A feed forward neural network for solving the inverse kinetics of non-constant curvature soft manipulators driven by cables

The solution of the inverse kinetics problem of soft manipulators is essential to generate paths in the task space to perform grasping operations. To address this issue, researchers have proposed different iterative methods based on Jacobian matrix. However, although these methods guarantee a good degree of accuracy, they suffer from singularities, long-term convergence, parametric uncertainties and high computational cost. To overcome intrinsic problems of iterative algorithms, we propose here a neural network learning of the inverse kinetics of a soft manipulator. To our best knowledge, this represents the first attempt in this direction. A preliminary work on the feasibility of the neural network solution has been proposed for a conical shape manipulator driven by cables. After the training, a feed-forward neural network (FNN) is able to represent the relation between the manipulator tip position and the forces applied to the cables. The results show that a desired tip position can be achieved quickly with a degree of accuracy of 0.73% relative average error with respect to total length of arm.