運用基因演算法求解全供給快速配送問題

隨著快速配送的興起，有越來越多的研究探討快速配送相關的議題。在其中，一個深具潛力的應用模式是全供給 (Omni Supply) 模式，此模式提議「整合一區域內的供給點，讓消費者僅需一次下單即可一次購足所需商品，並獲得單一配送」。先前有關全供給的研究雖已有初步的成果，然而，仍存在若干議題亟待更進一步地探討。譬如，前述的研究只進行單目標的求解，不利管理者進行多目標決策的制定。此外，其所提供可行解的數量有限，面對消費者與管理者各種不同的需求，應該要有更佳的方法來提供更多的解決方案。為改進前述研究不足之處，本研究運用「雙目標基因演算法」(Bi-objective Genetic Algorithm)來求解全供給快速配送問題。本研究參考ESG (Environmental, Social, and Governance)的精神，同時考慮環境面的「最短總距離」(因此產生最少的碳排放量，對環境有利)和社會面的「最低總價錢」(因此消費者可用較低的價錢購物，對消費者有利)兩個目標來進行求解。由於上述兩個目標是互相衝突的，本研究也嘗試從治理的角度來探討各種解決方案。研究結果顯示，採用「雙目標基因演算法」能夠有效求得許多柏拉圖妥協解，使管理者更容易進行多目標決策制定。另外，只需透過簡單的權重設定，本研究所提出的方法可以快速地根據管理者的需求尋找出適當的解。在基因參數方面，不同的交配率和突變率對結果的確會有影響，但影響的程度有限。

Keywords

快速配送；全供給；雙目標基因演算法；柏拉圖解； ESG

Parallel abstracts

With the rise of express delivery, more and more research is being conducted to explore issues related to express delivery. Among them, an application model with great potential is the Omni Supply model, which proposes to integrate supply points within a region so that consumers only need one-stop ordering to purchase all necessary items and receive one-trip delivery (One-stop Ordering and One-trip Delivery). One previous study on Omni Supply has obtained some good results; however, several issues still require further investigation. For example, the previous study only dealt with single-objective optimization, which disadvantaged managers in making multi-objective decision formulations. Furthermore, the number of feasible solutions provided is limited. There should be a better method to provide more solutions when facing the various needs of consumers and managers. To improve the shortcomings mentioned above, this study employs a Bi-objective Genetic Algorithm to solve the express delivery problem for Omni Supply. This study refers to the concept of ESG (Environmental, Social, and Governance) and considers two objectives: the shortest total distance from the environmental viewpoint (resulting in the least carbon emissions, beneficial to the environment) and the lowest total price from the social viewpoint (enabling consumers to shop at lower prices, beneficial to consumers). Since the above two objectives conflict with each other, this study also attempts to explore various solutions from a governance perspective. The results from a variety of experiments show that employing the Bi-objective Genetic Algorithm can effectively obtain numerous Pareto-optimal solutions, making it easier for managers to make multi-objective decisions. In addition, the method proposed in this study can quickly find the appropriate solutions according to the needs of managers by simply setting weights. As for the genetic parameters, different crossover rates and mutation rates do influence the results, but the extent of the influence is limited.