January 23, 2018, 07:47:33 AM
bigger smaller reset 800px Wide width Full width Reset * *

NRTracks

 
Welcome, Guest. Please login or register.

Login with username, password and session length
« previous next »
Pages: [1] Go Down Print
Author Topic: PASSING ON THE OUTSIDE @ SuperSpeedways  (Read 3529 times)
fortine_oo
Legacy
Sr. Member
*****
Posts: 188



« on: June 20, 2012, 04:55:18 AM »

  PASSING ON THE OUTSIDE @ SuperSpeedways
             Experiment and Conclusions

Like the cycle of the moon, the question of "PASSING ON THE OUTSIDE" crops up with regularity.
I understand the desire to have the ai do this, you want the racing to reflect what you see on Sunday.
The responses vary and are generally re-cycled from thread to thread. Some of the suppositions show a knowledge of what the various
   track and track.ini adjustments do, but I've never experienced the "proposed changes" actually effecting the desired result.
Just for my own curiosity, I tried a whole lotta things, multiple surface types with different grip values configured numerous ways over
   the whole circuit along with a variety of locations on the track for the race.lp, minrace.lp and maxrace.lp.
                               ---                          ---                          ---

What I experienced:

Definitions:
(CCW track rotation implied.)
1. superspeedway: a track where the driver NEVER lifts (the throttle), or partially lifts for a fraction of a second.
2. most inside line: As close as you can possibly run to the inside of the track and still be on the racing surface.
3. outside line: The high line, right next to the wall or no more than one car width from the wall.

The Parameters:
    There is one absolute that must be remembered, the fastest way around is to run as close to the inside (the shortest distance) as
      possible, and this is exponentially exploited by the NR2003 exe. When trying to get the ai (or any driver) to run another line,
      you've got to make that line exponentially better (faster). ***1
    The track MUST have various grip levels throughout the turns, highest on the outside, lowest on the inside, and something in between in
      the middle. Taking this a step further, the surface properties should to be taken into account. ***2
    The race.lp must run no lower than the second lane down from the outside wall through the corners. ***3
    While min/maxrace lp placement and track.ini values have an affect, it's just not that cut and dried that you'll get the desired results.
      There is no exact rule on track.ini values and lp placement, only generalities. It's strictly trial and error, and dependent on the track,  
      the track.ini ([ track ] and [ track_ai ] parameters), the quality of the race.lp, and the placement of the min/maxrace lp's to bonafide
      secondary racing lines and their relative location to the race.lp.

The Results:
    The leader is going to try and run the race.lp (given the space), even after he takes the lead from an alternate racing line.
    Any ai looking to pass will jump out of the line following the race.lp. Whether they choose the inside or outside depends on the location
      of the minrace and maxrace lp's from the race.lp, the space to move to another line, the speed that racing line is capable of producing,
      and other considerations (that I haven't thought of or don't know about).  
    The ai can be enticed to move from the race.lp by a greater (than 0.00) ai_squeeze_pcnt value. However, increasing this value can
      cause the ai to leave the preferred racing line (the race.lp) even when doing so provides no benefit, and will in all probability cause
      them to lose momentum and fall back. Too high a value will cause the ai to turn off the track for no reason and to run into walls when
      the lp's run close to walls (especially at racing line clip points). With plate racing, or any track where the throttle is always floored and
      where the field tends to run in a pack, using ai_squeeze_pcnt = 0.00 provides the best racing.
    When the race.lp stays close to the outside wall, the outside line will maintain the pace, but the leader always comes from the lower line.
      When the race.lp is close to the wall in the corners but runs closer to the inside down the straights, the whole inside line will dominate,
      even if the outside can maintain the pace.
    The closer the maxrace.lp is to the inside, the tighter the ai will run to the bottom, even if the race.lp is right next to the wall. The leader
      may run there periodically during a lap, but will quickly return to the race.lp.
    In all the configurations tried, the outside line of cars can do no better than maintain the pace. The leader always comes from the
      inside line. If the layout of the various grip sections, lp's, and grip levels are less than optimum, the outside line won't even be able to
      keep up with the pack.

Conclusions (somewhat redundant):
1. The lead (ai) car will always try and drive on the race.lp, or at least have one wheel in contact with the race.lp.
    Even if the ai is 3 or more car widths off the race.lp when it becomes the lead car, it will try and get to the race.lp as soon as
      clearance allows.
2. The most inside line is always the fastest way around. If the race.lp is close to the most inside racing line, that line will be the dominate
      line, period. This is true even if the other racing lines have far better grip.
    Regardless of the location of the race.lp, ai (other than the leader) will gravitate to the inside as far as the maxrace.lp will allow (that's as
      far to the left as possible with the right-side wheels on the line). No extra track.ini help is necessary, i.e. ai_drafting_distance, ai_dlat_pad,
      ai_squeeze_pcnt, to get the ai to run the fastest line. Using an ai_squeeze_pcnt value > 0.00 can actually inhibit the ai from staying on
      the fastest racing line because it allows the ai to vary from the race.lp when it's not actually beneficial to do so.
    To have any chance for the ai to pass successfully on the outside, the grip level for that line will need to be significantly greater than
      any other racing line. {I tried ridiculously high grip levels on the outside and ridiculously low grip levels on the inside line. At no time was  
      the outside line ever able to "dominate" the inside line.}
3. At least due in part to the NR2003 exe, ai overtaking on the outside is not possible, at least not on superspeedway tracks. Sad
      The ai are programmed to stay in control and on the track and are not as adversely affected by the track's physical limits as the Player is.

*** To keep the main narrative focused, I'm including the more detailed background information here.
More info:
1. This is always true for the Player because a Player is only constrained by the physical properties (of the car and the track) and physical
      boundaries. Ai, on the other hand, are "told what do do" by the exe, which evaluates and uses the information from the track
      configuration, the race.lp (and secondary lp's), the track.ini, the fast setup, the ai ratings, and the papy_ai.ini. The importance of
      each of these inputs is calculated by the exe in a way that does not always yield results that are comparable to a human driver, so
      in one instance the ai may follow a racing line that is nowhere near reality and are consequently way off pace, and in another instance
      they may follow a shorter path (within the contraints of the lp's) because the exe cannot translate the advantage of "more grip" into
      more speed.
2. In decreasing order of grip, based on the coefficient of friction, concrete (ever so slightly grippier than asphalt), asphalt, and paint.
      Asphalt and concrete are mostly inter-changeable. Paint has a definite slippery quality that is not noticeable in line widths, but it is a
         factor when it is used as a major racing surface (even with a high grip value). Use paint, with a relatively high grip value (e.g. 1.10),
         for the upper line since it tends to be a little more dirty there and there's less rubber build up. Use concrete for the middle (1.04-1.045)
         and asphalt on the lower line (1.03).
      [I'm not including "invisible", which is like concrete or asphalt in grip properties. While it is useful to have four property types to get the
         maximum grip variance, invisible does not produce any sound so you can't hear the tires sliding, which is critical feedback in the
         "sim driving" world.]
    Addendum 3-6-14:
      While I think the original order, aphalt (low), concrete (middle), paint (high), is good for the Player vs. Player in an online situation, this
         order does not translate as well for the ai when the objective is to get the high line ai as fast or faster than the middle or low line ai.
      Inhibited only by the lp's, the exe will choose to run the ai as low as possible. The ai aren't subject to the same physical limitations as
         the Player, so the ai can run efficiently on low grip. This makes it hard (nigh impossible) to make the high line as fast as the low line.
         Using the lowest coefficient of grip surface for the fastest racing line is more helpful in reducing ai efficiency than just a lower grip value.
3. The race.lp must run where you want the ai to predominately run. The ai won't run wide to the outside just because the minrace.lp allows
      them to, you have to force them to run a farther distance with the race.lp and keep them there with superior grip. This only works
      with ai close together, a single ai will follow the path of the race.lp.
    Addendum 3-6-14:
      I'm not disputing my original findings, I still don't think you'll get ai to "pass" on the outside, but you can get the high to run AS FAST as the
         the low line and maintain (or even retake in some instances) the lead in the high line. See: ai Racing the High Line



An added note on racing incidents at SS tracks.
The ai ratings are going to have an effect. It's not so much how fast they'll be, although a poorly rated ai in "aerodynamics" and "engine" will
   be decidedly slower, but it's more on how well they can stay out of trouble. In close racing, any ai with less than a 50 minimum rating in
   "track type" and "chassis" is going to cause problems at some point.
 


« Last Edit: February 17, 2015, 04:19:43 PM by fortine_oo » Logged
Pages: [1] Go Up Print 
« previous next »
Jump to:  

Powered by SMF 1.1.15 | SMF © 2006-2008, Simple Machines
Leviathan design by Bloc | XHTML | CSS