Optimised robot calibration using a vision-based measurement system with a single camera

摘要

Robot calibration plays an increasingly important role in robot production as well as in
\nrobot operation and integration within computer integrated manufacturing or assembly
\nsystems. The production, implementation and operation of robots are issues where
\nrobot calibration results can lead to significant accuracy improvement and/or cost-
\nsavings.
\nThe thesis describes techniques for modelling, optimising and performing robot
\ncalibration processes using a 3-D vision-based measurement system for off-line
\nprogramming. The identification of the nonrlnal kinematic model is optimised using
\nnumerical methods to eliminate redundant geometric parameters in the model.
\nCalibration based on the optimised model shows improvement in robot accuracy when
\ncompared to the non-optimised model.
\nThe basics of the measurement system consist of a single CCD camera mounted on the
\nrobot tool flange, image processing software, and algorithms specially developed to
\nmeasure the end-effector pose relative to a world coordinate system. Geometric lens
\ndistortions are included in the analytical technique. The target consists of two identical
\nclusters of calibration points printed on photographic paper, and mounted on the sides
\nof a 90-degree angle plate. Experimental work was performed to assess the
\nmeasurement system accuracy at different distances from the camera to the target. An
\naverage accuracy from O.2mm to O.4mm was obtained at distances between 6S0mm to
\n9S0mm.
\nTests were also performed on three different robots to assess the improvement in the
\noverall robot accuracy. The robots tested were: PUMA-SOO, IRB-2400 and IRB-6400.
\nThe errors before calibration for the three robots were approximately in a range from
\nSmm to lSmm if measured in a large volume. The best average accuracy obtained after
\nthe calibration of the three robots was O.3Smm, O.60mm and O.4Smm respectively.
\nThis study shows that many different variables are involved in the calibration process.
\nThe influence of these variables was studied both experimentally and by means of
\nsimulation.