[ODE] Has anybody had any luck with dJointSetAMotorAngle?

[Patrick McColgan]

Thanks Andrew.

You are right about the normalizations, the dot product should be of 
two unit vectors, I do do that I just neglected to write it.

I think I also understand you're point about the negative angles, I 
wrote my email from memory but I do recall at the beginning the angles 
were negative, even at rest one or two may have been negative zero.  To 
fix it I think I started evaluating the rotation about the y about the 
relative right axes and then evaluated both the x and z rotations by 
looking at the difference in up axes.  I think this may have been based 
on the fact that ideally the up axes will remain the same.

Hope that helps, apologies if it was misleading.

Andrew Schroeder wrote:

>I have looked this over and learned from it, I only have a couple
>small insights for now, but I plan to keep looking into this....
>
>First, when using the dot product to find the angle between the two
>axis, I think you need to normalize them after "flattening" them.
>
>Second,  when "flattening" an axis by zeroing out one of the
>coordinates, there is a potential problem when the axis goes negative.
>Imagine finding the rotation around the x axis by "flatening" the y
>(remove the x coordinate). so, rotation around the z axis does not
>show up as rotation around the x because the x is flattened out, but
>when the z rotates past PI/2, the y coordinate goes negative and the x
>suddently thinks it has rotated all the way upside down.
>I hope that makes sense, it's hard to explain.
>
>anyway, I plan to continue testing out aMotors and will post any more
>relivent learnings....
>
>