Re: Problem with Euler Angles

Thanks Mike,

Great! This routine works perfectly on my test walk.
(I will have to check a few values formatted as "?")

I have seen another question about the inclinometer yesterday, and
have not seen yet any such routines from Webots.
Do you care to publish the routine on the Wiki pages? I can make the
page if you don't have time.

Frans

--- In [email protected], "Mike Jost" 
<webotsContest@m...> wrote:
> Frans,
> 
> The problem with the mixing of the X and Z angles is probably
because you
> are using one of my older getAngles( ) functions.  The most current
> getAngles( ) function is at
> http://www.mikejost.com/cyberbotics/RotationMatrix/Judoka0_AngleTest.zip
> The function will return X=0 and Z=0 when the robot is standing
vertical.
> 
> The previous functions were based on the two rotation matrix
compositions:
>      R = Rx * Rz * Ry  and
>      R = Rx * Ry * Rz
> Both of these compositions mix the X and Z angles, since they are
not the
> way the controller actually calculates the composition.
> The lastest getAngles( ) function uses a composition based on the
heirarchy
> of the robot from the judo.wbt VRML file(see previous post Nov. 30,
2003).
> This composition is
>     R = Ry * Rx * Rz
> This function returns correct X and Z angles with no mixing for -90
< X <
> 90.  For X angles beyond +/-90 the Z angle is locked at +/- PI.  The
Y angle
> is not correct.
> This function will work as an inclinometer, but cannot be used to
determine
> the correct X,Y, and Z angles all the time.  There is a problem with
using
> Euler angles for rotation composition because the order of the
composition
> determines the final matrix and is not unique.  This has the effect of
> mapping a rototion axis onto another axis (the mixing problem) and for
> intermediate elementary rotations of +/-Pi/2 it will be impossible
to rotate
> around the final elementary axis (the locking problem).
> 
> Mike Jost
> 
> 
> ----- Original Message ----- 
> From: "fransklip" <frans@v...>
> To: <[email protected]>
> Sent: Monday, December 15, 2003 18:20 PM
> Subject: [webots-contest] Re: Problem with Euler Angles
> 
> 
> > Hi Mike,
> >
> > I tried your angle calculation routines in my Kimaita robot for fall
> > detection. That seemed to work at first, with the initial dive-attack.
> > But later on in the match it went wrong: falls were detected when
> > there was none.
> >
> > This weekend I logged sensor data every 64 msec while the robot was
> > walking a small circle, and doing some body rotations after every
> > (about) 90 degrees: tilt forward/back, tilt left/right and rotate ,
> > each + and -10 degrees.
> >
> > Near the Z-axis, the angles routines return values that can be related
> > to the actual robot posture. But when moving away from the Z-axis, I
> > cannot relate the returned angles to the robot anymore.
> >
> > I have made a summary of the test rotation values (angletest 01
> > summaray.txt). You can donwload this file from
> > www.van-der-klip.demon.nl/webots/angletest01.zip. It also contains the
> >  sensor data for each of the 4 test movements in the walk (set 1, 2,
> > 3, 4.txt) and the complete sensor and action log files).
> >
> > You can download the webview file of the test walk from:
> > www.van-der-klip.demon.nl/webots/angletest01.wva
> >
> > Are the calculated angles something else then a rotation of the robot
> > body around its center? If so, could you please axplain what they are?
> >
> > Regards,  Frans / Kimaita
> >
> > --- In [email protected], "Mike Jost"
> > <webotsContest@m...> wrote:
> > > Olivier,
> > >
> > > I found some problems with the Euler angle calculations this
> > weekend.  I tested the previous equations by rotating the robot with
> > the mouse, since I did not have a robot controller capable of turning
> > on its own.   Apparently the rotation composition used by the mouse in
> > the Webots program is different than the rotation composition used
> > when a robot actually moves itself.  The current gps_euler( ) function
> > in Webots 4.0.17 also returns wrong X and Z angles for some Y
> > orientations of a robot.
> > >
> > > Solution:
> > > I looked at the VRML file and used a rotation composition based on
> > the heirarchy of the robot.
> > >    Judoka_Robot {
> > >       rotation 0 1 0 0 (rot about Y)
> > >       children {
> > >          back_1 servo {
> > >             rotation 1 0 0 0 (rot about X)
> > >             children {
> > >                back_2 servo {
> > >                   rotation 0 0 1 0 (rot about Z)
> > >                }
> > >             }
> > >          }
> > >          gps_device
> > >       }
> > >    }
> > >
> > > The resulting composition should be  Rgeneral = Ry * Rx * Rz instead
> > of the previous  Rx * Ry * Rz.
> > >
> > > The routine to compute the euler angles based on this
composition is:
> > >    public static float[] getAngles(float matrix[]) {
> > >
> > >       float[] angles = new float[3];
> > >       double cosX;
> > >
> > >       angles[0] = (float)Math.asin(-1*matrix[9]); // find X angle
> > >
> > >       cosX = Math.cos(angles[0]);
> > >       angles[1] = (float)Math.atan2( -1*matrix[8]/cosX,
> > matrix[10]/cosX  );  // find Y angle
> > >       angles[2] = (float)Math.atan2( matrix[1]/cosX, matrix[5]/cosX
> > );       // find Z angle
> > >
> > >       return angles;
> > >    } // end of getAngles()
> > >
> > >
> > > I have generated a Judoka0 test robot that will rotate itself about
> > Y and stop at 45 Degree intervals and bend forward,backward, left, and
> > right, and display the computed angles from this new routine.
> > > The back_1 and back_2 servos are set to a rotation of +/- 0.5 rad at
> > each test point.
> > > Hopefully, this routine will work in all cases, such as after Coyote
> > does a summersalt and rotates X around a full 2*pi radians.
> > >
> > >
> > > Judoka0 Test Robot:
> > >
> >
http://www.mikejost.com/cyberbotics/RotationMatrix/Judoka0_AngleTest.zip
> > >
> > > HTML version of Maple worksheet of General Rotation Matrix for
> > R=Ry*Rx*Rz:
> > >
> > http://www.mikejost.com/cyberbotics/RotationMatrix/RotationMatrix.html
> > >
> > >
> > > Mike Jost (FreeTIme)
> >
> >
> >
> > To unsubscribe from this group, send an email to:
> > [email protected]
> >
> >
> >
> > Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
> >
> >
> >


To unsubscribe from this group, send an email to:
[email protected]

 

Yahoo! Groups Links

To visit your group on the web, go to:
 http://groups.yahoo.com/group/webots-contest/

To unsubscribe from this group, send an email to:
 [email protected]

Your use of Yahoo! Groups is subject to:
 http://docs.yahoo.com/info/terms/