||Multi-Device Indoor Positioning System based on wireless fingerprinting system
||Institute of Computer & Communication
本篇論文基於藍芽無線訊號，使用多裝置的合作來進行偕同室內定位，我們提出的方法為先進行指紋數據庫的建立，然後再利用機器學習進行初步的位置預測，這裡使用機器學習的方法為KNN模組。但若單純使用KNN模組進行室內定位，它沒辦法根據先前的位置與地圖上的資訊來進行更精準的預測。因此，我們還加入了能夠紀錄過去位置與地圖資訊的粒子濾波器，透過這些方法來讓室內定位更加準確。此外，在我們的系統中，提出Tightly-Coupled Fusion、Loosely-Coupled Fusion及Joint Particle Filtering三種方法，這些方法是基於多裝置來改善室內定位的方法。Tightly-Coupled Fusion是在做決定KNN模組的K值前，就已經將多個裝置併在一起來決定K值，再放入KNN模組中並決定出粒子濾波器的參考位置；Loosely-Coupled Fusion與Joint Particle Filtering的方法為多個裝置先分開進行決定KNN的K值並放入KNN模組中，最後在一起匯入粒子濾波器中進行位置預測。其中，Joint Particle Filtering與Loosely-Coupled Fusion不同的地方是，由KNN模組決定出來的參考位置不只有兩個，而是KNN模組的所有候選位置都參考。我們提出的方法可以根據不同情境來決定要使用哪一種方法，讓我們的系統更加靈活。在結果展示中，我們系統的性能和定位準確度幾乎都優於單個設備。
In this paper, we use multi-device cooperation for indoor positioning based on Blue-tooth wireless signals. The method we propose is to establish the fingerprint database first, and then use the machine learning which choose KNN model to make a preliminary actual location prediction. Since only use the KNN model for indoor positioning, it cannot know the previous position and the information on the map to make more accurate predictions. Therefore, we have also added particle filter which can record the past position and map information to make indoor positioning more accurate. Moreover,in our system, we propose the Tightly-Coupled Fusion, the Loosely-Coupled Fusion and the Joint Particle Filtering. These three methods are based on multiple devices to improve indoor positioning. The Tightly-Coupled Fusion has combined multi-devices to determine the K value of KNN model, then do KNN model and particle filter; The methods of the Loosely-Coupled Fusion and the Joint Particle Filtering are that multi-devices are separately determined to determine the K value of KNN model and then do particle filter. The Joint Particle Filtering difference from the Loosely-Coupled Fusion is that all candidate positions of the KNN model are referenced instead of only two reference positions determined by KNN model. We can decide which method to use according to different situations and let our system be more flexible. The performance and positioning accuracy of our system is almost better than only a single device.
List of Figures iii
List of Tables vi
1 Introduction 1
2 Related Work 3
3 System Implementation 4
3.1 Fingerprint Database 4
3.2 Tightly-Coupled Fusion 4
3.3 Loosely-Coupled Fusion 7
3.4 Joint Particle Filtering 8
4 Experimental Study 10
4.1 Testbed Setup 10
4.1.1 Experimental Devices 10
4.1.2 Experimental Environment 12
4.2 Data Collection and Database Creation 13
4.3 Moving Strategies and Device Assignment 14
4.3.1 Scripted Walk 14
4.3.2 Free Walk 16
4.4 Experimental Result 18
4.4.1 Stationary and Experiment-fast 18
4.4.2 Experiment-slow 21
4.4.3 Free Walk 24
5 Influence of the Number of Beacon 27
5.1 Testbed Setup 27
5.2 Experimental Result 28
5.2.1 Tightly-Coupled Fusion 28
5.2.2 Loosely-Coupled Fusion 33
5.2.3 Joint Particle Filtering 37
5.2.4 Proposed Method Comparison and Moving Strategy Analysis 41
6 Conclusions 54
List of Figures
3.1 Tightly-Coupled Fusion 5
3.2 Schematic Diagram of Particle Update 6
3.3 Loosely-Coupled Fusion 7
3.4 Joint Particle Filtering 8
4.1 Android Phones 10
4.2 Beacon Simulator 11
4.3 LAB environment 12
4.4 Beacon Deployment 12
4.5 Data Analysis 14
4.6 Stationary 15
4.7 Move forward 15
4.8 Move forward and backward 16
4.9 Application of record time 17
4.10 Setting of Free Walk 17
4.11 Result of KNN + Particle Filter 19
4.12 Result of KNN + Particle Filter in different moving strategies (Group1 in Experiment-fast) 20
4.13 Result of KNN + Particle Filter in different moving strategies (Group2 in Experiment-fast) 21
4.14 Result of KNN + Particle Filter 22
4.15 Result of KNN + Particle Filter in different moving strategies (Group1 in Experiment-slow) 23
4.16 Result of KNN + Particle Filter in different moving strategies (Group2 in Experiment-slow) 24
4.17 Result of KNN + Particle Filter in Free Walk 25
4.18 Result of KNN + Particle Filter in Free Walk 26
5.1 Deployment of different numbers of beacons 28
5.2 Compare different numbers of beacons in Tightly-Coupled Fusion 29
5.3 Different numbers of beacon in Tightly-Coupled-Fusion (Group1) 30
5.4 Different numbers of beacon in Tightly-Coupled-Fusion (Group2) 31
5.5 Compare different numbers of beacons in Free Walk in Tightly Coupled Fusion 32
5.6 Different numbers of beacons in Free Walk in Tightly-Coupled Fusion 33
5.7 Compare different numbers of beacons in Loosely-Coupled Fusion 33
5.8 Different numbers of beacon in Loosely-Coupled-Fusion (Group1) 34
5.9 Different numbers of beacon in Loosely-Coupled-Fusion (Group2) 35
5.10 Compare different numbers of beacons in Free Walk in Loosely-Coupled Fusion 36
5.11 Different numbers of beacons in Free Walk in Loosely-Coupled Fusion 37
5.12 Compare different numbers of beacons in Joint Particle Filtering 37
5.13 Different numbers of beacon in Joint Particle Filtering (Group1) 38
5.14 Different numbers of beacon in Joint Particle Filtering (Group2) 39
5.15 Compare different numbers of beacons in Free Walk in Joint Particle Filtering 40
5.16 Different numbers of beacons in Free Walk in Joint Particle Filtering 41
5.17 Comparison of different numbers of beacons in different methods in Group1 (BTrack chooses d2 for analysis) 42
5.18 Comparison of different numbers of beacons in different methods in Group2 (BTrack chooses d4 for analysis) 43
5.19 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in 4-Beacon Deployment in Group1 (BTrack chooses d2 for analysis) 44
5.20 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in 3-Beacon Deployment in Group1 (BTrack chooses d2 for analysis) 46
5.21 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in 2-Beacon Deployment in Group1 (BTrack chooses d2 for analysis) 47
5.22 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in 4-Beacon Deployment in Group2 (BTrack chooses d4 for analysis) 48
5.23 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in 3-Beacon Deployment in Group2 (BTrack chooses d4 for analysis) 49
5.24 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in 2-Beacon Deployment in Group2 (BTrack chooses d4 for analysis) 50
5.25 Comparison of different numbers of beacons in different methods in Free Walk (BTrack of Group1 chooses d2 for analysis and BTrack of Group2 chooses d4 for analysis) 51
5.26 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in different numbers of beacons in Group1 (BTrack chooses d2 for analysis) 52
5.27 Comparison of Tightly-Coupled Fusion and Loosely-Coupled Fusion in different numbers of beacons in Group2 (BTrack chooses d4 for analysis) 53
List of Tables
4.1 Specifications of the mobile devices 11
 I. F. Akyildiz, Weilian Su, Y. Sankarasubramaniam, and E. Cayirci, “A survey on sensor networks,” IEEE Communications Magazine, vol. 40, no. 8, pp. 102–114, 2002.
 J. B.-Y. Tsui, Fundamentals of global positioning system receivers: a software approach. John Wiley & Sons, 2005, vol. 173.
 W.Xue, W.Qiu, X.Hua, andK.Yu, “Improvedwi-firssimeasurementforindoor localization,” IEEE Sensors Journal, vol. 17, no. 7, pp. 2224–2230, 2017.
 S. Papaioannou, H. Wen, A. Markham, and N. Trigoni, “Fusion of radio and camera sensor data for accurate indoor positioning,” in 2014 IEEE 11th International Conference on Mobile Ad Hoc and Sensor Systems, 2014, pp. 109–117.
 S. Yang and B. Wang, “Residual based weighted least square algorithm for bluetooth/uwb indoor localization system,” in 2017 36th Chinese Control Conference (CCC), 2017, pp. 5959–5963.
 C. Huang, L. Lee, C. C. Ho, L. Wu, and Z. Lai, “Real-time rfid indoor positioning system based on kalman-filter drift removal and heron-bilateration location estimation,” IEEE Transactions on Instrumentation and Measurement, vol.64, no.3, pp. 728–739, 2015.
 W. A. Cahyadi, Y. H. Chung, and T. Adiono, “Infrared indoor positioning using invisible beacon,” in 2019 Eleventh International Conference on Ubiquitous and Future Networks (ICUFN), 2019, pp. 341–345.
 S. Sou, W. Lin, K. Lan, and C. Lin, “Indoor location learning over wireless fingerprinting system with particle markov chain model,” IEEE Access, vol. 7, pp. 8713–8725, 2019.
 Senion, “How accurate are indoor positioning systems?” https://senion.com/ insights/accurate-indoor-positioning-systems/, accessed January. 16, 2019.
 A.Noertjahyana,I.A.Wijayanto,andJ.Andjarwirawan,“Developmentofmobile indoor positioning system application using android and bluetooth low energy with trilateration method,” in 2017 International Conference on Soft Computing, Intelligent System and Information Technology (ICSIIT), 2017, pp. 185–189.
 Y. Wang, Q. Yang, G. Zhang, and P. Zhang, “Indoor positioning system using euclidean distance correction algorithm with bluetooth low energy beacon,” in 2016 International Conference on Internet of Things and Applications (IOTA), 2016, pp. 243–247.
 P. Sthapit, H. Gang, and J. Pyun, “Bluetooth based indoor positioning using machine learning algorithms,” in 2018 IEEE International Conference on Consumer Electronics - Asia (ICCE-Asia), 2018, pp. 206–212.
 N. Li, J. Chen, Y. Yuan, and C. Song, “A fast indoor tracking algorithm based on particle filter and improved fingerprinting,” in 2016 35th Chinese Control Conference (CCC), 2016, pp. 5468–5472.
 C. Chiu, K. Feng, and P. Tseng, “Spatial skeleton-enhanced location tracking for indoor localization,” in 2017 IEEE Wireless Communications and Networking Conference (WCNC), 2017, pp. 1–6.
 H. Xu, Z. Yang, Z. Zhou, K. Yi, and C. Peng, “Tum: Towards ubiquitous multidevice localization for cross-device interaction,” in IEEE INFOCOM 2017 - IEEE Conference on Computer Communications, 2017, pp. 1–9.
 L. Bjornsson, “Ble beacons for indoor positioning ‒beacon limitations,” https: //locatify.com/blog/ble-beacons-no-bull-beacon-review/,accessedSeptember.20, 2016.
 dburr, “linux-ibeacon,” https://github.com/dburr/linux-ibeacon, accessed December. 10, 2014.
 estimote, “Ranging beacons,” https://developer.estimote.com/android/tutorial/ part-3-ranging-beacons/, accessed August. 16, 2019.
 A. Wang, G. Chen, J. Yang, S. Zhao, and C. Chang, “A comparative study on human activity recognition using inertial sensors in a smartphone,” IEEE Sensors Journal, vol. 16, no. 11, pp. 4566–4578, 2016.