W.I.P. Goodwood Circuit 0.99

[mantasisg]

Great

Yes, they must be similar cars. I still hope that LM had more suspension adjustments, because it is weird if racing car would have some kind of universal suspension kit for all tracks.

In the end I could possibly have to do further smoothing for few potential areas. But I don't bet it all on one car. Especially not too Goodwoodish.

 

[luchian]

Agreed.
However, T2 or 3 (?) seems very rough and makes for a very bumpy ride . Let's see what the comparison video will show. But honestly, it makes me wonder:
Let's say lidar (or even LS) is 100% accurate.
But what if AC has a problem with "bumps" simulation ? And for example 1 cm bump irl has an augmented effect in AC. And in fact, in order to simulate correct behavior, maybe one needs to use sort of a "smoothing" factor. The 1 cm irl, to be used as 0.7cm in AC in order to have the same impact on the suspension ? . Just babbling here, but maybe focusing too much on the data accuracy we miss other important variable from the equation..

We need somebody to go measure irl one of the tracks modeled by Kunos and check .

 

[mantasisg]

The very exit of Madgwick should be bumpy. As the rest of the straight till the Fordwater - maybe should not be too bumpy.

I wonder if tarmac at theose spots could have been colder, where bumps are potentially too intensive. I suppose it could be that LiDAR is working in similar principle to thermal vision. At least intensity data looks similar to thermal vision in gray scale.

I also had this thought about bumps in AC being augumented but I don't think so. After all thats more like small undulations than actual sudden bumps. I also don't have any complains for most of the track. Just few spots. More over, track is perfectly fine with most of the cars for my taste. So I wonder if it could also be cars physics thing. kunos don't have many very bumpy tracks. Even Zandvoort is surprisingly smooth, I alsways thought that Zandvoort is too smooth. So perhaps very stiff cars has something just not adjusted for more intense bumps, which I like to call small scale undulations.

At the very end those are few cars which are kinda "too bumpy" and those are not the main ones to take for goodwood. Also there is no proof that such cars would have ever driven in goodwood (honestly) flat out. Also track foundation is cancrede refueling road for fighter planes from WW2. There was a repavement in 1998, but that really was just to maintain the tarmac nice and smooth, but not to actually smooth out the undulations. Also the intention of Goodwood is to keep the soul of the sixties track. Thats an important point.

So it will never be billiard table smooth almost ready for F1 to race, for F2 for sure, at least not by me. But I can agree about some spots being too bumpy.

I also find this as interesting discusion when it is a discussion and not some kind of accusation. So i really want to thank you for that

 

[mantasisg]

Good to watch externaly. Even mentions harsh ride at 11:55

Mentions bumps at 10:40

Mentions them here, also shimering gives out some undulations, because thats where shimmering is going on

Shadows caused by headlights in some cam angles shows some:

More clips like this could be awesome also shows well how cars are following the surface, very easy to see how they moves up and down:

 

[mantasisg]

Good to watch externaly. Even mentions harsh ride at 11:55

Mentions bumps at 10:40

Mentions them here, also shimering gives out some undulations, because thats where shimmering is going on

Shadows caused by headlights in some cam angles shows some:

More clips like this could be awesome also shows well how cars are following the surface, very easy to see how they moves up and down:

 

[Willy Wale]

Regarding the accuracy of LIDAR the environment agency have information on their site

"Accuracy

We ensure all our LIDAR data meets high accuracy specifications.

Vertical (height) accuracy
Our specifications require the absolute height error to be less than ±15cm. This is the root mean squared
error or RMSE. It quantifies the error or difference between the Ground Truth Survey and the LIDAR data.
With our more recent surveys we see this fall to about ±5cm.
We expect the relative height error (random error) to be no more than ±5cm.

Horizontal (planar) accuracy
The absolute spatial error in our LIDAR data is ±40cm. For our datasets at 2m, 1m and 50cm resolution,
this error is effectively absorbed in the pixels of the raster image.
The relative horizontal accuracy of the LIDAR sensors we have been using over the past decade, as stated
by the Instrument manufacturer, is 1/5500 x altitude (m Above Ground Level). For example, if the average
survey height is 1000m AGL relative horizontal accuracy is 0.1818m."

Based on the above I wouldn't assume that any bump smaller than 5cm (possibly up to 15cm) is a real physical bump. It could just be measurement error.

 

[mantasisg]

I have smoothed the mesh probably more than 40 times. At first I was smoothing whole mesh, later started smoothing more localized spots. Some parts of the track needed less smoothing, for example Lavant became rather smooth quite early.

I assume by smoothing I should have had reduced the error quite a bit. When observing the original point cloud mesh I didn't notice any significant artifacts on the track surface.

The part from Madgwick exit till NoName braking point had the most smoothing.

Also I wouldn't assume that any bump is off by 5cm, I would assume that some could be. Why should it apply only for bumps smaller than 5cm ?

I would like to focus on particular zones which seems too bumpy, and not whole circuit. If you dislike bumpyness drive other cars or forget the track. I will not be ruining authenticity of the whole mesh for the sake of a few cars. Also I will be smoothing those zones only when I'll be totally sure about that they should be smoother.

 

[mantasisg]

 

[mantasisg]

I like this because it is something like AC locked to horizon camera. Shows the movement very well.

 

[Capzor]

Thanks mantasisg

Just wondering if you can quicky share your workflow to make this track with lidar?

Cheers ;P

 

[mantasisg]

Your welcome

It was first try with LiDAR. It went like this:

1. Got information about LiDAR data and the link.
2. Found the data and downloaded it.
3. Downloaded CloudCompare and captured the mesh by using tutorial from this site made by DanTDBV;
4. Got mesh into a Blender, it was far from world center, I think I have used Origin to Geometry, and then Geometry to Center commands, don't remember exactly.
5. There were about 2-4million vertices IIRIC, and blender was crazy slow;
6. Started removing unnecessary detail, slowly blender was getting faster to use;
7. Began looking for the road, and couldn't tell where it is because I didn't know about intensity map at the time. I was color coding where I think the road was, by applying material with color. Soon it began to get funky and I was starting to realise that I'm lost. I have posted here and there, and then LilSki informed me about the intensity data. With intensity data I could easily tell where is what, and I could really start to go.
8. I wanted to try the track real quick, so I quickly assigned road material, and grass material. Mesh was spiky, so I have smoothed it about 20-25x till it visually appeared not to be spiky anymore IIRIC
9. Tried the track, and it was very believable, elevations, cambers, curves.... It was very bumpy but drivable.
10. I needed to decide if I'll do stuff in normal way - just project road mesh over lidar mesh, but the mesh of lidar looked very tempting and I just continued smoothing it a lot of times.
11. To get the road edges straight I have Lifted grass edges about 2-3cm up and made them straight and overlaped the road. Also I have tried to make road physical mesh to go in not too much. This was quite demanding, even though Goodwood is not very big, though there are two edges, four in total.
12. Projected visual mesh for road over Lidar mesh which is used as physical mesh.
13. Used Lidar mesh for terrain.
14. Lidar mesh was great for buildings and placement, and their dimentions too.

If I would do it again I would probably just project heavily subdivided mesh over lidar mesh. I just would have to find out how to randomise and triangulate it in a way how LiDAR mesh naturaly is, because it creates some great feeling while driving, FFB noise without pattern.

I thought I will loose the noise when smoothing so much, but seems like it is still there.

 

[Capzor]

Thank you very much mister

 

[mantasisg]

3D Hortensia. ~950vertices with doubled faces. Needs a bit texture tweaks. Planes on the sides are slightly giving away but not much. Probably could use to replace some of the cardboard ones.

Couldn't get to look right with alpha blend and transparency on. Alpha test seems to work. Perhaps would work with alpha blend if it would be a single object, now it consists of three objects.

 

[mantasisg]

Begining to work with new terrain texture. Tweaking texture, shaders, colors, cleaning, improving...

Also have done one new camera set. Can't to improve so many things

 

[Mr Whippy]

Looking good.

On smoothing the aerial lidar data, it's a can of worms.

Isn't it polled every 1m? Literal scans are 1m point intervals, so very sparse.

The best you can do is probably smooth the data so it averages no more than +/- 15cm up/down any given point, otherwise you're artificially pushing a point out of it's error range and making it become increasingly wrong.


I love this course, and I like bumpy courses. Real cars are very much characterised by their compromises on comfort vs road holding, or ability to match certain track types from firm to soft for smooth and rough tracks.

AC has been short on rough roads. Even their Highlands fantasy road course is very NFS with it's wide bends and smooth curves and undulations.


I'll grab this tomorrow for a drive in my favourite road cars!

 

[luchian]

I think it's a very good idea to give some love to the grassy areas. Judging by the screens, will definitely make the track look even more coherent.
In the version I have, grass looks a bit "faded" when driving (in the sense that it looks like a green mass, but you cannot observe some detail - like you would see for the road).
So yeah, very good step forward, great work.

Sent from my phone using Tapatalk

 

[mantasisg]

@Mr Whippy I don't remember exactly, but I think some tiles were 0.5m and some were 1m resolution. I think road naturally needed less smoothing where resolution was lower. I also had this idea about ruining track by smoothing it. I have started to smooth it in segments soon after, because some places seemed like smooth enough, and also to prevent smoothing out real undulations. I don't think that there are much danger to smooth out high resolution mesh into totaly smooth plane, with each smoothing action mesh changes less and less. But perhaps after extreme number of smoothing applications it would smooth out.... Math could be useful there

I agree about AC tracks being "not too rough". I think Magione is one of the roughest. Well probably Monza Oval is the roughest, but thats different thing. I think that Laguna Seca is surprisingly smooth.


@luchian Thank you Wasn't totaly happy about previous texture. Now there I have different challenges. have to align everything, tweak colors and shaders again, then tweak road to grass fade texture and shader. When I added this new texture for terrain, everything seemed to drop down in quality somehow, strange.

 

[luchian]

When I added this new texture for terrain, everything seemed to drop down in quality somehow, strange.

Maybe it's a question of scale ?
If you stretch the texture to cover an area that is too big, it will loose detail and look meh. For surfaces this big, I think multi-layer shaders are the only solution (..unless one makes a ton of hand painted textures )

 

[mantasisg]

No. I think that because terrain got more detailed, and everything else stayed the same. So relatively everything else dropped down.

 

[mantasisg]

Here is how hills in distance should look like. But Kunos use way too strong default natural "distance fog", so had to be modded to look like that.

Still quite far from this real look, but that hill really lacks when it is desaturated by natural fog even in clear weathers

Tweaking Hortensia, I think I still have to adujst it.

Hills in the distance cut down by rendering distance. Grasstotarmac fade object is a bit off.