Inverse kinematic at acceleration level using neural network

Abstract

The inverse kinematic of a constrained redundant robot manipulator is considered. An optimization procedure using neural network is formulated. It produces position, velocity and acceleration trajectories in joint space from position and orientation trajectories in Cartesian space and guarantees a good tracking of the desired end-effector trajectory. The redundancy is solved by minimizing a performance function. This new method gives an accurate solution with only a few iterations. The application of this scheme to a 3 degrees-of-freedom redundant manipulator is demonstrated through simulation results.