Page 2 of 3

Re: Erode status/update

Posted: Tue Feb 05, 2019 10:43 pm
by checkerz
jlv wrote:Are they around two feet apart? The game merges erode entries that are within a foot to save bandwidth. It usually averages out but at a high erode setting and small radius if they lined up the same way on consecutive runs they could reinforce instead.
It does it with a single rut.

Re: Erode status/update

Posted: Wed Feb 06, 2019 2:24 am
by jlv
One rut would be in phase with itself by definition. But you didn't answer the question. Are they around 2 feet apart? Actually might be 1 foot. I'll have to double check the averaging code to be sure.

Re: Erode status/update

Posted: Wed Feb 06, 2019 1:43 pm
by checkerz
jlv wrote:One rut would be in phase with itself by definition. But you didn't answer the question. Are they around 2 feet apart? Actually might be 1 foot. I'll have to double check the averaging code to be sure.
Didn't realize if it did it with a single rut that the distance question applied. My bad.

I didn't have the track saved with the first erode, so I did some new passes. This is on the actual San Diego track, not with mud traction or any changed erode settings:

Hard to get a good picture because the erode is tough to see - again set at 4.0 and this is one of the straights that isn't getting much erode:

Image

Here is the rhythm that erodes well each time:
Image

As a test, I rotated the terrain 90 degrees and when and rode. What I found was pretty substantial IMO.

Image

The sections that were eroding so poorly before, are now great:
Image

Image

Image

The sections that were so great before, now are getting the bumpy, barely eroding stuff:

Image

Re: Erode status/update

Posted: Thu Feb 07, 2019 3:33 am
by jlv
Can you reduce this to a tiny test case? Which direction do you think is not working right?

Re: Erode status/update

Posted: Thu Feb 07, 2019 6:39 am
by checkerz
jlv wrote:Can you reduce this to a tiny test case? Which direction do you think is not working right?
I don't have a clue honestly.

I could create create a terrain with a section going every 15 degrees or something to test on? Would that be suitable?

I'm a bit tied up at the moment building east coast terrains and traveling to races for my real job, so not sure how much time I can devote to testing this, but can provide some type of terrain you need if you give me an idea what you need.

Re: Erode status/update

Posted: Fri Feb 08, 2019 2:31 am
by jlv
Does it happen on flat ground? If you can get it down to "park in this direction and dump the clutch and it erode less than this other direction" it'll be easier to figure out.

Re: Erode status/update

Posted: Wed Feb 20, 2019 4:20 am
by Jeremy150
jlv wrote:Does it happen on flat ground? If you can get it down to "park in this direction and dump the clutch and it erode less than this other direction" it'll be easier to figure out.
Think I found the issue
Image
I took off from all 8 directions, with these bales.
Image

Re: Erode status/update

Posted: Wed Feb 20, 2019 4:29 am
by Jeremy150
DP
Changed the angle of the sun and got this.
Image

Re: Erode status/update

Posted: Wed Feb 20, 2019 8:34 am
by checkerz
What would it take to simulate dirt density?

We can simulate the initial dig that way, but there isn't really a way to make the top "fluff" that you dig through fast to stay fluffy and not packed and affect the physics less than the deeper and more dug in stuff.

Re: Erode status/update

Posted: Thu Feb 21, 2019 3:22 am
by jlv
Jeremy150 wrote:
jlv wrote:Does it happen on flat ground? If you can get it down to "park in this direction and dump the clutch and it erode less than this other direction" it'll be easier to figure out.
Think I found the issue
https://gyazo.com/800aaa637f56b5728567b4616f4c8e28.png
I took off from all 8 directions, with these bales.
https://gyazo.com/232e25e40381c82d231a8e0ae243ec05.png
That's just visibility though. I believe Chris is talking about actual differences in depth.
checkerz wrote:What would it take to simulate dirt density?

We can simulate the initial dig that way, but there isn't really a way to make the top "fluff" that you dig through fast to stay fluffy and not packed and affect the physics less than the deeper and more dug in stuff.
I think the best way to do surface fluff would be for me to add a purely cosmetic rut layer to the terrain engine. If I just make ruts instantly appear under the wheels and not affect the physics I think most people would be happier. Since it'd just be cosmetic, there would be no need for network consistency or anything like that so it'd be way easier to do compared to what erode does.

On this subject, I was thinking about a possible traction update and wanted your take on it. Currently, the main problem is when you want to make a loose but forgiving friction curve, you usually want a lot of high slip speed friction but have to compromise between too much for acceleration or too little for turning. People have suggested a separate lateral friction, but the rationale for that seems limited. What I was thinking is the acceleration slip isn't sweeping fresh dirt, so maybe the slip vector should be limited to the contact patch velocity vector. So for the back wheel accelerating, there should be no ramp up since the slip vector and contact patch vector are in opposite directions. But for braking and turning they'd at least partially line up, meaning it has to sweep fresh ground to slip more. Does this sound good? Feels like it'd help for forgiving traction and also have a decent rationale behind it.

Re: Erode status/update

Posted: Thu Feb 21, 2019 3:46 am
by Jeremy150
jlv wrote: That's just visibility though. I believe Chris is talking about actual differences in depth.
Yes it looks like it's just visibility. But I believe it's what is also causing ruts too appear to be chop? Maybe this is a prime canidate for multi point lighting in some sections? I have some ideas on it too, so it does't blow out the rest of the track. I'll save it for now unless you'd like to hear it and are interested in possibly moving that direction.

Re: Erode status/update

Posted: Thu Feb 21, 2019 10:37 am
by checkerz
jlv wrote:
Jeremy150 wrote:
jlv wrote:Does it happen on flat ground? If you can get it down to "park in this direction and dump the clutch and it erode less than this other direction" it'll be easier to figure out.
Think I found the issue
https://gyazo.com/800aaa637f56b5728567b4616f4c8e28.png
I took off from all 8 directions, with these bales.
https://gyazo.com/232e25e40381c82d231a8e0ae243ec05.png
That's just visibility though. I believe Chris is talking about actual differences in depth.
checkerz wrote:What would it take to simulate dirt density?

We can simulate the initial dig that way, but there isn't really a way to make the top "fluff" that you dig through fast to stay fluffy and not packed and affect the physics less than the deeper and more dug in stuff.
I think the best way to do surface fluff would be for me to add a purely cosmetic rut layer to the terrain engine. If I just make ruts instantly appear under the wheels and not affect the physics I think most people would be happier. Since it'd just be cosmetic, there would be no need for network consistency or anything like that so it'd be way easier to do compared to what erode does.

On this subject, I was thinking about a possible traction update and wanted your take on it. Currently, the main problem is when you want to make a loose but forgiving friction curve, you usually want a lot of high slip speed friction but have to compromise between too much for acceleration or too little for turning. People have suggested a separate lateral friction, but the rationale for that seems limited. What I was thinking is the acceleration slip isn't sweeping fresh dirt, so maybe the slip vector should be limited to the contact patch velocity vector. So for the back wheel accelerating, there should be no ramp up since the slip vector and contact patch vector are in opposite directions. But for braking and turning they'd at least partially line up, meaning it has to sweep fresh ground to slip more. Does this sound good? Feels like it'd help for forgiving traction and also have a decent rationale behind it.
I just re-ran the same test (Jeremy did the same with a flat terrain). The erode is not broken like I suspected, it's simply lighting (yea...simply haha).

Here is the same terrain (just saved and then changed lighting direction and re-took screens)

Image
Image

Image
Image

Erode doesn't look bumpy on the actual terrain, it's just the lighting angle not capturing an angle.

So that's good news.

The fluff layer sounds very promising, as I think we can cut down the erode numbers even more so they don't make jagged stuff and mess with the physics as much but get the visual effect we're chasing.

I'll have to read more on the traction stuff when I'm not half asleep to answer that one. I'll get back to you.

Re: Erode status/update

Posted: Thu Feb 21, 2019 6:38 pm
by Racers52
jlv wrote:I think the best way to do surface fluff would be for me to add a purely cosmetic rut layer to the terrain engine. If I just make ruts instantly appear under the wheels and not affect the physics I think most people would be happier. Since it'd just be cosmetic, there would be no need for network consistency or anything like that so it'd be way easier to do compared to what erode does.
Can this get some more explanation?

I know we previously touched on some different mappings, but I think if you could work out the issues with displacement mapping that could also add to the illusion of surface fluff.

Also would allow riders to see that someone has ridden there before and follow that line which would then allow erode to come in and start digging those ruts out.

I'm imagining something similar to what Milestone has done with their games

Image

Image

It looks like they just use flat 2d decals with displacement mapping.

I have worked on the rF nationals and supercross for many years, and with the inclusion of erode it seems you have to take a step backward in terms of visuals with erode because you can not see what is forming in front of you.

I feel like the erode system is fine but we arent able to unlock its true potiental unless there is a visual update/lighting update.

Example from this week's supercross track, I have a set of extremely nice 3D scanned textures that look incredible but when paired with erode become a hazard.

I run these at thirty percent opacity on top of a base decal with little no contrast
Image

Here is an example in game, great decals camouflage erode.
Image

Same decals with a solid color overlay applied, taking all contrast out - moto minecraft but exposes erode so much better
Image

more examples
Image

motominecraft - solid decal color, no detail, no contrast
Image

Is there a solution to have erode visible and still have detail in the decals?

Re: Erode status/update

Posted: Fri Feb 22, 2019 3:13 am
by jlv
Racers52 wrote:I have worked on the rF nationals and supercross for many years, and with the inclusion of erode it seems you have to take a step backward in terms of visuals with erode because you can not see what is forming in front of you.

I feel like the erode system is fine but we arent able to unlock its true potiental unless there is a visual update/lighting update.

Example from this week's supercross track, I have a set of extremely nice 3D scanned textures that look incredible but when paired with erode become a hazard.
Instead of removing contrast, try duplicating the layer, then blur the resulting layer with a radius that would equate to around 6 inches in the game, desaturate, and set it to subtract mode. You'll want to put a solid gray addition layer between the two layers and adjust the brightness of that layer so the overall brightness stays the same.

The idea is you want to remove low frequency information from the texture that confuses/competes with the shading but leave the mid to high frequency details.

Re: Erode status/update

Posted: Fri Feb 22, 2019 4:08 am
by checkerz
So...is this "fluff" layer something we can expect to see? Near future as in before SX is over? Outdoors? Not at all?