Similarity measuring strategy of image patterns based on fuzzy entropy and energy variations in intelligent robot’s manipulative task

Abstract

A similarity measuring strategy of image patterns based on fuzzy entropy and energy variations, using an intelligent robot's part macro-assembly (part-bringing) as an example, is presented. A part macro-assembly, locating various shaped assembly holes (targets) in a workspace corresponding to shapes of parts and then bringing a part to a corresponding target for the purpose of part mating despite existing obstalces, is introduced. This is accomplished by cooperating a neural network system with a fuzzy optimal control. Fuzzy entropy and energy functions, which are useful measures of variability and information in terms of uncertainty, are introduced to measure its overall performance of task execution related to the part-bringing task. An interrelation among learning, fuzzy entropy, and energy variations used as a measuring tool for a degree of similarity of image patterns is described. Through variations of fuzzy entropy and energy, a degree of similarity between input and desired output image patterns of neural network can be measured. The proposed technique is not only a useful tool to measure a degree of similarity between image patterns, but applicable to a wide range of robotic tasks including motion planning, manufacturing, maneuvering around workspace, and part mating with various shaped parts and targets.