The TORCS Racing Board
Username: Password: Remember Me?
Lost Password Register
Author: firechief | Created: 2017-11-20 03:54:17
Subject: Race Review, Street-1
As per usual, my pictorial race review: https://imgur.com/a/ALblm

Quite an interesting race in many ways. The Tiger team continue to get faster & narrow the gap to USR, especially with Tiger's superior navigation past backmarkers. The Droids were also in the mix early on (and quite fast too) but lost places due to poor collision avoidance.

Although USR has the championship in the bag now, next year is looking like it'll be very competitive indeed.
Last Edited: 2017-11-20 03:54:17 by firechief
    Author: dummy | Created: 2017-11-20 07:43:47
    Subject: Re: Race Review, Street-1
    Congratulations for winning the championship again! The dominance is simply astonishing, especially in consideration of the names that were participating this season. It's now fair to say that you are the champion of champions ;)
    Last Edited: 2017-11-20 07:43:47 by dummy
    Author: phi | Created: 2017-11-20 09:26:29
    Subject: Re: Race Review, Street-1
    Congratulations for the championship.

    ------

    In my race I noticed the ease and security of the Mouse & Tiger in overtaking, particularly with regard to the Klins: they show themselves, Klin opens, and they follow without hesitation, only stoping when they actually see that they have no chance of success; usually, when they show up, they pass. On the other hand, the other cars are hesitant and waste time with this, often leading to touches and losing opportunities: they appear on one side and Klin opens that side, but they decide to go to the other and Klin opens that other side: this ballet happens even in the straight, as in the case of the great straight in Street-1 (counting with the straight and the fast right that precedes it); and for the USR this ballet was aggravated by the unfortunate instability it brought to this race.

    On the other hand, in my race Klin was overlapen several times in the area of great technicality of the circuit, and the last left-right of this area was the most problematic due to the excessive impetuosity that the Klin showed there, sometimes leading to accidents of the Klin's own responsibility; but these situations were fortunately rather rare (twice, I believe), and fortunately damaging the Klin, more than the other cars.
    Last Edited: 2017-11-20 09:26:29 by phi
    Author: timfoden | Created: 2017-11-20 20:43:44
    Subject: Re: Race Review, Street-1
    Congratulations for securing the win with one race to go. Personally I thought the writing was on the wall for this by only halfway through the season.

    At that point my mission instead became to try to secure 2nd place with the Tigers... which as you noted is now looking quite probable, provided the Tigers don't do anything stupid in the last race. :) Hopefully I'll have time to come up with some speed improvements before next year.

    And racing on this track also reminded me of the time when I was trying to get my robot Sparky to drive around Adelaide in the old RARS competition. It made me think that it'd be interesting to see how well a Sparky based bot would do in TORCS. It wouldn't be fast enough to trouble the best TORCS robots, but Sparky was always a good passer in RARS and still respectable speed-wise (it was the fastest non-clothoid based bot.) I make no promises, but I'll certainly think about it.

    Cheers, Tim.
    Last Edited: 2017-11-20 20:46:05 by timfoden
      Author: firechief | Created: 2017-11-21 08:14:00
      Subject: Re: Race Review, Street-1
      Hey Tim, I'd be very interested to see Sparky in action! Variety's always a good thing, and there'd be a lot of kudos to be gained from beating clothoid-based robots with it.

      I'll certainly run 2 robots again next year, but am not sure which way I'll go about it. One will obviously be USR based, but I'd like to explore other ideas with the second ... my adventure with the Mouse code has come to an end, so I'm not sure which way I'd go here. I've looked at the source for a bunch of old RARS robots, but I don't really understand how that API worked, it doesn't seem to map very well to TORCS's, and I struggle to see where/how the robots are providing their driving inputs to the simulation.

      Alternately, I could try to copy W-D's approach and develop a branch of USR that doesn't need any human guidelines, calculating its own setup based on its analysis of track characteristics and not needing to be told which areas of the track are dangerous & need to be given more space, as I currently have to do with my bots.

      Chances are I'll have too little time and probably just end up running 2 USR clones, but at least one of them using one of the less competitive TRB cars. If I feel particularly masochistic, I might even go with car1 which in my testing is about 2 seconds a lap slower than everything else - shame on any clothoid robot that finishes behind it at the end of the season!
      Last Edited: 2017-11-21 08:17:14 by firechief
      Author: dummy | Created: 2017-11-21 09:59:59
      Subject: Re: Race Review, Street-1
      >At that point my mission instead became to try to secure 2nd place

      I just discovered a serious one line bug in my collision braking code introduced six races ago. This explains at least some of Droids decline. Well, well... there's always hope to do it better the next year :)
      Last Edited: 2017-11-21 09:59:59 by dummy