[Some details of the Robot Auto Racing Simulator - RARS - are
presented below, culled from a RARS announcement.]
Starting on Monday 17 April I will be posting a digest of the most
significant messages from the rars-list to rec.autos.simulators and
to any individual who requests the digests. Initially I will repost
the RARS tutorials which have been posted to the list recently. I am
doing this because traffic on the list is now quite high and many
people do not wish to handle the volume of mail involved in day to
day discussions.
Send requests to be added to the rars-digest list directly to me,
--- extracts from announcement ---
The Robot Auto Racing Simulation (RARS) is a simulation of auto
racing in which the cars are driven by robots. Its purpose is
two-fold: to serve as a vehicle for Artificial Intelligence
development and as a recreational competition among software
authors. The host software, including source, is available at no
charge. It currently runs under MSDOS and UNIX, including Linux.
There is a RARS anonymous ftp site: magdanoz.mcafee.com in
directory /bin/ftp/rars. Anyone can get any RARS stuff there,
code, announcements, car controllers, documentation, and tracks.
There is a listserver so that interested parties may discuss
RARS by e-mail. To subscribe to the list service send e-mail to
subscribe rars-list
(To unsubscribe, send "unsubscribe rars-list" to the same address.)
Rex Sheasby is the volunteer who is handling this.
RARS consists of a simulation of the physics of cars racing on a
track, with a simple bird's-eye view of the race. The unique feature
is that each car is controlled by a separate and independent control
program. Each car is "driven" by its own control program, which
receives information from the simulation telling it about the car's
local situation. The "driver" (control program) adjusts the steering
and throttle, and then the physics simulation moves the car a little.
This happens many times per second, of course. Every car has exactly
the same physical characteristics, only the "drivers" are different.
Hence, the result is a competition between the control programs.
Furthermore, the competition is visible as an auto race, with
acceleration, passing, cornering, braking, etc.
It is intended that many users will write their own robot "drivers".
Thir*** similar, but not identical, examples are supplied. These
are meant to serve as examples for programmers wishing to develop
their own. The control programs may be written in other languages if
they are linker-compatible on the the intended platform. To date
the robot "driver" programs have used C++ or ANSI C.
For genetic programming, the races will be between several programs
selected from an evolving population of programs. The racing may
take place continuously for long periods of time, with the graphic
display disabled for faster execution. Of course losers will be
eliminated and winners will breed. Genetic Algorithm proponents will
probably design robot drivers with a vector of parameters to be
determined by evolution. Neural nets are also candidate "drivers".
It will be up to the experimenter to decide if human-designed robots
are allowed to compete with the evolving population.
-- Dave "Gizmo" Gymer Telephone +44 (0)1268 534228
-- General DataComm Advanced Research Centre Ltd, Basildon, UK
