結合M2M相遇機制的室內定位技術

Abstract

近年來，由於以所在位置為基礎，所提供的地理位置服務的需求持續上升，使得行動定位技術近期內被深入且廣泛的研究。不像室外定位已經以全球定位系統(GPS)為大宗， 室內的解決方案必須考慮不同的傳感器輸入，各種環境的觀察，並透過融合技術將這些資訊相結合，來產生較為精確的位置結果。隨著移動/可穿戴技術漸趨成熟，合作式定位已經被提出並試著突破傳統定位的限制。在本文中，我們觀察到當多台裝置彼此碰面，我們稱之為“M2M相遇”，此時我們可以透過裝置間的即時數據交流，合作並完善彼此的位置資訊。利用可能在開放式空間常常發生的M2M相遇的機率性事件，我們可以頻繁的校準那些參與M2M相遇事件的裝置的粒子機率分佈。因此，通常用於融合單一裝置的環境觀察的粒子濾波器技術(稱為intra-PF)，被延伸到跨裝置的融合粒子濾波器的狀態的技術(稱為inter-PF)。此外，我們考慮了計算的複雜度，進一步的分割出近似的解決方案(稱為approximated inter-PF)。我們運用電腦模擬，以及利用智慧型平板的系統原型，在建有藍芽低能耗(BLE)基礎設施的辦公室環境中實地測試，來驗證了我們的intra-PF和inter-PF演算法。最後，我們證實透過額外的M2M相遇事件，具有較少的傳感器或更低的計算能力的裝 置，仍然可以享受能夠與那些擁有更豐富的感測信息的裝置相匹敵的良好定位品質。

Mobile localization has been intensively studied recently due to the exploding demands on Location-based Services (LBS). While outdoor localization is dominated by Global Positioning System (GPS), indoor solutions have to consider various sensor observations and apply fusion techniques to merge them. As mobile/wearable devices become popular, inter-device collaboration provides another opportunity to fuse data. In this thesis, we observe that when multiple devices meet up, which we call “M2M encountering”, there is an opportunity to collaboratively refine their locations via instant inter-device data exchanges. We apply the particle filter (PF) by taking inertial data, RF signals, indoor floor plan, and M2M encountering events as inputs for location tracking. With opportunistic M2M encountering events, which may happen frequently in public spaces, we are able to calibrate the distributions of particles amongst those encountered devices. Hence, the PF technique, which normally fuses observations of a single device (called intra-PF here), is extended to a cross-device fusion (called inter-PF) technique. Moreover, we consider the computational complexity and derive an approximated inter-PF calibration scheme. We validate our intra-PF and inter-PF schemes by simulations as well as a prototype system with smart-pads and Bluetooth Low Energy (BLE) as the infrastructure in an office environment. It is verified that with the additional encountering indicators, those devices with fewer sensors or lower computing power can still enjoy good localization quality as those with richer sensors.