Special Stage Forums banner

1 - 20 of 45 Posts

·
N43 37.126' W-115 15.273'
Joined
·
1,284 Posts
Discussion Starter #1
Since I can see this converation going long, let's start a new thread.

This thread captures the orginal posts:
http://specialstage.com/forums/showthread.php?t=42480

-----------------

I found lat/lon resolution is an issue when combined with the noise in the elevation data. Every dataset I've seen has a what appears to be a variance of about 5m (width of the noise) if not more. Meaning, the elevation data may not be too useful at the scales we're interested in. I bet a good length scale for a rally road is close to 30m if not shorter. To capture a crest 30m seems about right. You have a somewhat sparse data set to work with for filtering if you are trying to resolve something at less than 30m with a dataset that is 30m in resolution. 10m might be better, 3m might work great!

I haven't looked yet but I wonder if there is some statement on what the bias and variance is in the sat data we're looking at. I wouldn't be suprised if it was a function of slant angle to the sat or other paramaters too.

There is also some really good LIDAR data but it doesn't cover many of the regions of interest yet.

Also, I have a GPS/APRS unit in the rally car, I might leave it runnign during recce next time to collect high res data.

-e
 

·
Registered
Joined
·
236 Posts
Hey Eddie,

I have .kml (I think that is the extension,) files for the 2008 Mt. Hood Rally: Fir Mountain, FS 17, and Gilhouley, if you or anyone else wants to give that a try.
 

·
Registered
Joined
·
3,475 Posts
I probably have all of the stages for LSPR this past year, and some of STPR, maybe all of Ojibwe, too.

I usually forget to dump the tracks, then on the drive home, the GPS overwrites it with such interesting roads as I-75 or I-80.

I should get a PC so I can play these games. Or someone should port them to OSX. :D
 

·
Registered
Joined
·
3,063 Posts
Nice gorman stage 1, btw, :D
 

·
RallyX Weenie
Joined
·
672 Posts
nice work!

now comes the real time consuming part - making it perfect :D

The only problem I've found w/ zaxxon's scripts is that his algorithm for mapping a KML into nodes and control points isn't quite as good as BTB's KML import - mainly with KMLs generated by Google Earth. On most things (particularly most rally roads) they both work really well. However, if there's something like a long-ish straight, followed by a square turn, followed by another straight, zaxxon's method will turn the square turn into a big sweeper.

I'm going to hack together a script to take a KML, and 'fill in the gaps' so that there's points every N meters - I think it's the sparseness of points in this situation that causes zaxxon's code to do something wacky with the spline control points.
 

·
N43 37.126' W-115 15.273'
Joined
·
1,284 Posts
Discussion Starter #7
Thanks Bob.

Yeah, getting it right will take time. :) I'm ok with moving nodes around to fix problems. Even with minimal filtering I'm finding that the elevation changes are not as dramatic as they are in real life. I find myself moving nodes around vertically quite a bit.

Gorman has a two stages that share part of the same road so I implemented the multitrack feature. I got it loaded into BTB and RBR ok but I'm having issues with track/terrain height mismatch at the intersection of one track and the other. Basically the nodes/lines/surfaces in gmesh was bad (as predicted by the documentation) and I cleaned it up manually. However, seems like the road/terrain info didn't update after change the mesh. I think I get 60% of what's going in with these scripts (the Spanish doesn't help either), is it the btb06 script that needs to be rerun? I have the documentation but I've been more apt to tinker than rtfm :D.

-eddie
 

·
N43 37.126' W-115 15.273'
Joined
·
1,284 Posts
Discussion Starter #9 (Edited)
Actually, something I didn't do was make sure that the separate KML files share common nodes at the intersections... That would have been a good idea to do, huh? But even still if the scripts are interpolating nodes independently of each other the two tracks probably wouldn't line up anyway...
 

·
RallyX Weenie
Joined
·
672 Posts
getting a decent approximation of just the road itself is pretty easy (and surrounding terrain, if you use the zaxxon scripts) - a little bit of a learning curve, but not too bad IMO (but I'm a computer geek for a living, so ymmv...)

As I alluded to earlier, it's the fine tuning and 'dress up' of a stage that really makes it good or bad, and is also the time consuming part.

A quick tip for those playing with this stuff-

If you want your track to show up under the 'BTB Tracks Test' window in RSRBR, you need to do the following:
* Export your track to RBR - should be somewhere like {RBR}\RX_CONTENT\Tracks\TrackName
* Put Plugins\rbr_rx.dll and RX_CONTENT\Tracks\TrackName into a 7-zip file TrackName.7z (doing 7-zip->open on the file should show the Plugins and RX_CONTENT folders) [Also important to have the names of the track folder and the 7z file match]
* Put TrackName.7z into {RBR}\RXRSRBR

Easy, right?
 

·
N43 37.126' W-115 15.273'
Joined
·
1,284 Posts
Discussion Starter #13 (Edited)
Needs more waterbars.

I wonder how hard that would be to do for NNR...
Waterbars are a bit beyond the resolution of the data we have. :) But, don't worry the bumps will be there. For high fidelity, imo in-car footage is the way to go. Luckily there is no shortage of it, e.g. Welker and Sombrano have some great HD Gorman footage for me to play with.

Given the tools available, any US rally is probably doable. Depending on the fidelity it could be quick (like that Gorman vid which took 2 hours) or not. ;)
 

·
N43 37.126' W-115 15.273'
Joined
·
1,284 Posts
Discussion Starter #14
Ok here is an update. I got the multitrack working so both sections of N. Boundary are included (that makes 4 stages total including forward and rev). I also added satellite imagery for the terrain. I'm about 40% through elevation adjustments. Then I need to do the objects (I already started that too).

Current Progress:
http://www.youtube.com/watch?v=o57m1IdS27E
 

·
pressing on tirelessly
Joined
·
2,225 Posts
That's spookily accurate. At least at the beginning.

I think some of the elevation changes are conservative. There're a bunch of other little things I noticed but jeez, I'm impressed. We might have to add a clause to the 90-day rule to prohibit playing RBR.
 

·
So long & thx for all the fish
Joined
·
362 Posts
That's spookily accurate. At least at the beginning.

I think some of the elevation changes are conservative. There're a bunch of other little things I noticed but jeez, I'm impressed. We might have to add a clause to the 90-day rule to prohibit playing RBR.
Or add a virtual championship to the CRS like [shudder] NASCAR has.

Jonathan
 

·
N43 37.126' W-115 15.273'
Joined
·
1,284 Posts
Discussion Starter #18
LiDAR Data is available for all of PA and WV, and NC for free. PA is at: http://www.pasda.psu.edu/

USGS supplies ~3m LiDAR derived terrain models for the rest of the US here:

http://seamless.usgs.gov 1/9-arc second data

Coverage for rest of US is spotty: http://seamless.usgs.gov/data_availability.php?serviceid=Dataset_19

Jeremy
Yup, that's the data source I used (1/3 arcsec). 1/9-arc second is very limited right now but 1/3 arcsec is readily available.

If someone wanted to STPR the 1/9 arcsec data would be a good set to use. :)
 
1 - 20 of 45 Posts
Top