- 學位論文
無線感測網路下使用混合式神經網路室內定位之研究
Research on Hybrid Neural Network Based Indoor Positioning in Wireless Sensor Networks
摘要
近年來，無線感測網路(Wireless Sensor Network, WSN)隨著物聯網的廣泛應用而逐漸熱門，無線感測網路中定位服務(Location-based Service, LBS) 的需求也日漸增長，無論室外或室內對於目標定位且有效取得準確的位置資訊一直是被熱門探討問題之一，室內定位系統相較於室外定位系統來說，精確性要求相對較高，室內環境空間較小，若誤差超過數公尺則失去定位效果，因此本論文將於無線感測網路中藉由混合式人工智慧演算法讓室內定位可以達到快速且高辨識精確的效能，本研究將使用接收訊號強度指標(Received Signal Strength Indicator, RSSI)判斷目標點與參考點距離，且利用數學模型建立通道模型並模擬各種不同的室內環境，在通道模型中加上高斯分佈隨機變數 來模擬實際室內環境中的複雜性，如遮蔽效應、路徑損耗、多路徑效應、雜訊及干擾等影響，並藉由調整標準差參數來定義不同的室內環境。 本論文主要提出兩種定位模式，分別為區域定位及座標定位，區域定位將使用倒傳遞類神經網路演算法、適應性網路模糊推論演算法及基因結合倒傳類神經網路來判斷比較目標點所在的區域位置準確度。為了提高定位解析度進而探討座標定位，我們採用兩種方式，第一種座標定位方法還是利用區域定位三種演算法，第二種則是先利用區域定位估算的區域結果，再進一步使用調整式三邊定位法，與原本演算法本身來比較定位目標點座標精確度，在模擬結果得知三種演算法分別加入調整三邊定位法的座標準確度比演算法本身座標來的較精準。
並列摘要
In recent years, a wireless sensor network (WSN) with the Internet of things widely used and are gradually popular. Demands for Location-based Services in the WSN is alo growing. Therefore, how to effective obtain accurate positioning information has been one of the most popular research issues whether in a outdoor or indoor environment. Compared to outdoor positioning system, accuracy requirement of the indoor positioning system is relatively high. If the positioning error exceeds the number of meters, it would lose the meaning of positioning. This thesis will use hybrid artificial intelligence algorithms in wireless sensor network for indoor positioning to achieve fast and precise locatization. This study will use the Received Signal Strength Indicator (RSSI) to determine the distance between the target point and the reference point, and use the mathematical model to establish the channel model to simulate the various indoor environments. In the channel model, we would add the Gaussian distribution random variable to simulate the complexity of the real indoor environment such as shadowing effect, path loss, multipath effect, noise and interference. We adjust the standard deviation parameters to define various indoor environments. This paper uses two positioning modes, namely, zone positioning and coordinate positioning. In zone positioning, we use Back Propagation Neural Network (BPNN), Adaptive Network-based Fuzzy Inference System (ANFIS), and Genetic Algorithms combine with Back Propagation Neural Network (GA-BPNN) algorithms to determine the positioning of the target node in which correct zone. In order to improve the positioning resolution, we explore the coordinate positioning. We adopt two ways to implement coordinate positioning. First, coordinate positioning still uses the three algorithms in previous zone positioning to locate coordinate position. Second, we use two phase to locate coordinate. In the phase one, we still use the zone positioning to obtain the zone position, and then we further use the adjustment of the trilateral positioning method to locate coordinate in the second phase. From the simulation results, the performance of two phase coordinate positioning of three algorithms with the adjustment of the trilateral positioning method have higher accuracy than that of coordinate positioning of three algorithms.
參考文獻
延伸閱讀
- Weng, T. C. (2002). 無線網路系統之混合式位置追蹤策略：適用於具馬可夫鏈結移動模式使用者 [doctoral dissertation, National Taiwan Normal University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0021-2603200719125071
- 吳鴻志、張振明（2008）。類神經網路應用於室內通道定位之研究。載於樹德科技大學（主編），2008資通技術管理與應用會議（頁617-628）。樹德科技大學。https://www.airitilibrary.com/Article/Detail?DocID=c0000064-200812-x-2008-617-628-a
- Liu, J. C. (2013). 應用於混合式無線感測網路之偵測與估計演算法 [doctoral dissertation, National Taiwan University]. Airiti Library. https://doi.org/10.6342/NTU.2013.00001
- 嚴毅（2016）。Research on Two-stage Fuzzy Inference Based Indoor Positioning Algorithm in Wireless Sensor Networks〔碩士論文，國立虎尾科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0028-1205201615194700
- 許巽堯（2007）。Development of an Automatic Indoor Positioning System in Wireless Sensor Network〔碩士論文，國立臺灣師範大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0021-2910200810575223