LiDAR circuit models

Discussion in 'General Discussion' started by toodaft, Sep 28, 2011.

Tags:
  1. fanlebowski

    fanlebowski Registered

    Joined:
    Oct 19, 2010
    Messages:
    132
    Likes Received:
    1
    ah sorry ;) some modders have sometimes CAD files, by the way.
     
  2. DeDios

    DeDios Registered

    Joined:
    Oct 5, 2010
    Messages:
    298
    Likes Received:
    1
    Most of times are not enough detailed for give u perfect elevations but..better than nothing :D
     
  3. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    Shouldn't worry about what? Most of the rF1 tracks were fictional, the 'Barcelona' was obviously not right (elevation), Monza has fewer elevation changes so maybe it's closer but nothing said it was perfect. I'm guessing most rF2 tracks are either fictional or historical (so any sort of scanning isn't feasible).

    The question of whether we need super-realistic renditions of real tracks is obviously a contentious one, unlike a question of "'flat' vs 'realistically bumpy'" which I think we'd all agree on, but rF1 (and probably 2) tracks surely aren't a part of the discussion.
     
  4. ZeosPantera

    ZeosPantera Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,397
    Likes Received:
    14
    If the track is a modern well know circuit I think it should be accurate. Old tracks you have to give a pass because... no time machines. A current track should at least be surveyed and video'd for recreation later. I am one of those people who usually argues against the advantages of Lidar tracks. Still doesn't mean I want real tracks based only on guesswork.
     
  5. DeDios

    DeDios Registered

    Joined:
    Oct 5, 2010
    Messages:
    298
    Likes Received:
    1
    ISI's Brianza is IMHO the best version available for rFactor. Looks really similar to real one.
     
  6. juls

    juls Registered

    Joined:
    Oct 13, 2010
    Messages:
    5
    Likes Received:
    0
    I do not ask for laser scan, but rFactor 2 should be able to replicate believable track texture with random tiny bumps (not only big bumps), especially for vintage tracks, because it is not only a fancy gimmick: when all 4 wheels get closer from the grip limit, bumps radically change the driving experience, because a tyre will slip and regain grip on every tiny bump and the car rotates and you have to be very reactive and control this.

    That's why IRL racing videos the driver acts very rapidly on the steering when cornering, and this art of maximizing grip and keeping control over a rough surface is totally lost in most simulations.

    IMO the influence of track surface is very strong and impacts every aspect of a simulation. For example it impacts tires and the entire driving experience:
    With the usual smooth tracks the only possibility to reach a realistic level of difficulty is to punish the drivers with a brutal collapse of grip above the limit and unrecoverable spin. This way keeping the car at the limit is difficult, but for wrong reasons.
    But with a believably bumpy texture, you do not need such unforgiving tires. Even with forgiving tires, keeping the car at the limit is difficult, because it requires to pay careful attention and react very fast to track surface as the grip changes widely several times per second per wheel resulting in yaw/speed loss.
    You open the possibility to have a sim where driving is intuitive, car is easy to keep on track, and at the same time reaching top level is a very difficult task. And that's way more realistic :)

    Is rFactor able to replicate such detailed surface, even if not scanned (randomly generated) and it's effects on handling ?
     
    Last edited by a moderator: Oct 27, 2011
  7. Revvin

    Revvin Registered

    Joined:
    Oct 5, 2010
    Messages:
    40
    Likes Received:
    0
    I'm hoping Tim might be able to shed some light on this as I asked something similar a day or so ago:

     
  8. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    Great post!

    That greater variability is something sorely lacking with rF1's sine-wave material 'bump' system, because instead of the randomness manifesting itself as bigger microbumps (for want of a better word) in various parts of a corner or the track as a whole, you have one fairly smooth material here and then maybe a bumpy one there - which, in practice, just feels more like a grip level change.

    But, when people talk about scanned tracks and how good they feel they tend to be referring to how bumpy it is in the sense of those bumps that you notice unsettling the car or moving it around - which might occur up to, say, 5 times a second at speed. The resolution required for this stuff is much more coarse than is talked about with various scanning techniques, and it seems to me that with a suitable 'randomness' system (seems possible a 2d multiple wave system would be a 'simple' solution, where the addition of the peaks and troughs give a fair amount of 'random' variability) and a fair bit of work on placing suitable larger 'bumps' in key spots on a track you could give a non-scanned track a lot of character and nearly effectively as close as a quality scanned version - close enough that you could say it's scanned and people would believe it, and close enough to make the extra cost and process of actually scanning for more realistic results something to seriously consider before undertaking.

    Of course, scanning will definitely give more accurate results (4 waves split between X and Y on a material, for example, might produce a peak where the real track has a hollow, and there'll obviously a pattern, albeit less obvious than the current linear sine one especially when the driving line has you moving through/across the pattern most of the time), but in terms of both cost and processing power it's more demanding as well.
     
  9. blakboks

    blakboks Registered

    Joined:
    Oct 20, 2010
    Messages:
    843
    Likes Received:
    30
    That's because you keep teasing us with screenshots and stuff! Excitement for your product is a good thing, Tim. You should worry if people AREN'T asking you every day when rF2 will be ready! Not venturing into a possible business-expanding opportunity because people might be anxious about it's release doesn't sound like a very good reason--just the opposite.

    There are plenty of businesses that are perfectly viable relying on licensing. Typically, there are two pricing options--exclusive use, and licensed use. Fonts, music, pictures, etc.--all sorts of media work this way. You could have a 'lower resolution' or 'out-of-date' (i.e. last year's scan of a track vs. this year's) that the average consumer would be quite happy with--at least knowing that sight-lines and major bumps and a more 'organic' surface (i.e. not sine-wavy) are present--and then the full-resolution, as accurate as possible version that the big players pay big money for. Of course, there is the possibility of lower-budget teams willing to subsidize the cost of the lower-resolution data as well. As I'm sure you're aware, it's not uncommon for companies to charge more for the same product just because it's used for commercial purposes. Take ISP's for instance. A business account is MUCH more money than a residential account--and probably slower speeds to boot--just because they can charge more because they know the business is making money off that internet connection.

    I think it's clear from this whole discussion (despite its 'inelegance' at times) that a large amount of people desperately want to know that what they're racing on is going to be within perceptible levels of accuracy. It's a little disheartening to be met with so much resistance to the subject. It doesn't feel like there's much effort on ISI's behalf to look into it. Maybe ISI has, but so far all we've gotten is "pay us beaucoups $$$, and we'll do it".

    As far as the investment goes, since you still have to create the geometry, textures, surface properties, test it, etc., the only REAL additional investment at risk of creating LS tracks comes from the actual scanning and data integration. Let's suppose that a track costs $50,000 USD to scan (equipment costs are going to be spread out over multiple jobs, and have depreciation and all that other accounting mumbo-jumbo)...well let me first ask: Am i even in the right ballpark here, or are we talking more like $500,000 USD? I don't know what ISI predicts rF2's sales figures to be, but it doesn't seem like 10,000 people purchasing the track is an unrealistic number. If everyone pays $15 for the track, we're talking about $150,000--a 200% profit margin! If my assumptions are way off-base, please correct me.

    I implore ISI to give it a try with one track. Test the waters--do Lime Rock-a nice short track with a lot of character. Some people will certainly buy it--I know I will.
     
  10. DeDios

    DeDios Registered

    Joined:
    Oct 5, 2010
    Messages:
    298
    Likes Received:
    1
    +1 :)
     
  11. Old Hat

    Old Hat Registered

    Joined:
    Feb 22, 2011
    Messages:
    163
    Likes Received:
    8
    Well, a lot of people don't realise how varied the cambers are. Here's a vid showing a plugin for iRacing that locks the driver's view to the horizon so you see more of the changes. Even iRacers were surprised to find that stretches of straight were actually on big slopes. Some people drive with this effect on all the time now because it gives useful information about the shape of the track surface that you can't normally see. You only experience the effects of more or less grip than you instinctively expect. In RL you'd feel gravity, g-force and orientation, but in a sim the only way of knowing this is the effects it has on grip levels on different lines when driving on the limit - i.e very fast; actually faster than most drivers can around corners most of the time.

    http://www.youtube.com/watch?v=saCDumR56Bg&feature=player_embedded#

    This is the problem IMO with people saying that they can tell there's not much difference between this and that track based on vids, photos and possibly even driving on the LS track in some cases.
     
  12. CdnRacer

    CdnRacer Banned

    Joined:
    Feb 4, 2011
    Messages:
    1,894
    Likes Received:
    31
  13. ZeosPantera

    ZeosPantera Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,397
    Likes Received:
    14
    I found the same to be the case due to high FOV. It is usually set too wide forcing all the curves to be squeezed onto your screen which is usually something like 20" wide. In real life seeing that turn come up across your whole vision indicates it is a VERY big hairpin. The elevation changes also get watered down with a high fov. Just look at the difference in these two identical runs at bathurst.

     
  14. markshires

    markshires Registered

    Joined:
    Oct 28, 2011
    Messages:
    9
    Likes Received:
    0
    Hi Everyone,

    I have read a bit of this thread, not it all (it's too long)
    Basically I have been producing LIDAR race tracks now for the last 4 years for a number of customers.

    I have been supplied side by side graphs showing the output of each wheel, showing the bumps from the real circuit and the bumps from the LIDAR circuit and they are very very close, it has to be otherwise the track would not be accepted! It's amazing really! I know that the data produced is used for car setups etc.

    I think from a driving point of view it all comes down to your dynamic model that you are using and whether it makes use of the high density mesh that you are driving on.

    Anyway if anyone needs a hand in generating a LIDAR track I am more than happy to help out.

    Mark
    www.markshires.com
    (Don't forget to Facebook Like my page)
     
  15. CdnRacer

    CdnRacer Banned

    Joined:
    Feb 4, 2011
    Messages:
    1,894
    Likes Received:
    31
    Sorry bud but facts and sense don't apply here. Fanboyism rules.
     
  16. ZeosPantera

    ZeosPantera Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,397
    Likes Received:
    14
    Ignore him mark. Some people don't quite get how a sensible forum goes. Any insite into your profession is a great help to this discussion.

    If you don't mind me asking mark. What do you figure the Average cost of Lidar'ing a 4-5 km track with lets say a 3-10 cm accuracy would be? Assuming road surface only scans.
     
  17. markshires

    markshires Registered

    Joined:
    Oct 28, 2011
    Messages:
    9
    Likes Received:
    0
    If I'm honest I don't know about the cost side of these things, I'm just the artist that uses the data to create the final product. What I can say is that you would have one cost to actually buy the data from a supplier in the first place, which I know can be cheaper if the data could be sold on rather than being specialised areas which will only be sold to you (Generally not race ciruits, more like if you want your street scanning for instance). You would then have the actual cost of processing the data, and actually making the model of the track, it is not an entirely automatic process.

    Another point to make is that you wouldn't be able to run a 10cm LIDAR track inside rFactor it just wouldn't cope, you would have to lower the resolution of the data down, it would be more like somewhere between 25cm to 50cm data.

    Mark
     
  18. Pandamasque

    Pandamasque Registered

    Joined:
    Oct 5, 2010
    Messages:
    160
    Likes Received:
    2
    Even if you have an invisible higher detail mesh to drive on? (as in iR)
     
  19. DeDios

    DeDios Registered

    Joined:
    Oct 5, 2010
    Messages:
    298
    Likes Received:
    1
    Hi markshires :) welcome and thanks for posting!
     
  20. blakboks

    blakboks Registered

    Joined:
    Oct 20, 2010
    Messages:
    843
    Likes Received:
    30
    Hey Mark, despite what CdnRacer seems to believe, we are actually quite hungry for factual answers around here...we just haven't gotten many so far, so speculation and hearsay presides.

    Even if you're only getting a mesh resolution of 25-50cm, the positional accuracy of each point is going to be MUCH greater, correct (i.e. it's location in the Z-axis, most importantly)? If this sort of resolution is what we'd realistically be working with in the game, would it then make more sense to do a moving scan to cut down on the expense of scanning? Or does a moving scan decrease both the resolution AND the accuracy; and by what degree?

    As I am also a 3D artist, and Lime Rock and Watkins Glen are both within a few hours from my home (to be able to collect reference); if anyone is able to get scan data and wants a helping hand creating the artwork, I'd be more than willing to lend a hand if we want to make this a community project :D
     

Share This Page