Microassembly of complex and solid 3D MEMS by 3D vision-based control by CNRS

CNRS works concern the use of vision feedback control in aim to assemble MEMS with a high precision. These MEMS can be assembled without any external fixing (glue, wending …). The assembly task concerns the assembly of 3D solid structures performed on a high precision five degree of freedom (dof) microrobotic cell, which is equipped with a four dof gripping system and a single optical videomicroscope tilted at 45° from the vertical axis.

In this work, we target the assembly by insertion of several silicon microparts of dimensions 400µmX400µmX100µm. A notch of 100µmX100µmX100µm is engraved in every side of each micropart. We have considered a 3D model-based tracker that allows the computation of the object pose during the assembly process in real-time. The 3 D measurements delivered by this algorithm are used to implement accurate 3D vision control (pose-based visual servoing).

The control law developed in this work consists in a 3D pose-based visual servoing (PBVS). The association of the high accuracy 3D model-based tracking algorithm with pose-based visual servo developed allows to obtaining promising results concerning robustness and precision in the MEMS microassembly field. The precision of the techniques developed in this work reaches 0.3 µm for the positioning tasks and the 0.2° for the orientation in space tasks. These precision results allow a solid MEMS assembly with only 3µm of tolerance insertion (mechanical play).

The mean cycle time required to assemble two silicon microparts is estimated by performing several times the assembly times. It can be note that the faster assembly is performed in 34 seconds while the slower is achieved on 47.7 seconds (mean cycle time is 40.8 seconds).

Contact:  Nadine Piat
FEMTO-ST Institute, UMR CNRS 6174 – Besançon
Email: Nadine.piat@ens2m.fr