Measuring rotation with the OFS

8 views
Skip to first unread message

Jason Garland

unread,
May 25, 2012, 5:20:47 PM5/25/12
to robo-magel...@googlegroups.com
2.2 Vehicle Motion Compensation
Vehicle rotation rates and optic flow are highly coupled
since the cameras are rigidly mounted (or strap-downed)
to the vehicle body. Any rotation about the x or y axes
will manifest as optic flow. This is true even if the vehicle
is not translating relative to the navigation frame. As
the coordinate systems of the flow sensors are aligned
with the body frame, a rotation about the x axis will
appear as a change in the optic flow of the y axis, and
vice versa.
The rotational motion can be described by the socalled rotation vector φ and its differential equation
[Bortz, 1971]...

Jason Garland

unread,
May 25, 2012, 5:35:37 PM5/25/12
to robo-magel...@googlegroups.com
http://www.ri.cmu.edu/pub_files/2009/7/DilleFSR09.pdf

This one does a good job of explaining the problem:

Paul Danset

unread,
May 25, 2012, 8:29:27 PM5/25/12
to Robo-Magellan Project on behalf of Jason Garland
I haven't read the paper yet; just skimmed a few figures and equations.

On the IJCAS paper they used a two wheeled, dual motor drive robot model.  This one is easier to estimate the center of rotation, which would always be on the axis of the two wheels, with a bias toward the wheel that is moving the least.

However we have  a 4-wheeled robot with single motor drive plus "rack & pinion"-like steering which I believe is harder to estimate a true center of rotation.  For simplicity we'd probably just pick the center of the 4 wheels as the center of rotation, though I don't know how one could put an OFS at that location on our robot unless we lift up the chassis by a few inches.

TL;DR:  I don't think we have enough time to accurately model and compensate for rotation using the methods described in the paper, using the OFS alone.  I think an easier way (admittedly flawed and less accurate) would be to use the X/Y flow values from the OFS + the knowledge of the angle of steering to ROUGHLY ESTIMATE the odometry.  For example, estimate length of movement at  dr = sqrt(dx**2 + dy**2) and the heading based on the angle of the steering servo.  We all know this is not accurate, but it may be enough to get our robot within striking / line-of-sight distance to the pylon.



-- You received this message because you are subscribed to the Google Groups Robo-Magellan Project group.
 
Check out our project page: http://www.jigsawrenaissance.org/donation/robo-magellan-project/
 
To post to this group, send email to robo-magel...@googlegroups.com.
To unsubscribe from this group, send email to robo-magellan-pr...@googlegroups.com.
For more options, visit this group at https://groups.google.com/d/forum/robo-magellan-project?hl=en

jerrymcm

unread,
May 25, 2012, 8:55:36 PM5/25/12
to robo-magel...@googlegroups.com
Does the phone have a compass?  Or would the accelerometer give us rotation info?
To post to this group, send email to robo-magellan-project@googlegroups.com.
To unsubscribe from this group, send email to robo-magellan-project+unsub...@googlegroups.com.

Paul Danset

unread,
May 25, 2012, 9:05:33 PM5/25/12
to Robo-Magellan Project on behalf of jerrymcm
Many Android phones (including the Galaxy Nexus) have a 3 axis compass (i.e. the compass works regardless of how the phone is oriented).  But like other sensors, it is not super accurate.  Also, the servos / motors that generate magnetic fields near the compass may affect the readings.  For example, the phone sits on a turret with two servos in close proximity.

The MEMS gyro chip my give us some additional rotational info, but it is difficult to do correctly.  My guess is that we shouldn't bother with this one given our time constraints.



To post to this group, send email to robo-magel...@googlegroups.com.
To unsubscribe from this group, send email to robo-magellan-pr...@googlegroups.com.

jerrymcm

unread,
May 26, 2012, 3:04:23 AM5/26/12
to robo-magel...@googlegroups.com
I was thinking maybe some combination of compass/gyro/accelerometer with OFS.  I think you mentioned upthread that accuracy isn't critical as long as it gets us within shooting distance of the "pylon".  Either way, perhaps mounting the phone on the turret isn't such a good idea?
To unsubscribe from this group, send email to robo-magellan-project+unsubscri...@googlegroups.com.

Jason Garland

unread,
Jun 1, 2012, 6:51:27 PM6/1/12
to robo-magel...@googlegroups.com
Pierce was at Metrix last night and I had a discussion with him about measuring the turn angle using the OFS. He suggested that our OFS may support measurement of curl. It does not unfortunately. :(

Budi Mulyo

unread,
Jun 1, 2012, 7:02:29 PM6/1/12
to Robo-Magellan Project on behalf of Jason Garland
Good that you finally met Pierce.

Now, we know what to use for the next iteration: curl enabled OFS. How much and where can we get that?


To post to this group, send email to robo-magel...@googlegroups.com.
To unsubscribe from this group, send email to robo-magellan-pr...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages