無線感測網路中同時考慮配置與排程之和聲搜尋演算法

Abstract

物聯網(internet of things; IoT)的逐漸興起改變了資訊傳遞的方式，進而帶動智慧都市的推行。其藉由將無線感測網路(wireless sensor networks; WSN)架設於都市環境中來收集與建立資料，再經由雲端中心計算，使IoT能持續運作，提供居民方便的服務。在WSN的研究中，有關其存活時間的議題一直深受人們熱烈的討論，這牽涉到有關感測器覆蓋的配置問題和感測器休眠的排程問題，然而，在過去的研究中，針對有關配置與排程的問題，往往都是分開來各別探討或採用各自獨立的兩階段(配置階段和排程階段)方法處理，如此容易忽略兩者間緊密的關聯性。因此，本文將針對聯合感測器配置與排程的問題進行探討，不同於過去各自獨立的兩階段處理，而是在配置階段中同時考慮配置與排程的關係，並透過所提出的改良式幾何選擇和聲搜尋演算法(improved geometric selective harmony search; IGSHS)來解決問題。該演算法主要是以幾何選擇和聲演算法(geometric selective harmony search; GSHS)為基礎，並結合了交配機制和動態調整機制，前者可以有效解決編碼中不同決策變數的問題；後者可以增加求解的穩定性與多樣性。在分析上，將考慮實驗不同的覆蓋需求(simple、k、Q覆蓋)和感測器數量組合，並考量更實際的參數設定，用以比較本研究所提的方法(在配置階段考慮配置與排程)和過去的方法在WSN存活時間上的差異，以確認此方法在實際問題運用上的價值。

The rise of internet of things (IoT) changes the way of information transfer and then drives development of smart city. By construction of wireless sensor networks (WSN) in city, it can collect data and establish data. Then through the cloud computing center, which enables IoT to continue to operate and provides convenient services. In the WSN research, the issue related to the lifetime has been a lively discussion by people. This involves deployment problem about sensors coverage and sensor scheduling problem about sensors sleep. However, in the previous studies, these two problems were discussed separately or using two separate stages (deployment stage and scheduling stage) to solve, so that their close relation were neglected. Therefore, we focus on the joint sensor deployment and scheduling problem are discussed. Unlike the past two separate stages, we taking into account the deployment and scheduling in the deployment stage. This research further proposes an improved geometric selective harmony search (IGSHS) algorithm to solve this problem. The algorithm mainly uses geometric selective harmony search (GSHS), combined with the crossover mechanism and dynamic adjustment mechanism. The former can be an effective solution of different decision variables in coding and the latter can increase the stability and diversity of solving. In the experiment, considered experimenting with different coverage requirements (simple, k, Q coverage) targets and different number of sensors, and consider more practical parameter settings. Compare our methods and past methods differences in lifetime.