A real time solution to the forward kinematic problem of a general spherical three-degree-of-freedom parallel manipulator

In this paper, a real time solution to the forward kinematic problem of a general spherical three-degree-of-freedom parallel manipulator is presented using polynomial learning networks. These networks learn the forward kinematic problem based on a database of input-output examples. After the learning process has been achieved, the networks exhibit very little error when presented with inputs which were not part of the learning database. The computation time required to compute the forward kinematics is very small since the networks consist only of additions and multiplications, The results are compared with a Newton-Raphson scheme.