Botpathing request thread

[sektor2111]

All right... Original Mover won't deal that much with A.I. as it might use ONE and TWO TriggerActors and Movers here are using even more than 3 keys and multiple triggering places. FerBotz using XC24 (probably latest version) are not having many troubles with navigation - these might be exactly the sort of maps which they like. Anyway... we do have several paths and CleanUp-s in these two. When a Pawn has bugs this doesn't means that 2 GB of maps must be edited instead of causing pawn to work properly. Messing with a new sort of mover here might not be good because:
- it takes a lot of time testing;
- maps have no brushes and then... pressing rebuild button it's not advisable.
Here they are it is:

DM-AeonFlux_Net_JM.7z

(1.33 MiB) Downloaded 28 times

Second one will be solved at some triggers and I'll add it back (with name changed)...
These are not including stunts a la UT Bot, maps were aiming Unreal and their "Bots" not "Bot". They are having the same inventories, purpose was pathing it for general roaming not only for one single type of pawn. In AeonFlux they might go for armor placed on that rope - yes, main paths were executed and completed with hand-made paths. Rebuiding paths here might deliver unpredictable results - I don't think I recommend more touches here.
If these are not what were you looking for, delete these files from your install.
Let me know if something else must be pathed - this can be done without to screw up the rest of data.

[a9902957@nepwk.com]

Hey, you rock! Thank you for doing this.
I will have to try these right away now.
EDIT: Yes, these are much better than before, one could tell that with ease from just one BRUT match on each.
EDIT2: On RealCity they appear to pile up/get stuck on that dysfunctional (it seems) elevator kind of thing in a building of a slightly rounded shape (which is located more in the center area of the map).

[sektor2111]

It would be helpful an image of that spot, perhaps I can setup a second combo for the same lift where it has to be placed or something else. FerBotz do look like are interested about paths but... these paths have to be executed/placed based on their movement capabilities (a bit aggressive I have to admit...).

Now days it's easy to connect navigation nodes but you need to make sure if movement it's physically possible - it depends on what can do that Pawn.

Either way I have to solve and test a few features which I added in my "paths builder" tool because last time I saw various options in messing up with automated pathing task which I solved right now, but... testing a wide range of maps is mandatory.

Then... Ferbotz and XC_Engine 24 to me are more suitable because prior versions like XC21 are doing something which I will never understand.
By example checking paths for a particular navigation node in XC 21 the function "GetReachSpec(R,RI)" doesn't do anything in Editor - with builder compiled in the same XC 21 environment, if I'm taking this in account, I can expect other things to work in the same way: Not Working more exactly.
This is report for a node in Xc 21

Code: Select all

ReachSpecsLoad: Paths List for PathNode417:
DataBelow: ___________________________________________
Out: Paths[0]-Index 1320
Start=	NONE
End=	NONE
Radius=0
Height=0
Flags=0
Dist=0 UU
Pruned=0.
...

And the same thing in XC 24:

Code: Select all

ReachSpecsLoad: Paths List for PathNode417:
DataBelow: ___________________________________________
Out: Paths[0]-Index 1320
Start=	PathNode417
End=	PathNode416
Radius=115
Height=110
Flags=1
Dist=385 UU
Pruned=0.
....

It's from that City map...

[UnrealGGecko]

I'll try to update the 1st post when I can with the newest maps in here, do remind me if I don't get around to it in a week or so.

[a9902957@nepwk.com]

Oh, you said the movers are trouble. I checked it, it is indeed meant to act as an elevator but as you pointed out UT pawns seem to not trigger this.
However I now also saw a stock bot get stuck there.
https://ibb.co/jGcGYdT
(Across the street. View is centered on the dark grey elevator box, maybe the sister building is affected, too. What is for sure is the elevator in the "clinic"/silo looking building where the bus starts [that building also has a red cross on its roof. Saw a bot getting stuck in on that elevator as well.])
I guess it may be too much of a trouble then. However this already plays way better as I said. Thanks again!

testing a wide range of maps is mandatory.

So, you would take more requests?
If so you could make e.g. https://www.mapraider.com/maps/search/v ... sT-Valley-][
shine more with your botpathing know-how.
However for BattleRoyale games improving the suggested maps in
viewtopic.php?p=131143#p131143 might be better. The 2ndSpaceBeacon map would need it the most of those three and SkirmishInLostAtlantis and AssiniboineCavern are doing quite OK already in comparison.

[sektor2111]

I have to look at it first because it says "huge map". I know such maps where DevPath is not doing anything because space calculations do a lock-down in internal natives, but it might work with XC24 and above (XC 25b + UT 469b), but... in 469b I did not check FerBotz yet.

CTF-AYeti is a sample and MH-GardenOfDeath where first time I did some triggers for Bots or else none of them was roaming in map.

[TankBeef]

This one has similar issues to the Cynosure][ map, so I figured, maybe Sektor would like to give this one a shot. https://unrealarchive.org/maps/unreal-t ... 48bd3.html
Apparently these sort of maps are made mainly for leagues, and so bot pathing may not be priority.

[UnrealGGecko]

There's actually a few versions of this ine, latest one seems to be this one
https://unrealarchive.org/maps/unreal-t ... 2ad24.html
Still no botpaths though

[sektor2111]

I'm going to remove that City because there I forgot about some Triggers and their events when I pathed Lifts. They need to match lift or else mover won't move ever. I was sure about forgetting something but I resumed work next day and I failed to recall what I needed to look at. I'll solve these but still lifts with more triggers won't help Bots.

[TankBeef]

There's actually a few versions of this ine, latest one seems to be this one
https://unrealarchive.org/maps/unreal-t ... 2ad24.html
Still no botpaths though

Thanks Gecko. If this is the latest version, then I guess it would be better to work on this one.

[sektor2111]

In account...

Previous DM map (that "huge" map) will be here too... I think I finished solving routes in RealCity map then Lost-Valley will be next as long as I see it a good looking one, then CTF-Alcanize whatever.
I only need a bit of time for testing in game moves because paths aren't the challenge, the challenge it's making them to be useful... and not the last thing is to figure if DevPath won't go crazy in processing...

Resuming:
RealCity was having a few isolated spots without a problem in theory. In practice some Bots were jumping Over lifts and not Inside Lifts and suddenly jumping of the lift in such a spot. My decision is a bit ugly - one way paths to LiftCenter doesn't matter where is the lift in that moment. Bot trying to get there might die falling. This way Bot will respawn and it keeps game flow instead of sitting in building all day long out of war-zones. This is a quick deal which I find suitable when Lifts are over Engine capabilities toward A.I. "brain" - I might study extra options but it takes too much time. All testing is executed with original Bot and original UT - no add-ons.

[a9902957@nepwk.com]

Thank you again! This is exciting.
Now I can also test with 469b XC_Engine 25b (+FerBotz 18).

[sektor2111]

Something new as an information. It looks like whatever "PointReachable" is another sort of lousy coding result. First time I thought that map has a bugged stair which is not having a normal collision, and I went to check map in a game session using weapons.
This is a demonstration why sometimes working in certain range for making paths has a lot of advantages.
Look what Paths is doing Editor through stair:

Cough, Epic boys, cough. If pawn wants that weapon it will climb stair as long as it's impossible to go through stair. Here FlakAmmo will get connected With SniperRifle for reasons which it's hard to understand.
After moving it away and using 512 and wrapping points in certain order I got this...

Why a restrictive range ? Because at testing with PointReachable even my builder using original checks will do a stupid connection through stair when actually this is NEVER a valid path. At this point manipulating reachspecs for normal results it's highly recommended. Instead of doing moves with items and changing design and testing these X times, it's more easy to dictate how paths needs to be done. It's hard to rip out this stage in a testing environment as long as map has no brushes and I have to do other stunts for capturing assets... Probably I have to reinforce other checks in my builder because what Engine says it's utterly Sci-Fi for kids.

[a9902957@nepwk.com]

Heh, so that PointReachable appears to do not much more than check LOS and distance but obviously not check if a regular UT playerpawn can walk there and not get obstructed or end up at the right X/Y but wrong Z e.g. (if only checking A to B but not B to A as well). Thanks for pointing out this pitfall. Good thing you spotted it and could work around this case with the range tweaking!

[sektor2111]

I was thrilled about these calculations in first minute. You have a stair perfectly climbable not allowing any failure like a solid ground and later Devs are saying that you can move through the stair as well ... In theory an invisible collision hull could solve problem but... it is not advisable to do this in such maps without brushes making things more difficult than expected. Map is pretty well executed - those TeleZones can actually teleport spectators so you can track A.I. events much better.

Resuming - On the way...

RealCity has now a bit of tinkering...

DM-RealCity_JM1.7z

(1.33 MiB) Downloaded 17 times

Bot dying in an attempt to catch a lift will respawn and I think it's better than doing nothing.

Next stage - indeed a big map - can be called huge but... idk, DevPath responds well - even has fake responses as I pointed but...

DM-DoCdOoM_LosT-VaLLeY_rS1525.7z

(2.29 MiB) Downloaded 18 times

This is a version with paths - Bots are not supposed to reveal shooting secrets as additional goal. Stairs and their paths are operated with a bit of love to the code from tester scout using some traces. For other maps having similar stairs, grates, etc. things can be done so far using two toys: MapGarbage and updated XC_PathsWorker which will get out soon.
Techs unexplained yet anywhere but are doable:
- MapGarbage can deploy InventorySpots linked with their Inventories;
- MapGarbage can create navigation chain properly after solving zones for nodes and adding these markers for items because VOID nodes are not taken in account;
- After using XC_PathsWorker with newer option bKeepOldPaths True, navigation chain and current paths (or no paths) are no longer destroyed and processing additional needs;
- Because chain is available and all Items are marked properly, we can adjust positions for losing the trace with bad nodes around or even setting them bHiddenEd or bOneWayPath for excepting them from being linked - links are doable later manually using normality not DevPath's animality because bot is not a jelly thing sliding through stairs/grates.
- Tracing can be checked with another option which I forgot to add in Builder in first stage because at a moment I needed it - now it will be available.
- If builder has no deal with Invisibility from Lost-Valley map paths are doable hand-made and it's what I did.

Good... Let me look at that CTF map, and let me know if something goes into an endless loop in above maps. Lifts are not my doing but now they are pathed even for other Levels where Bot might evade for resuming roaming.