bad benson 1733 Posted September 4, 2015 I think it's just a limitation of the way AI behaviour is implemented (I'm not arguing for/against anything here). AI only have varying degrees of combat awareness, other types of information aren't a consideration. So when they hear fire, they know it's hostile because to them, there's no other possibility. agreed. i think often times in games it's similar to magic. it's about creating the illusion. the "game/engine" knows everything anyways so it's about simulating knowledge by setting limits. i agree with cosmic10r that they should always go investigating instead of magically reacting even if they cross checked with friendly units that might have actual line of sight. i don't really see there being a problem with complexity. we're not talking about reality. do we really want to have AI go investigate friendly gunfire? i don't think so. and if someone wants that he/she can find a scripted solution. so suppressor/enemy weapon used or not. they should only consider enemy gunshots anyways since this is about creating meaningful gameplay and not simulating friendly fire accidents (imho). so i'd like a routine where they, as mentioned, investigate (seemingly) by changing their position, after hearing an enemy gunshot or footsteps, with the "secret" agenda to get actual line of sight to then, when they get it, engage in a justified manner. and the cross checking with friendlies might have to get limited in some way. because who knows maybe the fact they do that already is exactly the reason they shoot at stuff they don't see. it has to be considered that they should only get a "he's somewhere around that wall there" from a friendly instead of a pinpoint 3d position to create the illusion of human error and limited perception. so i'd even go as far as to remove shooting through walls entirely (squad mates having visual or not) unless they had line of sight a certain short amount of time ago (few seconds). 5 Share this post Link to post Share on other sites
SilentSpike 84 Posted September 4, 2015 That's exactly the kind of behaviour I was going to suggest, but I couldn't figure out how to eloquently explain it. ^_^ Agree on all points. Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 4, 2015 Agreed... if you watch Greenfists mission you can see that as the AI lose line of sight, they follow an estimated path based on speed and trajectory. The only blind firing I would like to see (since we will likely never get full blown blind firing) is following that last known trajectory. The AI seem to only have this knowledge for 5 seconds or so. So as you go behind a small house the ai may fire following your speed and trajectory and may even hit you if you maintain same speed. That seems plausible... As for the rest... gaining a better position and investigating sound seems like the ticket... may have to try some scripting on Mikey's framework later Share this post Link to post Share on other sites
domokun 515 Posted September 7, 2015 I think the realistic solution is that the AI doesn't blindly fire but as said... Investigates gun fire while attempting to gain a positional advantage to the known shooter location QED https://forums.bistudio.com/topic/184331-the-ais-reaction-timeability-is-insane/page-2#entry2909473 Share this post Link to post Share on other sites
KeyCat 131 Posted September 7, 2015 From today's change log... Removed: Class-based differences in AI spotting abilities (it is covered by other systems instead) Sound interesting! Anyone got a chance to test yet and noticed any difference in AI spotting? /KC Share this post Link to post Share on other sites
R3vo 2652 Posted September 7, 2015 AI seems to be able to look through wall and floors of the large watch tower. Share this post Link to post Share on other sites
msy 22 Posted September 8, 2015 AI seems to be able to look through wall and floors of the large watch tower. Not seem, it always does in the tower. http://steamcommunity.com/sharedfiles/filedetails/?id=514754783 http://steamcommunity.com/sharedfiles/filedetails/?id=514754113 http://steamcommunity.com/sharedfiles/filedetails/?id=514753745 You or AI can go through the wall and you or AI can shoot the body that is seen. Share this post Link to post Share on other sites
KeyCat 131 Posted September 8, 2015 As promised here are my updated test missions using Greenfist's scripts and two different skill settings for all soldiers in the OPFOR squad (0.1 and 0.5). I used the following values in my *.Arma3Profile... skillAI=0.5; precisionAI=0.34999999; Link: https://www.dropbox.com/s/yang4mei4lsxl5y/A3_AI_Terminators_II.zip Repro steps:1. At mission start go prone directly and remove the silencer.2. Slow crawl ~2 m to the bush in front of you and make sure you don't select the "Look at you" option by mistake.3. Take a nice prone position in/under the bush on the right side and look into the base NW with 4x scope, you should see the AI group standing with their backs against you at distance ~200 m4. Try kill the guy at the corner of the white building using 1-2 shots in rapid fire then just lay comlpetly still. 5. Result without/with silencer ??? I tested both versions today with the latest build and as soon as the AI squad leader turns in your direction after killing one in the squad he spots and engage you in a way that is quite impossible for a human. I wouldn't mind if this was the result at very high/max skill settings but as it is it seems impossible to scale down AI using lower skill settings, at least not in this scenario. If anyone tries please share the result and your settings and if anyone have any other test mission in similar situations (i.e hiding at semi long ranges after firing a few shots) that gives different results please share or maybe the dev's have suggestions on how to test it in a better way... /KC Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 9, 2015 As promised here are my updated test missions using Greenfist's scripts and two different skill settings for all soldiers in the OPFOR squad (0.1 and 0.5). I used the following values in my *.Arma3Profile... If anyone tries please share the result and your settings and if anyone have any other test mission in similar situations (i.e hiding at semi long ranges after firing a few shots) that gives different results please share or maybe the dev's have suggestions on how to test it in a better way... /KC Had a look This Video uses aceAI.pbo and therefore does not represent unmodded gameplay in Dev branch... Apologies for the confusion That video is on the next page of this thread. Thanks Video moved to this thread https://forums.bistudio.com/topic/184708-ai-discussion-any-branch/ 1 Share this post Link to post Share on other sites
KeyCat 131 Posted September 9, 2015 Thanks for making the video! It's really interesting since even if you are less concealed than in my tests it looks like the AI have harder time detecting you. What settings do you use in your *.Arm3Profile file? I watched it on my phone but will take closer look tomorrow.... /KC Share this post Link to post Share on other sites
Dwarden 1124 Posted September 9, 2015 important explanation for disableRemoteSensors true; mission/scripts/addons must have code/actions/triggers etc. written in way, where is executed/handled/operated with everything in locality where AI is local (runs) this covers script-commands: knowsAbout, nearTargets, targetKnowledge Example: 1. you setup HC group with AI (guard_grp) 2. guard_grp notice player trying sneak to base situation on HC (AI is local there) knowsAbout will be tested to raise alarm > Success ! situation on dedicated server (AI not local there) knowsAbout will be tested to raise alarm > Fail ! group explanation: Squad leader is player #42, all group AI are local to client #42 member of group is player #67 (beacuse he is player, he is remote, non-local) this case is covered so player #67 share what he sees with player #42, there is condition on whole group and not just unit note: High Command (complex groups) is WIP and should be adddressed in future it's not per AI command, it's for whole "simulation" locality so you run it once (e.g. mission init) on each client, server, each HC ... unless you have code which awaits something from remote locality for those 3 commands which could break e.g. in init event scripts https://community.bistudio.com/wiki/Event_Scripts so just add into mission init.sqf : if (!getRemoteSensorsDisabled) then {disableRemoteSensors true;}; yet repeated re-enable and disable is possible, as/if needed additional details: http://dev.arma3.com/post/sitrep-00122 1 Share this post Link to post Share on other sites
R3vo 2652 Posted September 9, 2015 Here's a ticket and another showcase for AI spotting and shooting through the watcher tower floors. I needed 6 attempts to clear that tower because of that. http://feedback.arma3.com/view.php?id=25523 Share this post Link to post Share on other sites
danil-ch 165 Posted September 9, 2015 AI refuses to cross bridge while moving backwards Still true Share this post Link to post Share on other sites
KeyCat 131 Posted September 9, 2015 Had a look Looking at your video again, this time on a real screen and it's really interesting since the AI reactions differs a lot in your test... Things I notice... you fire 6 shots and are less concealed than I use to be (I hide in the bush to your left), I only fire 1 and max 2 shots and when the squad leader turns in your direction he doesn't spot you (when I test it take 1-2 seconds after he looks in my direction). You even roll on the ground a couple of times after firing, I lay completely still. A few questions. 1. What version are you using in the video, 0.50 or 0.10? 2. Would be great if you could check what AI settings are at in your <name>.Arma3Profile is or maybe upload it somewhere if you are unsure. Now the question is why are the result is so different? Looking at your HW specifications i see you are using a 6 core 3930k @ 4.2 GHz. Myself use a ancient 2 core E6850 @ 4.22 GHz. I know 2 vs 4-6 shouldn't make any difference and I'm just stabbing in the dark here but maybe it is??? Videos are great and next time I will try to make a video showing the result when I test... /KC Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 9, 2015 I used 0.10 in that test. I'll check my arma profile next and see what that says but it is more likely the ace ai pbo that I have further edited... will try to isolate later today by disabling that and see if they re... terminate... ;) Share this post Link to post Share on other sites
KeyCat 131 Posted September 9, 2015 I used 0.10 in that test. I'll check my arma profile next and see what that says but it is more likely the ace ai pbo that I have further edited... will try to isolate later today by disabling that and see if they re... terminate... ;) That AI addon probably explains the different results... for tests everyone should run stock A3 without any addons and preferable use the same AI settings in the <name>.Arma3Profile so we compare apples to apples ;) /KC Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 9, 2015 That AI addon probably explains the different results... for tests everyone should run stock A3 without any addons and preferable use the same AI settings in the <name>.Arma3Profile so we compare apples to apples ;) /KC I'll disable it and do a vid tonight... see if we get same results then Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 10, 2015 I'll post the vid later but it may not be relevant as I realized I had the ai set to custom .05 on the skill slider lol So I set them to expert preset... but still no terminators... that will be the more relevant test so I'll do that later today altho I may go straight to vanilla for the test and disable all mods to be sure... I have a hard time believing that the extra threads contribute to a different ai experience. .. Is that even possible? Share this post Link to post Share on other sites
old_painless 182 Posted September 10, 2015 As for todays entry about AI being able to give medical aid to units outside their team: Does that mean that in SP I could be treated by another teams medic on the AI commanders order, or does it "only" mean that I can order my medic to heal units in other teams ? So to what extent is the AI healing across teams by themselves ? -OP Share this post Link to post Share on other sites
seba1976 98 Posted September 10, 2015 I have a hard time believing that the extra threads contribute to a different ai experience. .. Is that even possible? The only time when it would be legit to leave an addon loaded during testing AI, would be when you have made that addon, otherwise better safe than sorry. Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 11, 2015 The only time when it would be legit to leave an addon loaded during testing AI, would be when you have made that addon, otherwise better safe than sorry. That confused me... what does having made the addon have to do with extra cpu threads changing the way the ai behaves? Share this post Link to post Share on other sites
seba1976 98 Posted September 11, 2015 That confused me... what does having made the addon have to do with extra cpu threads changing the way the ai behaves? It just means you can't be 100% sure of what an addon changes, unless you have made it, or studied it by opening it and looking every line of code for yourself. Share this post Link to post Share on other sites
cosmic10r 2324 Posted September 11, 2015 It just means you can't be 100% sure of what an addon changes, unless you have made it, or studied it by opening it and looking every line of code for yourself. We may be talking about different things here as KeyCat was analyzing hardware and wondering whether the 4 extra thread on my processor were leading to the difference in results in our tests ... To both your points this would be testing apples to apples... Somewhere in the difference to those results lies the answer... ;) Share this post Link to post Share on other sites
andersson 285 Posted September 11, 2015 We may be talking about different things here as KeyCat was analyzing hardware and wondering whether the 4 extra thread on my processor were leading to the difference in results in our tests ... I'm sure KeyCat wasn't analyzing hardware, he was analyzing AI in devbranch. As he gave his settings and wanted others to try it's assumed you would have the same settings. As the result was so different and the only thing he could see was that you had more cores he was probably trying to find if that was the reason why. Until you told us you 1 - use a different setting (0.1) which will change the test scenario and make it hard to compare. And 2 - You use an AI addon, specifically made to alter the AI behaviour (ace AI)! Then your test is of no use in testing the dev branch AI. So please disable all addons, specially all AI altering addons. Set you settings to what KeyCat use and try the same scenario again. Then KeyCat can compare and we can get some conclusions of the dev branch AI :) Share this post Link to post Share on other sites
Greenfist 1863 Posted September 11, 2015 10-09-2015 EXE rev. 132304 (game) Fixed: AI was throwing grenades at unknown targets Sometimes the AI would seemingly randomly throw the grenade at a position where there was no enemy. I bet these guys are Arma 2 vets, and were just trying to open their backpacks... This is probably the case where the AI detects an enemy and instantly throws nade, regardless of his direction? 2 Share this post Link to post Share on other sites
