The TORCS Racing Board
Username: Password: Remember Me?
Lost Password Register
Author: firechief | Created: 2017-08-01 03:35:12
Subject: I knew I was overdue for a screw-up :\
In USR's car3-trb1/race/hidden-valley.xml is this:
<attnum name="chkpitstop" val="1"/>

This makes the cars want to pit, every single lap. I turned it on for testing of course, hadn't realized that I'd used that version of the XML for the upload. That's the problem with starting a new job & having too little time for TRB...

Oh well.

If you want to see a proper race, you could change that value to 0 - but if you'd rather see a comical race instead then by all means keep it at 1 :)
Last Edited: 2017-08-01 03:35:12 by firechief
    Author: dummy | Created: 2017-08-01 05:49:57
    Subject: Re: I knew I was overdue for a screw-up :\
    Hmmm.. let me think. What is better to make the championship interesting again? Leave it or fix it? .. I will fix it, it's all about good racing after all ;)
    Last Edited: 2017-08-01 05:51:03 by dummy
      Author: firechief | Created: 2017-08-01 06:04:13
      Subject: Re: I knew I was overdue for a screw-up :\
      Thanks - that's most kind of you! I'll be sure to return the favour some day, not that you ever screw up like I do ;)

      I ran one lap, which is all I had time for (and which is how I noticed the pit problem) ... and I don't think I'm giving anything away when I say that the wheel to wheel action looked EXCEPTIONAL on this track. Low friction surface + high speed corners, seems to be the perfect recipe for TRB races.

      Its made even more interesting by the varying approaches we've taken to setup. Mid-downforce for my cars, yours are very low downforce, W-D went the other direction with max wings, really mixes things up.
      Last Edited: 2017-08-01 06:04:13 by firechief
        Author: dummy | Created: 2017-08-01 06:38:00
        Subject: Re: I knew I was overdue for a screw-up :\
        Looking forward to this one.

        >Low friction surface
        Oh, that's why I had trouble to stop my cars in the pits.
        Last Edited: 2017-08-01 06:38:00 by dummy
          Author: firechief | Created: 2017-08-01 07:02:09
          Subject: Re: I knew I was overdue for a screw-up :\
          Yep most road tracks have a friction of 1.08 (e-track-6) up to 1.2 (alpine-1). Hidden Valley's is 1.04.

          Pits however are usually 1.0, and they are on this track as well. Not sure why your car would have trouble stopping - maybe its a combination of your lack of wing angle and the narrow pit lane?
          Last Edited: 2017-08-01 07:02:09 by firechief
        Author: wdbee | Created: 2017-08-01 11:05:13
        Subject: Re: I knew I was overdue for a screw-up :\
        If car6 would be able to drive 300 km/h with max downforce, it would win the championship ;)

        In fact the used wing angles are

        <params name="car6_trb1">
        <section name="Front Wing">
        <attnum name="angle" unit="deg" val="3.43162"/>
        </section>

        <section name="Rear Wing">
        <attnum name="angle" unit="deg" val="3.43454"/>
        </section>

        Compared to the droids

        <section name="Front Wing">
        <attnum name="angle" unit="deg" min="0" max="12" val="6"/>
        </section>

        <section name="Rear Wing">
        <attnum name="angle" unit="deg" min="0" max="17" val="3"/>
        </section>

        it is similar, but they drive 317 km/h, so they just pass at the straight.
        And with max wings Los Lobos would need more pit stops.

        Last Edited: 2017-08-01 11:05:13 by wdbee
          Author: firechief | Created: 2017-08-01 11:38:45
          Subject: Re: I knew I was overdue for a screw-up :\
          Hmm. That's odd.

          In the wdbee_2016 tarball I downloaded for this race, in hidden-valley-race.xml it says this:-

          <section name="Front Wing">
          <attnum name="angle" min="0" max="12" unit="deg" val="12"/>
          </section>

          <section name="Rear Wing">
          <attnum name="angle" min="0" max="17" unit="deg" val="17"/>
          </section>
          Last Edited: 2017-08-01 11:38:45 by firechief
            Author: wdbee | Created: 2017-08-01 14:32:51
            Subject: Re: I knew I was overdue for a screw-up :\
            You are completely right, but if you look to the private section there is the balance and the downforce defined which is used by the bot (and it results in the angles I posted before).

            Last Edited: 2017-08-01 14:32:51 by wdbee
              Author: firechief | Created: 2017-08-01 14:56:33
              Subject: Re: I knew I was overdue for a screw-up :\
              Ah that explains things, thanks :)
              Last Edited: 2017-08-01 14:56:33 by firechief
              Author: timfoden | Created: 2017-08-01 15:45:59
              Subject: Re: I knew I was overdue for a screw-up :\
              Yes, this has confused me in the past when looking at settings files for your cars :) I did eventually figure it out.

              I too have an optimisation tool (which I assume you do) that I run across many of the .xml file settings, but it really doesn't seem to work at all well for the wing angles. As you can see in this race, the starting angles for my car were 10.5, 7 and the finishing ones are 10.595, 7.105... not much different at all, and surely not really optimal. I've pondered changing the way the wing angles are optimised but haven't yet got around to trying any other way. Changing the values as a pair as you appear to seems like a good idea to try.

              Cheers, Tim.

              Last Edited: 2017-08-01 15:45:59 by timfoden
                Author: firechief | Created: 2017-08-01 15:58:32
                Subject: Re: I knew I was overdue for a screw-up :\
                I don't have an optimisation program yet (its on a very long todo list), but yes changing them as a pair does seem a good idea. I remember reading somewhere that the main purpose of the front wing is to balance the downforce generated by the back wing. Hence it makes sense to find the balance that best suits your car, and increment/decrement them both together - if you change them individually then a lot of testing will be wasted on configurations where the car's fighting understeer or oversteer.
                Last Edited: 2017-08-01 15:59:01 by firechief
                  Author: fa1 | Created: 2017-08-01 16:50:18
                  Subject: Re: I knew I was overdue for a screw-up :\
                  We don't have any optimisation tool on the wing angles either but we we have some code that calculates the needed rearwing angle to a preset frontwing angle in order to have a balanced car.
                  Last Edited: 2017-08-01 16:50:18 by fa1
                    Author: wdbee | Created: 2017-08-01 17:34:50
                    Subject: Re: I knew I was overdue for a screw-up :\
                    Hi Fabian, yes, that is a part of it. With this code you can adjust the rear wing to the front wing, once the decision about the needed downforce is done and the corresponding front wing angle is known, or vice versa.


                    Last Edited: 2017-08-01 17:34:50 by wdbee
                Author: wdbee | Created: 2017-08-01 17:03:35
                Subject: Re: I knew I was overdue for a screw-up :\
                Yes Tim, I normally use a separate optimization tool for the setups. I also see this cases, where the optimization is stuck at a local instead of the global min (Lap time). The code used here is remaining from the wdbee_robotics, a bot that autonomously did setup the wings and the differential ratio depending on the estimated lap times and the calculated top speed. Using balance and downforce as parameters helped to adjust the needed downforce but keep the car balance work.

                As side effect it did block some of the setup copy actions ;)

                The setup optimization result is quite good for qualifying, but the car6 has a bad weight distribution. To drive fast it needs a lot of brake pressure at the rear, but in traffic this often makes the driver loose the grip at this axle :(
                So it needs too much downforce at the rear to be fast at straights and together with the bad aero parameters, area too large, cw is realistic compared to the wind tunnel results, but all the other cars had cw values much higher in the wind tunnel resuts than in the car setups :(, it increases the fuel consumption. This results in not be able to drive the car as fast as possible or use additional pit stops.

                For this race I finished the rework of the code to allow to handle the grip for the different drivers separately. Before there was one racingline for both drivers only. This allowed a common use of learned data while race. But with the wheel model changed it did not work any longer. But it took too much time to realize it and so I had to do the setups manually this stime.

                From the result I have to say that a manual setup is still better, at least if parameters depending on the track position are allowed.




                Last Edited: 2017-08-01 17:12:20 by wdbee
                  Author: firechief | Created: 2017-08-01 17:20:12
                  Subject: Re: I knew I was overdue for a screw-up :\
                  > From the result I have to say that a manual setup is still better, at least if parameters depending on the track position are allowed.

                  Yes, manual setup is still superior but the gap is narrowing. The work you guys are doing on optimisation is, I think, going to soon reach a point where it equals & perhaps betters manual setups. I hope so, it'll certainly make the racing closer & the outcome less predictable.
                  Last Edited: 2017-08-01 17:20:12 by firechief
    Author: phi | Created: 2017-08-01 08:10:22
    Subject: Re: I knew I was overdue for a screw-up :\
    It's already fixed.


    EDIT: I also canceled the "brake debug" writing in the console.
    Last Edited: 2017-08-01 11:33:50 by phi
      Author: firechief | Created: 2017-08-01 11:40:34
      Subject: Re: I knew I was overdue for a screw-up :\
      Thanks for that, I'll make sure to remove that #define for the next race. Got to get myself better organized...
      Last Edited: 2017-08-01 11:40:34 by firechief
        Author: timfoden | Created: 2017-08-01 12:01:21
        Subject: Re: I knew I was overdue for a screw-up :\
        I have a #define that is only defined for the project I use here on my development machine, and is automatically stripped from the project file when creating the bundle to submit. This is by a Python script that does the whole packaging into a tar-gzip job. This define is used to control all the debug output, and thus guarantee that it's always switched off for everyone else by default. This python script also checks that the most up-to-date version of the settings files for the race are bundled (by comparing dates in various directories), and also can generate the bundle for Tiger directly from the Mouse source (so at the moment they are still identical robots... still struggling to make Mouse fast enough though, so no spare time for doing the extra things I wanted to try for Tiger yet.) The script could easily be used to check or edit the settings file to make sure certain settings are not set, which would solve problems like this for you.

        Cheers, Tim.
        Last Edited: 2017-08-01 12:02:18 by timfoden
          Author: firechief | Created: 2017-08-01 12:10:32
          Subject: Re: I knew I was overdue for a screw-up :\
          That's actually an excellent idea! I'll write a similar script (my python could do with some practice) this weekend, hopefully will prevent problems like this in future - thanks!
          Last Edited: 2017-08-01 12:10:32 by firechief
            Author: timfoden | Created: 2017-08-01 15:22:07
            Subject: Re: I knew I was overdue for a screw-up :\
            No problem. I'll send you a copy of mine when I get back home if you like, just to get you started or to use as a reference.
            Last Edited: 2017-08-01 15:22:29 by timfoden
              Author: firechief | Created: 2017-08-01 15:25:30
              Subject: Re: I knew I was overdue for a screw-up :\
              Sure ok, thanks :)
              Last Edited: 2017-08-01 15:25:30 by firechief