Decentralized adaptive force/position control of reconfigurable manipulator based on soft sensors

Abstract

Two soft sensor control methods are proposed to deal with force/position control of reconfigurable manipulator without using wrist force sensors. First, modeling uncertainties and coupled interconnection terms between the subsystems are approximated by using adaptive radial basis function neural network, and the soft sensor model of the contact force is established by means of adaptive radial basis function neural network to design hybrid force/position controller. Then, a decentralized explicit force controller based on impedance inner control is designed. The reference trajectory of impedance inner controller is provided by explicit force controller based on the fuzzy prediction, and the soft sensor model of the contact force is established by the fuzzy system. The proposed soft sensor models do not request the exact mathematical relationship between the contact force and auxiliary variables and provide a feasible method to replace the wrist force sensors which are expensive and easily influenced by the external factors. Compared with the observer method, the proposed soft sensor methods do not depend on the knowledge about the model of reconfigurable manipulator, so provide better position and force tracking precision.