A recurrent neural network, called the Lagrangian network, is presented for
the kinematic control of redundant robot manipulators, The optimal redunda
ncy resolution is determined by the Lagrangian network through real-time so
lution to the inverse kinematics problem formulated as a quadratic optimiza
tion problem, While the signal for a desired velocity of the end-effector i
s fed into the inputs of the Lagrangian network, it generates the joint vel
ocity vector of the manipulator in its outputs along with the associated La
grange multipliers. The proposed Lagrangian network is shown to be capable
of asymptotic tracking for the motion control of kinematically redundant ma
nipulators.