Vision Based Automatic Calibration of Microrobotic System
Julkaisun pysyvä osoite on
During the last decade, the advancement of microrobotics has provided a powerful tool for micromanipulation in various fields including living cell manipulation, MEMS/MOEMS assembly, and micro-/nanoscale material characterization. Several dexterous micromanipulation systems have been developed and demonstrated. Nowadays, the research on micromanipulation has shifted the scope from the conceptual system development to the industrial applications. Consequently, the future development of this field lies on the industrial applicability of systems that aims to convert the micromanipulation technique to the mass manufacturing process. In order to achieve this goal, the automatic microrobotic system, as the core in the process chain, plays a significant role. This thesis focuses on the calibration procedure of the positioning control, which is one of the fundamental issues during the automatic microrobotic system development. A novel vision based procedure for three dimensional (3D) calibrations of micromanipulators is proposed. Two major issues in the proposed calibration approach - vision system calibration and manipulator kinematic calibration - are investigated in details in this thesis. For the stereo vision measurement system, the calibration principle and algorithm are presented. Additionally, the manipulator kinematic calibration is carried out in four steps: kinematic modeling, data acquisition, parameter estimation, and compensation implementation. The procedures are presented with two typical models: the matrix model and the polynomial model. Finally, verification and evaluation experiments are conducted on the microrobotic fiber characterization platform in the Micro- and Nano Systems Research Group (MST) at Tampere University of Technology. The results demonstrate that the proposed calibration models are able to reduce the prediction error below 2.59 micrometers. With those models, the pose error, compensated by the feed-forward compensator, can be reduced to be smaller than 5 µm. The proposed approach also demonstrates the feasibility in calibrating the decoupled motions, by reducing the undesired movement from 28 µm to 8 µm (For 4800 µm desired movement).