不同貨架儲存模式對RMFS物就人揀貨系統運作效能之影響

機器人智動揀貨系統（Robotic Mobile Fulfillment Systems, RMFS）的相關研究，近年來受到廣泛的重視。回顧文獻可以發現，RMFS作業層面的決策問題可大致分為五大類：揀貨訂單指派（Pick Order Assignment, POA）、揀貨貨架選擇（Pick Pod Selection, PPS）、補貨訂單指派（Replenishment Order Assignment, ROA）、補貨貨架選擇（Replenishment Pod Selection, RPS）以及貨架儲位指派（Pod Storage Assignment, PSA）。先前研究指出POA決策對績效影響最大，且建議應深入探討PSA問題。本研究亦認為PSA決策將影響後續商品儲放的布局，對於預測式商品儲放的動態布局乃是重要的機制；亦即此決策具有逐步調整藉以動態適應訂單改變的潛力。因此，本研究聚焦於PSA決策，針對分級儲存策略提出彙集式分級（Aggregated Class）並設計三種計算規則，包含總量占比、品類占比與空間占比；就彙集式分級之PSA指派與其他PSA指派規則進行總移動距離之比較，以及針對訂單長短期趨勢之參數進行敏感度分析。本研究結果發現： 1.在POA採用『訂單與貨架匹配規則』(Pod-Match)、ROA採用庫存量與再訂購數量判斷補貨數量之簡單規則、而PPS、RPS採取不同組合規則時，亦即PPS採用『貨架品項總數量+最近規則』且RPS採用『最近規則』時，總移動距離表現最好，且約僅為『隨機規則』-『隨機規則』組合之20%~25%。 2.當PPS採用『貨架品項總數量+最近規則』、RPS採用『最近規則』時，PSA採取本研究提出的各種分級指派規則，經統計檢定均顯著優於隨機指派、最近指派與固定指派規則。 3.本研究提出的各種分級指派規則之間並無顯著差異；總體而言，總量占比之彙集式分級加上最近法則之成效最好。 4.以實際案例進行測試，總量占比之彙集式分級加上最近法則成效亦為最好。 5.以長期（半年）與短期（一月）訂單趨勢之加權變動而言，本研究之測試資料所呈現長期（半年）與短期（一月）訂單趨勢之加權組合，可明顯發現短期訂單趨勢決定較佳的運作表現，且當短期加權增加時仍維持結果的一致性。綜上可知， PPS與RPS之組合對於貨架移動距離有顯著影響；在PSA方面，本研究所提出之分級指派規則，均顯著優於其他指派法則，能降低貨架移動距離。此外，長期及短期訂單趨勢之權重組合分析並未對於貨架移動距離產生影響。本研究也針對實驗中所遇到模型太小無法有產生足夠的效果提出建議，並提供未來可持續研究之方向，以期在本研究基礎上能產生更豐富的研究成果，為實務帶來貢獻。

Keywords

機器人智動揀貨系統；分級儲存策略；貨架儲位指派；動態適應

Parallel abstracts

In recent years, research on Robotic Mobile Fulfillment Systems (RMFS) has garnered significant attention. A review of the literature reveals that decision-making issues at the operational level of RMFS can be broadly categorized into five types: Pick Order Assignment (POA), Pick Pod Selection (PPS), Replenishment Order Assignment (ROA), Replenishment Pod Selection (RPS), and Pod Storage Assignment (PSA). Recent studies indicate that POA decisions have the most substantial influences on performance and suggest that the PSA problem deserves further investigation. Given such understanding, this research focuses on PSA decisions, proposes an aggregated class-based storage policy, and designs three computational rules including total volume ratio, category ratio, and space ratio to investigate their influences on the total movement distance. The study compares the total movement distance of PSA assignments using the aggregated hierarchical strategy with other PSA assignment rules and conducts a sensitivity analysis of parameters for short-term and long-term demand trends. The research findings are as follows: 1.When POA adopts the “Pod-Match” rule, ROA uses simple rules based on inventory levels and reorder quantities to determine replenishment quantities, and PPS and RPS adopts the “Pile-on + Nearest” rule and the “Nearest” rule repsectively, the total movement distance is the shortest, accounting for about 20%-25% of those from the “Random-Random” combination. 2.When PPS uses the “Pile-on + Nearest” rule and RPS uses the “Nearest” rule, PSA adopting various hierarchical assignment rules proposed in this study significantly outperforms the “Random”, the “Nearest”, and the “Fixed” assignment rules. 3.There are no significant differences between the various hierarchical assignment rules proposed in this study. In brief, the aggregated hierarchical rule based on the total volume ratio combined with the “Nearest” rule performs the best. 4.A number of tests show that the aggregated hierarchical rule based on total volume ratio combined with the “Nearest” rule also has the best performance. 5.Using both the long-term (six months) and short-term (one month) orders, results from some tests clearly show that the short-term orders lead to better operational performance. Furthermore, the results show no differences when the short-term weighting increases. In summary, the combination of PPS and RPS has a significant influence on the total pod movement distance. Regarding PSA, the hierarchical assignment rules proposed in this study significantly outperform other assignment methods, reducing pod movement distance. Additionally, the weights of long-term and short-term demand trends does not affect the total movement distance.