leader is not too far away. This will allow the AI to start flanking if they can't see the enemy and have a good shot.

[Windwalking 18]

Ok I added code to have the AI stick to their cover unless the leader moves too far, they will then leave the cover and return to formation automatically. This is not thoroughly tested so please do not download this unless you want to test it for feedback. Consider yourself warned!!

https://dl.dropboxusercontent.com/u/3047338/%40WW_AICOVER_test.zip

Feedback is NEEDED :D

[GShock 10]

I'm thinking of a list of checks but I don't know the mechanics so I don't know if AI can be programmed to follow these tasks and if it can be made to follow them in order.

Any in Squad has LOS:

1) Break Formation.

2) Sprint for the closest cover.

3) Turn towards threat.

4) LOS AI engages.

5) Non-LOS AI waits for suppression, then sprints to cover closer to threat.

6) Back to 4) or repeat 5) until "Clear".

7) Clear.

8) Regroup.

More or less sound threat without LOS can be handled the same way. AI should still rush for cover and search for LOS until threat is over.

(Search for LOS is what I'd call an S&D route in proximity of sound source)

Very concerned about the AI not using such logics in the open field, Showcase Armed Assault (2nd part) shows how useless friendly AI is without specific orders and a "self-equipped" logic. This AI can't handle simple buildings and would get lost in cities. Player would really have a huge advantage here.

---------- Post added at 17:13 ---------- Previous post was at 16:51 ----------

Ok I added code to have the AI stick to their cover unless the leader moves too far, they will then leave the cover and return to formation automatically.

But that's the point. The leader CANT move far because he himself is in LOS with enemy. :)

Engaging FRAPS! :)

[Windwalking 18]

I'm thinking of a list of checks but I don't know the mechanics so I don't know if AI can be programmed to follow these tasks and if it can be made to follow them in order.

Any in Squad has LOS:

1) Break Formation.

2) Sprint for the closest cover.

3) Turn towards threat.

4) LOS AI engages.

5) Non-LOS AI waits for suppression, then sprints to cover closer to threat.

6) Back to 4) or repeat 5) until "Clear".

7) Clear.

8) Regroup.

More or less sound threat without LOS can be handled the same way. AI should still rush for cover and search for LOS until threat is over.

(Search for LOS is what I'd call an S&D route in proximity of sound source)

Very concerned about the AI not using such logics in the open field, Showcase Armed Assault (2nd part) shows how useless friendly AI is without specific orders and a "self-equipped" logic. This AI can't handle simple buildings and would get lost in cities. Player would really have a huge advantage here.

---------- Post added at 17:13 ---------- Previous post was at 16:51 ----------

But that's the point. The leader CANT move far because he himself is in LOS with enemy. :)

Engaging FRAPS! :)

Baby steps :p

For now I don't apply the effect on leaders because I am not checking LOS just yet. It's a fail safe. Next step is to implement a good way to find alternative cover first.

[GShock 10]

Be advised I think there's a new patch (appears on DEV build in BETA tab but I am not in). Visible in main menu right hand side.

EDIT: No it's just updates on imminent release of a hotfix.

Tested your latest file but I don't know what you're doing so just have a look at the youtube vid as usual (if you keep the links you'll see tests of your versions in progress, I think we started on .54 and I think that is the last surely working version you should keep in thread starting link).

http://www.youtube.com/watch?v=lRO22O72WqY

The way I see it, they are too close to one another, the leader is still spamming lots of commands (overriding probably whatever you did) and they are all completely in the open. :)

Considering I might be doing something wrong DO test yourself and see if you get my same results. Tampering with files... you never can tell I'm n00by :)

Edited October 1, 2013 by GShock

[lucadena 10]

Be advised I think there's a new patch (appears on DEV build in BETA tab but I am not in). Visible in main menu right hand side.

EDIT: No it's just updates on imminent release of a hotfix.

Tested your latest file but I don't know what you're doing so just have a look at the youtube vid as usual (if you keep the links you'll see tests of your versions in progress, I think we started on .54 and I think that is the last surely working version you should keep in thread starting link).

http://www.youtube.com/watch?v=lRO22O72WqY

The way I see it, they are too close to one another, the leader is still spamming lots of commands (overriding probably whatever you did) and they are all completely in the open. :)

Considering I might be doing something wrong DO test yourself and see if you get my same results. Tampering with files... you never can tell I'm n00by :)

While I do really appreciate your efforts in testing this mod, I would like to give you a small suggestion.

In my humble opinion the showcase is not a good way to test this mod, because while it is repetable and is equal for all users, it might contains scripts or orders/waypoints that change the normal behavior of units (as you suggested, leader is spamming orders).

You should open the editor, place 2 squads in a place with some covers (like some city, or in the open with some fortifications added by you), then save the mission (maybe as ww_AIcover_test).

To test the mod you should then use that mission. I would appreciate if you could share that mission here (so i could do some tests with the same conditions) and the results you have.

I do not mean to be blunt or anything bad, I would like just to help :)

[froggyluv 2136]

Ok, messed around abit with the latest and theres something strange going on.

I have my Blufor Fire squad of 4 vs Opfor 4 in a somewhat hilly, woodsy area with starting positions 75m between with trees in the way to block outright contact.. I have my guy set to allowdamage false and watch the whole thing in Commander mode while moving around to see if I pull my guy's leash or not. There are a few rocks, bushes and plenty of trees around as well as a cabin to our rear some 50m away. As soon as I initiate contact, 50% time 1 of my guy takes off for that house and generally leads another guy with him via that split fireteam thing going on. It wouldn't be a terrible move except its quite a little jog, putting himself at risk but he generally makes it there regardless. Problem is he tends to go unresponsive once there, he had a perfect window LOS with the enemy but just chose to stand facing a wall EVEN after I gave him a Watch direction command. Sometime the other 3 would follow him. I took same scenario to nearby town and same thing happened, 1 guy roams a good 30-40m to a garage for cover and won't respond.

Another problem is when we did stay on the ridgeline, they weren,t really using the closest cover for some reason and sometimes just stood to the side of a rock or else, tended to go prone a lot. It also seems they would leave cover still quite easily if I pulled 10m or so left or right, it seems after their target is down they grab the leash even though other enemies are still in the same direction (50m or less away) -maybe once they lose LOS with their one target they get pulled?? We still had known targets in the menu list so maybe Knowsabout needs to get involved..? Also, I'm guessing we never got back the quicker turning AI from Beta as a lot of times they stand up from cover to do god-knows-what, turn their orientation sideways and get killed :(

I'm starting to realize the biggest problem of Arma AI -it's their wonky sense of orientation. Meaning, if all known enemies are at 12 oclock and under 50m, they need to stay locked on 12 oclock at all times, looking briefly left to right for flankers. It almost makes the game completely random as it boils down to -which AI squad does the LEAST amount of random, goofy shit -wins. I don't know if you can override the orientation parameter but that would be huge.

[GShock 10]

Froggy publish this mission please so we can test the same thing and show via youtube what happens when test mod is running.

[ru55lee 2]

Froggy publish this mission please so we can test the same thing and show via youtube what happens when test mod is running.

Yes please do. Im still debating whether to use this mod yet or not. Would be nice for a comparison :D

[lucadena 10]

I agree, I would test the same mission too :)

[GShock 10]

Luca if you are proficient with the mission editor, create a simple 3 vs 3 mission with the neutral player in a good observation spot and I'll test it.

All that matters is that WW has that mission to counter-test the results.

[lucadena 10]

Yes, I would gladly do that, as I am able to use the editor. Unfortunately I will only be able to do that this friday, as now I am at university and I have not my gaming pc with me :(

[froggyluv 2136]

Ok, I will soon as get home from work today.

[kremator 1065]

... boils down to -which AI squad does the LEAST amount of random, goofy shit -wins.

Sad but true!

[mikey74 186]

I love yours and TPW's mods firggin awesome. I have a request though. Could you add an option in the user config to turn on and off mortar support? Thanks for the great mod. :)

[froggyluv 2136]

Well heres the simple mission: https://dl.dropboxusercontent.com/u/62320088/WW%2520cover%2520test.Altis.rar

It's slightly changed as I had saved it previously in the Town scenario and seems to play out a little different (haven't seen guy run to house).

So just dump the folder in your Documents/Arma3/Profiles and open in editor. Also, try copy/pasting all units to the nearby town to the West and start them out of sight like around the corner. When playing don't fire yourself, just go into commander mode by pressing the "." on the numpad then press the numpad "-" to scale all the way back. Practice getting your guys near cover then initiate firefight and start moving around. Civilian is playable if you want to see how the opfor squad is doing since they have less formation worries than the player.

[Windwalking 18]

Well heres the simple mission: https://dl.dropboxusercontent.com/u/62320088/WW%2520cover%2520test.Altis.rar

It's slightly changed as I had saved it previously in the Town scenario and seems to play out a little different (haven't seen guy run to house).

So just dump the folder in your Documents/Arma3/Profiles and open in editor. Also, try copy/pasting all units to the nearby town to the West and start them out of sight like around the corner. When playing don't fire yourself, just go into commander mode by pressing the "." on the numpad then press the numpad "-" to scale all the way back. Practice getting your guys near cover then initiate firefight and start moving around. Civilian is playable if you want to see how the opfor squad is doing since they have less formation worries than the player.

That's really helpful thanks.

I am making some good progress. Not sure how it will play in the grand scheme of things though.

[GShock 10]

Beginning tests, will start with Vanilla AI, record, then enable WW+TPW and retest.

Oh uhm... the only commander mode I can access to is with "\" while numpad "." has no effect and numpad "-" unzooms.

I was more thinking of an aerial view if it can't be figured the best thing to do would be to setup a drone in hot on cold view on the area. Can't see through the trees and need to be on a higher spot.

Edited October 2, 2013 by GShock

[lucadena 10]

Beginning tests, will start with Vanilla AI, record, then enable WW+TPW and retest.

I was more thinking of an aerial view if it can't be figured the best thing to do would be to setup a drone in hot on cold view on the area. Can't see through the trees and need to be on a higher spot.

During editor preview mission, press esc, in the menu click spectator, it should go back to game, press space, it will go in 3rd person, press * on numpad, to get free flying camera, press n twice to go in thermal mode and evaluate the ai. Remember you can even zoom a lot with +,-

Hope it helps

Edited October 3, 2013 by lucadena
did a mess with the quote

[sttosin 67]

Ok I added code to have the AI stick to their cover unless the leader moves too far, they will then leave the cover and return to formation automatically. This is not thoroughly tested so please do not download this unless you want to test it for feedback. Consider yourself warned!!

https://dl.dropboxusercontent.com/u/3047338/%40WW_AICOVER_test.zip

Feedback is NEEDED :D

I ran into this error using this version along with Dev Version: 1.03.110653:

Error in expression <er (group _unit)) > 25 || (_destination select 1) == "LEADER PLANNED" || (_desti>

Error position: <select 1) == "LEADER PLANNED" || (_desti>

Error Zero divisor

File WW_AICOVER\ww_aicover.sqf, line 417

I also got the following error while replaying the same mission

Error in expression < < 50 ) && ((expectedDestination _unit) select 1) != "Leader Planned") then

{

if>

Error position: <select 1) != "Leader Planned") then

{

if>

Error Zero divisor

File WW_AICOVER\ww_aicover.sqf, line 312

I wonder if the error happens briefly after a team leader dies just before the new leader takes over.

---------- Post added at 01:24 ---------- Previous post was at 01:20 ----------

For the mission posted, would it be better to test with a mission in urban setting with a lot of cover? Is it possible that in order to simulate realistic AI behavior WindWalking will end up modifying behavior depending on type of cover/terrain is applicable? It seems urban settings would be easier, once perfected, then do wooded/vegetation terrain? What do you all think?

Edited October 3, 2013 by sttosin

[sttosin 67]

I threw a couple of vids together for an urban mission to compare with/without AI Cover

My observations:

AI tends to watch the direction of possible fire better.

They sometimes run back and forth between open areas without trying to use cover or move along walls.

They still seem to still use the buddy system. They are not as bunched together as in vanilla.

They also flank as you can see towards the end of the video.

Some members still run a bit further away from the rest of the team unnecessarily.

Thanks for the hard work on this mod. The difference between vanilla and this is night and day. Hope this helps in some way as well.

[GShock 10]

The only way to assess exactly what each soldier is doing is by using an aerial drone view. Heat signature makes soldiers visible through trees and you can observe the same mission from a different angle of the battlefield easily because the drone moves a lot faster than a civilian.

I think it is easier for the sake of testing to put just a 2vs2 fight and try in rugged, flat, terrain.

This ensures no LOS problems so that all 4 soldiers fight seeing the enemy, leadership is present but no target switching happens, easy steps first.

Is there a page to check about the details of the DEV build?

[ru55lee 2]

The only way to assess exactly what each soldier is doing is by using an aerial drone view. Heat signature makes soldiers visible through trees and you can observe the same mission from a different angle of the battlefield easily because the drone moves a lot faster than a civilian.

I think it is easier for the sake of testing to put just a 2vs2 fight and try in rugged, flat, terrain.

This ensures no LOS problems so that all 4 soldiers fight seeing the enemy, leadership is present but no target switching happens, easy steps first.

Is there a page to check about the details of the DEV build?

You mean this?

http://forums.bistudio.com/showthread.php?149636-Development-Branch-Changelog/page14