Generating Performance Criteria for Sensorimotor Trainer Sets of a Robot Manipulator
Mert Uneri · 1992
We consider intelligent wrist positioning control for dextrous manipulation in unstructured, uncertain, changing environment. Robot environment is modeled as a cellular multiwell vector field generated by attractor (repeller) cells identified as task relevant (irrelevant) graspable parts. Uncertainty in identification is assigned as entropy measure to the each corresponding cell. The paper analyzes the response surface of robot dynamics excited by the attractor-repeller vector field using the Cell-to-Cell-Mapping concept. We generate basins of attraction of an attractorrepeller multiwell vector field configuration and develop its performance loss measures based on basin erosion in control space. These measures are needed as design criteria in the space of system parameters (control space) for an optimum trainer vector field configuration in an intelligent control. Numerical and simulation results are given for a 2 DOF planar manipulator dynamics.