A solution to the inverse kinematic problem in robotics using neural network processing

A solution algorithm is presented using the T.T. Hopfield and D.W. Tank (Biol. Cybern., vol.52, p.1-12, 1985) analog neural computation scheme to implement the Jacobian control technique. The states of neurons represent joint velocities of a manipulator, and the connection weights are determined from the current value of the Jacobian matrix. The network energy function is constructed so that its minimum corresponds to the minimum least square error between actual and desired joint velocities. At each sampling time, connection weights and neuron states are updated according to current joint positions. During each sampling period, the energy function is minimized and joint velocity command signals are obtained as the states of the Hopfield network. The network dynamics are analyzed using a closed-loop control structure. Simulation shows that this method is capable of solving the inverse kinematic problem for a planar redundant manipulator in real time.<>