透過您的圖書館登入
IP:44.202.128.177
  • 學位論文

以位元層分群演算法進行低功耗藍牙之室內定位

A Bit-plane Clustering Technology with Bluetooth Low Energy Beacon for Indoor Positioning

指導教授 : 魯大德
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


隨著物聯網(Internet of Things, IoT)與智慧型行動服務(Location base services)的發展,不只室外有著定位的需求,近幾年來將定位系統用於室內的需求也越來越多,例如醫療、室內導覽與室內巡航等。 訊號紋比對法為常見的一種室內定位方法,須要在離線階段收集相當數量的Received Signal Strength Indicator (RSSI)訊號值儲存於資料庫,由於不同的廠牌的手機所使用的晶片與擺放位置不盡相同,造成在接收訊號時也會有所落差,所以必須個別建立不同的廠牌的手機的訊號紋,在連線階段進行比對,才可以獲得較高的精準度。本論文中提出了位元層分群定位法,位元層分群定位方法將收集到的RSSI訊號值分成N 個位元層,由最高有效位元(Most Significant Bit, MSB)層至最低有效位元(Least Significant Bit, LSB)層,尋找位元層的共同特徵後進行分群訓練,其目的在於減少離線階段資料庫RSSI數量以及加快連線階段定位的時間。實驗以ASUS、APPLE、SAMSUNG等不同手機測試,並以RADAR、KNN(K-Nearest NeighborNearest)、Correlation、訊號紋定位法、位元層分群定位法等不同的定位方法做比較。 實驗結果得知,在環境一的實驗環境下,三款不同手機一米內的定位累積誤差,位元層分群定位法與RADAR相比提高了約13.7% ~ 33% 的定位精準度,與KNN(K=2)定位法相比提高了約22%~44%的定位精準度,與訊號紋定位法相比提高了約44%的定位精準度,與Correlation相比提高了約16.7%~100% 的定位精準度。在實驗環境二,在不同發射間隔的定位累積誤差,位元層分群定位法與RADAR相比提高了約22% 的定位精準度,與Correlation相比提高了約13% 的定位精準度,與訊號紋比對法相比提高約62%的定位精準度。在演算複雜度中,位元層分群定位法與訊號紋比對法的演算複雜度最為相似,但在位元層分群定位法的離線階段經過分群後,資料量減少了約58%~59%,與訊號紋比對法相比加快了比對的速度。

關鍵字

藍牙 KNN BLE RSSI Beacons

並列摘要


In recent years, positioning system can provide location based services (LSB) such as indoor navigation and tracking. Pattern matching method is able to provide accurate positioning services. However, the positions of sensor’s chips are different for smartphones. A sufficient number of Received Signal Strength Indicator (RSSI) values are required to be stored into the database for each smartphone at the offline stage. In this paper, we proposed a bit-plane clustering technology to divide the collected RSSI values into N bitmaps. The goal of this work is to reduce the numbers of database RSSI at the offline stage and to speed up the connection time at the online stage. The bitmap separated many layers from the most significant bit (MSB) layer to the least significant bit (LSB) layer. After training, the same characteristic bitmaps were clustering into one group. In this study, bit-plane clustering technology compared with different positioning methods, such as RADAR, K-Nearest Neighbors algorithm (K=2), correlation method, and pattern matching, the positioning accuracy at 1 meter in environment 1 increased about 13.7% ~ 33%, 22%~44%, 16.7%~100%, and 44 %, respectively. In environment 2, the positioning accuracy compared with RADAR, correlation method, and pattern matching at 1 meter increased about 22%, 13%, and 62%, respectively.

並列關鍵字

Bluetooth KNN BLE RSSI Beacons

參考文獻


[1] 李彥勳,「以TOA為基礎應用於衛星及基地台混合定位方法之效能改進研究」,國立屏東教育大學,碩士論文,民國一百零三年。
[2] Zhu, X., Feng, Y., “RSSI-based Algorithm for Indoor Localization”,Communications and Network, vol. 5, pp. 37–42, 2013.
[3] Faragher, R., Harle, R., R., R., Fingerprinting with Bluetooth Low Energy Beacons”, Selected Areas in Communications, pp. 2418 – 2428, 2015.
[4] Livinsa, Z. M., Jayashri, S., “Performance Analysis of Diverse Environment based on RSSI Localization Algorithms in WSNs”, Information & Communication Technologies (ICT), pp. 572-576, 2013.
[5] Mistry, H. P., et al., “RSSI based Localization Scheme in Wireless Sensor Networks: A Survey”, Advanced Computing & Communication Technologies (ACCT), pp. 647–652, 2015.

延伸閱讀