虛擬網路蟻群最佳化於儲位重整問題之研究

本研究是期望能輔助產業在其倉儲空間範圍有限的條件下，將目前存貨配置混亂的狀態，透過儲位重整方法，使散佈於不同棧板之存貨分類，恢復至原先存貨分類配置方式或是預先所擬定的分佈狀態，讓企業可享有良好完善的儲位規劃，使管理者能迅速瞭解並掌握存貨的情況與安排，進而提升整體作業流程之效率。
儲位重整問題乃屬於組合最佳化問題，問題本質適用於蟻群最佳化演算法，故採用巨集啟發式演算法中的蟻群演算法加以求解，但由於本研究中的儲位重整問題其網路架構為「樹狀結構」，為避免記憶體的使用過於龐大，造成求解上的瓶頸。因此本研究發展出「虛擬網路」之構想並將費洛蒙濃度紀錄於各棧板上，來大幅降低存取記憶體所需的容量，藉由蟻群最佳化演算法結合虛擬網路之架構，發展一套求解儲位重整問題的方法，進行「最小化總棧板搬移次數」的求解搜尋與收斂。本研究測試了數個不同性質的測試例，並探討問題規模對於求解效果之影響；在測試結果中發現虛擬網路蟻群最佳化在與其他演算法相互比較下，在問題規模為50、100及200棧板時，其改善題數較為顯著；而在問題規模為30棧板時，平均差值改善較佳，証明本研究虛擬網路蟻群最佳化演算法是有其效益的。

The study is aimed to assist firms in reforming inventories which are presently in great chaos. By using storage recombination methods, the mixed-up storages can be classified from different pallets and placed into the original or other stocking arrangements. Hence, the enterprises can have better storage planning, which allows the managers to have efficient control of the stocks. Even, the whole operating process can be more efficient as well.
Storage recombination problems belong to the problems of combination optimization. It is adequate to use the ant colony optimization of meta-heuristic algorithm to solve this type of problems. However, because the storage recombination problem in this study is constructed as “Tree Structure”, to avoid the huge ram usage while doing the calculation, “virtual network” concept is developed and pheromone is recorded down on every pallet. The combination of ant colony optimization and virtual network is developed to process optimum searching of “minimum total pallet movements”. The study examines several examples of different nature and discusses the effect of problem scale on optimum searching. The findings show that compared with other algorithms, the method of virtual-network ant colony optimization has better results when there are 50, 100 and 200 pallets. Especially when the problem scale is 30 pallets, the improvement is significant, which proves the method of virtual-network ant colony optimization is effective.