The TORCS Racing Board
Username: Password: Remember Me?
Lost Password Register
Author: berniw | Created: 2017-01-28 09:37:50
Subject: Time to set up 2017 soon:-)
Hi all

Which tracks would you like to race 2017 in the real championship, please give a list. Conditions for candidate tracks are as usual: The must have at least 15 pits.

Kind regards

Last Edited: 2017-01-28 09:37:50 by berniw
    Author: firechief | Created: 2017-01-28 11:22:05
    Subject: Re: Time to set up 2017 soon:-)
    Ones I'd really like to see:-

    Alpine-1 or Alpine-2, both are interesting
    E-track-1, because I haven't seen it for a long time and it always promoted good racing
    E-track-3, as I've never raced on this variant of the track
    E-track-4, as its been way too long since I last raced there, and imho we need more fast tracks to counterbalance all the twisty ones
    Street-1, another track that promotes good racing, and that I haven't seen for too long
    Dirt-3 - if we have to have a dusty-hellhole track I'd like a change from dirt-2 ;)

    Tracks that are very cool but tend to result in processions, though this would change if there was some form of reverse grid or a "no blue flags" rule introduced:-
    e-track-2, e-track-6, wheel-1

    Tracks that are too tight & tortuous for any kind of safe overtaking, or too short to be interesting:-
    aalborg, cg-track-1, ruudskogen

    However, I should mention that if this year has the same cars and the same rules as previous years, where everything pretty much depends on setting fast laptimes, I don't know if I'll be competing. The main robots (USR/Axiom/DanDroid/Wdbee) are all reaching the point of diminishing returns with laptime, and I think something needs to be done to focus attention on other aspects of robot development in order to shake things up & keep it interesting.

    So far we have 4 possibilities. First is a reverse grid of some kind, several ideas for which have been suggested, though W-D has a strong objection in that some of the backmarkers, including berniw_2004, are quite dangerous to overtake - and on twisty circuits a reverse grid could be a nightmare. Secondly, a "no blue flags" rule, where still the faster cars have to find a way to get past those irksome slow drivers, but at least everyone isn't queued up behind them from the outset. Third is a learning-based championship, however the proposal for that was not accepted for a variety of good reasons. And fourth, switching to a single-car tournament, preferably using the car2-ow1 that W-D and I developed last year.

    Just staying with the current format with the trb1 carset ... we've been there & done that for too long now in my opinion, its getting stale.
    Last Edited: 2017-01-28 12:44:02 by firechief
      Author: berniw | Created: 2017-01-28 13:49:09
      Subject: Re: Time to set up 2017 soon:-)
      What about fuel consumption times 3 and pro mode, enough entertainment?

      Right now im implementing tire wear btw., would that be interesting to enable?

      Kind regards

      Last Edited: 2017-01-28 13:50:24 by berniw
        Author: firechief | Created: 2017-01-28 14:23:26
        Subject: Re: Time to set up 2017 soon:-)
        Pro mode + tire wear, especially tire wear - that'd be a whole new challenge :)

        Even better if you took it a step further and allowed robots to choose two types of compound. Fuel doesn't really worry me, though a faster rate of burn could even things up a little as at the moment some cars are at a handicap on some tracks.

        Speaking of cars, why not give the car1 a bit of a boost? By any measure its by far the worst of the trb1 set at the moment - poor acceleration, unimpressive top speed, bad cornering...

        Edit: Thinking about it, tyre wear (especially with a choice of compounds) would make things different enough that I'd be enthusiastic to go around again. Pro mode & fuel changes, while they're a step in the right direction they don't change the current dynamic of "lap as fast as possible and you'll win". Whereas tyre wear & strategy, depending on how wear is implemented of course, could add a whole new layer of strategy and force improvements to how robots drive the cars. Its no good if a robot laps a second faster than everyone else if it burns through tyres twice as fast - that kind of thing.
        Last Edited: 2017-01-29 07:31:29 by firechief
        Author: dummy | Created: 2017-01-28 16:14:51
        Subject: Re: Time to set up 2017 soon:-)
        Entertainment is always good ;)
        Last Edited: 2017-01-28 16:14:51 by dummy
        Author: firechief | Created: 2017-02-01 12:51:20
        Subject: Re: Time to set up 2017 soon:-)
        Just out of interest Bernhard, and because I'm too impatient to wait for the code to be finished, I have a few questions about how you're implementing tyre wear...

        1. Will there be a linear loss of tyre Mu, or more like a curve with a cliff at the end?

        2. Will tyres wear faster if abused by locking up the brakes, burnouts when accelerating, drifting or understeering through corners?

        3. I assume that robots will be able to access the current Mu of each tyre at any time?

        4. Will robots need to specify replacement of tyres when pitting, or will they be changed automatically?

        5. Should I change the name of the USR Daneel robot, seeing its too close to Daniel Ricciardo? I don't want his tyres to be lost somewhere out the back of the garage when he pits ;)

        Also I was thinking about asking if you'd think about a maximum fuel limit for the race in order to force the robots to conserve fuel if they want to finish? Or would that be something for future years? It'd only be fair if each car had the same fuel consumption, or if the thirstier cars were balanced by having higher performance when running at full throttle.
        Last Edited: 2017-02-01 12:55:07 by firechief
          Author: berniw | Created: 2017-02-01 21:15:14
          Subject: Re: Time to set up 2017 soon:-)

          I am currently still experimenting, but mainly the model will be based on recent research, lets first talk about the

          1 temperature model:
          - There will be an environmental temperature
          - The heat in the tire is generated by the energy wasted in rubber deformation (-> so as the tire wears down it produces less heat because of the thinner rubber layer and is more subject of graining, keyword: hysteresis)
          - More deformation and more dynamic friction will lead to more energy (heat)
          - The tire is cooled by convection (radiation and surface contact does not contribute enough to take it into account for "ordinary" TORCS)
          - Undecided: The heat transfer through the tire itself takes some time, maybe I will include or neglect this in the model, I have to do some calculations and experiments about that first (could be important for burn outs, there the outer surface gets cooked, but the tire inner temperature/pressure does not go up in the same pace... we will see)

          2 wear:
          - From the tire gas temperature you can calculate the pressure, from that an approximation of the tire contact patch
          - The stress in the rubber depends on the patch size, which will be the major factor for wear

          3 grip:
          - The grip will be dependent on the temperature and on the surface condition, so if you have a half worn down tire and the temperature and surface is still perfect, it will have still perfect grip (but you will probably not able to achieve this because of the temperature model)
          - There will be probably be a simulation of graining which can hit the grip hard
          - If the tire is worn down the grip will drop dramatically (transition from rubber to kevlar) or the car is taken out of the race, to decide).

          4 pit stops:
          - Tire change might "cost" significant extra time

          What I will leave out for sure in the first implementation:
          - Damage localization (discretizing the arc e.g. in 64 segments to simulate flatspots)
          - Chemical processes (run in tire 24h before use for optimal grip and such)
          - Chemical processes during the ride (vulcanisation of the inner layers before they get exposed some milage later)
          - Instant tire destruction because of too high stress
          - Dynamic contact patch deformation
          - Discretization over the tire width
          - Various compounds
          - Just changing a subset of the tires

          So to your points:
          1. No such simple model in place, it depends on your setup and driving style.
          2. Wear is a function of material stress in the contact patch. Higher stress -> higher wear, no stress -> no wear
          3. Probably not, otherwise it would be boring, wouldn't it? But I plan to make wear and eventually other data accessible, but you can guess on the grip regarding your trajectory error
          4. I plan that you have the choice to change or not, but no different compounds in the first implementation. Changed tires/tires at the start will be cold.

          Eventual Addons:
          - Environment temperature: The TRB will randomly choose the temperature for an event, it will not be known beforehand.
          - Eventually the air density of the aerodynamics will also take into account the temperature

          For final 1.3.8, but probably disabled in trb:
          - Dynamic track: I plan to put a grid of 0.2*0.2 Meter patches over the whole asphalt to record the tire contact and rub in the track (or make it "dirty"). This will be then taken into account for the friction
          - Wind

          Additional point: I plan to have the initial tire life for manual test runs (driven by me in pro mode somewhere around 200km and 250km).

          Kind regards

          Last Edited: 2017-02-01 22:15:02 by berniw
            Author: firechief | Created: 2017-02-02 00:58:24
            Subject: Re: Time to set up 2017 soon:-)
            Ah ok, you're starting out with an extremely in-depth modelling then!

            There are some things that I'm a bit concerned about. I'll start with some questions:-

            1. Will tyres be pre-warmed at the start of the race, and for new sets fitted at pitstops?

            2. Will robots be able to see the temperature in each tyre at all times? This would be realistic as its the kind of info that engineers constantly monitor and notify the drivers about.

            3. Will there be any information available about the percentage of wear to each tyre, and degree of damage they may have suffered due to graining or blistering (if implemented)?

            Speaking of graining, which as I understand it is abrasion to the surface of a tyre due to repetitive over-exertion when its too cold on a rough surface (smooth tracks don't tend to do this) - as distinct from blisters which form when tyres are too hot and the rubber starts to de-vulcanize. Graining also only occurs if the tyres are overworked in one direction - if they're sliding around in all directions then the abrasive pattern doesn't get a chance to form, but obviously the tyre wears down faster.

            This is distinct from the graining on Formula 1 (Pirelli) tyres from 2014-2016, which appeared at higher temperatures as well as low, and regardless of the direction of the tyre abuse. This was caused by plastic elements added to the rubber compound, which were liable to melt if the tyre wasn't maintained at the ideal temperature - and once they melted they deformed the tyre's surface resulting in permanent loss of grip. That's how Pirelli tried to achieve the goal of degrading tyres, but imho it was a poor solution as it didn't allow drivers to push hard & penalized them for following opponents too closely. Pirelli are apparently abandoning this concept in 2017 and returning to racier tyres.

            I get the feeling that in implementing graining you might be trying for too much, too soon - and I do hope you're not taking Pirelli's artificial 2014-2016 approach as a model.

            I also query whether a tyre should provide the same grip level halfway through its lifespan provided the temperature's ok. In my understanding, there's always a gradual loss of grip throughout a racing tyre's lifespan even if they're treated perfectly. If they did steadily lose grip, it'd introduce the strategic element of whether a robot should limp home with aging tyres or pit for new ones & put in a series of hot laps, though that would require only a few seconds penalty for new tyres in the pits. If the delay is 10 seconds or more, in addition to the time lost through pitting, then the strategic considerations might be ruled out.

            Another benefit of a steady decline is that the robot would be able to calculate the expected remaining lifespan rather than just monitor the tyre state and have to perform an unplanned panic pitstop when the grip suddenly disappears. From a robot author's point of view, allowing planning & strategy is better than introducing chaos.

            On another point, I think once the tyre has worn to the kevlar the car should still be driveable, as it is IRL at least until there's a puncture, with the robot having access to data telling it that the tread's gone. It'd be up to the robot then to manage to limp home to the pits - if they couldn't do that they'd probably end up off-road somewhere and being removed in a more natural manner. I think this would be preferable to an immediate removal once the tread's gone.

            So to summarize this lengthy & much-edited post, I asked if the robots would have access to tyre Mu information, and your response was no, as that'd make things boring and that we'd have to look at trajectory error.

            For myself, I'd really prefer it if there were temperature and tyre-wear & tyre-damage information available, as its the kind of thing drivers would know about IRL. They wouldn't have to suddenly realize there's a problem when they find their cars sliding off the track on corners, and then take a blind guess at how severe the problem might be.

            I guess what I'm saying is that, when I agreed that tyre wear could make things interesting, I was thinking in terms of authors working out how to make the robots drive in order to minimize wear while maximizing speed, as well as adjusting to a deteriorating grip level - and in addition to that, planning pitstops and tyre changes from a strategic point of view. That, to me, sounds like fun.

            If instead, "tyre wear" means the robots having to constantly guess how much grip's in the tyres from how the cars are behaving, I'm not sure that strikes me as interesting or fun at all. It could easily result in cars carefully trundling around, not pushing to the limits (as the limits are unknown) and not using all of the track - and if cars do push hard they could find themselves suddenly off-track & into a wall, which would be chaotic to say the least. All of which sounds not like a simulation of racing but of blind exploration similar in some ways to Daniele's championship.

            If I've gotten the above wrong in any way, please set me straight :) Its up to you where you want to take TORCS & the TRB Bernhard, all I can do is give you feedback from the point of view of a potential participant in the championship.
            Last Edited: 2017-02-02 15:42:05 by firechief
              Author: berniw | Created: 2017-02-02 20:45:53
              Subject: Re: Time to set up 2017 soon:-)
              Hi Andrew

              - Fresh tires will be cold (environment temperature)

              - I just said "no" to the mu, as stated you will see the wear and more, more means at least temperature, pressure and graining level. I will not implement blistering at this time, the chemistry department might follow later.

              - Yes

              - Too cold tire with too much stress will start to grain, which will then not be able to gain temperature and grain more, I will do my best to simulate this death spiral;-) But there is a way out of it: do some slower laps to gently drive of the grained layer, then gently increase pace to get temperature in the tire without graining. Of course this will not be possible anymore if the tire is worn down too much (if the model turns out as I want it to), then you have to live with it or change tires.

              Grip level:
              - As far as my understanding goes, tires do NOT loose grip through wear, this is just correlation, not causality. The problem is that the worn down tire will not be able to produce the required heat level, and because it is colder it will loose grip (there is a heat <-> elasticity correlation which is the cause for that). But if you cut out a piece of the half worn tire and put it on a material test bench, it will still be fine (if not overheated or something).

              Remember the formula one season where drivers went out for up to 6 qualifying laps in a row and the last laps were the fastest? So it is not that simple as "new == fast", "worn == slow".

              Steadily decline:
              So from some point on you will start losing steadily grip, but due to temperature drop, which you can not avoid. When this happens is dependent on the car setup (toe in, suspension) and driving style (clean, under/oversteer). The most difficult problem here will be balance the wear over all 4 wheels, e.g. e-track-3 is designed to kill the left front tire.

              Kind regards


              Last Edited: 2017-02-02 20:50:55 by berniw
                Author: firechief | Created: 2017-02-02 23:35:16
                Subject: Re: Time to set up 2017 soon:-)
                Ok that sounds promising. Thanks for answering my questions, and I very much look forward to testing this :)
                Last Edited: 2017-02-02 23:35:16 by firechief
                Author: wdbee | Created: 2017-02-04 12:00:47
                Subject: Re: Time to set up 2017 soon:-)
                > The most difficult problem here will be balance the wear over all 4 wheels
                As we did not use it in the past, the balance of the cars is not checked regarding this feature.

                So I would like to enable the weight distribution (front-rear) and the ground effect distribution (front-rear) as parameter with reasonable limits. This way the championships will help to fix issues of the existing setups regarding the new tyre features.



                Last Edited: 2017-02-04 12:01:41 by wdbee
                  Author: berniw | Created: 2017-02-05 14:51:46
                  Subject: Re: Time to set up 2017 soon:-)

                  I did not mean that, I mean the balancing regarding driving style and car setup, so the pure diver/engineer aspect. Of course I will do my best to set the "base model" properly, but special track features can then still be challenge, e.g. in reals racing this happens all the time.

                  Kind regards

                  Last Edited: 2017-02-05 14:51:46 by berniw
    Author: dummy | Created: 2017-01-28 18:24:15
    Subject: Re: Time to set up 2017 soon:-)
    Last Edited: 2017-01-28 18:24:15 by dummy
    Author: jspenn | Created: 2017-01-30 15:29:42
    Subject: Re: Time to set up 2017 soon:-)
    I like Spring and Street 1.
    Last Edited: 2017-01-30 15:29:42 by jspenn
      Author: firechief | Created: 2017-01-30 16:28:25
      Subject: Re: Time to set up 2017 soon:-)
      Spring is awesome, but unfortunately its too long for racing purposes. Cars will just take off in qualifying order, spread out, and hardly ever see another car. That's why this track is rarely ever used.
      Last Edited: 2017-01-30 16:28:25 by firechief