Re: Torcs-NG Newer Generation

I'm perhaps the least familiar with Torcs on this list, since I
haven't had time to build it on OS X and play with it yet, but I think
the Torcs system is very interesting, and your Virtual Darpa effort
sounds like a good challenge (for both you and the developer-users).

Comments below...

On Wed, Feb 11, 2009 at 5:07 AM, Jean-Christophe Hoelt
<[email protected]> wrote:
> ...
> The first stone of the simulator is Torcs. We are going to improve it
> to give AI plugins access to "virtual cameras" and other sensors (like
> GPS), which should be the only way those plugins can access the track
> and cars positions.
> ...

Several years ago I did a little survey term paper of just this --
"artificial vision" -- which was basically a vision system for an
autonomous avatar. As the avatar moved about the world, the system
would render a low-res image of whatever the avatar was looking at,
and the avatar built a rough, voxel-based model of its world. This
system wasn't for path planning or other aspects of autonomous robots,
it just focused on building a map of the world through simulated
vision.

That said, "virtual cameras" would be interesting. Autonomous vehicles
such as those used in the DARPA challenge use a mix of video and
telemetry. The video is highly processed, either from cameras or from
laser-range scanners. Sooo, mapping real-world data sources like this
to a virtual world could mean supplying regular RGB images (as the
artificial vision system I mentioned above did), or supplying other
image types, such as the depth buffer of a scene rendered from the
car's camera. For simulation and algorithm development, you can do a
lot with just this: send only red channel, inject noise into the image
to simulate a faulty camera, etc.

Good luck with your project.

-Brent

p.s. Anybody know of a working OS X (10.5.x) build of Torcs? The ones
I've tried so far don't work.

------------------------------------------------------------------------------
Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
software. With Adobe AIR, Ajax developers can use existing skills and code to
build responsive, highly engaging applications that combine the power of local
resources and data with the reach of the web. Download the Adobe AIR SDK and
Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com