Solving inverse kinematics of a planar dual-backbone continuum robot using neural network

Abstract

The inverse kinematics of multiple-backbone continuum robots is a highly non-linear problem. Traditional methods for solving such kinds of problems include the inverse transformation, geometric approach, etc., which are relatively complex and have multiple solutions. On the other hand, the pseudo-rigid-body model (PRBM) is one simple approach for solving the forward kinematics of multiple-backbone continuum robots, but currently, it is still not applicable for solving the inverse kinematics problem. In this paper, we present a strategy for solving the inverse kinematics problem of a dual-backbone continuum robot using PRBM and the Artificial Neural Network (ANN). The strategy firstly computes the forward kinematic solutions of the dual-backbone robot via the PRBM approach. The obtained solutions are then used to build up an ANN model for solving the inverse kinematics of the dual-backbone continuum robot. Based on the Bayesian Regularization training algorithm, the accuracy of the ANN results after linear regression is 99.99%. Finally, we compared the errors between the inputs of the forwards kinematics problem and the output (solutions) of the inverse kinematics problem solved by the ANN model.