Search

• A hybrid manipulator for optomechanical modules assembly in an inertial confinement fusion facility is presented. • The analytic inverse kinematics solution is obtained by decomposing the coupled kinematics model into two simple sub-models. • An improved back propagation neural network is proposed to increase the kinematic accuracy. A novel hybrid manipulator is proposed in this work, in order to assemble optomechanical modules in an inertial confinement fusion facility (SG-Ⅲ). Differ from general serial and parallel manipulators, both the inverse and forward kinematics problems of hybrid manipulators are difficult to be solved. Based on an analytic method, the position and orientation mapping functions between the optomechanical module and the parallel manipulators are established, and then the method of transformation matrix is employed to obtain the inverse kinematics solution. The forward kinematics of this hybrid manipulator is highly nonlinear and coupled, so a back propagation neural network using all class one network strategy is trained to approximate the mapping relation. Aiming to improve the accuracy, the network is split into two sub-networks corresponding to the natural property of the outputs. In the end, an improved back propagation neural network is proposed to optimize the network, results illustrate that the kinematic accuracy is increased significantly. [ABSTRACT FROM AUTHOR]