一個能夠滿足群體需求的兩階段資源分配機制---以智慧家庭房間電力分配為例

Abstract

隨著無所不在的運算興起，環境狀態可以精確、即時的捕捉，加上人性對於便利生活的需求，人們開始關注如何利用情境資訊、察知人的需求，提供自動化服務協助日常活動，現代社會步調快速，大至國家、小至家庭，都有能因應使用者需求與環境變化，立即對握有的資源作出適切分配決策的需要，然而跨越多個目標群體的分配決策與協調工作繁雜，因此，本論文提出一個能滿足群體需求的兩階段資源分配機制作為解決方案，此機制具備快速的整合多個使用者意見、分析處理情境資訊、與快速對有限資源做出優化分配決策的能力。組成此機制的兩階段功能包含：第一階段以修改Fuzzy TOPSIS為方案評價機制，提出貼近群體利益的可接受方案集，結果顯示前五名評價方案中，鄰近排名方案平均效用差為+6.71%、平均認同差為+27.36%，此方法確實給予兼具群體效用與認同者較高的評價；第二階段是具情境感知且能快速達成整體利益優化的資源分配子機制，實驗結果顯示本論文之方法與Akbar等人的方法解價值差平均在+0.06%到+7.67%之間。智慧家庭案例模擬結果顯示，啟用兩階段資源分配機制犧牲了5.40%的群體舒適評價為代價可以節省高達19,48%的電費，能夠有效統籌運作家電，在綠能目標下達成整體成員舒適優化的效果。

With the advent of technologies and the needs of convenient life, people began to focus on how to employ contextual information and provide automated services to support daily. How to coordinate across multiple group and make a good resource allocation is a complex problem. However, fast-paced modern society has made organizations of any size in search of such capability to utilize the resources fast and appropriately depending on changes in user needs or environment state. This thesis proposed a two-phase resource allocation mechanism as a solution. The first is an altered Fuzzy TOPSIS to evaluating alternatives for reaching consensus in each group. Simulation results present that for the top five alternatives the average utility difference to the next alternative in percent is +6.71%, the average approval difference to the next in percent is +27.36%. The second is a context-aware resource allocation process, mainly solving a MMKP. Simulation results show that the proposed algorithm performs average 0.05%~7.67% better than Akbar et all's method. In the smarthome scenario simulation, enabling the proposed mechanism with 24 NTD can save 19.48% cost but only lose 5.40% comfort evaluation for total users.