Map assessment and pathing enhancements

[evilgrins]

What was the title saying ? Unlocked !?

I may have nominate you for sainthood.

[Nelsona]

Done ?

CTF-ClarionUltimateV5_rS488.7z

(2.04 MiB) Downloaded 830 times

Edit Notes:
- Aligned better a few textures;
- Completely redone Paths Network - it has a split network loaded with no purpose;
- Added weapons, a few MedBoxes - ammo aren't connected into paths, just weapons;
- Adjusted VacuumZone settings - this crap was doing Accessed Nones - now it's more domestic;
- Added defenses, alternate paths, extra-paths - excuse me, I could not resist... helpful for CTF games smarter than UT's stock one;
- replaced useless Boulder1 projectile with Boulder1 decoration.
Using:
- UT Editor with Engine 440 - base tuning and paths creating - final checks;
- UGold227h Editor - pathing tasks;
- UT Editor with Engine 451b - last build post texturing modifications and re-zoning.

[evilgrins]

Give it a few days of testing, will let you know how it goes.
- Put a redeemer in each well.
- 2 Warlords to a team.

Did you notice there's a secret room under each base? I saw it when I was placing the redeemers, seems like it's just there for ambient sound.

[Nelsona]

I did not connected those rooms because I did not see Bots falling there and then I considered to leave these alone... of course, any time can be added routes, those roofs are... non-colliding portals or such, one of them had annoying miss-aligned textures.

[evilgrins]

As you know, I love the HallOfGiants maps in all their variety, which why it bothers me so that this map I just found is not well pathed at all.

Have a look · https://unrealarchive.org/maps/unreal-t ... 11f96.html

[Nelsona]

which why it bothers me so that this map I just found is not well pathed at all.

Ahem... not entirely true. I would like to see people learning and making aths as this map is having - talking serious here.
My views:
Map has combos oriented as uni-directional exactly as needed for plain devs which don't allow us to remove junk reachspecs not needed;
Map has PathNodes placed separately from InventorySpots which is what Polge said and wanted to see.
However - probably because of map's size we do have:

Code: Select all

NumReachSpecs: This map has 1785 ReachSpecs shown as direct paths in Navigation Network.
NumReachSpecs: Also this map has 1731 ReachSpecs as shortcuts aka PrunedPaths in Navigation Network.
NumReachSpecs: These are not visible in Editor but they are used by Pawns.
NumReachSpecs: After calculation we have 3516 ReachSpecs.
NumReachSpecs: Technically this map is overloaded and this is not that healthy for A.I.

What exactly happens in my opinion ? It's a flawed movement timing of Bot in LowGravity vs no response of DevPath when pawn doesn't have required physics for navigating certain part of path. LandsOfNapali by example hits boundaries when falling pawn it's looking for a path, I believe devs are scanning entire network for a potential way to go. When map has many reachspecs, cycles are filling all hard-coded capabilities causing DevPath to not return anything.
Nodes placement here are demonstrating KNOWLEDGE not guessing, if plain Devs are stupid and Bot cannot understand low gravity that's not bad Pathing.
What I don't like here is the exit out of tubes. Those "Exits" have incoming connection which Might cause pawn to move against tube's gravity - that's stupid. Like in other cases from the same map, these should be ONE-WAY - pushed higher for preventing creating incoming connections.

What I would try to do here ?
Causing a removal of reachSpecs (all junks must be removed and not added others over them). Then rebuilding paths using XC_EditorAdds (not UT469). Examining well combos and/or getting rid of those triads using 4 internal reachSpecs and over 4-5 other links to outside - ONE it's needed and ENOUGH. If XC will help dev to gain a minimal charge (under 3000 reachSpecs - direct reachSpecs and not all sort of prunes) perhaps we can figure another behavior of Bot - which it's not the last thing with problems - including your Bots, Sir, because we are speaking about the same root called BotPack.Bot.

[evilgrins]

So essentially, difficulties with tubes aside, the bots maneuver their ream side of the map but they don't have any means to get to the other team's side.

[Nelsona]

Problem with bases is solved... but the other problem was sitting around.
At kickers, when they were switching direction (sudden carrier) funky stuff was happening, and then I had to add other kickers for deviating... So far they got unlocked. I added several blockers - not really AlternatePaths, it looks enough harder for them this gravity but... at least they play CTF...
Paths Network is aggressively reduced at 1406 reachSpecs, better than what it was. I think it can be even more simple but... we are inside Engine boundaries so I won't change these. I'm looking at how goes the game, perhaps it's ready...

[Nelsona]

Allow me to explain a bit what was the major problem or... less problem but definitely against movement logic.
Initially map had these routes impossible to follow against gravity rules - pure physics so to speak:

Original_Problem.PNG (1.32 MiB) Viewed 144842 times

Other combos from this kind are just mitigated correctly given resources from those times. But now days we do have another technology allowing deleting ALL useless and less logic reachSpec like here:

Problem_Solved.PNG (1.3 MiB) Viewed 144842 times

All I have to do it's to think if map needs more routes and/or AlternatePaths - main route is on higher bridge, perhaps I don't have any alternate option to follow - too far for translocator even if low gravity expands this option but... we do have hard-coded calculations here.

[evilgrins]

Yeesh!

When you get a chance, check this one. It's massive... running 8 bots, 2 teams of 4, showed its pathing probably could use some serious work.