There is a flaw in the test suite. It tests the values in rpy2r rather than the results rpy from the solver's function. Also, the tolerance needs to be widened a bit.
Actually that's not a flaw, William. Since it is not always possible to determine the exact three original angles. In the second test case for instance, there are infinite solutions due to a singularity. Therefore, he tests the rotation matrix since it will be equal for all possible solutions.
It's interesting that a precision error allows you to recover the original angles: atan2(cos(pi/2)*sin(pi/3),cos(pi/2)*cos(pi/3)) amazingly returns pi/3, which obviously should not be true since atan2(0,0) = 0 (?). The issue seems to be that cos(pi/2) is not zero (due to an approximation error), but just a really small value close to zero.
2240 Solvers
Rotate and display numbered tile
239 Solvers
Matrix indexing with two vectors of indices
486 Solvers
229 Solvers
324 Solvers
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!