AI Driving - Feedback topic

[Tankbuster 1733]

4 hours ago, fn_Quiksilver said:

 

AI tickets do not get assigned to a dev, and have not been for years.

 

example:

 

https://feedback.bistudio.com/T124620

 

I speak to them directly occasionally too, but there is little evidence to suggest anything has been done to address AI driving issues since June 2016.

That and having to get your head around the dog's dinner that is Phabricator.

[mad_cheese 592]

At the risk that this was noted before:

 

I was really happy wih AI driving since the last improvements (apart from bridges etc), however I have now started to experiment with large convoys and the results have been equally impressive as frustrating.

 

One thing that keeps happening to me is that the convoy gets stuck, but it's actually the leading vehicle that refuses to move. If I give the vehicle a gentle push, all the other vehicles in the convoy bravely follow up, but the leading vic keeps having problems - sometimes pushing it eventually makes it recalculate the path and move. Since ungrouping the lead-vic makes him very reliable again, i assume he is waiting for something indefinitely. Best footage I have is this, just ignore the rest of the video.

 

 

Otherwise I have also seen following vics go off-road causing the leader to slow down drastically. Interestingly that is not always the case. This happened if more than one vic went offroad, but this could easily be me connecting the dots wrong.

 

However, if it does work... It looks and feels really amazing. The bugs feel as if the main task has been completed and there's some problems in the fineprint. But what do I know.

 

Still love this game. Foreva!

[SuicideKing 233]

Yeah i had zeus-spawned tanks refuse to move and/or skip waypoints after getting into formation. Seemed to be random, and maybe influenced by where they were placed on the map. In one case the tank went in the opposite direction to the waypoint...

[krycek 349]

Translation from BI devs "Next one will be better guys, don't worry!"

Meanwhile my ticket where crew abandons vehicles after group switch has gathering 2m of dust like so many other AI driving issues, they can't even fix stuff that worked in A2....but somehow things will be better in Enfusion.

See you in A4 with "Don't worry guys, our team moved to A5/Trulyfixedfusion but here take some dlc".

[dragon01 902]

Do note that Enfusion is a new engine. I don't know if they carry over anything from RV, but in all likelihood, it isn't much. Any bugs found in RV are irrelevant to Enfusion, because it's new code and new AI. It will have bugs, of course, but those will be new bugs and will likely manifest in a completely different manner. ArmA3 still has holdovers dating all they way back to OFP, but ArmA4 won't. They're not fixing it because they're throwing it out. Given the fundamental problems with AI behavior in ArmA, that is probably the best course of action here. Enfusion is a chance to do it right from the start. 

[The Man Without Qualities 110]

1 hour ago, dragon01 said:

Do note that Enfusion is a new engine. I don't know if they carry over anything from RV, but in all likelihood, it isn't much. Any bugs found in RV are irrelevant to Enfusion, because it's new code and new AI. It will have bugs, of course, but those will be new bugs and will likely manifest in a completely different manner. ArmA3 still has holdovers dating all they way back to OFP, but ArmA4 won't. They're not fixing it because they're throwing it out. Given the fundamental problems with AI behavior in ArmA, that is probably the best course of action here. Enfusion is a chance to do it right from the start. 

 

Those bugs are not related in the engine itself, it is simply BIS way of distributing attention and care.

The problem with disembarking crew we will see again if the very same devs will work on and with Enfusion again.

[pierremgi 4737]

Sure. there is no one magic solution, but a lot of work to make things better.

[dragon01 902]

28 minutes ago, The Man Without Qualities said:

 

Those bugs are not related in the engine itself, it is simply BIS way of distributing attention and care.

The problem with disembarking crew we will see again if the very same devs will work on and with Enfusion again.

It is related to the current AI system. Again, AI on Enfusion (I hope) will be a completely different system. There's literally no reason why an exact same issue would pop up in two game engines that only share a few similarities. Unless your problem is with something that was done on purpose, that is, but even that can change if the new engine offers a better way of modeling the offending situation. 

[pierremgi 4737]

3 minutes ago, dragon01 said:

It is related to the current AI system. Again, AI on Enfusion (I hope) will be a completely different system. There's literally no reason why an exact same issue would pop up in two game engines that only share a few similarities. Unless your problem is with something that was done on purpose, that is, but even that can change if the new engine offers a better way of modeling the offending situation. 

 

https://www.reddit.com/r/arma/comments/5vfzua/whats_the_hype_with_the_enfusion_engine/

 

[dragon01 902]

Make sure to read the response, too. 

Quote

[...]Enfusion (at least, in Take on Mars) is more like a generic game engine; like Unreal. It provides a bunch of basic features (models, physics, UI tools, etc) and lets you build almost anything from that via the scripting language (Enforce).

The idea behind it is you could implement anything; a card game, driving simulator, shooter, etc.

In contrast, much of Arma 3 is hardcoded in C++, like damage handling, ballistics, vehicle models, inventory handling, editing and so on. Scripts can interact in very limited ways, but you couldn't for instance implement a rocket ship or pinball machine in Arma 3.

With Enfusion you could. Since BIS would be implementing almost everything in Enforce, modders would be able to do almost anything BIS can do. For example many of the C++ changes BIS is currently doing to implement new radar systems in A3 could, in the Enfusion engine, be done by modders instead.[...]

The problems with AI (and many other things in ArmA) lies in those hardcoded parts. BIS can, in theory, change them, but it's a risky business. Enfusion lacks those restrictions. It's a long road from Enfusion to ArmA4, but it gives a chance (actually, kind of forces them) to rebuild problematic components from scratch, which is what I'm hoping for. It also gives modders a lot more freedom with regards to customizing various aspects of the game, which means they could attempt to fix issues like that on their own.

[krycek 349]

43 minutes ago, dragon01 said:

It is related to the current AI system. Again, AI on Enfusion (I hope) will be a completely different system. There's literally no reason why an exact same issue would pop up in two game engines that only share a few similarities. Unless your problem is with something that was done on purpose, that is, but even that can change if the new engine offers a better way of modeling the offending situation. 

 

Actually it's not only about legacy bugs, but also about stuff that simply worked before and now it doesn't or they broke it but they don't actually care at all to devote manpower to fix it because "next one will be better" excuse.

Also, if I remember right the waypoint bug other users reported where it skips it or AI won't simply move anymore it's again something quite new. After they announced improvements to AI driving in 2015 or 2016 (not even sure anymore) the magic lasted only a few months because they broke stuff and waypoint bug showed up.

Regarding my ticket this is about a bug that was in A2 in the beginning and it was fixed (ironically group switch was considered a feature for A2). Here's a video I made in AI Discussion thread a while ago that shows exactly what I mean and to prove them that it actually worked in A2OA:

 

Basically even stuff that worked now it's broken so no, I don't have much hopes for their Enfusion engine even if it's from scratch, what stops them doing the same?

Maybe they release a patch that brakes parts of the AI in the later part of the A4 lifetime and they'll say like always do "yeah we moved our main team to the next project, tough luck".

[dragon01 902]

You still don't get the point. If the AI is good at the start, it will stay that way. Enfusion is modular, which is the main point here. Do you think they deliberately broke the AI (it's ridiculous if you do, BTW)? ArmA3 has expanded greatly upon OFP, but other than large naval vessels, there really aren't any major things that could be added at this point. Enfusion AI simply has to be designed to correctly handle everything that is in ArmA3, and it'll be mostly good for the rest of the game's lifetime unless they come up with something bizarre. Current ArmA AI, on the other hand, was iterated on with every single RV version since OFP, not to mention as a hardcoded part of the game, there's a chance that even changes in code not directly related to AI can break it. Enfusion mostly avoids that by being modular. That's not to say AI subroutines won't require fiddling with at some point, but it should be a lot more robust system than RV.

 

ArmA3 is nearing its end of life, like it or not. Post Tanks DLC, the only things added will be fixes and maybe the odd community DLC (an idea which doesn't seem to have taken off). Hopefully they'll get to fix AI after finishing the new armor system, LWRs (promised back when they did sensors) and other tank-related functionality.

[oukej 2910]

2 hours ago, krycek said:

Also, if I remember right the waypoint bug other users reported where it skips it or AI won't simply move anymore it's again something quite new.

 

The waypoint issue got fixed in rev. 144260. Some changelog entries got lost throughout the 'logistical issues' ;)

 

1 hour ago, dragon01 said:

Current ArmA AI, on the other hand, was iterated on with every single RV version since OFP, not to mention as a hardcoded part of the game, there's a chance that even changes in code not directly related to AI can break it.

 

This. Code or data :/ Thank you :)

 

1 hour ago, dragon01 said:

LWRs (promised back when they did sensors)

 

Hmm...did we? I'm sorry but it's not planned :/

[x3kj 1247]

@oukej

The "drive as commander in proxy" feature is very nice. One thing that i find inconvenient however is that the AI interrupts its automatic driving (engaged by double tapping forward) as soon as you give a steering command. It stops as soon as you release the steering key. It would be nice if it would keep going at that speed, until a forward/reverse button is pressed again, no matter what steering inputs are used.

[dragon01 902]

1 hour ago, oukej said:

Hmm...did we? I'm sorry but it's not planned :/

Well, I do remember it being talked about. In particular, the ability to detect when another vehicle/designator guy ranges or paints you. Pity that it isn't planned after all, it could add some tactical depth.

[krycek 349]

12 hours ago, oukej said:

The waypoint issue got fixed in rev. 144260. Some changelog entries got lost throughout the 'logistical issues' ;)

 

Then this is great news, from two serious issues for me that completely broke AI driving at least an important one got fixed. Hopefully the group switch bug can also be fixed.

 

@dragon01 I never said they broke stuff on purpose frankly don't know how it even crossed your mind, just AI driving parts(besides the legacy bugs) that they broke late in A3 life and it's by luck if it gets fixed or not.

Or (if) they pay enough attention to it to care to fix these issues.

For all the excuses on this planet and how much people are willing to accept all sorts of crap in gaming I like to think that for Arma series even though we agree on the complexity of it no player is masochistic enough or has such low standards to enjoy A3 if it will be left in a broken state.

[dragon01 902]

Well, you acted like BI devs were at fault for the bugs. Truth is, as @oukej mentioned, on a monolithic engine like RV, breaking the AI can be caused by changes in a seemingly completely unrelated area. Debugging this sort of thing is a serious problem, you have to understand just how maddeningly complex that sort of thing is. Be glad that it's not an open-source project dating back 1994 or so, I've modded a game that ran on that, so I know how it's like. You don't just "go and fix a bug" and any attempt to do that in a haphazard manner will inevitably break something else, if you can get the code to compile at all after your "innocent change". That's just how things are, and why I'm excited for Enfusion - it'll be somewhat less affected by that sort of crap. The modular design, if implemented and used right, should be much easier on both devs and modders alike (not to mention opening up a lot of functionality to the latter).

[pierremgi 4737]

Wait and see...

Only the result is important. No matter the engine for us.

You can change all engines you want, if there is "no engineer", you'll fail, not saying that for BI in general.

Example: if you have some HE shells for tanks, never used by AI crew on infantry, but rather crew spilling APFSDS shell for nuts... never mind if it's RV or Enfusion, you miss something.

 

[dragon01 902]

You still fail to understand. When a new AI system is developed for Enfusion, this will most likely be accounted for. In RV, it may or may not be. It might have worked in OFP and broke at some point, or maybe it was never implemented because OFP AI wasn't advanced enough and nobody found a good way of implementing this afterward. Stop assuming that anything about the current AI will carry over, because the simple fact that they're changing engines means that it will all probably have to be written from scratch. This is not because the devs are incompetent.

[.kju 3239]

DEL

[krycek 349]

33 minutes ago, dragon01 said:

Well, you acted like BI devs were at fault for the bugs. Truth is, as @oukej mentioned, on a monolithic engine like RV, breaking the AI can be caused by changes in a seemingly completely unrelated area. Debugging this sort of thing is a serious problem, you have to understand just how maddeningly complex that sort of thing is. Be glad that it's not an open-source project dating back 1994 or so, I've modded a game that ran on that, so I know how it's like. You don't just "go and fix a bug" and any attempt to do that in a haphazard manner will inevitably break something else, if you can get the code to compile at all after your "innocent change". That's just how things are, and why I'm excited for Enfusion - it'll be somewhat less affected by that sort of crap. The modular design, if implemented and used right, should be much easier on both devs and modders alike (not to mention opening up a lot of functionality to the latter).

  In that case if this ancient engine (which I've agree, there are some legacy bugs going to OFP like the AI switching between binocs/nvgs) that you can't drop a fart around it and AI will collapse, then work on Enfusion should have started much sooner, perhaps before A3 release.

You're forgetting a tiny detail, I've been here since OFP, I'm not mentioning this for some bragging crap but for me the "next one will be better" excuse grows extremely tiring after OFP-A1-A2(OA)-A3. A lack of competition probably caused this engine to live so long.

 

At one point there is a line where you stop, how much understanding for the devs are you capable after you invest lots of your time (and money) in their products. That investment also translates into "ok I trust the devs that the next project will indeed be better". So far all I'm readying from you are excuses and more excuses and how Enfusion is second coming. If that is enough for you, maybe you don't have any expectations about their games and you're happy no matter how good/broken a BI product is.

[dragon01 902]

7 minutes ago, krycek said:

  In that case if this ancient engine (which I've agree, there are some legacy bugs going to OFP like the AI switching between binocs/nvgs) that you can't drop a fart around it and AI will collapse, then work on Enfusion should have started much sooner, perhaps before A3 release.

Here I can agree. This ancient thing should've been ditched much sooner. As capable as RV is, aside from bugs, it has a number of fundamental limitations. Inability to have anything underground, "real" water above sea level or simulating Earth's curvature (with 20x20km maps you really do need that) come to mind. Ability to grow past those is at least as exciting as the possibility of old issues being gone. However, the decision to switch engines is not one to take lightly. Enfusion has actually been in dev for quite some time, and presumably is the reason for A3's longevity. They know next ArmA isn't gonna be on RV, but the replacement is nowhere near ready. 

[.kju 3239]

Quote

Tweaked: AI driving wheeled vehicles now prefer roads less when in Combat or Stealth mode

 

I am seeing three trucks in a convoy no longer follow the road - the leader is still fine in the center, but number two and three drive left and right besides the road...

They are set careless. So the supposed change should not apply but i am sure it started with this dev branch update.

 

This was working fine a few dev branch versions before. Also confirmed working fine in stable branch.

Its also fully repeatable/reproducible. Will try to make a non (WW2) mods demo later tonight.

[.kju 3239]

@oukej can you please revert this recent change ^, or fix the issues it appears to cause please

demo mission: https://cdn.discordapp.com/attachments/165841157350227968/416318519848992779/testTruckConvoy.Stratis.7z

repro: launch with 1.80 stable and with latest 1.81 dev branch and watch
obs: in stable convoy follows road well; in dev only lead vehicle follows roads - the other two are left and right from the road (seems like V formation)
exp: convoy working again following the road properly

stable: 

dev: 

(the music is unintentional - was from a stream running at the same time :D )

[mad_cheese 592]

On 2/22/2018 at 9:46 AM, krycek said:

For all the excuses on this planet and how much people are willing to accept all sorts of crap in gaming I like to think that for Arma series even though we agree on the complexity of it no player is masochistic enough or has such low standards to enjoy A3 if it will be left in a broken state.

 

On the other hand, what positive thing does this inability to accept this state at any given moment in time accomplish?