Bernhard Wymann wrote:
> hi all
> i would like also to put some comments/thoughts to the torcs future.
> > Simulation:
> > -----------
> > Tires:
> > - different types of tires for different surfaces
> > - more stability in tire reaction
> > - tire temperature and wear
> > - tire pressure
> > - camber
> i think the only things worth to implement are tire temperature and
> wear. why that? so, already now, you have a lot of parameters on the car
> you can change, and if we add more it will become very hard to find a
> near optimal setup. why? think of all free car parameters (say N) as
> parameters on an N-dimensional function, and consider the optimal setup
> as the minimum or maximum of this function. while it's easy to get
> minima or maxima for 2 dimensions it becomes REALLY difficult in higher
> dimensions, because for eg. N=10 you have one minimum, one maximum and
> if you have a "bad" function 2^10-2 = 1022 saddle points, and a huge
> number of local minima/maxima. so, the chance to get the optimal setup
> becomes really small, because we already have much more than 10 parameters.
It is possible by the use of the categories definitions to limit the parameters
modifications, so races can be restrained to specific categories allowing
very few parameters change.
But, by allowing the parameters modifications it is possible to play on the
two sides, the robot driving abilities and the settings of the cars, that's
two ways of improving the lap time.
I know that I have a HUGE documentation effort to produce about how to
configure the cars and every other points relative to the robots.
> > Transmission:
> > - better equations for differentials and better support of 4WD
> didn't have a look on them, no comment.
It's just that I'm not happy with the current implementation.
> > Suspensions:
> > - limit the suspension effect (no more jumps on michigan...)
> didn't have a look on it, no comment. it's just a bit funny if the cars
> fly around;-)
> > Aero:
> > - better configuration (try to find other name for wings...)
> i have a feeling that also the aerodynamics-equations need improvment:
> if you put the car on the front axis say 50 mm above the ground, and on
> the rear at 150 mm there should be kind of a "diffusor" effect, i miss that.
This is also a point to improve, I agree.
> > Damages:
> > - localisation of dammages
> > - effects on car 3D body
> > - effects on wheels, engine, ...
> i think that's very cool, but not really urgent.
> > Weather:
> > - rain, snow, ice, temperature, wind...
> this is also nice, but will be not too easy to handle with the bots. i
> think on changing conditions during the race. hmm, will the be a module
> called libracestewards.so (simulating the stewards, responsible for the
> rules and safety), which will send the pace car on the track, when the
> rain becomes too hard or too much wrecks are on the race track? (just as
> > Tracks:
> > - temperature, adherence modification on the cars path
> > - borders
> > - not only circuit but also open roads
> > - documentation on the track format (a tutorial ?)
> > Races:
> > ------
> > - flag management for the robots (blue, yellow, red, black...)
> > - race rules (like speed limit in the pit lane...)
> > - pace car
> > - different start procedures (warming lap, start from rest or behind pace car)
> > - championship modes and season management
> > - qualif sessions
> > - results and record saving
> > Teams:
> > ------
> > - group robots in teams
> > - play the "team manager":
> > The team manager can send orders to the robots
> > and can receive information messages from the robots.
> > you can take decisions like slow down / pit / don't pit...
> > Robots:
> > -------
> > - different classes of robots: from beginner to "wizard" (thx berniw)
> > to allow beginners to write new robots
> > - Communication with "team manager" (see network races)
> > - tutorials
> hmm, all the gameplay features, which would make the games nice for
> humans, will make it very hard for the bots... i think you should really
> think about, if you want to go towards a racing game for humans, where
> the bots are "faked" (like in commercial games), or towards a bot racing
> game, where the environment stays quite easy. aginst going toward a
> "normal" racing game is IMHO, there is already a great one, look at
> http://www.marketgraph.nl/gallery/racer. looks ok, doesn't it?
> > Telemetry:
> > ----------
> > - more easily customizable telemetry and more integrated to the program,
> > it currently uses gnuplot too draw the figures
> this would be of high value.
> > Network and replay:
> > -------------------
> > - network play for humans
> > - store replay of races
> > Network races:
> > --------------
> > - play team manager by network, so you can participate to robots races.
> > Graphic & Sound:
> > ----------------
> > - enhance graphic quality, framerate (LOD)
> > - multitexture / bump mapping for cars (PLIB)
> > - skid marks and smoke
> > - mirror in driver view (PLIB)
> > - configurable on-screen information (position and content)
> > - custom dashboard
> > - better sound (OpenAL) 3D (engine, rolling, transmission, wind, skid, collisions)
> all nice, but most important IMHO are improving speed (LOD) and better
> > GUI:
> > ----
> > - better menu navigation for player configuration
> > - display of the cars pictures and info on selection screen
> > - display of the tracks pictures and info on selection screen
> > - allow the player to choose the color of the car
> > - cars settings screen
> display of the cars in the selection would be nice, eg. rotating on a
> nice presentation plane, lots of reflections and some technical data
> like engine characteristics, etc. and similar the the tracks.
I agree that at least the characteristics of the cars or tracks
should be displayed on the selection screens.
> what i missed above:
> - to get high quality pictures, the texture size of the cars sould be at
> least 1024x1024 under use of s3tc where available. if there is lack of
> memory on the graphics board, one could generate automaticly lower
> resolution textures (perhaps you do already for mipmapping?)
> - enhanced graphics configuration menu, where you can switch eg.
> environment mapping on and off, you can choose a smaller visibility
> range, 16 or 32 bit textures, etc.
> - "real" shadows (everything on everything) with "shadow mapping" (needs
> just multipass rendering or some extensions, no projection matrices to
> the floor or something like that). disadvantage: CAN become quite blocky
> near the camera (depends on Z-buffer resolution and Z-far/Z-near).
> - better environment (looks very artificial), could we borrow the
> clouds/weather partially from flight gear? better trees, etc. i think
> this will be possible, if we get free recources from LOD (i expect
> really a boost from that).
I leave the graphic part to Christophe, he has a lots of ideas on the subject.
> - moving parts on the car (not just wheels) eg. suspension, for formula
> one cars.
> - a cool intro (not a film) rendered by torcs from a replay/replays with
> music (look at the parsec demo, it's quite cool IMHO).
Need to implement the replay first ;-)
> - tire changing in pits (obvious, but not availible yet, because not needed)
> - graphical configuration screen for robot parameter files.
> - use of multithreading, split of (at least) rendering/simulation.
May be shared with network client/server split.
> - small release cycles, so that torcs pops up often in news ( all 2 or 3
> ok, what would i put on the task list as project manager for the near
> 1. finish and integrate the windows port.
This point is currently in progress, this is the current task.
> 2. documentation, tutorial bot and track, to get more audience.
This one should be in high priority, and all the help is welcome.
> 3. testing, release 1.1.
> 4. tire wearing & temperature, tire changing in pit.
This will change the grip of the cars, so the robots will have to face
a new problem...
> 5. level of detail.
I am thinking about this as the next task.
> 6. better sound.
Christophe is in charge of the graphic and sound enhancements.
> 7. testing release 1.2.
> 8. decision where to go -> robot/human/network etc.
> 9. version 2.0 todo document.
> final remarks:
> the direction i would prefer is toward an even greater robot simulation,
> great sound, great graphics, etc. why i don't need another "human"
> racing game? simple, because there are already a lot commercial games
> for windows, PS2 (have you seen GT3 graphics? i can just say:
> ooooooooooooooooh, great), ..., which are really great, and there is an
> open source competitor, the racer, which i think looks also great. what
> do you think about that?
> bye, bernhard.
The direction is clearly towards the robot simulation.
The idea is not to be another racing game, but the ability to
drive was done to test the simulation itself.
When you drive yourself you found more easily the simulation
problems, because the robots tends to "compensate" when
the simulator is not accurate, and problems can be masked.
For example, the robots are really better that human when
driving unstable cars.
Maybe, it can convince new people after trying the human part
that it could be fun to program a robot, no ?
Thanks for the comments,
The Open Racing Car Simulator