DSpace collection: 博碩士論文

零工式排程中以最小化完工期與延遲階段作業數為目標之雙目標分支定界演算法;A bi-objective branch-and-bound algorithm for solving a job shop scheduling problem when minimizing makespan and total number of tardy stage-outs

Fri, 06 Mar 2026 10:19:47 GMT

title: 零工式排程中以最小化完工期與延遲階段作業數為目標之雙目標分支定界演算法;A bi-objective branch-and-bound algorithm for solving a job shop scheduling problem when minimizing makespan and total number of tardy stage-outs abstract: 本研究探討一個具有多層級作業結構的零工式排程問題（Job shop scheduling problem），同時最小化最大完工時間（Makespan）與層級延遲完成次數（Total number of tardy stage-outs）。其中，tardy stage-outs 為管理層常用之階段性進度評估指標。為處理此一特性，本研究建構分離弧線圖（Disjunctive graph），表示多層級作業間的先後順序與機台資源使用關係，並在指定層級設置虛擬終點以判斷是否如期完成。本研究提出一套分支定界（Branch and Bound）方法以解決此雙目標問題。在分支策略中，演算法會計算所有可排程作業的最早完成時間，據此選定機台，並從中篩選具最早啟動潛力的作業進行分支。分支前，根據目標層（Interest layers）與共同到期日（Common due date）條件，預先排除可能影響目標層準時性的非關鍵作業，以降低延遲風險。在下界估算部分，makespan 下界參考單機排程問題理論，評估每台機台在不可搶佔條件下的最早完成潛力，取最大者為全域下界。tardy stage-outs 部分，針對管理層關注的目標層，預估其最早完工時間，若已超出指定期限，即視為必然延遲，納入遲交下界計算。為提升搜尋效率，本研究設計兩項剪枝條件。第一，於分支前排除非目標層的作業；第二，比較節點下界與目前 Pareto 前緣中的可行解，若其被支配則直接剪除。實驗結果顯示，在多數測試實例中，節點數可減少逾九成，顯著提升效能。綜上所述，本研究方法可應用於多層級製程環境中兼顧長短期目標的排程問題，提供對應的部分 Pareto 前緣解集合，協助決策者於雙目標間進行權衡排程。;This study investigates a job shop scheduling problem with a multi-layered operation structure, aiming to minimize both the makespan and the total number of tardy stage-outs. Among these, tardy stage-outs are used by managers as a key indicator of short-term progress performance. To address this, a disjunctive graph is constructed to model the precedence relationships and machine resource constraints among operations across layers. Sink nodes are added at the end of designated layers to evaluate their on-time completion status. A branch and bound method is proposed to explore the solution space efficiently. During branching, the algorithm first identifies the schedulable operation with the earliest possible completion time, selects its corresponding machine, and filters operations on that machine with the potential to start earliest. Prior to branching, operations that may delay the completion of interest layers are removed, based on the due dates specified for those layers and the common due date constraint, thereby reducing the risk of short-term violations. For lower bound estimation, the makespan bound refers to a classical single-machine scheduling formulation. The lower bound for each machine is calculated under non-preemptive conditions, and the global makespan lower bound is taken as the maximum of these values. For tardy stage-outs, the evaluation focuses on layers of managerial interest. If the earliest possible completion time already exceeds the due date, the layer is considered inevitably tardy and contributes to the lower bound calculation. To improve search efficiency, two pruning rules are developed. First, operations not associated with interest layers are excluded during branching. Second, nodes whose lower bound vectors are dominated by current Pareto-optimal solutions are pruned. Experimental results show that these pruning strategies significantly reduce the number of nodes explored, with reductions exceeding 90% in some cases, thereby enhancing computational performance. The proposed algorithm is suitable for scheduling problems involving both long-term and short-term objectives in multi-layered environments. It is especially applicable to semiconductor manufacturing settings that emphasize stage-based progress control. Under the constraint of a user-specified allowance for tardy layers, the algorithm provides partial Pareto front solutions to support trade-off scheduling decisions.

Contract Design and Energy Allocation of Aggregators in Renewable Energy Supply Chain;Contract Design and Energy Allocation of Aggregators in Renewable Energy Supply Chain

Fri, 06 Mar 2026 10:19:32 GMT

title: Contract Design and Energy Allocation of Aggregators in Renewable Energy Supply Chain;Contract Design and Energy Allocation of Aggregators in Renewable Energy Supply Chain abstract: 隨著氣候變遷議題升溫，全球已有逾 130 個國家宣示「2050 淨零排碳」目標並提出行動藍圖。再生能源因而成為能源轉型的關鍵要角。台灣 2017 年通過《電業法》修正案後，開放民間企業參與再生能源之發電與售電，進而促使聚合商（Aggregator）等新興業者投入市場。聚合商透過整合上游再生能源發電業者的電力，除可協助電力供需平衡外，亦能為企業客戶提供購電服務與再生能源憑證（Renewable Energy Certificate, REC）交易平台。在企業碳中和壓力與國際供應鏈減碳要求同步升高的情況下，REC 與企業電力購售協議（Corporate PPA）已成為台灣再生能源交易的重要管道，也為聚合商創造嶄新的商業模式。本研究首先以單一固定用電需求企業為案例，剖析聚合商在再生能源供應不確定情境下，如何透過「整合再生能源＋REC 交易」創造價值；接續，再將模型擴展至供應不足下之公平分配機制（比例分配、受限等額補償、受限等損配分），比較不同機制對聚合商利潤之影響。研究結果可為未來台灣電力交易設計與政策制訂提供量化參考，並進一步說明當聚合商兼具整合與交易功能時，所能帶來的附加價值與投資誘因。;With the growing urgency of climate change issues, over 130 countries worldwide have declared their “2050 Net Zero Emissions” targets and launched corresponding roadmaps. Renewable energy has thus become a key driver of energy transition. In Taiwan, the Electricity Act was amended in 2017 to allow private enterprises to participate in the generation and sale of renewable energy, thereby promoting the emergence of new market entrants such as aggregators. Aggregators integrate electricity from upstream renewable energy generators, not only helping to balance electricity supply and demand, but also providing electricity procurement services and access to Renewable Energy Certificate (REC) trading platforms for corporate customers. Amid mounting pressure and international requirements for carbon reduction, the demand for RECs and corporate power purchase agreements (Corporate PPAs) has risen significantly. These mechanisms have become essential instruments in Taiwan’s electricity market, offering aggregators innovative business models. This study begins by taking a fixed-demand electricity-consuming enterprise as a case to analyze how aggregators can create value under uncertain renewable energy supply conditions through “integrated renewable energy and REC transactions.” Subsequently, the model is extended to scenarios with insufficient supply and compares the impacts of different fair allocation mechanisms (such as proportional allocation, capped compensation, and capped loss-sharing) on aggregator profitability. The results may offer quantitative references for future electricity market design and policy formulation in Taiwan. Furthermore, the study explains the additional value and investment incentives that can be generated when aggregators also serve as integrated trading agents.

分區揀貨倉庫下類Locusbots系統之揀貨機器人派送問題研究;A Study on the Dispatching Problem of Picking Robots in Warehouses under LocusBots System

Fri, 06 Mar 2026 10:19:11 GMT

title: 分區揀貨倉庫下類Locusbots系統之揀貨機器人派送問題研究;A Study on the Dispatching Problem of Picking Robots in Warehouses under LocusBots System abstract: 隨著資訊科技的飛速發展及行動網路的普及，電子商務已成為全球經濟的重要發展趨勢。市場需求逐漸轉向「少量、多樣化」的模式，這一變化使物流中心的運營複雜度提高。根據De Koster et al.（2007）的研究，目前大部分物流中心仍然依賴大量人力，其中與揀貨作業相關的勞動力成本占比超過50%，導致營運成本居高不下。因此，如何透過導入自動化設備與優化揀貨策略，以降低成本並提升效率，成為企業關注的重點。在此背景下，類Locusbots系統的運用顯示出優勢，其具備彈性調整機器人數量的能力，使其無論在訂單高峰期或淡季皆能適應變化。此外，區域揀貨方式能有效提升作業效率，讓揀貨人員僅需在指定區域內作業，無須長距離移動，大幅降低人力需求，同時提升揀貨精準度與作業速度。揀貨機器人則負責將訂單與揀貨箱運送至指定區域，使整體物流作業更加流暢。本研究以類Locusbots系統為基礎，聚焦於機器人在分區揀貨倉庫中的運作方式，並針對其揀貨流程提出兩種主要策略：「先選擇Zone再選擇Block」與「直接選擇 Block」。研究進一步探討這兩種流程所涉及的關鍵議題，即「機器人的Zone選取問題」與「機器人的Block選取問題」。最後，透過Arena模擬軟體進行測試與數據分析，評估不同策略在各項績效指標上的影響，並比較兩種揀貨流程的優劣，期望研究成果能對未來相關議題提供參考與貢獻。;With the rapid advancement of information technology and the widespread adoption of mobile networks, e-commerce has emerged as a major trend in the global economy. Market demand has gradually shifted toward a "small-quantity, high-variety" model, increasing the operational complexity of distribution centers. According to De Koster et al. (2007), most distribution centers still heavily rely on manual labor, with labor costs related to order picking accounting for more than 50% of total operational expenses. As a result, enterprises are increasingly focused on how to reduce costs and improve efficiency through the implementation of automated equipment and the optimization of picking strategies. Against this backdrop, systems similar to LocusBots have demonstrated distinct advantages. Their ability to flexibly adjust the number of robots allows them to adapt to both peak order periods and off-seasons. Moreover, zone-based picking can significantly improve operational efficiency by limiting pickers to specific areas, thereby reducing unnecessary travel distance, lowering labor demand, and enhancing both picking accuracy and speed. The picking robots are responsible for transporting order bins to designated zones, which streamlines the overall logistics workflow. Building on the concept of LocusBots, this study focuses on the operational approach of robots in zone-based picking warehouses. Two main strategies for the picking process are proposed: (1) selecting a zone first and then selecting a block, and (2) directly selecting a block. The study further investigates the two key issues involved in these processes: the robot′s zone selection strategy and the robot′s block selection strategy. Finally, simulation experiments are conducted using Arena software to evaluate the performance of different strategies across various metrics, aiming to compare the effectiveness of the two processes and provide insights for future research in this area.

在分區揀貨倉庫下類 Locusbots 系統的揀貨人員 Robot 選取問題

Fri, 06 Mar 2026 10:19:04 GMT

title: 在分區揀貨倉庫下類 Locusbots 系統的揀貨人員 Robot 選取問題 abstract: 隨著電子商務快速發展與倉儲自動化技術成熟，物流中心面臨愈趨複雜的揀貨挑戰。尤其在協作型移動機器人（如 LocusBots）導入後，人機協作的揀貨流程設計成為倉儲優化核心課題。本研究以分區式倉儲環境為情境，系統性探討作業流程設計（Type-I 與 Type-II）、資源配置（Robot 數量, RN）、與揀貨人員決策策略（Block 選取法則 PB 與 Robot 選取法則 PR）之交互影響。Type-I 流程為揀貨員先選擇區塊 (Block) 後選 Robot，Type-II 流程則跳過區塊選擇，直接由整體可用 Robot 中挑選。透過大規模模擬實驗與多因子變異數分析，本研究針對系統總執行時間（Total System Time, TST）與訂單在系統內總時間（Total Time In System, TTIS）兩項績效指標進行系統性探討。結果顯示，資源配置與決策法則設計對系統績效均具顯著影響。Robot 數量提升雖能降低整體完工時間（TST），但同時增加系統內排隊等待，導致訂單端停留時間（TTIS）上升，呈現產出效率與服務品質之間的績效權衡現象。法則設計方面，具備前置規劃的 Type-I 流程與以最短旅行距離為核心的SDPR法則，在多數情境下皆展現穩健優勢，惟在資源極度充裕且搭配頂尖法則時，Type-II 流程亦能展現潛在極限效能，突顯流程、資源與策略三者之整體匹配性。本研究成果除驗證各作業流程與決策法則對系統效益的穩定貢獻，亦為倉儲業者在流程設計與資源配置決策上提供具體參考依據，並為未來智慧倉儲人機協作優化研究奠定實證基礎。 ;As e-commerce grows and warehouse automation advances, distribution centers face increasingly complex picking challenges. With the introduction of collaborative mobile robots (such as LocusBots), designing efficient human-robot collaborative processes has become a core optimization issue. This study investigates a zone-based warehouse system, focusing on the interactions among process design (Type-I and Type-II), resource allocation (number of robots, RN), and picker decision rules (Block Selection Rule, PB, and Robot Selection Rule, PR). In Type-I, pickers first select a Block before choosing a Robot within it; in Type-II, pickers directly select from available Robots in the Zone, bypassing Block selection. Large-scale simulations and multi-factor ANOVA were conducted, evaluating system performance based on Total System Time (TST) and Total Time In System (TTIS). Results show that increasing robot numbers reduces TST but increases TTIS due to queuing effects, revealing a trade-off between throughput and order responsiveness. Type-I processes, combined with distance-based Shortest Travel Distance Selection (SDPR), consistently achieve robust performance across scenarios. However, under abundant resources and optimal decision rules, Type-II can outperform Type-I, highlighting the importance of aligning process, resource, and decision strategies. The study offers insights for warehouse managers in designing picking processes and resource planning, while contributing to future research on intelligent human-robot collaboration in warehouse systems.