3.0 Bug - Character & Follower Pathing Worse That Usual

Character and follower pathing and path correction are worse than usual in TestLive 3.0. In the later forms of 2.8 turning your back on them for a few seconds and then directing them to “Move” would almost always free them up - I guess by resetting their start point to somewhere slightly behind the character camera. Also about 95% of the time they would redirect themselves around path-blocking obstacles after about 30 seconds. These two pathing solution features seem to be MIA in 3.0.

3 Likes

Yeah, I was hoping that pathing would be one of the major focus points of 3.0 but sadly it is not. It’s a shame too because UE4 has some wonderful ai abilities that don’t seem to be getting taken advantage of by funcom. Hoping they can use a little of the store money they’re about to earn and spend it on getting the ai to work.

I know it is irrelevant, but what’s the music called? i like it, it gives me goosebumps, but I still don’t know the name! :smile:

That’s the OST from the game… I thought it sounded different and was going to ask if they changed or added some BGM… But since I usually play with the music volume set between 3 and 5 (out of 100) I didn’t wanna ask and look dumb like; “Where have you been, that was there since release - duh!”. :smiley:

1 Like

I know this from the game, it plays at night, but I could not fint it :slightly_frowning_face:
I searched on youtube Conan Exiles night music, and it gave me the Darfari one, and the Jungle one.

I have had my thrall get stuck IN my door twice already and I have had to move far enough away that it “ports” over to me in order to free it.

not in unreal engine 4.15

I don’t know about that. Even UE2 had great pathing & AI capabilities, epic games even won awards for it.

that may well have been the case back then. UE4.15 was certainly also good when it was current, i.e. at the beginning of 2017. but as of today, the systems are not really good and will not be anymore. seen in context, i.e. including all other systems, the 4.15 AI system is not an award winner. as soon as the server is a bit more under load, the AI is one of the first systems to stop working. there’s a reason why the servers are limited to a maximum of 40 players, and even that is still too many, depending on what they’ve already built on the server.

we’ve just read too many times about AI and pathfinding optimization in the patch notes without really noticing anything. npcs are still stuck at every corner and thralls refuse to fight as soon as the server is a bit more loaded

my personal opinion: not much will come with the original 4.15 AI system. minor improvements here and there, but no really noticeable improvements of the general AI. if we are lucky (and if it is possible), funcom will back port a more current version of the AI system from the UE 4.2x branch. I see significantly better chances with a more up-to-date AI system.

3 Likes

There’s a ton of game’s that are built on 4.15 that do not have the same issues. Surely there are plenty that have large server sizes. I’m not convinced it’s simply an engine issue.

1 Like

which ones? I don’t know a single one

edit: but basically it doesn’t matter. I don’t expect any noticeable improvement with the 4.15 AI. the patch attempts and the lack of results confirm my opinion. the AI and the pathfinding have been optimized for almost 5 years. result: marginal. even an external company specializing in AI had only moderate success.

1 Like

There’s not a large list of MMO’s. But it looks like AION and lineage II had a conversion to UE4. Not sure the exact build and a bunch of korean MMO’s. While not a significant list I would also constitute CE more of a sandbox game than a full fledged MMO.

no version numbers in this list

Aion is still CryEngine… the next-gen version based on Unrel Engine was never released as far as I know
Lineage II is Unreal 2
Lineage II Revolution is Unreal 4 and a mobile only game

It doesn’t list the engine patch version used, unfortunately I don’t know anywhere to find that detailed of info. But the fact the Lineage II is put under the Unreal 4 engine most likely means it had a conversion but I don’t really play those games so not 100% on that. But if your claim is 4.27 is a cure all for AI and pathing I strongly doubt it. If that were the case Funcom would have transitioned already.

I never claimed that. I never suggested a complete engine upgrade. there will never be more than a back port for individual systems

not really. every version after 4.15 is a major update and more a full reimplementation of the game than only a game engine update. this is already a confirmed statement directly from funcom. the initial implementation was based on UE 4.12 and during early access they upgraded the engine to UE 4.15.3. higher version where available at this time but not really feasible because of the massive changes introduced with 4.16 and above. it’s more like that we will see a Conan Exiles 2 based on UE 5 (or what ever is premium at this time) than we will see a full engine upgrade to something higher than UE 4.15.3

but we can really stop here. the endless list of UE 4.15 games with better AI doesn’t exist. patches from the last 5 years haven’t changed much. the options for engine upgrades are very limited, at least when money and time is an issue. all these points tell me that it will not get better in the short term, no matter what funcom invests. a more up-to-date AI system or a complete in-house development. neither easy nor small tasks

You’re insinuating that the problems strictly with 4.15. Can you provide example’s of other 4.15 games with the same pathing issues?

There are loads of games that go through engine updates. I’m not suggesting that they should do that here or even if it’s financially feasible, Because I honestly don’t think that’s the issue to begin with. NPC’s in this game have very basic pathing setup compared to comparable games built with UE4. I have never heard that 4.15 has significant pathing issues nor could I see as to why it would. So I’m still confused as to why you are so certain that it’s the engine that’s the issue and not the development team.

1 Like

Well, ummm… AI pathing in the last X number of release versions is very definable in the Behavior Tree. It’s not like you just drag and drop AI onto your NPC and it does some behavior. The basic Node Graph pathing is comprised of “Find Path Point” (I am here!) and “Move To” (I wanna go there) typically with multiple “Increment Path Index” (figure out a path) nodes each which runs a sort of sim and reports success or fail to a “Selector” (pick the most successful one) node. The number of IPIs you use matters a lot. That whole branch should be running below yet another “Selector” above the very first FPP - which will reset the point and then do every thing all over again for any number of iterations until success is actually achieved.

All this is user (Dev) definable and with tunable variables all along the way - so that it’s entirely possible to make a very very good pathing and patrolling system or to make a really really bad one, all in the same release version of UE. For example if you didn’t iterate with multiple nested Selector IPIs the NPC would very often run into a wall and just stop there until perhaps something hierarchically critical occurred. If you know UE you’ll know this is sort of a partial explanation but it’s accurate to the degree it needs to be to make the point - which is that this isn’t a UE or a UE version problem… this is developer related.

that’s only part of the story. the older systems are very heavy and require a ton of resources, and that’s the main problem here. you can’t scale it up until it works 99% of the time without affecting other systems. as a reminder, currently AI is at low priority and is one of the first things to be scaled back when the server is under load. the newer systems are far more optimized. that’s where i think backporting a newer system will yield better results than trying to make a resource intensive system work properly. as i said, optimization has been going on for 5 years. even an external company that specializes in it has tried. the result is rather unspectacular. of course, you can make life easy for yourself and rant about the incompetent devs. but it doesn’t even come close to reality. if i have a crappy system, and that’s where i would put the 4.15 AI and pathfinding systems, then i can’t perform any miracles with it.

and of course you can get something useful out of such a system with a lot of effort. but then you have to face the prioritization and effort is also a decisive factor. what will get more priority? something where i see a noticeable result in 10 developer days or something that even after 100 developer days just brings an improvement in the per mille range?

solutions are theoretically always easy if you ignore the complete context, but unfortunately it doesn’t work that way.

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.