Abstract: A 6-DOF hybrid precision assembly platform was designed to assemble the optical and mechanical modules of the huge laser facility SG-III. Based on analytic method, the position and posture mapping relationship between the pose module and pose adjusting mechanism was established, and then the invert kinematics was obtained. By employing BP neural network, the nonlinear mapping model from driving displacements to the pose module positions and postures was built, and the forward kinematics was acquired by training the inverse kinematics samples. In order to improve the accuracy of the forward kinematics, the network was divided into six sub-networks corresponding of each DOF, thus, the accuracy of the solution was increased by one order of magnitude. Finally, position compensation algorithm was proposed to optimize the network. The results demonstrate the displacement accuracy is enhanced to level 10^-3 mm, which meets the requirement of engineering application. The reliability of the proposed methods is validated by simulation in MATLAB software, which can be applied to solve the forward and inverse kinematics of common hybrid robots.

Key words: mechanical design, kinematics, neural network, position compensation, hybrid robot