WiFi室內定位技術已經被廣泛應用在缺少視距（LOS）傳輸的室內環境中。在這篇論文中，我們設計了一個多室內指紋定位系統（Multiple Indoor Fingerprints-based Indoor Localization System, MIFI）。MIFI是基於深度可分卷積網路技術並且運用無人機（Unmanned Aerial Vehicles, UAV）來收集指紋資料。透過使用無人機來收集信道狀態信息（Channel State Information）可以大幅減少人力的耗損。在訓練的步驟，實驗環境會被區分成數個小方格，並視為參考點（Reference Point）來收集信道狀態信息。這些收集的資料在經過前處理後會作為輸入，輸入至機器學習模型中。在測試步驟中，於測試點所發送的資料會由樹梅派(Rasberry Pi4)所收集並做為模型輸入，並且模型會輸出預測的目標位置。與其他方法相比，MIFI的定位精準度皆較佳，在位置分類的問題上可以達到99\%以上的準確率，對於未訓練過的位置也可達到0.9公尺的精準度。透過使用無人機所收集的資料進行訓練，人所收集的資料進行測試，我們也可以達到1.28公尺的誤差，證明無人機搭配我們的系統可以達到輔助室內定位，減少人力損耗的目標。

The WiFi-based localization approach has been widely used in the indoor environment, which lacks line-of-sight communication between the target and the satellites. In this paper, we propose a multiple indoor fingerprints-based indoor localization system (MIFI). MIFI is based on the depthwise separable convolution neural network technique and utilizes unmanned aerial vehicles (UAV) to help with collecting fingerprint data. With the help of UAVs, human effort can be decreased. In the training phase, the environment will be divided into squares as the reference points (RP) to collect channel state information (CSI). Then the data will be fed to our model as the input features after preprocessing. In the testing phase, CSI sent at test locations are collected by Raspberry PI 4 (PI4) as the input, then the system will output the location prediction. Compared to the baseline work, the experiment results show that MIFI can achieve a higher classification accuracy that can classify learned locations, upper than 99\%, and mean localization distance error of 0.9 m that localize target in none learned locations in the indoor environments. And we can achieve mean localization distance error of 1.28 m with the help of UAV, which means UAVs are suitable for MIFI to help with collecting data.

摘要 . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . i
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . ii
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . 1
1.1 Indoor Localization System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Related Work42.1 Wireless Signal Localization via Traditional Approach . . . . . . . . 4
2.2 Wireless Signal Localization via Deep Learning Approach . . . . . . . . . . . . . 5
2.3 Localization Using UAV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1 Channel State Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 System Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4 Multiple Indoor Fingerprints-based Indoor localization System (MIFI) . . . . . . . . . . 11
4.1 Data Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.1 CSI Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.2 Label Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 Convolution Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2.1 Convolution Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2.2 Depthwise Separable Convolution . . . . . . . . . . . . . . . . . . . . . . 15
4.2.3 Pooling Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2.4 Fully Connected Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2.5 Activation Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3 MIFI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.4 Training Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5 Localization Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.1 Experiment Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.2 Number of PI4 Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3 Classification Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.4 Localization Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.5 Performance Evaluation with Help of UAV . . . . . . . . . . . . . . . . . . . . . 36
6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

[1]J. Jiao, F. Li, W. Tang, Z. Deng, and J. Cao, “A hybrid fusion of wireless signals and RGB image for indoor positioning,” International Journal of Distributed Sensor Networks, vol. 14, no. 2, p. 1550147718757664, 2018.
[2]Z. Cong, D. Yunzhi, X. Jing, and C. Jienan, “Deep matching network based image auxiliary localization system for unmanned aerial vehicle,” in IEEE Global Communications Conference (GLOBECOM), IEEE, 2019.
[3]X. Wenjing, S. Lijuan, H. Chong, L. Zhaoxiao, and G. Jian, “Depthwise separable convolution based passive indoor localization using CSI fingerprint,” in IEEE Wireless Communications and Networking Conference (WCNC), IEEE, 2020.
[4]O. Chung-Ming, H. Jiawang, and T. Wenyuan, “Signal fuse learning method with dual bands wifi signal measurements in indoor positioning,” IEEE Access, vol. 7, pp. 131805– 131817, 2019.
[5]E.Karim, E.Mohamed, andM.Torki, “Deepcreg: Improving cellular-based outdoor localization using CNN-based regressors,” in IEEE Wireless Communications and Networking Conference (WCNC), IEEE, 2020.
[6]F. Anastasios, M. H. C. Garcia, and R. A. Stirling-Gallacher, “CSI-based localization with CNNs exploiting phase information,” in IEEE Wireless Communications and Networking Conference (WCNC), IEEE, 2021.
[7]Z. Zhipeng, Y. Jihong, Y. Zheng, and G. Wei, “Mobifi: Fast deep-learning based localization using mobile wifi,” in IEEE Global Communications Conference (GLOBECOM),IEEE, 2020.
[8]Z. Yonghao, W. Wai-Choong, G. Hari Krishna, F. Tianyi, Z. Zhixiang, and T. Lin, “Indoor position recognition and interference classification with a nested lstm network,” in IEEE Global Communications Conference (GLOBECOM), IEEE, 2019.
[9]C. Luan, A. Iness, and R. Didier Le, “Csi-based probabilistic indoor position determination: An entropy solution,” IEEE Access, vol. 7, pp. 170048 – 170061, 2019.
[10]H. Daniel, H. Wenjun, S. Anmol, and W. David, “Tool release: gathering 802.11n traces with channel state information,” ACM SIGCOMM Computer Communication Review, vol. 41, p. 53, 2011.
[11]X. Yaxiong, L. Zhenjiang, and L. Mo, “Precise power delay profiling with commodity wifi,” IEEE Transactions on Mobile Computing, vol. 18, pp. 1342 – 1355, 2019.
[12]G. Francesco, S. Matthias, L. Jakob, and H. Matthias, “Free your csi: A channel state in-formation extraction platform for modern wi-fi chipsets,” in ACM Workshop on Wireless Network (ACM WiNTECH), 2019.
[13]S. Abdallah, R. Eric, and M. Paul, “Csi-mimo: K-nearest neighbor applied to indoor localization,” in IEEE International Conference on Communications(ICC), 2020.
[14]O. Kanhere and T. S. Rappaport, “Position locationing for millimeter wave systems,” in 2018 IEEE Global Communications Conference (GLOBECOM), pp. 206–212, IEEE,2018.
[15]G. Wang, Y. Xu, F. Tong, J. Pan, and S. Shen, “Modeling and analyzing single anchor localization for internet of things,” in ICC 2019-2019 IEEE International Conference on Communications (ICC), pp. 1–6, IEEE, 2019.
[16]C. Guo, J. Yu, W.-F. Guo, Y. Deng, and J.-N. Liu, “Intelligent and ubiquitous positioning framework in 5G edge computing scenarios,” IEEE Access, vol. 8, pp. 83276–83289,2020.
[17]A. Booranawong, K. Sengchuai, D. Buranapanichkit, N. Jindapetch, and H. Saito, “Rssi-based indoor localization using multi-lateration with zone selection and virtual position-based compensation methods,” IEEE Access, vol. 9, pp. 46223–46239, 2021.
[18]Y. Qianwen, F. Xiaochen, F. Gengfa, B. Hongxia, S. Xudong, and S. Rajan, “Capsloc: A robust indoor localization system with wifi fingerprinting using capsule networks,” in IEEE International Conference on Communications(ICC), 2020.
[19]L. Fangmin, Z. Yubin, L. Xiaofan, and X. Cheng-Zhong, “Wimage: Crowd sensing based heterogeneous information fusion for indoor localization,” in IEEE Wireless Communications and Networking Conference (WCNC), 2020.
[20]P. Wang, T. Koike-Akino, and P. V. Orlik, “Fingerprinting-based indoor localization with commercial mmwave wifi: Nlos propagation,” in GLOBECOM 2020-2020 IEEE Global Communications Conference, pp. 1–6, IEEE, 2020.
[21]S. Fan, Y. Wu, C. Han, and X. Wang, “A structured bidirectional lstm deep learning method for 3d terahertz indoor localization,” in IEEE INFOCOM 2020-IEEE Conference on Computer Communications, pp. 2381–2390, IEEE, 2020.
[22]C. Zhou, J. Liu, M. Sheng, and J. Li, “Hybrid rss/csi fingerprint aided indoor localization: A deep learning based approach,” in GLOBECOM 2020-2020 IEEE Global Communications Conference, pp. 1–6, IEEE, 2020.
[23]W. Guangyu, “Uav-based interference source localization: A multimodal q-learning approach,” IEEE Access, vol. 7, pp. 2169–3536, 2019.
[24]P. Sixu, B. Zhongjie, S. Lu, K. Dimitrios, L. Shi, and R. Kui, “Automating csi measurement with uavs: from problem formulation to energy-optimal solution,” in IEEE International Conference on Computer Communications(INFOCOM), 2019.
[25]H. Pei-Yuan, L. Chi-Yu, C. Hong-Rong, H. YuanHao, and W. Kuochen, “Wbf-ps: Wigig beam fingerprinting for uav positioning system in gps-denied environments,” in IEEE International Conference on Computer Communications(INFOCOM), 2020.
[26]Z. Hongbo, D. Hongwei, Y. Qiang, and L. Chuang, “Utilizing csi and rssi to achieve high-precision outdoor positioning: A deep learning approach,” in IEEE International Conference on Communications(ICC), 2019.
[27]G. H. Andrew, Z. Menglong, C. Bo, K. Dmitry, W. Weijun, W. Tobias, A. Marco, and A. Hartwig, “Mobilenets: Efficient convolutional neural networks for mobile vision ap-plications,” arXiv:1704.04861, 2017.
[28]D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprintarXiv:1412.6980, 2014.
[29]V. Zhmud, N. Kondratiev, K. Kuznetsov, V. Trubin, and L. Dimitrov, “Application of ultrasonic sensor for measuring distances in robotics,” in Journal of Physics: Conference Series, vol. 1015, p. 032189, IOP Publishing, 2018