A SOLUTION TO THE END-EFFECTOR POSITION OPTIMIZATION PROBLEM IN ROBOTICS USING NEURAL NETWORKS

The paper presents an original application of the Hopfield-type neural network to a robotics optimisation problem. The robot considered feat ures an aml composed of three revolute joints, the last of which is th e end-effector: The robot is planar as the movement of the end-effecto r is limited ro one plane. The mechanical characteristics of the actua tors in the joints, the accuracy of the angle position sensors, and di mensional errors in the mechanical elements which make up the end-effe ctor all contribute towards an end-effector positioning error along an assigned trajectory. The computational complexity of the algorithmic solution to the minimisation of this error is at rimes incompatible wi th certain particularly critical industrial applications. To reduce th e calculation time, the author presents a neural approach based on a H opfield-type model. A detailed definition of the neural approach is gi ven, its capacity for solving the problem is demonstrated, and the com putational complexity is analysed. This analysis shows the drastic com putational reduction provided by the neural approach as compared with an algorithmic solution to the problem of end-effector position optimi sation.