Title Page
Contents
Abstract 10
Ⅰ. Part 1: PRACH Preamble Detection using Deep learning 11
Ⅰ-Ⅰ. Introduction 11
1. Initial Access, Up Link Synchronization, and Its Problem 11
2. Machine Learning / Deep Learning in Physical Layer Perspectives 13
3. Related Works 13
Ⅰ-Ⅱ. Preliminaries 16
1. PRACH Preamble Transmission 16
2. Conventional Threshold Based Method 18
Ⅰ-Ⅲ. PRACH Preamble Detection with Deep Learning 22
1. Proposed Receiver Structure 23
2. Simulation Results 30
Ⅱ. Part 2: RTT Fingerprint Positioning 35
Ⅱ-Ⅰ. Introduction 35
1. Indoor Positioning 35
2. Fingerprint-Based Positioning and Its Problems 35
3. Related works 36
Ⅱ-Ⅱ. Preliminaries 38
1. System Model 38
2. Classical Methods 39
Ⅱ-Ⅲ. Efficient Hybrid RTT Fingerprint Based Positioning with Boundary Effect Alleviation 43
1. Search Area Restriction 44
2. Boundary Weighting Procedure 46
3. Simulation Results 51
Ⅳ. Conclusion 60
References 62
국문요지 67
〈Table 1〉 A concise comparison of our work with the existing ML-based PRACH preamble detection schemes 15
〈Table 2〉 Structure of proposed FNN model 28
〈Table 3〉 Simulation environment and set parameters 30
〈Table 4〉 Lowest SNR satisfies 3GPP requirements (Pd ≥ 0.99, Pf ≤ 0.001) for each scheme[이미지참조] 34
〈Table 5〉 mean distance error(m) for All area 54
〈Table 6〉 distance error(m) of 90 percentile for each selected noise variance for All area 54
〈Table 7〉 Mean distance error (m) for Ill-conditioned region 56
〈Table 8〉 Distance error (m) of 90 percentile for ill-conditioned region 56
〈Table 9〉 Mean distance error difference (m) between PM and SA Only, and SA-EBW and SA-BWP in All area 58
〈Table 10〉 Mean distance error difference (m) between PM and SA Only, and SA-EBW and SA-BWP in Ill-conditioned region 58
〈Figure 1〉 Initial access in 5G communication system 12
〈Figure 2〉 PRACH preamble transmitter structure 17
〈Figure 3〉 Time-advanced preamble sequence 18
〈Figure 4〉 Conventional PRACH preamble receiver structure 21
〈Figure 5〉 Problem in conventional scheme 22
〈Figure 6〉 Time advanced SW 24
〈Figure 7〉 Proposed pre-processing step 27
〈Figure 8〉 Proposed PRACH preamble receiver structure 28
〈Figure 9〉 Performance in the detection perspective (Pd) for each SNR[이미지참조] 32
〈Figure 10〉 Performance in the false alarm perspective (Pf) for each SNR[이미지참조] 32
〈Figure 11〉 Pd and Pf of each SNR of the training data in - 10 dB AWGN channel[이미지참조] 33
〈Figure 12〉 Pd of each optimized scheme for each SNR[이미지참조] 34
〈Figure 13〉 SA restriction results for (a) Well-conditioned target, and (b) Ill-conditioned target 44
〈Figure 14〉 Boundary effect comparision (a) Pm with perfect radio map and (b) PM with imperfect radio map 46
〈Figure 15〉 First-order curve fitting results in the distance space 47
〈Figure 16〉 The structure of proposed method 50
〈Figure 17〉 Simulation environment assumption 51
〈Figure 18〉 Mean error distance(m) for all area in all set noise variance range 53
〈Figure 19〉 Cumulative error probability in All area cases in each set noise variances 54
〈Figure 20〉 Mean error distance (m) for ill-conditioned region in all set noise variance range 55
〈Figure 21〉 Cumulative error probability in ill-conditioned region cases in each set noise variances 56