Neural network solutions of forward kinematics for 3RPS parallel manipulator

摘要: This paper describes how neural networks are much capable to solve the forward kinematics of parallel manipulators. The solution to the forward kinematics involves highly nonlinear equations of motion, which can be solved by numerical methods with great accuracy but the time consuming calculations restrain it in implementing for real time kinematic control. Two types of neural networks namely multilayer perceptron (MLP) and radial basis function (RBF) are considered to solve the forward kinematics of 3RPS (revolute-prismatic-spherical) parallel manipulator. The performance and suitability of both the networks are evaluated for this specific application.