QUESTION Gathering real track data

[andyl]

So it seems my enthusiasm for LiDAR has been hampered by the fact that the data is not available. Short of contracting a company to scan the ground, what other options are there? I'm interested in making the track as accurate as possible so I can use it as a training tool.

There appear to be 3D scanners available that will probably take some time to scan a whole track as they produce measurements from a static location. For example, the Scanse has a maximum distance of 40 metres, which realistically means scanning every 20m and stitching the results together. However, it's accurate to 10mm, which is pretty reasonable!

There's the option of using elevation data and aerial photos in Google Maps, possibly improved by taking GPS measurements around the track to get more accurate elevation data.

I could also build a kit that rolls along the floor and records GPS data for lat/long/elevation. Roll that down one side of the track and connect it via a cable to another unit on the other side of the track with a flat board. A range sensor would read the distance to the other side of the track and the cable would connect to an IMU (3D accelerometer and compass) that determines the angle of the cable, thus determining the camber at that point. That would allow the required data to be recorded around the track at every point; latitude, longitude, elevation, track width and camber. Is that good enough to give a realistic track map? Is GPS elevation data a high enough accuracy? Clearly, the absolute accuracy is not good enough, but the data I have of Castle Combe has a resolution of 0.1m and doesn't seem to drift much over a short period as overlaying the elevation over 3 laps of my GPS data seems to show very little difference. Here's 3 laps of longitude vs elevation up Avon Rise:

Photogrammetry is also possible, but in my experience, it needs lots of detail to work properly, so does it work well with grass/tarmac? There's some pretty simple tools out there that seem to work well with cities and other objects without big areas of single colours. I tried it out on my sports racer a few years ago and it was less than successful thanks to reflections and a big body of a single colour. Newspaper covering a car is hard work but feasible. The same cannot be said for a race track...

 

[luchian]

Interesting.

You might be aware about this project, but just in case. It's something built by a passionate member, and the results can be a good alternative.

http://projectvideoscanner.blogspot.nl/

Some discussions along development:
https://forum.studio-397.com/index.php?threads/cheap-alternative-to-laser-scanning.41287/

And a video:

 

[andyl]

That's brilliant! Worth trying out, I think. £90 buys a couple of the recommended cameras and I've got some aluminium angle lying around...

 

[Willy Wale]

For UK tracks we have the benefit of the Ordnance Survey (OS) maps.. The most detailed map they have is called the mastermap and several companies will sell you a chunk of it. It's priced on the area you select. You'll likely get a colour image of the map a dxf or similar CAD file with all the lines on the map (track edges, barriers, building outlines etc) and you sometimes get building heights, maybe a geo-referenced aerial image. You can see the level of detail of the map is like by visiting the local council's planning site.

http://services.wiltshire.gov.uk/PlanningGIS/Mapping/MapSearch



The drawback of course is that an area the size of Castle Combe is likely to cost around £100 for a one year licence.

 

[Pixelchaser]

if you think about these adhoc scanning systems enough you`ll realise ... theres a reason we don't see many tracks made this way. its nice to try but... is it really a solution within track making ? I don't think so. I don't even think aerial drone scanning is even bringing good results and as always its only ever as good as the person managing and dealing with the data and their understanding. so its certainly not fool proof as some methods suggest.. not even lidar is fool proof, but ive seen comments recently that it cant be wrong etc. this is simply not true. I remember being given that data from the roads department scanned from the back of a vehicle. about 1 cm res and it was the hardest thing to work with.

 

[andyl]

What I'm realising is that the data isn't as available as it first seems. There's no two ways about it; the best tracks will come from the best data handled by the best people, but with no data, you can't make anything. I don't expect a one-click solution, but without accurate data of what the track is actually like, how do you know if your completed track is accurate? Getting the data for latitude and longitude is reasonably easy, but elevation makes a massive difference. Camber gets you 99% of the way there and then you're at the level of bumps, which I can live without, frankly. That's what a track walk is for as it changes year on year. As much as I enjoy simulation, I really want it to be a good training aid for the tracks I drive on. I can't drive many of them except on a competition day, so extra practice would be a real benefit.

 

[Pixelchaser]

recently lewis Hamilton commented that between simulations and your average sim-cade game, there is not much difference, not that much more to be gained from the sim etc. I understand his comments and I believe it relates here, because before any of this light based capture technology, real drivers were simulating the experience with out this tech. and like before all this track makers were building by hand. and whilst one could argue it was worthless in actual realm of factual realism. in 20 years we be saying the same thing all over again about working with lidar compared to what we doing at that time. just not quite as much post reference of it being worthless which it was. there was zero accuracy by todays standard perceived.

I believe non drivers of cars on tracks racing get the benefit of the deal due to the insight into what it might be like to race on a track etc. but reversing that notion. I don't think there is much benefit to driving a real track over and above driving a real track. if sims has any credibility, they`d all feel the same because nature has only 1 opinion on that.

having better data is certainly better. but I just don't think with all the simulated opinions out there coming from different titles there is any merit in there being one that can improve driving on real tracks with real cars. better data is better for sims. but only sims.

I have know boeing pilots that love flightisimming, I mean hardcore as possible. but they don't fly airbus`s on them, because that jeopardises lives working with "other " flight systems. and that's a dangerous problem. so that an extreme view upon the errors flightsims can and will introduce when they`re not as realistic as they should be. or in this case so realistic as to become a problem beyond simple entertainment.

ive come to believe its just a mere nice notion to involve real racing with sim racing. more romantically based in theory than fact but In 20 years with quantum computers, then oah yeah we probably no need to be doing anything real at all .

 

[mantasisg]

not even lidar is fool proof, but ive seen comments recently that it cant be wrong etc. this is simply not true.

Who said that it can't be wrong ? And what exactly fool proof means ?

If lidar will give slight error, if track will be tiny bit wide, tiny bit wrong camber, a curve will be off by a little bit, if a bump will be a little bit rougher (of course thats me for sure - a fool). It is still going to be extremely close. because those are going to be really tiny numbers. I don't see a point of nitpicking of tiny percentages of lidar data being off, there are many points around, and the average levels them down to smaller error, so smoothing should work. The error will be small and it will not have big practical impact, unless in the very proffesional aplications, such as F1 are using, I think.

But if someone is annoyed for slightly (possibly) too bumpy mesh, then one should complain for the opposite too - for mesh being slightly too flat. But it is not going to happen because flat is comfortable, easy. Bumpy is challenging and potentially uncomfortable with cars which are suited only to glass smooth roads. At the end it is heading towards sick need of having something abolutely identical, down to millimeters. And thats not normal. To my mind in simulators, at least the way we use them, tiny error doesn't mean anything. There must be a point when the model is close enough.

There is one thing which is data not related, that is your video device, is it flat screen, or is it VR. Because only VR will give correct depth perception, and only in VR track will look like in real life.

To my liking even Lidar is not necessary, if the track is not too big, then GE error is not going to be huge, and if nice maps are available with isohypses in small enough step. I have created nemuno Ziedas in such way. There was heights map with isohypses in 5m step. I drove the track IRL and with VR and it was very similar. For a simple use IMO thats enough.

But people are going to caomplain even for Lidar data accuracy in future. I saw Castle Combe is a bit bumpy...

 

[Pixelchaser]

well fool proof means, a fool can not go wrong with it. (I'm not calling anyone a fool per say by saying this) its a phrase. and I agree with much of what you say,

your comment about vr interests me more because the more wide spread the use of vr and the notion that as you say gives better correct depth perception, the eventual social commonality here will be that playing on my 2d screen is tantamount to simcade and not simulation. which is merely interesting me as to the future of the term sim with the software we dabble with.

 

[mantasisg]

Well yeah, wrong is subjective, it can be still wrong or not wrong anymore, depending on standard. I suppose it better be wrong in such way that it could still be fixed, than being in wrong with no way back. I agree that fools can not see anything wrong, and always too confident about their right.

Yes VR is interesting, I think technology is not yet good enough. But the depth perception just to experience how the track really elevates. The biggest con is that it is one more gadget, and also kinda separates you from real world completely. I like to still be able to look around in a room.

But yes, IMO it is necessary to really experience the track model, and to get most out of it.

Nemuno Ziedas elevates 32m up in about 220m of path. In flat screen it is impressive, but not too much. IRL and with VR it is crazy.

Also distances feels a bit different, road looks wider in the distance. When I tried VR it had not enough peripheral vision.

Though F1 drivers are practicing in their sims without VR AFAIK. So it might not be necessary to learn the track, but depth of field makes great difference.

 

[Mr Whippy]

I'm pretty sure F1 simmers will use dome projectors or big screens or whatever else to get a correct 'fov' for the driver.

The core part for them using the sims is scales and parallaxes etc.

Ie, when trackside feature looks X big, turn in.
Or when feature Y and Z line up start braking.


They'll probably go VR once it's better.

The driver for better simming in racing over time is based around budgets.
Why go doing real testing when you can do more sim testing to get needed data.

 

[andyl]

But people are going to caomplain even for Lidar data accuracy in future. I saw Castle Combe is a bit bumpy...

Castle Combe is bumpy. It's known for it and the commentators often remark on it. A single seater's rear wing collapsed over a bump earlier this year.

As far as VR headsets go, using an Oculus Rift has been my best driving experience yet in a sim. Better than a triple-screen Cruden, although the motion of the Cruden really adds that bit extra.

My theory is that each time I go on track, it's another £300, so I can get a lot of sim time for my track testing budget. On many of the tracks I visit, there is no opportunity to test, so my only options are to watch on-board videos, analyse data or walk the track. If I could run a realistic sim of those venues, I'd have a real tangible advantage.

 

[fbiehne]

Now that LilSki has started his great tutorial videos on how to mod a track I have a really hard time on actually getting hands on LIDAR data. Are there any recommended tutorials or sources available? It feels like an intransparent chaos going through all these height data / geodata sites…

The only site I found which is quite accessible is https://www.coast.noaa.gov/ but the areas with Lidar data is only in the US. I’m interested on trying my hands on Dijon-Prenois but getting Lidar data for france is nearly impossible or perhaps I just don’t know where to look?

 

[luchian]

@barf is summoned, I would say .

 

[garyjpaterson]

Plenty of UK lidar here: https://environment.data.gov.uk/DefraDataDownload/?Mode=survey
You can mess with the checkboxes to see the coverage, though im my experience its often a little ambitious - and when you actually download stuff it can be missing half the area you wanted etc, but its a good starting point.

I say UK, sadly its just England. I did find some scottish stuff elsewhere, and whilst it has high quality dense stuff, the coverage is very little, nothing of any real interest yet.
https://remotesensingdata.gov.scot/products?collections=scotland-gov/lidar/phase-2/laz

And for Spain there is this, seems to have really excellent coverage for everywhere I've looked so far: http://centrodedescargas.cnig.es/CentroDescargas/locale?request_locale=en#
I almost dont want to point out how good it is because I have a strong desire to do some tracks from that (valencia, aragon etc) but knowing me that may never happen What I like the most about LIDAR from here is it somehow has aerial imagery attached to the point cloud, so you dont have to mess with intensity etc to see the materials (of course you can if you want, and it has classification etc too usually for filtering out trees etc).

Sorry I dont have any info for france, its not something I've looked into.

 

[fbiehne]

Thanks for the link, Gary.
Getting data for the US and UK is far more easy than for the rest of the world it seems.

 

[Prototype]

@fbiehne Or you could try the photogrammatery route. Ive got telemetry data back from various teams doing the Kyalami 9 hour, and the method has been proven to be sound.