Abstract:In the robotic application

the error would existbetween the actual and desired initial positions of the endeffector In this paper

a faulttolerant motion planning scheme was proposed based on quadraticprogramming to reduce the initial position error and improve the trackingaccuracy during the endeffector executing the task By using the neuraldynamicsmethod

an erroreliminating velocity was designed based on the realtimeposition error

and was incorporated into the endeffector velocity togetherwith the task desired velocity Furthermore

a primaldual neural network solverbased on linear variational inequalities was used to solve the proposedquadratic programming scheme in realtimeTheMATLAB simulationresults of a twodegreeoffreedomplanar manipulator demonstratethat the initial position error is exponentialconvergent to 0

and the faulttolerance scheme is effective