Absolute bearing / compass
6 views
Skip to first unread message
Tennessee
Apr 13, 2012, 4:20:20 AM4/13/12
to Car Racing Competition
Hi all,
I'm a total newb to TORCS and to this competition, so I'm sure I don't
understand everything about the physical model, the environment and
the sensors.
It looks as though there is no way to determine an absolute bearing of
the car, only the bearing relative to the track axis, which is a
relative concept. Is that correct?
It looks like, given some point in time, one can map the track around
the car, and thereby map the track around you as the reference point.
However, I'm struggling to understand how to preserve information
between tics. If I set a steering angle, there appears to be no way of
verifying if the car's facing angle matches that afterwards. i.e., if
the car skids and yaws at the same time, you lose information. You
can't get this from speedY either, since the car might skid
differently at the front and the back, with a yaw.
So it seems hard to do much by way of mapping and shortest-path
analysis. It would be great to have a compass in there somewhere.
Or is there some other way to infer a compass direction from the other
information?
I hope all this makes sense!
Thanks,
-Tennessee
I'm a total newb to TORCS and to this competition, so I'm sure I don't
understand everything about the physical model, the environment and
the sensors.
It looks as though there is no way to determine an absolute bearing of
the car, only the bearing relative to the track axis, which is a
relative concept. Is that correct?
It looks like, given some point in time, one can map the track around
the car, and thereby map the track around you as the reference point.
However, I'm struggling to understand how to preserve information
between tics. If I set a steering angle, there appears to be no way of
verifying if the car's facing angle matches that afterwards. i.e., if
the car skids and yaws at the same time, you lose information. You
can't get this from speedY either, since the car might skid
differently at the front and the back, with a yaw.
So it seems hard to do much by way of mapping and shortest-path
analysis. It would be great to have a compass in there somewhere.
Or is there some other way to infer a compass direction from the other
information?
I hope all this makes sense!
Thanks,
-Tennessee
tankjob
Apr 25, 2012, 12:54:18 PM4/25/12
to Car Racing Competition
Hi,
> It looks as though there is no way to determine an absolute bearing of
> the car, only the bearing relative to the track axis, which is a
> relative concept. Is that correct?
That is correct, I think. The absolute bearings that Torcs can provide
are
not accessible in the competition client. No GPS here. :)
> It looks like, given some point in time, one can map the track around
> the car, and thereby map the track around you as the reference point.
> However, I'm struggling to understand how to preserve information
> between tics. If I set a steering angle, there appears to be no way of
> verifying if the car's facing angle matches that afterwards. i.e., if
> the car skids and yaws at the same time, you lose information. You
> can't get this from speedY either, since the car might skid
> differently at the front and the back, with a yaw.
Actually, the steering angle is not identical with the direction the
car
goes. The steering angle only controls the position of the front
wheels.
> So it seems hard to do much by way of mapping and shortest-path
> analysis. It would be great to have a compass in there somewhere.
That's probably the reason why an absolute coord. system is omitted.
> Or is there some other way to infer a compass direction from the other
> information?
Given the fact that the relative sensors give "noisy" signals in the
race,
too difficult.
> It looks as though there is no way to determine an absolute bearing of
> the car, only the bearing relative to the track axis, which is a
> relative concept. Is that correct?
are
not accessible in the competition client. No GPS here. :)
> It looks like, given some point in time, one can map the track around
> the car, and thereby map the track around you as the reference point.
> However, I'm struggling to understand how to preserve information
> between tics. If I set a steering angle, there appears to be no way of
> verifying if the car's facing angle matches that afterwards. i.e., if
> the car skids and yaws at the same time, you lose information. You
> can't get this from speedY either, since the car might skid
> differently at the front and the back, with a yaw.
car
goes. The steering angle only controls the position of the front
wheels.
> So it seems hard to do much by way of mapping and shortest-path
> analysis. It would be great to have a compass in there somewhere.
> Or is there some other way to infer a compass direction from the other
> information?
race,
too difficult.
Tennessee Leeuwenburg
Apr 25, 2012, 11:06:40 PM4/25/12
to racingco...@googlegroups.com
In terms of the competition format, you get to drive the track before you race. You can use cte to steer the middle track, and sideways sensors to map the course boundaries. With a good map of the track, you could use a particle filter to probablistically guess the actual car heading. I might try that at some point.
--
You received this message because you are subscribed to the Google Groups "Car Racing Competition" group.
To post to this group, send an email to racingco...@googlegroups.com.
To unsubscribe from this group, send email to racingcompetit...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/racingcompetition?hl=en-GB.
--
--------------------------------------------------
Tennessee Leeuwenburg
http://myownhat.blogspot.com/
"Don't believe everything you think"
Reply all
Reply to author
Forward
0 new messages