|
|
The complexity of the current championship is
|
 
too high (0, 0.00%)
|
perfect (1, 25.00%)
|
too low (3, 75.00%)
|
Total votes: 4.
|
|
Poll open for racer.
|
|
Start: 2009-10-26 00:27:51.
|
|
End: 2009-12-26 01:27:51.
|
Author: trbadmin | Created: 2009-10-26 00:27:45
Subject: Next Season, interest?
|
First I guess there is interest for another season, right?
I collected the "bullet points" from some threads regarding the next season. This should just serve as summary, not as list of actual changes, so don't panic;-) Here we go:
- different classes
- hopefully a class like the current one
- mistery tracks
- no manual tuning for racing lines
- all the same car, predefined standard setups to save time
- different championships
- "budget"
- different/unclear goals
- multiple classes in one race (like ALMS, GT Championships)
- track conditions (grip build up, dirt, weather, oil)
My plan for 2010 was:
- pro mode (would cars need rebalancing for this, please do test runs)
- more damage (I set it very quite low intentionally long time ago, when "wild west driving" was commonplace)
- some rules enforced by TORCS, penalties
- aerodynamics: reduce downforce when speed vector is not parallel to airflow
Some short comments to the above points:
- different classes (controversial):
I think there is not enough interest for this
- mistery tracks (controversial):
Would be possible, but generated tracks would be optically boring (at least without writing a great landscape/camera placement generator). Required changes: every player submits a seed (TRB), all seeds are taken to generate one seed to generate the track (TORCS).
- no manual tuning for racing lines (controversial):
For some too time consuming, for some "required". Anyway, I think people spending more time are potentially more successful here, so it does in the end not matter much where you spend the time. People could as well write "offline" optimizers if manual tuning is too tedious. I actually planned to ban manual tuning for 2010 and replace it with a training session, where the robots can learn by themselves, but this was controversial.
- all the same car, predefined standard setups to save time (not much comments/opposition):
For me this would be boring, I like to watch the races and it is more entertaining if people use different cars. If this is a desire this would be very easy to do.
- different championships (not controversial, but quite different ideas):
I would prefer to fill the actual grid first.
- "budget"
Nice idea, but this would be a different game, not easy to achieve.
- different/unclear goals (not controversial, but quite different ideas):
I think the goal is clearly stated in the "Home" link, is it outdated, should we really change the focus?
- multiple classes in one race (like ALMS, GT Championships)
I would prefer to fill the actual grid first. I think the best robot will still be the same when we switch classes/cars/etc. Would be quite some work o implement this on TRB/TORCS (but technically possible of course).
- track conditions (grip build up, dirt, weather, oil)
I would like to have this too, but with graphics, so this makes it quite some work to achieve. What we could have though is wind and track temperature, this has optically no effect.
|
|
|
Last Edited: 2009-10-26 00:29:54 by
trbadmin
|
Author: dz_duck | Created: 2009-10-26 17:05:42
Subject: Re: Next Season, interest?
|
Yes, I'm definitely interested in next season. There are a lot of possible changes to make; I think we should not change too many things in one season. What we have works really well, and changing too much may take away from that. That being said, here is my prioritized list of changes, based on the above list and a couple other things:
1. Realism: Anything to improve the physics and realism is a good thing. This would include going to pro instead of semipro, increasing damage (which is included in pro, I guess), improving aerodynamics, adding wind and track temperature (track surface changes too, but as you say, the benefit is not worth the work), and TIRE WEAR AND TEMPERATURE. If damage increases significantly, it would probably be a good idea to implement rules to penalize drivers that are at fault in collisions. These changes alone may be enough for next year.
2. Multiple classes in one race: I like kennet's idea at the bottom of the "Next season..." thread. Each team must have four drivers, two for TRB1 and two for another class. That way, we don't have to worry about finding enough entrants for a given class. To simplify the races, the second class must use the default car setups. Andrew's supercar set may be ideal for this class.
3. Mystery tracks: This could be a lot of fun. I don't mind if the track is not visually appealing; I usually ignore the scenery anyway. This would go hand-in-hand with implementing a training session.
Now, for things that I don't think we should change:
~Different championships: I agree that there is not enough interest.
~No manual tuning for racing line: I am of the opinion that robot authors should be able to make their drivers as fast as possible by any means necessary. Perhaps we could experiment with your idea of a training session with one or more mystery tracks, using the training session only on those tracks.
~All the same car, predefined standard setups to save time: While it may be interesting for a different championship, the different cars and setups are a big part of the fun for me.
~Budget: I agree, nice idea but not for what we have here.
~different/unclear goals: I think the goal on the home page is perfect: "The goal is to form a TORCS racing community and to have fun." For example, the goal is not to foster development of robots to race against humans.
As I stated above, I'd be happy to see only a few changes each year. The best changes for me would be #1, but I wouldn't mind seeing #2 as well.
|
|
|
Last Edited: 2009-10-26 17:05:42 by
dz_duck
|
Author: kilo | Created: 2009-10-27 11:52:06
Subject: Re: Next Season, interest?
|
I would like to enter for next year for sure.
The things I'd like to see changing:
-Pro mode, including more damage, rules and penalties
-Kennet's 2-class idea or call it multiple classes
-Some weather and tire wear
No-no-no's:
-Mystery track
-No manual tuning
-Same cars for everyone
cheers
kilo
|
|
|
Last Edited: 2009-10-27 11:52:06 by
kilo
|
Author: dummy | Created: 2009-10-27 18:41:48
Subject: Re: Next Season, interest?
|
I'm going with kilo on this one. The only thing I would not like too much is the weather, at least not for next season.
Danny
|
|
|
Last Edited: 2009-10-27 18:41:48 by
dummy
|
Author: k8962c | Created: 2009-10-27 19:23:00
Subject: Re: Next Season, interest?
|
- multiple classes in one race (like ALMS, GT Championships)
The problems are as above, the best robot will probably win both. And there's a strong incentive to not be kind to overlapping cars.
Why not have a set of default robots [like the infernos etc] driving the supercars.
This will give a spread of different slow traffic to overtake, thus testing the quality of the top Class racers overtaking.
I.e. we'll just race in the fastest class, but have some slower cars to navigate effectively as part of the race challenge.
|
|
|
Last Edited: 2009-10-27 19:23:00 by
k8962c
|
Author: dz_duck | Created: 2009-10-28 00:02:32
Subject: Re: Next Season, interest?
|
|
That's a good idea.
|
|
|
Last Edited: 2009-10-28 00:02:32 by
dz_duck
|
Author: dummy | Created: 2009-10-28 07:06:11
Subject: Re: Next Season, interest?
|
>there's a strong incentive to not be kind to overlapping cars
Not if there are penalties ;)
|
|
|
Last Edited: 2009-10-28 07:06:11 by
dummy
|
Author: kennet | Created: 2009-10-29 18:02:53
Subject: Re: Next Season, interest?
|
Hello.
That the best robot wins can't be much of a problem?
I fail to see why the incentive to block would change at all from what it's
now with my idea, could you elaborate?
For those who doesn't want to read my other post: My idea was to get flexible
robots that can race skillfully by having them drive both slow and fast cars
in a race. The robots now are not very good in a mixed field. If that is
desirable or not you have to decide.
The idea by k8962c makes them better at overtaking, it isn't a better or
worse idea, just a different one.
Regards
Kennet
|
|
|
Last Edited: 2009-10-29 18:05:25 by
kennet
|
Author: firechief | Created: 2009-10-30 01:47:08
Subject: Re: Next Season, interest?
|
Hi everyone,
FWIW I love Kennet's idea of having your robot drive one fast car and one slow. It would remove one of my pet hates - slipstreaming - as a tactic for two cars to work in tandem to post incredible laptimes & remove the opportunity for overtaking early in a race.
That said I wouldn't want to see the cars too far apart in quality. McLaren F1s versus VW Beetles would _not_ be fun, and would result in a ton of rear-end collisions.
Regarding pro mode and penalties, obviously the robots need to be told when a penalty is required, but also they'll need to add the ability to complete drive-throughs of the pit lane.
cheers
Andrew
|
|
|
Last Edited: 2009-10-30 01:47:08 by
firechief
|
Author: dz_duck | Created: 2009-11-02 05:10:21
Subject: Re: Next Season, interest?
|
"It would remove one of my pet hates - slipstreaming - as a tactic for two cars to work in tandem to post incredible laptimes & remove the opportunity for overtaking early in a race."
I happen to like slipstreaming... ;). But maybe there's a way to reduce it by adding some realism. How about adding engine cooling? Stay too close to a car in front, and your engine starts to overheat. I don't know how to model such a thing, but it would mean that your cars could only stay really close for so long before the trailing car would need to back off to cool his engine. That could also produce another setting to tweak: more cooling = more drag.
|
|
|
Last Edited: 2009-11-02 05:10:21 by
dz_duck
|
Author: firechief | Created: 2009-11-08 22:48:30
Subject: Re: Next Season, interest?
|
Actually I should have been more specific - slipstreaming where the car behind gains speed is great, I have no problem with it. What I don't like are two cars travelling close together where TORCS's physics make both cars, including the one at the front, faster. That's what allows teams to pull away at the front & makes overtaking early in a race very difficult, and usually disastrous for any car that attempts it. If you can get to the front of a grid and your robot's good at tailgating, you're 100% safe from challenges until you catch the back markers. Booorrring.
Making teams have two cars in different classes will separate them & so this won't be a problem anymore, with overtaking etc happening from the outset.
|
|
|
Last Edited: 2009-11-08 22:48:30 by
firechief
|
Author: dz_duck | Created: 2009-11-09 03:47:11
Subject: Re: Next Season, interest?
|
Right, I was actually addressing that issue as well. The teamwork slipstreaming requires the cars to stay very close, perhaps so close that the cooling of the tailing car is adversely affected. If the tailing car needs to back off to cool his engine, the leading car doesn't get as much of a speed boost.
Please note that the speed boost to the leading car is a real effect, not some bizarre, unrealistic consequence of "TORCS's physics." The tailing car reduces the drag due to the vacuum left by the leading car, since it fills that space. For the same reason, the rear wing of the leading car produces less downforce, and obviously the front wing of the tailing car produces less downforce as well.
|
|
|
Last Edited: 2009-11-09 03:47:11 by
dz_duck
|
Author: firechief | Created: 2009-11-09 05:42:03
Subject: Re: Next Season, interest?
|
Ok - that does make sense then. Do you know if TORCS models the rear/front wing loss of downforce as well? If it did, you'd have to imagine the front car would be inclined to fishtail a bit on corners, while the rear car could understeer and have trouble braking. I can't say I've noticed either behaviour, but its the kind of thing that could be hard to see unless the effect was significant.
Anyway, whatever encourages more overtaking will make the competition more interesting and entertaining. With polite backmarkers that move offline and slipstreaming drivers the contest risks becoming just a hot lap exhibition.
|
|
|
Last Edited: 2009-11-09 05:42:03 by
firechief
|
Author: dz_duck | Created: 2009-11-09 16:18:10
Subject: Re: Next Season, interest?
|
Taking a glance at the code (aero.cpp), it seems that drag is reduced for both cars, but downforce is not affected. Christos would be able to answer this for sure.
EDIT: I looked at it more carefully, and I'm sure that downforce is not affected.
|
|
|
Last Edited: 2009-11-10 02:33:27 by
dz_duck
|
Author: k8962c | Created: 2009-10-29 18:20:20
Subject: Re: Next Season, interest?
|
##I fail to see why the incentive to block would change at all from what it's
now with my idea, could you elaborate?##
Simply the number of times that they'd be passed. At the moment, the cars competing for the win are only very rarely overtaken, so there's no incentive to be cruel about it.
In the second class of car, they will be lapped continually, so the winner would be whoever's fastest despite being overtaken, i.e. the one who doesn't slow down much to let people pass.
It's also easy to say that we'd have rules to penalise naughty cars for not letting people pass, but it's a nightmare to enforce, even decide upon what's naughty. I don't think the car being lapped should do anything until the lapping car is at least alongside, wheras others thing they should get offline and blend out of the throttle.
|
|
|
Last Edited: 2009-10-29 18:20:20 by
k8962c
|
Author: kennet | Created: 2009-10-29 18:58:34
Subject: Re: Next Season, interest?
|
>In the second class of car, they will be lapped continually, so the winner would be whoever's fastest despite being overtaken
Very true if by fastest you mean shortest race time.
>the one who doesn't slow down much to let people pass.
For the moment this is not true. If you do not let a faster car pass safely it will
more likely than not crash into you while overtaking. You will have to make many
pitstops for repairs and you will not win your class.
Here is where the flexible part comes in: A given carset isn't always the fastest or
slowest one in a mixed field. It can be either or somewhere in the middle.
My idea is of course only interesting if TORCS is interested in racing with mixed
fields, otherwise just ignore it.
Safe overtaking: In the shipping world the rule is that the passed ship shall keep
its speed and course until the passing ship is well ahead. This would be an excellent
rule in robotracing too (yes, curves, I know, but keep your lane). Um, yes I have a
captains licence, for recreational purposes.
Regards
Kennet
|
|
|
Last Edited: 2009-10-29 19:02:05 by
kennet
|
Author: k8962c | Created: 2009-10-29 18:31:54
Subject: Re: Next Season, interest?
|
Also just saw:
#####
- mistery tracks (controversial):
Would be possible, but generated tracks would be optically boring (at least without writing a great landscape/camera placement generator). Required changes: every player submits a seed (TRB), all seeds are taken to generate one seed to generate the track (TORCS).
#####
I don't think that was the intention [to randomly create a new track]
I think it was intended to be a track chosen at random from a pool of selected tracks.
I.e. we select the calendar of races, but don't know which race will be next. This prevents any kind of manual tuning for the next race, and makes the test purely of the robot's skill, not the human tuner.
This also implies a predefined standard setup and the same car, and also would lead to no manual tuning of racelines, instead having a practice session on the host pc.
So the end result is reduced time by eliminating the human tuning stage, which is skillful, but detracts from the test of the robot.
Another point, I've not voted yet because the championship format is simple, however I find the time spent on tuning makes it quite complex. If there was less tuning I'd be able to spend more time improving the coding, instead of trying to get the setup & datafiles ready for the next race etc.
|
|
|
Last Edited: 2009-10-29 18:31:54 by
k8962c
|
Author: wdbee | Created: 2009-11-01 13:01:44
Subject: Re: Next Season, interest?
|
Hello all!
Next season? Yes!
But at the moment I have no time to discuss about it :-(
Real world work!
Will come back some weeks later.
Wolf-Dieter
|
|
|
Last Edited: 2009-11-01 13:01:44 by
wdbee
|
Author: dz_duck | Created: 2009-11-08 02:53:28
Subject: Re: Next Season, interest?
|
If you're talking about rebalancing the cars, perhaps we could use Haruna's new TRB1 car parameters (http://www.berniw.org/trb/forum/showthread.php?topicid=2680) or relax some restrictions on the current parameters. I haven't tried Haruna's parameters, and I don't know what logic is behind the current parameters, but the TRB1 parameters could be improved. In particular, I don't like the fact that for every race my ride height is set to the minimum, and one or both wings are usually set to the maximum. The TRB1 cars could go quite a bit faster if we relaxed some of the setup restrictions.
David
|
|
|
Last Edited: 2009-11-08 02:53:28 by
dz_duck
|
Author: firechief | Created: 2009-12-03 23:29:12
Subject: Re: Next Season, interest?
|
Hi Bernhard, just wondering if you've come to any conclusions based on the above discussions?
For me, the biggest change that'd renew my interest in the championship would be the adoption of a second (slower) car class. Each team still having two cars, but one in each class. The "supercars" set would seem ideal for this, as they've now been extensively tested for balance & driveability, plus they have fewer adjustable parameters so they wouldn't be an additional burden on the robot author to tune for each race.
The other suggested changes would be nice, but not have nearly as big an impact on the style of racing in TRB championships. Some, like pro mode & penalties, would require substantial coding work in the robots to recognise penalties & perform pit drive throughs.
Anyway, just interested to know what you're thinking & if there's any way I can help.
cheers
Andrew
|
|
|
Last Edited: 2009-12-03 23:29:12 by
firechief
|
|
|
