以改良集中式幾何演算法實現無線感測網路之室內定位

Abstract

無線通訊技術之發達促進遠端環境監測與健康照護系統的發展，而對監測所得之環境或生理資料而言，位置是極端重要的參考資訊。因此，多種非GPS 而基於傳遞時間延遲或接收訊號強度指標的定位機制被使用於室內定位。本研究於ZigBee 無線感測網路中設計並實現基於接收信號強度指標的集中式定位系統，使用之幾何定位演算法稱為三角重心法，並改良成為擴展三角重心法。由於接收訊號強度指標具有不穩定性與導因於晶片製程的差異性，故此文所提出之定位系統採用一維卡爾曼濾波器以及接收訊號強度指標校準的機制，此外，二維卡爾曼濾波器導入加速度資訊後用於即時位置追蹤。 本研究所提出之定位演算法與改良方法已透過模擬和實測驗證並評估效果。三角重心法和擴展三角重心法於25 平方公尺的虛擬空間中的模擬平均誤差分別為0.8164 和0.5347 公尺。實測中，接收訊號強度與距離之關係的數學模型藉由實際訊號收集和線性迴歸法建構後，擴展三角重心法配合訊號強度指標校準與濾波處理在35 平方公尺測試空間中的實測誤差為1.0687 公尺，此外，在同樣的測試空間中，擴展三角重心法結合二維卡爾曼濾波器與加速規追蹤矩形路徑達成1.2572 公尺的平均誤差。 集中式幾何定位演算法易於實現並提供合理的估測準確性，且此文所提出之改良方法─接收訊號強度指標校準與濾波處理提升了基於接收訊號強度指標之室內定位的準確性，同時，於ZigBee 無線感測網路中實現之定位系統與環境監測系統、遠端照護系統皆具有明顯的整合可能性。

The prosperous development of wireless communication techniques brings significant advances in remote environment monitoring and health-care systems and location is vital information for the sensed environmental or biomedical data. Thus, various non-GPS (Global Positioning System) localization mechanisms based on Time-Of-Arrival (TOA) or Received Signal Strength Indicator (RSSI) are utilized in indoor location estimation. This study designed and implemented a RSSI-based centralized localization system in a ZigBee-based Wireless Sensor Network (WSN). A geometrical localization algorithm, Triangular Center Of Gravity (TCOG), with its modified version, Extended Triangular Center Of Gravity (ETCOG), is presented in this thesis. Owing to the unsteadiness and the chip-to-chip variation in RSSI measurement, a one-dimensional Kalman filter for RSSI filtering and a RSSI calibration process have been applied on the proposed localization system. Moreover, a two-dimensional Kalman filter with acceleration information for real-time location tracking was also implemented. The proposed localization algorithms and improving methods were verified and evaluated through simulations and experiments. The average simulation errors of the TCOG and ETCOG algorithms are 0.8164 and 0.5347 meters respectively in a 25-square-meter virtual space. In the experimental phase, the RSSI-to-distance relation modeling was performed via realistic RSSI data collection and linear regression, and then the ETCOG algorithms with RSSI filtering and calibration induced an overall average error of 1.0687 meters in a 35-square-meter testing space. Additionally, in the same experimental space, the ETCOG algorithm with a two dimensional Kalman filter and an accelerometer for square route tracing resulted an average error of 1.2572 meters. The centralized geometrical localization algorithm is efficient in implementation and provides reasonable estimation accuracy. The improving methods, RSSI calibration and filtering, brought remarkable enhancement for RSSI-based indoor location estimation, and the localization system in a ZigBee-based WSN has apparent potential to merge with environment monitoring and telecare systems.