When extracting the pose vector [x y z qx qy qz qw] from the dual quaternion [q_r, q_t], the method DualQuaternion.to_pose() first gets the rotation quaternion q_r and inverts its components in case the scalar part is negative (q_r_minus). This should be fine as it describes the same rotation. However, when extracting the translation vector t from the "translation quaternion" q_t = 1/2 * t * q_r, we need to use the exact same quaternion q_r. Using the q_r_minus for instead of q_r for the computation of t leads to an inverted translation vector. Is this correct?
Summary
Compute translation vector using q_r (which is self.q_rot in the code).
Extend test_conversions unit test to also cover the case where the scalar part of the quaternion is negative.
Use integer instead of float for index accessing (unrelated, but this unit test failed on my machine).
Problem Description
When extracting the pose vector
[x y z qx qy qz qw]from the dual quaternion[q_r, q_t], the methodDualQuaternion.to_pose()first gets the rotation quaternionq_rand inverts its components in case the scalar part is negative (q_r_minus). This should be fine as it describes the same rotation. However, when extracting the translation vectortfrom the "translation quaternion"q_t = 1/2 * t * q_r, we need to use the exact same quaternionq_r. Using theq_r_minusfor instead ofq_rfor the computation oftleads to an inverted translation vector. Is this correct?Summary
q_r(which isself.q_rotin the code).test_conversionsunit test to also cover the case where the scalar part of the quaternion is negative.