Yaw rate

[Derk]

Is there a way to obtain the yaw rate of the car? There is no direct
sensor available for this, but maybe an estimation can be computed
from the other sensors? With yaw rate I mean the angular velocity
around the Z axis of the car (in rad/s), thus the change in
orientation over time. For orientation we have the angle sensor
availabe which measures the angle between the car's orientation and
the direction of the track axis. However, we cannot take the
difference for this sensor between two timesteps, because the track
axis can make a curve.

Thanks,
Derk

[Tennessee]

Hi Derk,

I have the exact same question, I just didn't realise until after I
finished asking it. I couldn't work out a way either. It "might" be
possible to infer it the track axis angle. If you take your first-
facing angle as "north", then you may be able to measure the curve of
the track ahead using the track sensors. If you then have an estimate
of your expected location, you can also then get your expected track
orientation and expected compass orientation, then the difference
between the expected track orientation and the actual track
orientation would equal yaw and could be applied back to your
estimated compass orientation.

A bit convoluted, though. It should be possible to test whether this
works using noiseless sensors.

Cheers,
-T

[Derk Mus]

Hi, I tried some things with noiseless sensors and maybe this is indeed possible. See  http://www.imgdump.nl/hosted/1388333c762d62a26dc176b52c6b2277.png  for an illustration. The green dot is your position at time t at (0,0) and heading north. Blue dots are points at the track edge found by your rangefinders (I excluded points with a distance of 100 meters or more).

Then the red dot is your position at time t+1. Because you know longitudinal and lateral speeds you know the location at time t+1, however you don't know the heading. The grey dots are positions at the track edge at t+1 if you assume the car is not rotated. However, you can see the car must have yaw rate because the grey dots are not on the same line as the blue dots. If you find the angle by which to rotate the grey dots, such that there are on the line formed by the blue dots you know the yaw rate at time t.

Maybe this approach will work, thanks for the ideas.

2012/4/13 Tennessee <tleeuw...@gmail.com>

--
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]

Another idea... assuming there is no such thing as tyre wear or different road surfaces, you might be able to use machine learning to learn the yaw rate for a given steering angle and speed, if you are on a straight section of track. I don't know if there are any artificial very long, wide straight tracks, but that would sure make things easier. If you had that, you could then conduct various turning manoeuvres and because the track angle is constant, you could actually use that measurement. You could then store that for a given car and maybe make use of it later.

Not as good as actually just having a compass sensor though.

--
--------------------------------------------------
Tennessee Leeuwenburg
http://myownhat.blogspot.com/
"Don't believe everything you think"