應用隨機演算法於動態式機器人裝配之研究

Abstract

機器人已普遍應用於裝配工作，而其裝配時間，深深影響生產成本及產能。因此如何縮短機器人裝配時間，為一個非常重要之課題。影響裝配時間之因素有三項：１. 機器人動作２.裝配順序３.零件位置排列。 機器人裝配動作可分為固定式 (Fixed Pick-and-Place，簡稱FPP) 與動態式(Dynamic Pick-and-Place，簡稱DPP)。過去在DPP研究中，零件槽排列採用隨機(random) 方式或啟發式 (heurstic) 方法，裝配順序之安排則利用TSP (TravelingSalesman Problem) 方法。基本上，TSP方法是適用在固定座標中安排行走最短之距離；但在DPP中，裝配板、零件槽或機器人等在不同移動速度之下，裝配點或零件槽位置間之相對座標隨時在改變，均有可能影響最佳裝配順序與零件位置之排列。換句話說，DPP模式為一動態問題，而TSP方法並未考慮到此點。 本文提出基因演算法、模擬退火法及塔布搜尋法等三個較適合解決動態問題之隨機演算法來對零件排放位置、裝配順序作更佳之安排，使機器人裝配時間更耗。模擬實驗結果顯示本文所提的方法優於傳統方法。

The industrial robot has been applied widely in manufacturing. The robot assembly time is heavily related to the production cost and capacity. Therefore, how to reduce the robot assembly time is a crucial issue. Two types of robot assembly motion have been characterized: (1) fixed robot motion between fixed pick and place (FPP) points and (2) robot motion with dynamic pick and place (DPP) points. Three factors highly influence the robot assembly efficiency: (1) robot motion control, (2) the sequence of placement point, and (3) the magazine assignment. In the robotics assembly problem, the coordinates of assembly point and magazine are reacted dynamically so that the evaluation function is extremely complicated. To route robotics travel, most investigations have utilized the fixed coordinate of insertion points and magazine of the Traveling Salesman Problems (TSP) method to sequence the insertion points after arbitrarily assigning the magazine. However, robotics travel routing should be based on a relative coordinate so as to obtain a better solution because the robotics, board and magazine are simultaneously moved at different speeds during assembly. To resolve such a dynamically combinatorial problem, this dissertation presents the Genetic Algorithm (GA), Simulated Annealing (SA), and Tabu Search (TS) based procedures. These approaches can simultaneously arrange the insertion sequence and assign the magazine slots by the computer and yield a better performance than in the conventional approach. Results presented herein also demonstrate that the larger the number of insertion points and/or part numbers implies a better performance. These approaches are also compared.