Intelligent jamming region division with machine learning and fuzzy optimization for control of robot's part micro-manipulative task

An algorithm for an intelligent jamming region division with machine learning and fuzzy optimization for the control of a robot's part micro-manipulative task is introduced. A comparison with existing works and the advantages of the proposed algorithm in this paper are described. A quasi-static part mating (micro-assembly) is accomplished using a fuzzy coordinator combined with a learning algorithm of the jamming region division while avoiding jamming. Depending on the positional relationships between a part and an assembly hole (target) in a workspace, a specific rule base for avoiding jamming is activated. The region division algorithm merges all adjoining subregions, of which the quad-tuple control values describe similar jamming states, into one region and the weights of the subregions are adjusted. A fuzzy entropy, which is a useful tool for measuring variability and information in terms of uncertainty, is used to measure the degree of uncertainty related to an execution of the part micro-assembly task. A degree of uncertainty associated with a task execution of the part micro-assembly is used as a criterion of optimality, e.g. minimum fuzzy entropy. Through a decision-making procedure, the most appropriate quad-tuple control value with the lowest fuzzy entropy in each region is chosen as a final control value to carry out an assigned task. The proposed technique is applicable to a wide range of the robot's tasks, including choosing and placing operations, manufacturing tasks, part mating with various shaped parts, etc.