臺灣博碩士論文加值系統

此論本基於在戶外環境中使用窄帶物聯網(NB-IoT)來實現降低設備成 本的目標，透過迭代構建內核空間，改變傳統的訓練信號方法。首先進行 室外位置叢集分類，然後提出一種半監督式轉移核心學習(Semi- Supervised Transfer Kernel Learning)，將標記、未標記和跨域數據後映射到三維平面空 間，然後使用梯度學習方法進行精細化且在映射過程中調整參數，使用未 標記和跨域數據的新標記數據，我們通過添加一些新的未標記樣本重複迭 代微調我們的目標模型，在幾次迭代後改進輸出，根據結果與現有相關文 獻的方法，有效得到較低的定位誤差及較高的定位精準度。

Using a general licensed band like 4G positioning and unlicensed band like Lora have made many results, this thesis uses narrowband internet of things (NB-IoT) to achieve the target of reduce equipment costs and iterated through and build multiple kernel spaces to change the traditional training signal method. First performs outdoor location cluster classification, and then proposes a method of kernel-gradient learning, the kernel learning maps the input labeled, unlabeled and cross-domain data to the three- dimensional plane space, then uses the gradient learning method to fine-tune the parameters during the mapping process. In our approach, we propose a cluster classification scheme that combines unlabeled and cross-domain data based on finding the relationship of existing labeled data from the source domain. Using the new labeled data of unlabeled and cross-domain data, we repeatedly iterative fine-tune our target model by adding some new unlabeled samples, improved the output after several iteration. Basically, the more labeled data are used and added to the fine-tuning process, the higher the positioning accuracy, in our actual implementation, the standardization process handles a number of different signal characteristics, such as RSSI, SNR and localization to increase the prediction accuracy of positional accuracy. Finally, the experimental results show that the proposed scheme effectively improves the location average accuracy up about 87% and reduces the localization average error about 4 m, compared with another existing localization results.

1 Introduction 1
2 Related Works 4
2.1 Related works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6
3 Preliminaries 7
3.1 System architecture . . . . . . . . . . . . . . . . . . . . . . . . . .7
3.2 Problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . .9
3.3 Basic idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12
4 An Iterative Kernel-Gradient Learning Approach for Outdoor Localization 14
4.1 Cluster classification phase . . .. . . . . . . . . . . . . . . . . . . .15
4.2 Kernel-gradient learning phase . . . . . . . . . . . . . . . . . . . .19
4.3 Iterative kernel-gradient learning phase . . . . . . . . . . . . . . .24
4.4 Estimated location uniting phase . . . . . . . . . . . . . . . . . . .29
5 Experimental Results 34
5.1 Localization error . . . . . . . . . . . . . . . . . . . . . . . . . . . .36
5.2 Data accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .37
5.3 Training loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39
5.4 MAPE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42
6 Conclusions 46

[1] H. S. A. Chiumento and S. Pollin, ”Localization in Long-Range Ultra Narrow Band IoT Networks Using RSSI”, in Proceedings of IEEE International Conference on Communi- cations (ICC 2017), pp. 1–6, Paris, France, May. 2017.
[2] D. Liu, S. Guo, Y. Yang, Y. Shi, and M. Chen, ”Geomagnetism Based Indoor Navigation by Offloading Strategy in NB-IoT”, IEEE Internet of Things Journal-Early Access, vol. 6, no. 3, pp. 4074–4084, 2019.
[3] A. Hoglund, X. Lin, O. Liberg, and A. Behravan, ”Overview of 3GPP Release 14 En- hanced NB-IoT”, IEEE Network, vol. 31, no. 6, pp. 16–22, 2017.
[4] W. Nakai, Y. Kawahama, and R. Katsuma, ”Adaptive Localization in Dynamic Indoor Environments by Transfer Kernel Learning”, in Proceedings of Wireless Communications and Networking Conference (WCNC 2017), pp. 1–6, San Francisco, USA, May. 2017.
[5] M. Xu, W. Xia, Z. Jia, Y. Zhu, and L. Shen, ”A VLC-Based 3-D Indoor Positioning System Using Fingerprinting and K-Nearest Neighbor”, in Proceedings of IEEE 85th Vehicular Technology Conference (VTC 2017), pp. 1–5, NSW, Australia, Jun. 2017.
[6] R. Wang, M. Liu, Y. Zhou, Y. Xun, and W. Zhang, ”A Deep Belief Networks Adaptive Kalman Filtering algorithm”, in Proceedings of IEEE International Conference on Soft- ware Engineering and Service Science (ICSESS 2016), pp. 178–181, Beijing, China, Aug. 2016.
[7] L. Li, C. Sunn, L. Lin, J. Li, and S. Jiang, ”A dual-layer Supervised Mahalanobis Kernel for the Classification of Hyperspectral Images”, Neurocomputing, vol. 214, no. 1, pp. 430– 444, 2016.
[8] B. Kulis, ”Metric Learning: A Survey”, Foundations and Trends in Machine Learning, vol. 5, no. 4, pp. 287–364, 2012.
[9] C. Xiao, D. Yang, Z. Chen, and G. Tan, ”3-D BLE Indoor Localization based on Denoising Autoencoder”, IEEE Access, vol. 5, no. 1, pp. 12751–12760, 2017.
[10] M. Mohammadi, A. Fuqaha, M. Guizani, and J. Oh, ”Semi-Supervised Deep Reinforce- ment Learning in Support of IoT and Smart City Services”, IEEE Internet of Things Journal, vol. 5, no. 2, pp. 624–635, 2017.
[11] F. Vita and D. Bruneo, ”A Deep Learning Approach for Indoor User Localization in Smart Environments”, in Proceedings of IEEE International Conference on Smart Computing (ICSC 2018), pp. 89–96, Taormina, Italy, Jun. 2018.
[12] G. Wu and P. Tseng, ”A Deep Neural Network Based Indoor Positioning Method Using Channel State Information”, in Proceedings of International Conference on Computing, Networking and Communications (ICCNC 2018), pp. 290–294, HI, USA, Mar. 2018.
[13] A. Wang, X. Hao, X. Zhang, A. Wang, and P. Hu, ”A Dynamic Target Visual Positioning Method Based on ROI”, in Proceedings of IEEE 3rd International Conference on Image, Vision and Computing (ICIVC 2017), pp. 592–595, Chongqing, China, Jun. 2017.
[14] C. Park, H. Shin, and Y. Choi, ”A Parallel Artificial Neural Network Learning Scheme Basedon Radio Wave Fingerprint for Indoor Localization”, in Proceedings of Tenth In- ternational Conference on Ubiquitous and Future Networks (ICUFN 2018), pp. 592–595, Chongqing, China, Jun. 2017.
[15] Z. Zhang, ”Improved Adam Optimizer for Deep Neural Networks”, in Proceedings of IEEE 26th International Symposium on Quality of Service (IWQoS 2019), pp. 1–2, Banff, Canada, Jan. 2019.
[16] N. Zhang, D. Lei, and J. Zhao, ”An Improved Adagrad Gradient Descent Optimization Algorithm”, in Proceedings of Chinese Automation Congress (CAC 2019), pp. 1–4, Xi’an, China, Jan. 2019.
[17] R. V. K. Reddy, B. S. Rao, and K. P. Raju, ”Handwritten Hindi Digits Recognition Using Convolutional Neural Network with RMSprop Optimization”, in Proceedings of Second International Conference on Intelligent Computing and Control Systems (ICICCS 2019), pp. 1–7, Madurai, India, Mar. 2019.
[18] N. Jean, S. M. Xie, and S. Ermon, ”Semi-Supervised Deep Kernel Learning: Regression with Unlabeled Data by Minimizing Predictive Variance”, Neural Information Processing Systems, arXiv:1805.10407, 2019.
[19] W. Nakai, Y. Kawahama, and R. Katsuma, ”Reducing Error of Positioning Based on Unstable RSSI of Short Range Communication”, in Proceedings of Advanced Information Networking and Applications (AINA 2019), pp. 572–578, Krakow, Poland, May. 2018.
[20] L. He and H. Zhang, ”Kernel K-Means Sampling for Nystrm Approximation”, IEEE Transactions on Image Processing, vol. 27, no. 5, pp. 2108–2120, 2018.
[21] A. Long, J. Wang, J. Sun, and P. S. Yu, ”Domain Invariant Transfer Kernel Learning”, IEEE Transactions on Knowledge and Data Engineering, vol. 27, no. 6, pp. 1519–1532, 2015.