marip8
commented
4 years ago Unit test results
Bad DH Guess, Good Transform Guesses
Correlation Coeff. > 0.500000: j2_r target_mount_to_target_y 0.997526 j1_d target_mount_to_target_z 0.997168 j1_theta j1_r -0.988434 camera_mount_to_camera_z camera_mount_to_camera_ry 0.911556 camera_mount_to_camera_y camera_mount_to_camera_rx 0.643034 j2_alpha j3_d -0.575622 j1_alpha target_mount_to_target_x 0.565983 camera_mount_to_camera_y camera_mount_to_camera_rz -0.532941
Difference between nominal and calibrated chain out to camera frame for 1000 random pose samples Position Difference Mean: 0.00148334 Position Difference Std. Dev.: 0.000560913 Orientation Difference Mean: 0.00154308 Orientation difference Std. Dev.: 4.37752e-06
Bad DH Guess, Bad Transform Guesses
Correlation Coeff. > 0.500000: j2_r target_mount_to_target_y 0.99589 j1_d target_mount_to_target_z 0.989064 j1_theta j1_r -0.987226 camera_mount_to_camera_z camera_mount_to_camera_ry 0.921216 j2_alpha j3_d -0.680302 camera_mount_to_camera_y camera_mount_to_camera_rz -0.61543 camera_mount_to_camera_y camera_mount_to_camera_rx 0.606904 j2_r target_mount_to_target_rz 0.527974
Difference between nominal and calibrated chain out to camera frame for 1000 random pose samples Position Difference Mean: 0.000740774 Position Difference Std. Dev.: 0.000215605 Orientation Difference Mean: 0.000776828 Orientation difference Std. Dev.: 4.12745e-06
This PR adds likelihood costs and variable masking to the dual DH chain kinematic calibration algorithm. It also adds a few new unit test cases where the DH parameters and transforms are modified.