Control of redundant robot and singularities avoidance based anfis network

In this work we exploit an anfis network to achieve the singularities avoidance of a redundant robot. This latter must carry out a trajectory tracking in the Cartesian space near a singularity point. The singularity avoidance without affecting trajectory tracking is involved via selfmotion method. The analytical determination of this self motion is obtained on the optimization of scalar function depending on the robot manipulability measure. In view to reduce the on line cumbersome computations due to the analytical method, a learning network based anfis is used to generate this self-motion. The learning process uses the input-output data coming from the analytical self-motion. The two methods of avoiding singularity (based on analytical method and on anfis one) are tested in the case of 3 dof planar robot performing, in Cartesian space, a trajectory near a singular point. The obtained results show that the proposed criteria ensure a good control when the robot operates near a singularity point.