WIP Beta released Valo City 0.79

Ngl, I'm on an old ass GateWay from 2011. Specifically the FX-6860 and yea that's basically my specs lol.

 

I know this thread is dead, but is this getting released soon? Download link please? Thank you.

 

Says "On Hold" So it's not dead, Because of your kind of rude lack of patience, I hope the creator takes a bit longer to make sure it's perfect..

 

I was just wondering, since there hasn't been any updates in over a month. I hope it turns out great too!

 

Looks like he was pretty new to the threads, should have been a bit more kind?

 

Great map. Releasing an "Early Access" version of this would be kickass!

 

Please don't bump.
IT'S A YEAR OLD FOR GOD'S SAKE

 

Hello,

oh well... where do I begin...

first of all...

THE PROJECT ISN'T DEAD.

So what happened to it?

First, let's address the past, then what's the current state, and then what the future hopefully holds for us.

Back when i started working on this project I was so focused on it I spent all my free time on it (6-9 hours a day after school and before sleep). This obviously put me in a bad situation regarding to my studies, resulting in me having to take a break, but now that it's all over I can address my free time better on various things, including this project too. At the beginning of the project I had very minimal knowledge of the technical side of things, both on with Blender and BeamNG. After 2 years of actively using Blender for various other things, I'm very confident I can do a better job on the 2nd go.
2nd go? What does that mean?
Well... when revisiting my project files it became clearly appearent for me that the project was deeply flawed from the beginning on my end. So now let's talk about the current state.
Cleanups... and I mean massive cleaning up and optimization. While performance was very good on Valo City with it's original assets it made me a bit lazy in the past actually optimizing it even further with the possibilites BeamNG could provide and actually provided since with the implementation of the walk mode. So currently I'm optimizing objects and meshes to maximize performance and clear up issues they could give with the free walk mode. This mainly affects interiors where player is expected to walk around mostly.

The other important things after collisions. Until now i've been using the old world editor, and the new editor takes a lot of time getting used to as I'm only getting familiar with it now.
That means things that i've been easily be able to do in the world editor, such as adjusting materials became a kind hit or miss thing to do as I come across more and more issues, such as materials not even showing up inside the editor, not even after manually adressing them in the materials.json files, so for now I can only adjust already imported materials but not new ones.
What new ones if I already imported everything? Well... that brings us to what the future is hopefully bringing us.
Custom LODs are going to be made for most, if not every object. As an example, for a wall object rendering a flattened picture of the surface then making the LOD model of it just a flat plane with a texture applied to it saving even more resources and improving performance.

But until I can't figure out how to add the materials making up this rendered texture I can't implement them. As for the scale of things there are hundreds of objects, so manually doing this on all (most of them aren't flat like this wall, so I need to render from multiple angles then slice together a texture and mesh that represents the base object

Another issue... AI.... this is what mainly set us back trying to figure it out.
In this thread we've tried to get some help about it however it went unnoticed
https://www.beamng.com/threads/specifying-lane-count-for-both-way-waypoint-based-ai-roads.85256/
While AI works, it doesn't work as intended with out current system/setup, we hope to find a solution to this but we haven't been able to so far.

The old version of this map served as a learning field, however using much more knowledge i picked up in the past 2 years I will spend time to clean up the models and optimize the map as much as possible. I feel like a major performance increase can be achieved, on top of how well the map runs at the moment.

(It actually used to run even better before PBR materials for vehicles but it's very scene dependant. Looking over the city the performance tanks as there aren't much LOD models)

Once the base map is finished I can start making scenarios and other gameplay elements.

I'm sorry for the silence for all this time. If you took your time to read all this mumbo jumbo, thank you, and thank you for everyone who kept waiting patiently on this project. If you have any suggestions or questions please don't hesitate to ask, I'm gladly answering anything.

Best regards,
Medek

 

Well hello there.

I feel like I should also showcase the current state of the map not just show a single screenshot of one of the garages. I added a bit of a gameplay to make the showcase a bit more interesting...
The gamemode is just the "Garage to Garage" as seen on other official maps, but I added it to here. It has an issue where the parking area corner mark thingys appear at 0,0,0 coordinate for some reason instead of around the parking itself, haven't figured it out yet. Like I said in the previous post, gameplay isn't my main focus right now, I mostly cleaned up the backend in the past few days which made the map much easier on the CPU however it seems that the map is now heavily GPU limited in comparison.
The cleanups were however enough for a ~25% performance increase in some scenarios so that's already a good improvement.


Now... onto the "showcase"



Again, it's not to showcase any gameplay really, it's just to show the current visuals of the map. Performance isn't AS good as mentioned above obviously because of the AI traffic and recording, but it may serve as a good reference in the future.

Thanks,
Medek

 

I'm glad to see a diverse urban map on the forum again. I feel like we've had a lack of those lately.

 

As SLRR is one of my favorite games, I'm excited to see this map back!

 

I'm glad this project isn't dead!

 

Nice to see Valo city on this game ,it reminds me of the mods i made in slrr

 

I'm putting the project on hold again as we wait for more node based AI options to be able to have good AI on the map, this would allow me to do the missions and scenarios as they currently don't really behave well with the current system

 

Create a terrain. Shape it the same as your city, and make sure it's just a few CM below the roadway and buildings so it otherwise doesn't get seen by the player. You can optionally color it in like colors to the models above it and nulldetail the models at distance to save on draw calls (and hence lower cpu bottlenecking and raise fps).
You can then draw the AI routes per-lane on said terrain. Provided you're not more than a ¼ meter of difference from the terrain or more it should be fine and the AI will drive fine. Otherwise, they'll get spawned in stuck in the road or stop while driving. You'll have to adjust the terrian accordingly so that the AI does not get stuck (a few CM should be fine, just enough that it doesn't flicker into view at medium view, you can also use terrain holes to hide it in some places you don't need it).
Also, you can use waypoints, but you'll have to script them in. That's very time-consuming but that's the way a lot of old and even modern games do it, if they don't just follow the textures (GTA SA's train just followed the track texture for example).
There is a 3D feature to decal roads when using OVER OBJECT (use invisible graphic on them for AI routes or just arrows while setting it up so you know which way they go and replace it with invisible later) features. This can allow you to place and position the lanes where you want them and the AI won't get stuck then. This is how I make AI on my Los Injurus map (I pioneered invisible roads used for AI routing in my maps, first seen on Nevada Interstate back in 2016~2017 or so, it's the quick-and-dirty-but-works method to AI).
Use more nodes for smoother corners, and when road grade changes from shallow to steep and back when mounting hills etc. Otherwise, the AI driving from point to point will hit walls and curbs on the inside of the corner, and on hills they might stop. Remember, the road nodes are what they drive point to point with. If you have bridges, tunnels, or other things of the like where the road doesn't always follow the intended path vertically, clear the box for 'use subdivisions' otherwise the AI will stop.
I hope some of this helps. It's pretty quick and easy to do AI when you have a hidden terrain under the map. You can as a bonus use the terrain for mountains around the map also to help with the view issues to prevent staring into the void too long.

EDIT:
You will find you are bottlenecked by draw calls, not triangles. 3 million triangles is easy. 6000 draw calls can slow things. Draw calls are incurred by too many different textured surfaces in view that aren't sewn/connected to other like-surfaces of the same textures.
That having been said, lowering triangles on models will not only take up less space in RAM and VRAM, but is obviously easier to draw. We used to be triangle-limited in the 90's and 00's, but now we are draw call limited, which is work done by the CPU to issue work orders in regards to drawing polygonal object's textures, for each frame, thousands of times. This is the weak point in this game engine (and in many commercial engines). Vulkan helps alleviate this, but can be crashy on some systems.
Learn about draw calls if you don't already know, and work to duplicate textures for windows, doors, roof sections, etc, when and where possible, and sew like-textured surfaces on UV for poly objects. This will help you the most. You can leave more detail on close-up objects but on far away ones, try to limit the different times it has to issue a draw call by duplicating things or making use of the nulldetail feature.

 

We used a script to automatically set up waypoints over the entire map using spline data from SLRR;
The entire problem is that Beam AI doesn't recognize lane count parameters for waypoint-based roads, so they drive as if there's only one lane per direction.
Having each lane as it's own 'road' surely isn't the proper way to set this up...

 

when can i play

 

when its released

 

Is this still being worked on or died?

 

Being worked on? No.
Dead? Neither.

On hold until I get the motivation to get it going again and it was mainly stopped being worked on because we cannot find a sufficient solution for the AI. We use a waypoint based AI system for traffic, however they don't understand how many lanes the road have. So If we have a 12 meter wide road with 6 lanes (determined by lanesLeft=3 lanesRight=3) the AI just don't understand these parameters meanwhile if it was decalroad it would understand it. This is particularly annoying with 2 lane roads which is the majority of the map, because on these roads the AI would just drive in the middle of the road not in a lane. But one day hopefully I can wrap up the map and publish it. Would need some major reworking regardless of the AI state.