Qualitative approach for forward kinematic modeling of a Compact Bionic Handling Assistant trunk

Abstract

This paper addresses the forward kinematic model of a class of hyper-redundant continuum robot, namely Compact Bionic Handling Assistant (CBHA). Inspired from the elephant trunk, it can reproduce some biological behaviors of trunks, tentacles, or snakes. Such systems, like the CBHA are subjected to a set of nonlinearities (flexibility, elasticity, redundancy,…) and uncertainties (parameters and modeling), making difficult to build an accurate analytical model, which can be used to develop control strategies. Hence, learning method becomes a suitable approach for such scenarios in order to capture un-modeled nonlinear behaviors of this continuum arm. The proposed approach makes use of Multilayer Perceptron (MLP) and Radial Basis Function (RBF) Neural Networks for the approximation of forward kinematic model (FKM) of CBHA trunk. The experiments have been conducted on the CBHA in order to validate the forward kinematic model where the arm trajectories are generated using a physical coupling with a rigid manipulator. A comparison of both qualitative approaches with a quantitative geometric approach, according to the model accuracy is given at the end of the experiment.