Abstract: Through the solution of inverse kinematics, a classified neural network was trained by genetic arithmetic combined with LevenbergMarquardt method to recognize the relationship between joint space and Cartesian space of parallel robot and the evaluation of forward kinematics(FKs) in a range with a little error was provided. Using QuasiNewton iteration method, accurate solution can be obtained. Applied the proposed solution in 6PRRS parallel robot, the result shows that this method can get high precision value and spends little time. Therefore it can be applied in real time task space control or in getting work space of parallel robot.

Key words: automatic control technology, parallel robot, forward kinematics, neural network, quasiNewton method, task space