Comparison of assembly-jam learning algorithms with fuzzy entropy measure for intelligent robot’s part micro-assembly

• Published in: International journal of advanced manufacturing technology Vol. 104; no. 9-12; pp. 3473 - 3488
• Main Author: Son, Changman
• Format: Journal Article
• Language: English
• Published: London Springer London 01.10.2019; Springer Nature B.V
• Subjects: Algorithms; Assembly; CAE) and Design; Computer-Aided Engineering (CAD; Engineering; Entropy; Industrial and Production Engineering; Machine learning; Mechanical Engineering; Media Management; Original Article; Robot control; Robots; Minimizing control motions; Fuzzy entropy; Task performance time; Comparison (assembly-jam learning algorithms); Measuring uncertainty; Regions minimizing algorithm
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-019-04100-7

Two assembly-jam learning algorithms are introduced for reducing the task performance time as well as getting out of an assembly-jam in a robot’s part micro-assembly. The two algorithms are then compared from the viewpoint of five factors. A comprehensive comparison of the results with other recent methods and discussions are also described. The two algorithms split or unify to simplify regions situated near a similar assembly-jam state on the assembly-jam location map. This allows the part micro-assembly task to be continuously reiterated such that the speed of task performance will be faster. The task for getting out of the assembly-jam is achieved by the fewest number of robot joint control motions. This is generated by a technique to minimize the number of control motions of the 1st algorithm, without wasting time and energy. Meanwhile, the number of the last formed regions is minimized by the region unifying process of the 2nd algorithm. The two assembly-jam learning algorithms significantly reduce the task performance time in micro-assembly processes. The degree of uncertainty (measured by a fuzzy entropy function) associated with the task for getting out of the assembly-jam is used as a criterion to determine the most valid plan for a present input. The results show that the task for getting out of the assembly-jam can be successfully achieved by the control plans generated by the two assembly-jam learning algorithms.