Zombies at boltons

[frankit]

Sweet Weapon! You're a sick man Gaz, but I guess you know that already ... I much prefer this vid over the others as its brighter and easier to get a feel for the game. Couple of suggestions since you said you were open to feedback:

1. The animation 'appears' to slow down as you chop, but I think the gore should speed up ...
2. Ambient zombie-attacking-sounds could be louder in the mix, even with the active chainsaw (i.e. something disturbingly horrifying sounding).
3. The facial expressions on the attacking zombies could be more ****ed up... Or maybe rampant zombie limbs coming at you faster would be better....

But I remember now that these are Romero zombies and so maybe you have different ideas.... BTW: What's that round green energy field @ 0:18?

[gaz661]

Moo Ha Ha,new weapon.enough said.

[Snipe34]

ah yeah your monsters are 'different.' You want them to go straight for the vitals.
Only 24 pathnode x unnecessary!!, omg you're doing something right!! I live learn, thanks for the explanation about what you did.

Pathnodes also ignored on mine when I use a navmesh. I dunno how to use both either, like you I hope someone does.
I'm also interested in getting the TriggeredPath working via kismet. It's for runtime pathing changes and useful to know.

[gaz661]

after a 15 hour build last night pathing was complete,but as I expected half the volume pathnodes didn't link.as I said its a lot of trial and error and with this build time I think its a no go.generally I place all pathnodes and select them all,copy and paste then replace the new ones with vpathnodes.as they are directly over the pathnodes they always work well.but in this instance the vpathnodes were not directly above and so some did not connect.i don't think copy/pasting the full 7000 pathnodes is a good idea.

less pathnodes is better?this is what I believed when I first started and stuck to epics guidelines of no closer than 2000 units.but through much,much experimenting I find no performance loss with more.i guess because hardware is so much faster than in 2004.

surprisingly,out of the 7000 pathnodes currently in there I only get about 24 'pathnode X is unnecessary!!' warnings.

im still continuing to fill one version with navmesh as it did show a lot of promise as well as lowering the map file size by nearly 30 mb.currently the pathnode version is at 270 mb and I know things can get weird after 300 mb.flying enemies work really well with navmesh.and zombies are crossing over between pylons,so I don't know what I did differently before.

im sure I read a thread ages ago where someone was using both navmesh and pathnodes but my udk ignores pathnodes if I place any pylons.this would be an awesome solution if its possible.anyone know about that?

ive used tactical move in other ai's but I don't want my zombies to dodge or think,zombie is hungry=zombie must eat.

[Snipe34]

Yeah, I'm no expert with navmeshes. They caused me similar problems so I keep to pathnodes. Usually I'd say 'send your map over' to look at paths, but it's not that simple with UDK, lol.
UDK must do a lot of whinging at you Gaz, 'pathnode X is unnecessary!!' etc etc etc It's the way UDK is programmed and I like it. I think with your 14 hour build and crash UDK is agreeing with me. I used to vet UT2004 maps before they went on a server(and UDK is basically a UT2004 massive update). 99% of dudes would saturate their maps with nodes. Often I'd delete the lot and re-path. It was quicker. Rather than 500 nodes on a road for example, I'd have 2, up/down and another dozen or so into buildings. Too many nodes slows AI thinking, slows the game, the server... AI goes from A to B. Having a choice of 2 rather than 100 is obviously quicker for the AI to think thru.

Beyond pathing there's code options, and I'm probably / maybe / perhaps as good a coder as you ^^ There's a lot of hopeful code in UTBot like 'state TacticalMove' that could be useful as my understanding of code improves. I'm saying besides pathing, I want my monsters to dodge, duck and generally avoid bullets. Least that's my way forward - pathing but more, bots with coded smart moves.

Crowd Agents, I intended them for survivor types too, but Frankit's knowledge on the subject is excellent, thanks man!

[frankit]

Frankit's ideas of crowd agents is interesting but they have no collision?

Crowd Agents have zero collision but the upside is a saving on clock cycles. They do have basic obstacle avoidance, but need clean lines of bounds to work well. If Gaz flies his Harrier into them, it won't have any effect. But if crowds walk into the player or the player walks into them while carrying a Dynamic Crowd Destination then it will trigger a handy event that can be used to maim the player or play gory animations etc. This is the basis behind the Jazz mini game (mobile). For PC / UTGame type games, its necessary to add a bool to the CA anim tree to get agent animations to fire ...

[gaz661]

last night I pasted 800 volumepathnodes from an earlier version of the map(because I knew they would work) and set it running.at 14 hours it crashed.so this morning I started messing with the navmesh.obvious advantage,in an hour or so I had done a quarter of the map.another advantage is the zombies head straight towards you with no detour.but its not getting into smaller places like the residential houses.

another problem is that the groundspeed of the zombies has been altered somehow(much slower).im using an older playercontroller for the navmesh but it should not mess with the groundspeed.the controller has 2 states,wandering and then attacking if the player is within a certain distance,but I think this kind of zombie isn't much use unless there are an awfull lot of them.the way my pathnode system works I only need 10 zombies in the whole map and you cant stand still for long before they are on you.

I will look into crowd agents,i have used them before for fleeing survivors.

[Snipe34]

Snipe34,my ai is based on the ut3 stuff and when a bot sees you he will travel to the nearest pathnode to the player before breaking off if needs be,so if I have few pathnodes a zombie might see you and head for your position and you move away making his behaviour very odd.with lots of nodes this behaviour isn't as noticeable.

Yeah the wander off and switch back. I always liked the way they do that, but it's different to what you want. I want flanking behaviour, so yeah, different

and btw like the new shine on your roads, indeed your whole Gothic atmosphere.

Frankit's ideas of crowd agents is interesting but they have no collision?

[frankit]

Did you ever give Crowd Agents a shot? You could add to what you have by populating some areas with slow moving CA Zombies... Attaching a CA Destination actor would 'attract' them to the player....

[gaz661]

yeah,the title and loading screen images are a little old and need updateing,my map has had a lot of different looks through the years.the map is way to complex for dynamic day/night cycles,ive tried and that was when the map was smaller.as it is half my kactors don't cast shadows for perfpmance reasons but its not noticeable anyway.

im a big fan of the original ramero films and would really have liked my zombies to be as slow as his but that would not work without hundreds of them and the engine just isn't up to that so im left with zombies that are just slightly slower than the player.not having the zombies show on the map does make a big difference and really does produce genuine surprises.but,saying that I could make it so when a zombie is at a certain distance it could randomly choose to rush or continue to stalk.that might liven things up a little.

pylons-again,i have experimented and even on a much less populated map it took as long as the pathnodes do now with poorer results.in my version of udk,bots will not cross over between 2 pylon areas(marked with the white surface)and ive found no way to tweek it.the latest vid only really show open areas but the map is full of nooks and crannies too.

thanks for the input and keep it coming,it does help.

Snipe34,my ai is based on the ut3 stuff and when a bot sees you he will travel to the nearest pathnode to the player before breaking off if needs be,so if I have few pathnodes a zombie might see you and head for your position and you move away making his behaviour very odd.with lots of nodes this behaviour isn't as noticeable.

[Snipe34]

Navmeshes never worked 4me on my big maps. I always went minimal on pathnodes and had nps.

If zs see you, they wander off nodes to get you. So on open areas, go totally minimal. But around trees, buildings, because nodes are line-of-sight probably 4 around a trunk so zs can circumnavigate. Buildings, 1 entrance 1 maybe 2 inside, another 1 if there's an exit. Test it thoroughly to find out where they have problems.

-Start up map size, removing the mid block you'd need only 2 nodes, because like I say, they see you, they go off nodes to get you. Test that and you'll see what I mean.

Looking at your maps, s/be 1000 max nodes, but personally I'd aim for 500 max.

[frankit]

Looks cool Gaz! Couple of suggestions if you're open to feedback :-

Visuals:
1. The stills for the clips hint at a Night / Day cycle, but the clips are always dark....
2. Head on zombies sometimes come across as a bit docile. At a distance I like how they can surprise you by jumping down in front of the player's path. But close up they don't hit home as being scary, not like those in other Zombie games such as the kismet-only one. I think it feels that way partially because you shoot off their lower arm limbs in the animations... So what if at close quarters they rushed the player instead?

Pathing:
Subconscious fear of Pylons Gaz? (10000 Pathnodes!)
Instead of adding more, flying enemies could be placed into flying mode, navigate using velocities, and have simple obstacle avoidance, because the level is large and open.

[gaz661]

But 14hours pathing, ouch.

yes,especially as I can build lighting in 30 mins.currently there are nearly 7000 pathnodes and I estimate I will need at least another 2000,and I really want to get volume pathnodes in there as well for flying enemies,but I know it takes a lot of trial and error to get them working right and if im looking at 16-18 hours a time it may not be possible.

and thanks for the comments.

[Snipe34]

udk never ceases to amaze me.a couple of weeks ago I decided to see how much of the surrounding landscape I could fill before hitting performance problems.my map now contains the entire village of froghall.not that there is much else there.
there is a pub,the railway station and a scrap yard as well as a dozen residential properties.the only problem ive encountered is that building paths now takes over 14 hours.

also I have made a start on the story mode-very basic at the moment but im just laying out the framework and covering all in game eventualities.

also the hud is changed again and the map is back but in the final game the zombies wont show up.they do just there for testing at the moment.

Goes without saying, nice to see an actual 1p game with menus and map ids actually working (but I don't want to see the bugs under the hood (I'm not being a smartA, everyone knows bugs are just part of gaming), I really mean, nice work!. Know what you mean about filling up a map and UDK coping all right, with the help of Massive LOD and render distance. But 14hours pathing, ouch.

[gaz661]

udk never ceases to amaze me.a couple of weeks ago I decided to see how much of the surrounding landscape I could fill before hitting performance problems.my map now contains the entire village of froghall.not that there is much else there.
there is a pub,the railway station and a scrap yard as well as a dozen residential properties.the only problem ive encountered is that building paths now takes over 14 hours.

also I have made a start on the story mode-very basic at the moment but im just laying out the framework and covering all in game eventualities.

also the hud is changed again and the map is back but in the final game the zombies wont show up.they do just there for testing at the moment.