Jump to content
Sign in to follow this  
dwringer

ANIMA - AI for advanced position analysis using genetic algorithms

Recommended Posts

ANIMA - Arma NSGA Interface for Milsim Applications

 

ANIMA is a collection of functions and scripts for providing AI groups with terrain/position analysis capabilities that can be used, for example, to dynamically create waypoints with tactical advantage (by criteria that can be programmatically defined), automatically land helicopters at suitably flat/clear locations, or find places to spawn units or objects with specific visibility requirements.

 

This release is essentially an open-ended set of scripts for getting units to do things in abstract terms like "land a helicopter nearby", find an overlook position with cover and concealment", "ambush nearby roads or intersections", "seek tactical advantage and engage targets", or "find a nearby place to unload troops". It does not affect unit AI in any way, but uses a genetic position-finding algorithm in various ways, many of which involve assigning different kinds of waypoints.

 

Sorry, but for now, ANIMA does not have the required conditions for multiplayer, but if you're an experienced multiplayer developer and want to take a crack at it then I'd welcome your pull request!

 

I've included a User Manual, which is essential for understanding ANIMA. There are also some example missions to accompany the manual which will hopefully elucidate some concepts. One of these (see below) comes with an additional guide that explains some of the development choices and basics of using the included applications.

 

Please check it out and let me know what you think! I'll try to release some more examples and explanatory content as time allows.

 

 

EXAMPLE MISSIONS (separate from release)

 

RELEASE (includes basic examples):

https://github.com/dwringer/ANIMA/releases/download/REL-220316-A/ANIMA-220316-A.zip
 

MANUAL:
https://github.com/dwringer/ANIMA/blob/main/ANIMA.pdf

 

GITHUB (Updated regularly, may contain bugfixes as well as new featuresbugs):
https://github.com/dwringer/ANIMA

 

Known Issues:
- Pausing the game while the algorithm is running sometimes causes it to fail.  It will either choose a wildly incorrect position or fail to do anything at all.

- Default landing zone search radius works best with only one helicopter.  If you want to land more than one at a time, try bumping up the radius, step count, and population parameters of the land_helicopters/land_transport methods [55/3/13 is a possible start].

- Headquarters attack/dispatch methods occasionally assign an extra waypoint to a group - the logic needs to be refined.

 

CHANGELOG

(Latest) REL-220316-A (Updated 16 March 2022):

- Added sort behavior to fnc_find_positions which can sort the final results by a given function (by default, uses the average across all objective scores)

- Added a setting to allow fnc_find_positions to work faster, but less accurately, by allowing undifferentiated results after the target number of generations (a single bin after the nondominated sort, no dominance ranks in the result set)

- The behavior of optional parameters for attack_targets and approach_targets methods of Headquarters was fixed

- Fixed behavior of wp_config parameter of internal Headquarters method dispatch_transports

- Added an optional start_offset parameter for approach/attack_targets which moves the search start location relative to the targets by [x, y, z]

- Revised and expanded user manual with section on fnc_find_positions and some classes

- New (WIP) example mission with all (non-airplane) waypoints generated from ANIMA positions (requires CUP addons)

 

REL-220311-A (Updated 12 March 2022):

- Added two parameters to fnc_find_positions: one is a sort-function (or "false") to sort the resulting positions (default sorts by average score).
- There are now two example missions.  The first one just follows the basic example in the manual, while the second employs a number of functions.

- Added a User Manual chapter on fnc_find_positions.  Combined with the chapter on LocationFinder, added in the last release, there should be enough information to customize the algorithm with new objective functions.

 

REL-220310-A: 

- Fixed a bug with the LocationFinder method for finding helicopter landing positions, which was causing the parameter for enemies to avoid to be ignored.

- Updated the example mission so that when OPFOR and BLUFOR units detect one another, their waypoints are removed and new ones are assigned using attack_targets with respect to the detected troops.

  • Like 9
  • Thanks 1

Share this post


Link to post
Share on other sites

Sounds riveting !

 

Edit: So i only learn things by trying the demo first and im not quite sure what to make of the Demo mission. No mods at all - first few times after calling 0-0-1 the helo shows up and kills Opfor with its Helo Machine Gun. Then the guys get out and run about a bit. I finally took all the Ammo from the Helo to try and  force an engagement but this time BluFor got out after Unloading at ruins near Opfor -but they pretty much immediately started sprinting back towards almost the beach they started at -like 700m away without ever checking for Opfor. Not sure i understand what your trying to do here -maybe its the Seek and Destroy Waypoints as they are pretty vague. Maybe you should give the BluFor a general idea of where Opfor is so they dont just up and run off in the opposite direction

 

Edit Edit: Just to mention the Blufor team finally ended up in a deep valley/Gulley in that 700m away -is that what they were pursuing -deep terrain type cover?

  • Like 1

Share this post


Link to post
Share on other sites
14 hours ago, froggyluv said:

Sounds riveting !

 

Edit: So i only learn things by trying the demo first and im not quite sure what to make of the Demo mission. No mods at all - first few times after calling 0-0-1 the helo shows up and kills Opfor with its Helo Machine Gun. Then the guys get out and run about a bit. I finally took all the Ammo from the Helo to try and  force an engagement but this time BluFor got out after Unloading at ruins near Opfor -but they pretty much immediately started sprinting back towards almost the beach they started at -like 700m away without ever checking for Opfor. Not sure i understand what your trying to do here -maybe its the Seek and Destroy Waypoints as they are pretty vague. Maybe you should give the BluFor a general idea of where Opfor is so they dont just up and run off in the opposite direction

 

Edit Edit: Just to mention the Blufor team finally ended up in a deep valley/Gulley in that 700m away -is that what they were pursuing -deep terrain type cover?

 

Thanks for taking a look!  The position that the infantry moves to is only calculated to be within sight of where the player was when the radio was called.  The OPFOR is not taken into account -  to be honest the OPFOR was just kind of an afterthought to make things play out a bit differently each time.  

 

As for the fact that they ended up so far away, I've seen it myself once or twice and it could have a couple of causes.  One is if you pause the game while the algorithm is working - unfortunately this seems to lead to unpredictable behavior.  The other cause might be if the waypoint finding algorithm couldn't settle on a position in sight of where you were standing that was even close to the one it ultimately chose in terms of nearby objects and vegetation.  Unfortunately they do seem to have a small chance of choosing kind of dumb locations sometimes.  In a real mission I would re-issue the order every minute or two, or in response to certain conditions, so they wouldn't get stuck somewhere like that for long.

I'll definitely put up a better example mission soon that takes all those things into account.  The current mission is little more than following the basic example in the manual.  Feel free to move the units around in the editor (along with the triggers next to them) and see how the waypoint finding behaves differently.

 

The ultimate goal is for every playthrough to unfold differently, it is a sad artifact of my imperfect code (and limited processing speed) that every once in a while it acts dumb.

Share this post


Link to post
Share on other sites
1 hour ago, dwringer said:

The position that the infantry moves to is only calculated to be within sight of where the player was when the radio was called. 

 

Oh well then thats my fault as I teleported locations to try and keep tabs on the action -that must have thrown off your code. After that i added some more opfor and did witness some combat nearby the ruins. The Blufor landed right next to the ruins with the Opfor pretty much in the spot you placed them -maybe slightly further away. Anyways seemed like Blufor chose the Lowground (gulley) and Opfor went higher up the hillside. I had tripled the amount of Opfor to 9 or 10 and they won out. 

 

I get that you are doing this more as a general favorable terrain rather than a tight AI cover system so wondering if this might play nice with an AI mod that also helps them quickly move to cover and/or stay there. The problem with Arma AI even with mods is -the AI tend not to STAY in good cover - like the low walls of those ruins. Blufor could have planted a defense there right away but these guys tend to always want to "Overly-move"

 

Anyways i still think its pretty cool what you  are attempting -but yeah maybe a tighter Demo mission to more clearly demonstrate what and how it might benefit the AI

 

Edit: Also nice job getting those helos to rapidly land even in Hot Zone

 

  • Like 1

Share this post


Link to post
Share on other sites
3 hours ago, froggyluv said:

I get that you are doing this more as a general favorable terrain rather than a tight AI cover system so wondering if this might play nice with an AI mod that also helps them quickly move to cover and/or stay there. The problem with Arma AI even with mods is -the AI tend not to STAY in good cover - like the low walls of those ruins. Blufor could have planted a defense there right away but these guys tend to always want to "Overly-move"

 

Indeed, my scripts don't actually touch the unit AI in any way, just attempt to find favorable terrain at more of a macro scale.  It should work fine with any AI mod that still allows groups to eventually follow their assigned waypoints.  I've found it is best to let units in Arma discover each other from a distance, as they really do seem to struggle at close range, running right out in the open and even at times ignoring enemies directly in front of them.  That said they do a way better job than original Arma 😅

 

And thank you for the kind words.  Getting helicopters to unload in a hot zone without having to hold their hand is probably the number one reason I wrote this system, hehe

 

Share this post


Link to post
Share on other sites

Wow, what a treat. The idea of using genetic algorithms for Arma have been on my mind for quite a while so it's nice to see a framework for it. Exploring the map to find best locations sounds like a great use case.

 

Adorable manual, btw 😀

  • Like 1

Share this post


Link to post
Share on other sites

I've updated the example mission so the OPFOR and BLUFOR units will get new orders when they detect one another, so they'll attempt to find positions relative to the detected units rather than the player.  So, the engagement should play out a little better and make more sense.

 

I also fixed a couple of minor bugs in the code.  The new release is in the main post above.

 

 

20 hours ago, _foley said:

Wow, what a treat. The idea of using genetic algorithms for Arma have been on my mind for quite a while so it's nice to see a framework for it. Exploring the map to find best locations sounds like a great use case.

 

Adorable manual, btw 😀


I'm glad you liked the 80's AI lab aesthetic of the manual, lol.

 

Hopefully the optimization code makes some kind of sense, I know how irritating SQF can be to work with in general.

EDIT: I've added some sections to the manual to make the code slightly more approachable

Share this post


Link to post
Share on other sites

I've created a new release of ANIMA, this time with two example missions (and a few bugfixes/feature additions).

 

The new example mission shows off the algorithm's ability to find multiple positions at a time, landing two helicopters at once and issuing move commands to multiple fireteams on each side.

 

CHANGELOG:
- Added two parameters to fnc_find_positions: one is a sort-function (or "false") to sort the resulting positions (default sorts by average score).
- There are now two example missions.  The first one just follows the basic example in the manual, while the second employs a number of functions.

Share this post


Link to post
Share on other sites

I have issued a new release to commemorate the 1-week mark, move some of the fixes and features from the repo into a release (along with the expanded manual), and package a WIP version of an example mission I'm working on.  The example mission requires the CUP addons, but I tried to stay away from any DLC content (please let me know if there's anything I missed!).

 

The example mission, which is a small assault on an airfield (rather one-sided in terms of equipment) is much larger in scope than the previous ones and is, at the moment, not much of a mission (no player character is set).  However, it should be playable from the perspective of just about any unit (if you start in a helicopter or plane, you're going to have to wait a few minutes for anything to happen).  You may note that all waypoints (except those of the airplanes) are generated exclusively by ANIMA's position finding functionality.  Occasionally the assaulting force doesn't move up enough at first (eventually they should get orders that are closer), but other times they move in way too recklessly, so it's a difficult balance.

 

OgpLwWb.jpg
Screenshots of the example mission


REL-220316-A: https://github.com/dwringer/ANIMA/releases/download/REL-220316-A/ANIMA-220316-A.zip

 

CHANGELOG:
- Added sort behavior to fnc_find_positions which can sort the final results by a given function (by default, uses the average across all objective scores)
- Added a setting to allow fnc_find_positions to work faster, but less accurately, by allowing undifferentiated results after the target number of generations (a single bin after the nondominated sort, no dominance ranks in the result set)
- The behavior of optional parameters for attack_targets and approach_targets methods of Headquarters was fixed
- Fixed behavior of wp_config parameter of internal Headquarters method dispatch_transports
- Added an optional start_offset parameter for approach/attack_targets which moves the search start location relative to the targets by [x, y, z]
- Revised and expanded user manual with section on fnc_find_positions and some classes
 

  • Like 3

Share this post


Link to post
Share on other sites

So i messed around with the Winter demo and have o say thats just a really enjoyable demo you put together there despite it being entirely vague. Looks and feels great tho. 

 

Something to remember, you seem highly intelligent and most of us trying this are dumber, far, far dumber people. We need a little guidance as to what exactly to watch out for in terms of what we might expect from this new dynamic waypoint creator. With a little (read:alot) of guidance, i  think there is something very cool to play with here

  • Like 1

Share this post


Link to post
Share on other sites

You're too kind... you may notice that the dates in some of the source code files go back to 2016 or even earlier. That's basically because I had only the most rudimentary grasp of the codebase myself, in spite of having written it. I finally decided that it's at least useful enough for throwing together quick editor skirmishes that it would be worth documenting and releasing.

 

This is in fact something that originated in 2012 for Arma 2, when I would actually spawn soldiers and check their knowsAbout to judge line-of-sight between positions (checkVisibility didn't exist yet and I wasn't trying to use it dynamically in missions). My primary motivation was to be able to quickly figure out where to place waypoints in the 2D editor where the AI would engage each other from a distance (it was hard for me to tell such things just looking at the topographic map).

 

The automatic waypoint creation code could definitely be better, I think it's overly complicated in some areas and lacking basic functionality in others. As I wrote the documentation and started working on example missions I encountered many bugs, and I expect there are many more I haven't gotten to yet. 

 

Also, the actual objective functions that I wrote for the genetic algorithm are pretty basic and often not what I'd really consider ideal. Aside from code optimization, I think this is where most of the potential for improved performance lies. I'd love to write a guide on the process of creating and combining these if I get a chance, but to me at the moment it's more of a black art than science.

 

I do plan to release an overview guide of how I put together the Winter example mission, how triggers and scripts interact, things to watch out for, etc., hopefully in the next day or two.

Share this post


Link to post
Share on other sites

ANIMA - Vybor Airbase example mission
YT16Cyn.jpg


SITUATION
In a surprise attack, Russian forces from the northeast have captured the Pobeda dam and are expected to launch a combined assault on Vybor airbase imminently.

What they don't know is that it wasn't actually a surprise, and in preparation we've received a large shipment of AA and AT missiles along with a state-of-the-art backpack-sized UAV control terminal from the CIA (who were also kind enough to grant us the use of an MQ-9 Reaper currently flying in from the south).

An airborne platoon will also be arriving by Ghost Hawk to reinforce, but they are still a few minutes out.

MISSION
It's imperative that we knock out the incoming armor and clear the skies before our reinforcements arrive.

Your task is to take control of the MQ-9 Reaper and ensure these goals are accomplished.


EXECUTION

Activate your UAV terminal and take out the Russian armor ASAP. There are some launchers beside you that can also be used to take out enemy aircraft and vehicles.


I've created a development overview guide to go along with the Vybor Airbase example mission, as well as expanded the mission significantly with more units on both sides and a winnable goal (destroy all the enemy armor). After the armor is destroyed, the player should get an action menu option to end the mission within 5-30s (sorry, there's no real notification).  With the guide I've tried to explain every aspect of how the triggers are used along with ANIMA, but it's still intended as a companion to actually looking at the mission in the editor.


DOWNLOAD LINKS:

The mission itself can be downloaded from:
https://github.com/dwringer/ANIMA/releases/tag/WIP-220322-A

 

Development overview guide written to accompany the mission:

https://github.com/dwringer/ANIMA/blob/main/ANIMA%20Guide%20-%20VyborAirbase.pdf

  • Like 1

Share this post


Link to post
Share on other sites

Hey all, I'm still far from anything resembling a new release, but I feel the ANIMA engine is in good enough shape where developing applications is an equally worthwhile use of time.

Here's a new example mission, showing a new application of ANIMA. Instead of creating group waypoints, this time ANIMA's position finding is used to place vehicles in a given trigger area.  The vehicles are placed near cover/concealment, near roads, with LOS to some specified targets, and hidden-from-view with respect to other specified targets.  The file ./classdef/VehicleTargets.hpp contains all the code that implements the new application (relying on the ANIMA system).

Open the mission in the editor:

cobzR4d.jpg

Now launch it, and immediately bring up the map. You'll see some orange dots as the ANIMA algorithm evolves some positions at which to place vehicles:
kCiqlZD.jpg

After the evolution process has yielded enough suitable positions, vehicles will be spawned at the found positions.  These vehicles are chosen from those which were placed inside the trigger at mission start.

Here you can see a few vehicles, all poised for ambush of the position at the intersection just above-right of screen center:
J30EmtF.jpg

An example of one such position, with a partially covered LOS to the ambush location:
Ed1ps1k.jpg

A BLUFOR tank is approaching, and you should see it enter combat around the time it reaches the targeted intersection.  OPFOR positions are randomized, so it will play out differently each time, but BLUFOR is all but guaranteed to lose due to the ambush (and being significantly outnumbered).

The example mission can be obtained here:
https://github.com/dwringer/ANIMA/releases/tag/EXM-220407-A

EDIT: Last night it was working pretty well, but today when I ran it there always seemed to be a tank that wanted to spontaneously explode itself 200m into the air.  I tried to use BIS_fnc_findSafePosition to avoid this but I'm not sure if it does more good than harm. The algorithm could be tweaked to find better/safer positions but it would get a lot slower.  Anyway, it works a lot better on flatter terrains in that regard.

  • Like 1

Share this post


Link to post
Share on other sites

Ok so tried out the Demo a few times -it works but basically anyone can spawn 7 AT vehicles against 1 tank and Auto-win. I think to fully demonstrate a superior AI positional fighting script -it needs to show how a 1vs 1  -the newer AI wins out well over 50% of the time. For instance -i knew AI accuracy subskills were WAAAYY off when i tested a Crewman vs a Marksman at 200m and it was going down as 50% probability rate for either guy -even when I gave Marksman the highest 1 at skill while giving the Crewman the lowest (standard) setting of .2. Its only when i took those subSkills down to really small fractions like 0.005, that I could create proper disparity and have the Marksman win with a 95% probability rate.

 

Now sometimes in this script the AI gets some really ideal level looking AI placements, like just below Blufor sightline where with turrets are aimed and ready to go at the Blufor tank's underbelly -an ideal ambush placement. But again, are these Opfor AI basically given forewarning of Blufor's makeup (1 enemy armor)  and intended direction of travel? Will Opfor ideally try to gain a position for a Flanking attack on the Armors weakside? 

  • Thanks 1

Share this post


Link to post
Share on other sites
1 hour ago, froggyluv said:

Ok so tried out the Demo a few times -it works but basically anyone can spawn 7 AT vehicles against 1 tank and Auto-win. I think to fully demonstrate a superior AI positional fighting script -it needs to show how a 1vs 1  -the newer AI wins out well over 50% of the time.


Hmm, I should perhaps not have used "artificial intelligence" as a tag for this because it doesn't affect unit AI in any way, but I still think of it as more of a generalized disembodied AI.  Literally the only thing the VehicleTargets example does is pick places to spawn the vehicles.  There are alternative ways of doing this including selectBestPlaces and silola's amazing DAC functions, but this is just a variation on that which uses a genetic algorithm to try to find the positions by evaluation a series of functions.  You're right that if it can't win over 50% of the time in a 1v1, then it's not "superior", but that's not the goal here.  The goal here is to be "better than randomly placing the tank somewhere in the area" or "better than the result you get using selectBestPlaces".  It would be great to have such good positions as to win in a 1v1 fight, but at this time that's not the intention.

 

1 hour ago, froggyluv said:

Now sometimes in this script the AI gets some really ideal level looking AI placements, like just below Blufor sightline where with turrets are aimed and ready to go at the Blufor tank's underbelly -an ideal ambush placement. But again, are these Opfor AI basically given forewarning of Blufor's makeup (1 enemy armor)  and intended direction of travel? Will Opfor ideally try to gain a position for a Flanking attack on the Armors weakside? 

 

The OPFOR are not given any information at all.  The position finding algorithm is given two lists: one list of GameLogics which are placed along the approach BLUFOR will be taking.  Positions are selected to be hidden from these gamelogics if possible.  The second list is a list of GameLogics which are "kill zone" indicators.  Positions are selected to have partial LOS to these logics when possible (and they are turned to face the first one to which they have LOS).

 

As you say, sometimes the placements are ideal and other times they are imperfect.  One use of this might be to run it a few times, take note of the most effective positions, and then use that knowledge to manually set up mission events.  But I still enjoy the randomness for setting up skirmishes.

ANIMA is really intended for people to mess around with the genetic algorithm to try creating their own position finding algorithms, or perhaps think up novel uses that I haven't implemented (of which I'm sure there are plenty). Things like the VehicleTargets and Headquarters classes, and other things from the example missions, are just examples of the applications and definitely leave room for improvement.

 

Final Edit:

I should say again, though, I appreciate your feedback and taking the time to check it out. You have definitely given me helpful things to think about. The real idea behind the VehicleTargets script was not actually to set up ambushes, but it emerged as a possible use.  As the name hints, it was written to create (empty) vehicles to use as target practice, and I developed it originally on the flat tank ranges of the Hebontes map.

 

Share this post


Link to post
Share on other sites
On 4/7/2022 at 12:01 PM, froggyluv said:

 I think to fully demonstrate a superior AI positional fighting script -it needs to show how a 1vs 1  -the newer AI wins out well over 50% of the time.


Okay, I said I wasn't trying for that, but I couldn't say no to a challenge 😄. Now I can say I have an example mission which (might) satisfy that criterion.  At least in the few tests I've done, the ambushers wins a majority of the time.

 

I added some improvements to the VehicleTargets class, so now it makes the LOS requirements mandatory before spawning vehicles, and it sorts the positions with a weighted average calculation similar to those used with selectBestPlaces

The result of those efforts is demonstrated by this mission, in which a convoy of 4 BLUFOR tanks drives along a road and is ambushed by 4 OPFOR tanks which are placed by a genetic algorithm.


Since Arma's tank AI is insane and will choose to turn its back to enemy tanks in response to a threat, I disabled the OPFOR's ability to move, and used TOV_fnc_SimpleConvoy (from https://forums.bohemia.net/forums/topic/226608-simple-convoy-script-release/) to drive the BLUFOR tanks down the road.

Here's an example of one of the ambushes it generated during my tests:

e0m1uHh.jpg

After a couple of minutes the remaining two red tanks worked up the courage to move forward, did manage to take out the blue tank in the road, but were promptly destroyed.

Edit: On starting the mission, pull up the map.  It will take a minute or two for the OPFOR to spawn, but you can watch the positions evolve as orange dots on the map.

You can get the example mission here:

 

"Tank Trials" Example Mission: (added 9 April 2022)
https://github.com/dwringer/ANIMA/releases/tag/EXM-220409-A

 

  • Like 2

Share this post


Link to post
Share on other sites
5 hours ago, dwringer said:

Okay, I said I wasn't trying for that, but I couldn't say no to a challenge 😄. Now I can say I have an example mission which (might) satisfy that criterion.  At least in the few tests I've done, the ambushers wins a majority of the time.

 

Nice man ! I dig the oneUpsManShip in the name of progress 😄

 

Tried it about 5 times so far and yes Opfor is winning out each time -nicely done.

 

5 hours ago, dwringer said:

Since Arma's tank AI is insane and will choose to turn its back to enemy tanks in response to a threat, I disabled the OPFOR's ability to move

 

This I find absolute tragic and has been probably my biggest gripe (i am a premier griper btw) against Arma 3 forever now. I just dont get how you can make an Army/Military game and not have Combat Orientation absolutely The number #1 priority above all else!? I mean ALL ELSE. It dont matter how many Islands maps to traverse, how many colorful snakes and butterflies, how many cool tac outfit from all nationalities -if you cant aim in the proper direction...like i just never got Bohemia's blaise attitude toward not correcting this. As far as Infantry goes, the LAMBS mod has made some pretty nice Orientation fixings -that being soldiers staying oriented on premier target regardless of formation - but vanilla is just a dice roll. It just saddens me that people spend all this time LARPing military protocol to figure out some clever way to ambush an enemy -only to get there and have half their units aiming towards the open sea, or getting off a shot, standing turning, laying prone turning only to be killed in the midst of that chicken dance. 

 

Sorry for the rant -but its just a shame this isnt at the top of their priority lists whilst they still release DLC's etc..

  • Like 3

Share this post


Link to post
Share on other sites

Hi dwringer - been keeping an eye on this as Im interested in stuff like genetic algorithms. Ive tried out your recent tank trials example mission and it works as indicated. A thought though - the OFOR tanks appear to spawn in quite late as the Blufor have already arrived and I wondered  - what exactly is it that triggers the completion of the algorithm and the spawning in of the OPFOR tanks? ie what is triggering the end of the algorithm and the selection of the final spawn points? Is there a "fitness"criteria that has to be reached? ( Obviously the genetic algorithm is sorting possible points by fitness as it runs but I mean what triggers the conclusion that the final selection has been reached?). 

 

Edit - having run tank trials a number of times  Im getting the following - 1) fairly often the algorithm doesn't "complete", no OPFOR are spawned and orange blob action seems to continue. 2) sometimes it did work though 3) the placement of the trigger area and its size seem very sensitive - again moving the trigger area and setting its size say to 300m usually resulted again in the algorithm not completing and no spawning in.  4) when the algorithm did complete and OPFOR tanks spawned in this was generally well after the bluFOR tanks had arrived and generally had reached their final move points. I mve the BLUFOR  tanks a further distance away to start to give the algorithm more time to complete but generally it often still didnt complete.

Share this post


Link to post
Share on other sites
On 4/10/2022 at 2:47 AM, froggyluv said:

This I find absolute tragic and has been probably my biggest gripe (i am a premier griper btw) against Arma 3 forever now. I just dont get how you can make an Army/Military game and not have Combat Orientation absolutely The number #1 priority above all else!? I mean ALL ELSE. It dont matter how many Islands maps to traverse, how many colorful snakes and butterflies, how many cool tac outfit from all nationalities -if you cant aim in the proper direction...like i just never got Bohemia's blaise attitude toward not correcting this.

Amen brother!  Dev responsible for Tank AI and Tank driving does need to be held responsible.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×