無線感測網路下基於二階段模糊推論室內定位法之研究

Research on Two-stage Fuzzy Inference Based Indoor Positioning Algorithm in Wireless Sensor Networks

嚴毅(Yi Yan)

指導教授 ： 鄭佳炘

國立虎尾科技大學/電機資訊學院/電機工程系碩士班/碩士(2016年)

摘要

近年來，定位服務(Location-based Service, LBS)隨著無線感測網路的廣泛應用而逐漸熱門，無論在室外或室內，對定位服務的需求將會快速成長。如何快速且方便取得準確的位置資訊將是一門重要的研究課題，許多學者也展開積極的研究。在許多定位技術中，全球定位系統(Global Position System, GPS) 因其有著高可用性、可靠性、可提供多種定位精確度且擁有廣泛的應用領域而被廣泛應用。但對於室內環境而言，一方面衛星信號難以穿透建築物而失去定位作用，另一方面，室內環境存在著多路徑效應、遮蔽效應及人員走動所帶來無可避免的干擾，使得GPS在室內定位的效果很難同時兼顧精確度與穩定度，因此尋找其他適用於室內環境的定位系統，已成為學術界的研究重點。
本論文於無線感測網路中使用人工智慧室內定位演算法，藉由人工智慧演算法讓無線感測網路之室內定位達到解析度高且辨識精確的效能。我們利用接收訊號強度指標(Received Signal Strength Indicator, RSSI)判斷室內環境中目標點與參考點之間的距離，並藉由數學模型的方式建立通道模型並模擬RSSI值，在通道模型中加上高斯分佈隨機變數 來模擬現實室內環境中的遮蔽效應、路徑損耗、多路徑效應、雜訊及干擾等影響而造成接收端所收集到RSSI值的不穩定性，並藉由調整標準差參數來定義不同的室內環境。
為了精準的找出目標點位置，本論文提出兩階段的定位方法，第一階段定位為初估，第二階段定位為細估。第一階段定位以區域的概念將3公尺 3公尺的室內空間範圍進行區域劃分，並結合模糊推論系統判斷目標點所在的區域位置，提高區域定位之正確判斷率，從模擬結果得知，結合模糊推論之矩形區域劃分法在不同標準差參數的情況下有73%至83%的正確判斷率。為了進一步將定位的解析度由區域提高至座標，我們在第二階段根據區域判斷的位置來進行三邊定位法，為了降低計算複雜度，本論文提出線性插值測距法來估算目標點與參考點之間的距離。然而，估測距離的誤差將會影響到定位的精準度，我們考慮到傳統的三邊定位法無法應用在估算距離有誤差的情況，進一步提出調整式三邊定位法，從模擬結果得知，以結合模糊推論之矩形區域劃分法進行第二階段定位，其誤差範圍約在7公分至48公分左右。在本論文提出的二階段定位法中，不論是在區域判斷正確與否，皆使用多組定位參考點選擇機制進行運算，即便在區域判斷錯誤時，此機制也可以通過額外選擇鄰近的參考點組合進行定位來平均誤差，達到縮小定位誤差的效果。
本論文所提出的兩階段模糊推論定位法只適用於3公尺 3公尺的小尺度範圍。為了解決此限制，我們提出擴展式定位法，將3公尺 3公尺的小尺度範圍視為一個單位拓樸並向外擴展至6公尺 6公尺，利用判斷式找出較適合的定位拓樸，接著利用該拓樸內五個參考節點為基準進行二階段模糊推論定位。從模擬結果得知，當取樣點落在非重疊拓樸之範圍時可達到接近100%的判斷拓樸效果；若在拓樸重疊之範圍時，則具有拓樸擇優之效果，在中央及角落擇優區域的誤差平均值約11至在13公分；二擇一之區域的誤差平均值約在19至21公分。由此得知，本論文提出的擴展式定位法除了能應用在較大的室內面積下，同時也能選擇適合的單位拓樸進行二階段定位，並且保有原本的定位解析度。

關鍵字

RSSI ； 模糊推論系統 ； 三邊定位法

並列摘要

With the development of Wireless Sensor Networks (WSNs), the demand for Location-based Service (LBS) grows rapidly. The provision of location information is very important for LBS and has drawn much academic interest. Among the existing localization techniques, Global Positioning System (GPS) is widely used in many applications. However, this technique is not applicable to the indoor environments. Due to the blockage, multi-path, interference, etc. Therefore, it is urgent to indoor localization techniques.
This thesis aims to utilize the computational intelligence based indoor positioning algorithm to achieve high positioning accuracy and resolution for WSNs. The received signal strength indicator (RSSI) is used to estimate the distance between the target node and the reference nodes in the indoor environment. A propagation channel model generated form real indoor environments, it considered to describe the effects of shadowing, multipath effect, path loss, noise and interference. We also vary the standard deviation of the model to identify various indoor environments.
We propose a two-stage fuzzy inference based indoor positioning algorithm. In the first stage, a zone-based positioning method is applied in the indoor space and divided into several zones. The fuzzy inference based method is given to find out the correct zone. Where the target node is located and thus improve the performance of zone-based positioning. In the simulation results, the correct judged rate is 73% to 83% with fuzzy rectangular split method in various indoor environments.
In the second stage, the positioning resolution will promote to coordinate-based positioning. A trilateration technique is applied to position the target node by using the judged zone determined in the first stage. We also propose a linear interpolation-based distance measuring method to decrease the computational complexity. However, the error from distance estimation obviously affects the accuracy of positioning. In this regard, an adjustment trilateration technique is proposed to improve the existing trilateration technique. The simulation results show the estimated errors is 7 cm to 48 cm with fuzzy rectangular split method. The multi-groups with reference nodes selection mechanism is adopted in our proposed method to decrease the positioning errors.
The proposed two-stage positioning method must be applied in indoor space. In order for the larger indoor space, the area is regarded as a unit topology, which is used to expand the positioning range to indoor space. Moreover, an expanded positioning method is proposed to select the suitable positioning topology for the larger indoor space. With this, the two-stage positioning method is applied to measure the coordinate of target node. In the simulation results, the judged rate of topology is approximately 100% in non-overlapping range. In addition, the most suitable topology can effectually selected in overlap range. The estimated errors between 11 cm to 13 cm in the central and corners area. Furthermore, the estimated errors between 19 cm to 21 cm in the alternative area. Consequently, we have proposed an expanded positioning method to applied the larger indoor space, and select the suitable topology, the original position resolution will be preserved in this method.

並列關鍵字

RSSI ； fuzzy inference system ； trilateration technique

參考文獻

[37] 吳伯彥，2013，”模糊邏輯技術在XBee無線感測網路之RSSI定位演算法”，國立虎尾科技大學第電機工程碩士班碩士論文，頁1~96。

[2] F. K. Santoso, N. C. H. Vun, “Securing IOT for Smart Home System,” Proc. of Int. Conf. on Symposium on Consumer Electronics (ISCE), pp. 1-2, 24-26 June 2015.

[7] T. B. Karamat, M. M. Atia A. Noureldin, “Performance Analysis of Code-Phase-Based Relative GPS Positioning and Its Integration with Land Vehicle’s Motion Sensors,” Proc. of IEEE Sensor Journal, Vol. 14, no. 9, pp. 3084-3100, Sept 2014.

[8] E. Wang, S. Zhang, Z. Zhang, “Research on the Incomplete Constellation GPS Positioning Algorithm Aided by Altitude,” Proc. of Int. Conf. of Electrical and Control Engineering (ICECE), pp. 2974-2977, June 2010.

[11] Z. Y. Luo, L. H. Xu, R. H. Wei, “A positioning System Design Based on RSSI Partition of Greenhouse Environment,” Proc. of Int. Conf. on Wireless Communication and Sensor Network (WCSN), pp. 371-377, 13-14 Dec. 2014.