基於感測資料融合及粒子濾波機制之協同室內定位技術

Abstract

在本研究中，提出了基於貝氏近似法理論的室內定位系統框架，在此框架下，系統將用戶在追踪區域中的位置視為一機率率分佈。 然而，連續性的機率分佈很難在複雜的室內建築中用數學公式呈現，而計算機也無法處理連續的資料。 所以本研究使用粒子濾波器實現貝氏近似法，而是在追踪區域中產生離散的粒子對分佈進行加權採樣。 因此，系統可以根據由傳感器提供的觀察通過位移和加權粒子來追蹤用戶的位置分佈。

典型的追踪系統大多使用行人追踪機制以取得粒子位移，以及無線測距法取得環境中無線定位的結果以用於加權。 在本研究中，追蹤系統將擴展到2.5D空間，其將室內結構視為多個樓層的 2D 平面圖和一組樓層變換設施(如電梯、樓梯等)。 本研究提出的追踪系統能夠使用由慣性訊號判斷電梯以及樓梯等室內地標設施、和利用大氣壓力變化來判斷樓層和轉換。 對於 2D 定位，系統可利用室內的路標或是門牌作為視覺地標。 此外，系統亦可透過使用者間的相遇事件而在短距離進行協同測距，可以在由於無限設施覆蓋率不足的區域防止使用者位置因為沒有環境觀察或是由於劇烈運動造成的位置確定性降低的結果。

最後，本研究提出使用信標間距來感測由於環境更動造成的信號傳播特徵變化，以建立具自我調整能力之無線環境設施。 該方案使用藍牙信標來實現，並且感測的環境動態可以簡單地使用BLE廣告分組傳遞到使用的設備。

In this research, an indoor localization framework based on Bayesian approximation theory is proposed, which treats the users' locations in the tracking area as a probability distribution. However, it is difficult to model such distribution in well-formed mathematical formulas in indoor buildings. The approximation is realized using the particle filter algorithm, in which the distribution is sampled with discrete weighted particles in the tracking area. The system can therefore track a user's location distribution by shifting and weighting the particles according to the observation provided by the sensors.

A typical implementation of particle filter tracking system would consists of a pedestrian tracking mechanism to obtain motion displacement indicators for particle shifting, and a wireless ranging scheme to obtain wireless location indicators for weighting. In this research, the tracking scheme is extended to 2.5D, which models the indoor structures as multiple 2D floor plans and floor transitional facilities. The proposed tracking scheme is capable of detect floor levels and transition using the inertial landmark indicators formed by inertial pattern recognition, and floor/altitude indicator with atmospheric pressure detection. For 2D localization, a signboard recognition scheme is designed to generate visual landmark indicators. Also, an M2M encountering scheme enables the users to conduct collaborative ranging at short distances to create accurate inter-device encounter indicators, which can be applied when a user with high location uncertainty caused by insufficient infrastructural coverage or highly motion dynamics.

Finally, an adaptive ranging mechanism is designed to use inter-beacon ranging to sense the signal propagation dynamics caused environmental changes. The scheme is implemented using Bluetooth beacons, and the sensed environmental dynamics can be delivered to the used devices simply using BLE advertisement packets.