Indoor localization with smartphone using BLE ibeacon

VIETNAM NATIONAL UNIVERSITY UNIVERSITY OF ENGINEERING AND TECHNOLOGY Duong Ngoc Son INDOOR LOCALIZATION WITH SMARTPHONE USING BLE IBEACON MASTER THESIS IN ELECTRONICS AND TELECOMMUNICATIONS MAJOR : COMMUNICATION ENGINEERING CODE : 8510302.02 SUPERVISOR : PHD DINH THI THAI MAI HANOI - 2020 Publication thesis option This thesis would consist of the following six articles: Paper 1: Thai-Mai Thi Dinh, Ngoc-Son Duong, Kumbesan Sandrasegaran, “Smartphonebased Indoor Positioning Using BLE iBeacon and Reliable Lightweight Fingerprint Map”, IEEE Sensors Journal, 2020 In press https://doi.org/10.1109/JSEN.2020.2989411 Paper 2: Ngoc-Son Duong, Thai-Mai Dinh, “Develop a true real-time iBeacon-based indoor positioning system using smartphone”, to be submitted to IEEE Transactions on Instrumentation and Measurement Paper 3: Ngoc-Son Duong, Thai-Mai Dinh, “Indoor Localization with lightweight RSS Fingerprint using BLE iBeacon on iOS platform”, in 19th International Symposium on Communications and Information Technologies (ISCIT), Vietnam, Sept 2019 Paper 4: Thai-Mai Dinh, and Ngoc-Son Duong “Smartphone Indoor Positioning System based on BLE iBeacon and Reliable region-based position correction algorithm”, in International Conference on Advanced Technologies for Communications (ATC), Vietnam, Oct 2019 Paper 5: Ngoc-Son Duong, Tuan-Anh Trinh Vu, and Thai-Mai Dinh, “Bluetooth Low Energy Based Indoor Positioning on iOS Platform”, in IEEE 12th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC), Vietnam, Dec 2018 Paper 6: Ngoc-Son Duong, and Thai-Mai Dinh, “Smartphone Indoor Positioning Based on Enhanced BLE Beacon Multi-lateration”, TELKOMNIKA, submitted, in revision Authorship “I hereby declare that the work contained in this thesis is of my own and has not been previously submitted for a degree or diploma at this or any other higher education institution To the best of my knowledge and belief, the thesis contains no materials previously published or written by another person except where due reference or acknowledgement is made.” Hanoi, 2020 Student i Acknowledgment This thesis would never have been done without the support of many colleagues, friends, and my family Firstly, I would like to thank my advisor, PhD Dinh Thi Thai Mai, who has given me all the support and guidance I needed as a master student I am very grateful to have had her trust in my ability, and I have often benefited from her insight and advice during the time I conducted my thesis work I am grateful to other teachers and friends in Communication Systems Laboratory, Faculty of Electronics and Telecommunications, University of Engineering and Technology I would like to also acknowledge my family and my beloved ones for cheering and supporting me during my six years at the university Your sentimental values mean a lot to me This work has been supported by Vietnam National University, Hanoi (VNU), under Project No QG.19.25 ii Abstract Nowadays, in large cities, human activities tend to shift from outdoor to indoor environments This has led to a growing need for services related to the indoor environment such as Location-Based Services (LBSs), Social Networking Services (SNSs), etc Location accuracy is a measurement of service quality GPS has done this well for outdoor environments However, due to the obstruction of building materials, GPS signals can not work well in indoor environments Therefore, many technologies are exploited to deploy indoor positioning systems (IPS) such as Wifi, RFID, Zigbee, etc To overcome the limitations of previous technologies, a Bluetooth-Low-Energy-based (BLE-based) technology, iBeacon was introduced as a appropriate solution for IPS requirements due to the advantages such as low energy consumption, wide-coverage, easy deployment, and potential high accuracy To achieve high location accuracy, this thesis proposes a real-time indoor positioning system which combines iBeacon technology and smartphone sensors Two main techniques are used for positioning, i.e, Pedestrian Dead Reckoning (PDR) and Range-based using Least Square Estimation (LSE) These two methods help each other create a highly accurate system Firstly, we offer a solution for Received-Signal-Strength-based (RSSbased) continuous positioning problem by investigating heterogeneity in RSS Secondly, we propose a method of improving accuracy for LSE We consider PDR-based position and improved LSE-based position both have a Gaussian uncertainty that comes from initial position plus drifting and RSS-to-distance conversion, respectively Then, two kinds of Normal distribution will be fused by the Kalman filter to produce more precise positions The method is intended to design a real-time system for locating moving target The results show our proposed solution is not only highly accurate but also feasible in actual deployment iii Contents Abbreviations vi List of Figures vii List of Tables viii Introduction 1.1 Motivation 1.2 Approach 1.3 Contribution 1.4 Outline Background 2.1 Positioning Technology 2.1.1 Bluetooth Low Energy 2.1.2 Inertial sensor 2.2 RSSI-based Positioning Techniques 2.2.1 Fingerprinting Method 2.2.2 Range-based Method (Lateral) 2.3 Bayesian Filtering - From Kalman Filters 2.3.1 General Bayes Filtering problem 2.3.2 Kalman Filter 2.3.3 Particle Filter Proposed System 3.1 System overview and architecture 3.2 PDR subsystem 3.2.1 Embedded Sensor Block 3.2.2 Sensor–based positioning method 3.2.3 Step Length Estimation 3.3 LSE subsystem 3.3.1 RSS Uncertainty Analysis 3.3.2 RSSI-to-Distance Conversion 3.3.3 Location Estimation 3.4 Kalman Fusion Evaluation 4.1 Experiment Setup 4.1.1 Device and Software 4.1.2 Experiment Setting 4.2 Results and Discussion iv 1 3 to Particle Filters 6 10 10 11 12 12 13 14 18 18 19 19 19 19 20 20 25 26 27 30 30 30 30 30 4.2.1 4.2.2 4.2.3 4.2.4 Ground Truth and Accuracy Comparisons Performance evaluation under impact of different velocity Performance evaluation under impact of different beacon density Compare to Fingerprinting 30 32 32 34 Conclusion 36 5.1 Conclusion 36 5.2 Future Work 36 v Abbreviations Order No Acronyms 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 a.k.a AOA BLE Eq Fig FM GPS i.e ID IMU INS IPS KF LBS LOS LS MD PAN PDR PF RF RFID RP RSS(I) Tab SIR SIS SNS SSID TOA UUID UWB Wi-fi Description as known as Angle of Arrival Bluetooth Low Energy Equation Figure Frequency Modulation Global Positioning System that is Identification Inertial Measurement Unit Inertial Navigation System Indoor Positioning System Kalman Filter Location-Based Service Light of Sight Least Square Mobile Device Personal Area Network Pedestrian Dead Reckoning Particle Filter Radio Frequency Radio Frequency Identification Device Reference Point Received Signal Strength (Indicator) Table Sequential Importance Re-sampling Sequential Importance Sampling Social Networking Service Service Set Identifier Time of Arrival Universally Unique Identifier Ultra Wide Band Wireless Fidelity vi List of Figures 1.1 Comparison of different signals for smartphone-based indoor localization [18] 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 Channel configuration of BLE BLE iBeacon protocol architecture INS axis system on iPhone (source: Apple) Accelerometer measures changes in velocity along the x, y, and Gyrocopter measure rotation rate in the x, y, and z axes Fingerprint Concept Least square position algorithm of three beacons Comparison of raw RSS and KF-filtered RSS The estimated position using Kalman filter Illustration of importance sampling method 3.1 3.2 3.3 System overview and architecture Change of acceleration as the user moves RSS uncertainty at different distances Legend: The bar charts represent observed data histograms at at different distances Each environmental case at each distance includes 400 samples Blue bar, light orange bar, purple bar, green bar denote LOS, wall blocked, column blocked, wall blocked situation, respectively; The dashed lines represent fitted line from data specified by Normal distribution; The solid lines represent the fused distribution of possible cases Linear approximations of distance path loss model Visual view of our proposed method Fusion of LSE-based position and PDR-based position 3.4 3.5 3.6 4.1 4.2 4.3 4.4 4.5 z axes 8 9 10 11 15 15 15 18 20 23 25 27 28 The position of the iBeacons and true path on the experiment area Ground truth and accuracy comparisons a) Distribution of corrective points b) Trajectories of true path, PDR path and proposed method path c) Cumulative localization error distributions of our proposed method Cumulative localization error distributions in cases: running and walking Average localization error given different number of iBeacon Comparison between different positioning methods a) Box-and-whisker plot of localization error for specific cases b) Trade-off between positioning accuracy and efforts of calibration time 31 vii 32 33 33 34 List of Tables 1.1 1.2 Comparison between Wi-Fi or BLE Beacons for indoor location Pros and cons of the positioning methods 2.1 Classic Bluetooth vesus BLE 3.1 3.2 Mean RSS and its standard deviation at different distances 24 Distance calculation model for each RSSI range 26 viii herein αh and βh are real coefficients of h-th piece and = δ0 < δ1 < < δm = ∞ To find an approximation, we use Newton’s method [48] Let f : R −→ R be a differentiable function We seek a solution of f (d) = 0, starting from a reference distance d0 At the nth step, given dn , compute the next approximation dn+1 by: dn+1 = dn − f (dn ) f (dn ) (3.11) We repeat (3.11) until n reach to maximum of number iteration The tangent lines that found at {d0 , d1 , , dnmax } create an approximation for non-linear function In our case, the distance model is adapted by 3-piece-wise linear function, i.e:   −5.11d − 58.814, ≤ d < 1.63 Γ(d) = −2.182d − 63.58, 1.63 ≤ d < 4.05 (3.12)   −0.8d − 69.177, 4.05 ≤ d < 10 An adaptive strategy is then applied for input as RSSI The details are depicted by Tab 3.2 Table 3.2: Distance calculation model for each RSSI range RSSI range -67.13 to -60 -72.42 to -67.13 -80 to -72.42 3.3.3 RSS-to-distance model Γ = −5.11d − 58.814 Γ = −2.182d − 63.58 Γ = −0.8d − 69.177 Location Estimation To locate a point in 2-D space, we use Least Square Estimation as range-based method To ensure reliability, our algorithm only works with beacons that have the strongest signal strength instead of all observed beacons Proposed Method for LSE As mentioned above, the level of RSSI uncertainty decreases as RSSI increases Therefore, it would be unfair to treat all nearby beacons as the same iBeacon with highest RSSI should be assigned with higher weight than the others To this, we move the estimated LSE-based position to a new position that belongs to the coverage of the nearest beacon This idea comes from true range lateral method It said that if the estimated distance is absolutely accurate, the estimated position must be on the intersection of circles which are created by satellites and their own estimated distance Since we consider that there is only one trusted circle, an estimated position must lie on it Fig 3.5 is the visual view ← → of the proposed method BT and a circle which has radius of dB , centered at B and is denoted by (B; dB ) Let P be an intersection of the infinite line and the trusted circle We wish to find P that satisfies the following conditions: ← → P = BT ∩ (B; dB ) PT is minimum 26 (3.13) RSSI Sensitivity of RSSI to d Uncertainty in RSSI d Uncertainty in d T P B Improved LSEbased position LSE-based position Figure 3.5: Visual view of our proposed method Indoor Context-based strategy correction Because iBeacons are uniformly distributed throughout the map, LSE might not work well for long and narrow environments In this case, we use single beacon as reference point instead Having inspiration from above method, we move PDR-based position to circle of the nearest beacon 3.4 Kalman Fusion Assume that both P and PDR-based position follows 2-dimensional Gaussian distribution, let us consider the fusion problem of combining position derived from PDR and P for resulting a new position has less uncertainty Denote, za = [zax , zay ] , zb = [zbx , zby ] and u = [ux , uy ] are positions of PDR, P and estimated result, respectively A rational way to fusion is to use Bayes law: P (u|za , zb ) ∝ P (u)P (u|za )P (u|zb ) (3.14) herein, P (u) is a prior density and P (u, z) is the likelihood of u given the position z Eq 3.14 can be written as follow: P (u|za , zb ) ∝ × exp u − za 2σa2 27 × exp u − zb 2σb2 (3.15) 0.4 Fused Position 0.35 0.3 Improved LSE-based position (a.k.a Corrective Point or Control Point) PDF 0.25 0.2 PDR-based position 0.15 0.1 0.05 Y (m) -2 -2 -3 -1 X (m) Figure 3.6: Fusion of LSE-based position and PDR-based position From Bayes Theorem, the fused MAP estimate is given by: uˆ = arg max P (u|za , zb ) u = arg [− log P (u|za , zb )] (3.16) u = arg u u − za 2σa2 + u − zb 2σb2 ˆ2 = The best estimation for (3.16) is: uˆ = (za σb2 + zb σa2 )(σa2 + σb2 )−1 with variance: σ (σa2 σb2 )(σa2 + σb2 )−1 This can be done in recursive form of Kalman filter, to update cur2 rent estimated position (ˆ ut , σ ˆt2 ) with improved LSE-based position (zb,t , σb,t ) to produce (ˆ ut+1 , σ ˆt+1 ) σ ˆ2 Kt = t (3.17) σ ˆt + σb,t uˆt+1 = uˆt + Kt (zb,t − uˆt ) (3.18) σ ˆt+1 = (1 − Kt )ˆ σt2 (3.19) herein, Kt is known as the optimal Kalman gain Fig 3.6 visualizes fusion in dimension space In our environmental case, the initial variance of PDR-based position, σa2 , is set by initial error that equals The variance of LSE-based position, σb2 , is calculated via uncertainty propagation with each RSS input, i.e: σb2 = 5.11−2 σ if −60 ≤ Γ < −67.13 or σb2 = 2.182−2 σ if −67.13 ≤ Γ < −72.42 or σb2 = 0.8−2 σ if −72.42 ≤ Γ < −80 We can see that improved LSE-based position does not always have low variance, so fusion should only be performed when conditions between iBeacon and smartphone are good Since we wish to achieve a meter level system, the variance of corrective points (control points) must be less than or equal to (i.e σ ≤ 1) Then, we can find a good condition via uncertainty propagation using Eq 3.12 and Eq 3.9 The good condition herein is to 28 get RSS value greater than: Γ= ∗ | − 2.182| + 10.067 σ|m| + 10.067 = ≈ −70(dBm) −0.1752 −0.1752 (3.20) where m is gradient of the line Γ(d) = −2.182d − 63.58 In summary, fusion is only performed in the case of having at least one scanned beacon that has RSSI greater than −70 dBm 29 Chapter Evaluation 4.1 Experiment Setup 4.1.1 Device and Software We implemented the whole system that encompasses the coordinate of iBeacons, the distance path loss model, the iBeacon ranging scheme, the localization algorithm in an iPhone SE running on iOS 12.0 In details, we use CoreLocation framework for RSS ranging This framework allow us read RSS at approx Hz We use CoreMotion framework for sensor reading and simd module in Accelerate framework for matrix computing For PDR, IMU is sampled at 60 Hz When a step is detected, the application records the time stamp, estimated distance and then sent data via mail Ultimately, the data is used to plot figure using MATLAB The anchor node used in our experiment is Estimote Beacon They are 5.0 BLE beacons which are configurable by using a smartphone application To ensure reliability, beacons are set to generate BLE signal at 10 Hz and at transmitting power of dBm All beacon have the same technical configuration as well 4.1.2 Experiment Setting The experimental testbed is a typical indoor environment with medium open space and small hall space Acreage of area is approximately equal to 350 m2 To evaluate the effectiveness of the system, we deployed 11 iBeacon nodes uniformly distributed on the wall at a height of 1.6 m The maximum distance between adjacent iBeacons in the interested area is about 6-8 m The iBeacon placement follows a strategy introduced in [27] On the receiver side, smartphone is kept horizontally in hand and close to the body The user then walks along a true path to evaluate the system’s accuracy in several situations The initial position in each experiment is estimated by LS The map, iBeacon position, and true path are shown in Fig 4.1 4.2 4.2.1 Results and Discussion Ground Truth and Accuracy Comparisons Fig 4.2b shows the positioning results obtained in a single loop with normal speed, say, steps per second In this figure, blue, green, and brown lines represent the true path, our method path, and PDR path, respectively As we can see, the PDR path shows a distortion, far from the true path This can be explained by error derived from two 30 O x STORE HOUSE 10 m2 COMPUTER CENTER 11 8.1 m G2B Area MULTIMEDIA ROOM 10 x Up Up Up Up R 101 Start Point y Figure 4.1: The position of the iBeacons and true path on the experiment area 31 sources: initial position and sensor noise Since no corrections are made, initial error exists in the whole moving process and becomes even more serious if the user starts from a position where lack of iBeacon signal When combined with the error derived from sensor noise, the system accuracy may further decrease as the user carries out more loop In the critical case, say, two parallel aisles separated by a wall, the true position may turn into wrong ones that belongs nearby aisle, as a result, the user may misunderstand their position and then make the wrong decision In the case of using the proposed method, the created path is nearly identical to the true path The reason is that the user position is regularly corrected by corrective points that are derived from LSE or proximity iBeacon Since corrective point lies in the true path (as shown in Fig 4.2a) and they have less uncertainty as well, fused position tends to be pulled closer towards them Thanks to the adjustment of iBeacons, the proposed algorithm achieves a high localization accuracy In Fig 4.2c, we can see that the probability of having localization error less than m is 60 % The mean localization accuracy of our proposed approach is 1.04 m a b c 25 25 20 20 15 10 0.8 15 CDF y (m) y (m) 0.4 10 True Path Proposed Method (Corrective Point) iBeacon Node True Path Proposed Method PDR only iBeacon Node 0 10 x (m) 20 0.6 0.2 Proposed Method PDR only 0 10 x (m) 20 Localization Error (m) Figure 4.2: Ground truth and accuracy comparisons a) Distribution of corrective points b) Trajectories of true path, PDR path and proposed method path c) Cumulative localization error distributions of our proposed method 4.2.2 Performance evaluation under impact of different velocity In this experiment, we want to assess how velocity affect the accuracy of the system The test speed is about 1.25 - 1.3 m/s for the walking case and 2.5 m/s for the running case The completes path of the experiment includes 531 steps with the distance equals 339 m for loops The results is shown in Fig 4.3 In general, the walking case has done better than running case For PDR standalone, the INS subsystem successfully detected 522, 518 steps and the calculated track distance is 323.67, 321.19 m, achieving a 98.30 %, 97.55 %, and 95.47, 94.74 % accuracy for step detection, distance estimation in the walking and running case, respectively When using our proposed method, the average localization error for walking case and running case is 1.04 m and 1.45 m, respectively Our method is 54.6% better than PDR in walking case and 42.4% in running case 4.2.3 Performance evaluation under impact of different beacon density To evaluate the effect of the number of iBeacons on the system accuracy, we, in turn, remove iBeacons from the map in the way that they are uniformly distributed and at least beacons must be detected in open spaces We consider that 11 iBeacon is reasonable 32 0.8 CDF 0.6 0.4 PDR Only (Walking) PDR Only (Running) Proposed Method (Walking) Proposed Method (Running) 0.2 0 Localization Error (m) Figure 4.3: Cumulative localization error distributions in cases: running and walking 2.5 Avg Localization Error (m) 1.5 0.5 PDR only 10 11 Number of iBeacon Figure 4.4: Average localization error given different number of iBeacon 33 for our environment Beyond this amount of iBeacon, the system is no longer considered as low cost The system with 10, 9, 8, 7, iBeacons corresponding to the case of iBeacon number 4, and 8, and and 11, and and 10 and 11, and and and 10 and 11 is/are discarded, respectively From the trend is shown in Fig 4.4, we can see that positioning accuracy increase as the number of iBeacon increase The system has the highest accuracy of 1.1 m with 11 iBeacons and the lowest one is 1.6 m with iBeacons This is understandable because the more iBeacons we have, the more corrective points we obtain We also realize that the change of accuracy between cases was insignificant as the number of iBeacon increased This indicates that we can have an inexpensive system with nearly equivalent accuracy For example, iBeacons is good as well In this case, the number of iBeacon which serves for each area is adequate, i.e, iBeacons for open spaces and iBeacons for corridors 4.2.4 Compare to Fingerprinting Fingerprinting is one of the common techniques for indoor localization and tracking Thus, we compared our proposed approach with Fingerprinting under the constraints of criteria, i.e localization accuracy and calibration time The studied cases that we choose for comparison include: i ) conventional Fingerprint (FP) for discrete positioning with multiple measurements in fixed positions; ii ) combination of PDR and conventional Fingerprinting (PDR+FP), and iii ) lightweight Fingerprint (LW-FP) [43] for continuous positioning For conventional Fingerprinting, we use 28 uniformly distributed reference points (RP) Each RP include 200 vectors at four different directions The map-matching algorithm is the nearest neighbor base on major voting For LW-FP, the number of vectors for each RP is similar to FP but only RPs were selected for data collection In this experiment, PDR is considered as well The results are shown in Fig 4.5 Proposed Method PDR+FP b 1.5 1 0.5 LW FP [10] FP PDR P R 0] od PD eth P [1 R+F F D M P ed LW os Calibration Time (h) 2.5 Avg Localization Error (m) Localization Error (m) a FP op Pr Figure 4.5: Comparison between different positioning methods a) Box-and-whisker plot of localization error for specific cases b) Trade-off between positioning accuracy and efforts of calibration time As presented in Fig 4.5a, our method, PDR+FP, and LW-FP give equivalent results that reach to 1-meter-level The variance of our method indicates that the fused position depends on LS-based positions, which is returned by an unstable way The smaller variance of PDR+FP, and LW-FP, in another way, come from fixed RPs, which are considered stable and have no uncertainty Although PDR+FP and LW-FP achieve high accuracy as well, the price to pay, in this case, is the need to collect data and maintain for database of RP The conventional FP and LW-FP take 3.7 and 1.3 hours for data collection tasks, 34 respectively In actual deployment, time for data collection could be enormous, even using LW-FP While our proposed method only needs to calibrate for RSS-to-distance model, which requires 0.7 hours In this aspect, our method supports scalability better than Fingerprinting approach 35 Chapter Conclusion 5.1 Conclusion On the Internet of Thing (IoT) system, iBeacon promises to bring back many benefits not only for indoor positioning but also for many other fileds In this thesis, we propose a realtime indoor positioning system in smartphone via BLE iBeacon signal In which, we used embedded sensors for displacement calculation and BLE iBeacon signal as a calibrated opportunity for sensor-based IPS Firstly, we investigate the problems associated with the uncertainty of RSS, then offer solution for RSS-based locating moving target under low sampling rate Secondly, we proposed method of improving accuracy for LSE method Improved LSE-based position then fused with PDR-based position using Kalman filter to produce more accurate positions The accuracy of the proposed approaches proved high persuasion for service providers to deploy this low-complexity system for various location-based services 5.2 Future Work This thesis opens several avenues for future research: • Combination with data processing techniques: Noisy environment has always been a problem that takes a lot of effort to solve Advanced data processing techniques can help to counter the effects of signal fluctuations, fading, interference, and so on • LOS/NLOS detection: The influence of obstacles is very heavy Therefore, this problem needs solving to improve the accuracy of the system • Combination with different technology: It is reasonable to have a more accurate hybrid system with more technology, for example, Wi-fi and more RF measurement methods such as ToA, TDoA, and AoA 36 Bibliography [1] F Zafari, A Gkelias and K Leung, “A Survey of Indoor Localization Systems and Technologies”, IEEE Communications Surveys & Tutorials, vol 21, no 3, pp 25682599, 2019 [2] H White, “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity”, Econometrica, vol 48, no 4, p 817, 1980 [3] W Zhao, S Han, R Hu, W Meng and Z Jia, “Crowdsourcing and Multisource Fusion-Based Fingerprint Sensing in Smartphone Localization”, IEEE Sensors Journal, vol 18, no 8, pp 3236-3247, 2018 [4] Chung-Hao Huang, Lun-Hui Lee, C Ho, Lang-Long Wu and Zu-Hao Lai, “RealTime 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 [5] Q Tian, K Wang and Z Salcic, “A Low-Cost INS and UWB Fusion Pedestrian Tracking System”, IEEE Sensors Journal, vol 19, no 10, pp 3733-3740, 2019 [6] Z Chen, Q Zhu and Y Soh, “Smartphone Inertial Sensor-Based Indoor Localization and Tracking With iBeacon Corrections”, IEEE Transactions on Industrial Informatics, vol 12, no 4, pp 1540-1549, 2016 [7] H Zou, Z Chen, H Jiang, L Xia and C Spanos, “Accurate indoor localization and tracking using mobile phone inertial sensors, WiFi and iBeacon”, in 2017 IEEE International Symposium on Inertial Sensors and Systems (INERTIAL), Kauai, HI, USA, 2017 [8] H Xia, J Zuo, S Liu and Y Qiao, “Indoor Localization on Smartphones Using Built-In Sensors and Map Constraints”, IEEE Transactions on Instrumentation and Measurement, vol 68, no 4, pp 1189-1198, 2019 [9] R Yadav, B Bhattarai, H Gang and J Pyun, “Trusted K Nearest Bayesian Estimation for Indoor Positioning System”, IEEE Access, vol 7, pp 51484-51498, 2019 [10] A Belmonte-Hernandez, G Hernandez-Penaloza, D Martin Gutierrez and F Alvarez, “SWiBluX: Multi-Sensor Deep Learning Fingerprint for Precise Real-Time Indoor Tracking”, IEEE Sensors Journal, vol 19, no 9, pp 3473-3486, 2019 [11] Estimote, Inc (2019) Estimote (Version 2.42.5) [Mobile application software] Retrieved from https://apps.apple.com/us/app/estimote/id686915066 [12] Developer.apple.com (2019) Core Motion — Apple Developer Documentation [Online] Available at: https://developer.apple.com/documentation/coremotion 37 [13] Developer.apple.com (2019) Core Location — Apple Developer Documentation [Online] Available at: https://developer.apple.com/documentation/corelocation [14] Pachi, A., and Ji, T “Frequency and velocity of people walking”, The Structural Engineer, vol 83, pp 36-40, 2005 [15] Getting Heading and Course Information — Apple Developer Documentation”, Developer.apple.com, 2019 [Online] Available: https://developer.apple.com/documentation/corelocation/getting heading and course information [16] R Liu, C Yuen, T Do and U Tan, “Fusing Similarity-Based Sequence and Dead Reckoning for Indoor Positioning Without Training”, IEEE Sensors Journal, vol 17, no 13, pp 4197-4207, 2017 [17] S Pradhan, Y Bae, J Pyun, N Ko and S Hwang,“Hybrid TOA Trilateration Algorithm Based on Line Intersection and Comparison Approach of Intersection Distances”, Energies, vol 12, no 9, p 1668, 2019 [18] S He and K Shin, “Geomagnetism for Smartphone-Based Indoor Localization”, ACM Computing Surveys, vol 50, no 6, pp 1-37, 2018 [19] E C L Chan, G Baciu and S Mak, “Wireless Tracking Analysis in Location Fingerprinting”, in IEEE International Conference on Wireless and Mobile Computing, Networking and Communications, Avignon, France, 2008 [20] N Yu, X Zhan, S Zhao, Y Wu and R Feng, “A Precise Dead Reckoning Algorithm Based on Bluetooth and Multiple Sensors”, IEEE Internet of Things Journal, vol 5, no 1, pp 336-351, 2018 [21] V Cant´on Paterna, A Calveras Aug´e, J Paradells Aspas and M P´erez Bullones, ”A Bluetooth Low Energy Indoor Positioning System with Channel Diversity, Weighted Trilateration and Kalman Filtering”, Sensors, vol 17, no 12, p 2927, 2017 [22] W Kang and Y Han, “SmartPDR: Smartphone-Based Pedestrian Dead Reckoning for Indoor Localization”, IEEE Sensors Journal, vol 15, no 5, pp 2906-2916, 2015 [23] S Naghdi and K O’Keefe, “Trilateration With BLE RSSI Accounting for Pathloss Due to Human Obstacles,” in 2019 International Conference on Indoor Positioning and Indoor Navigation (IPIN), Pisa, Italy, 2019, pp 1-8 [24] H Hashemi, “The indoor radio propagation channel”, Proceedings of the IEEE, vol 81, no 7, pp 943-968, 1993 [25] C Gomez, J Oller and J Paradells, “Overview and Evaluation of Bluetooth Low Energy: An Emerging Low-Power Wireless Technology”, Sensors, vol 12, no 9, pp 11734-11753, 2012 [26] Faragher R., Harle R., “An Analysis of the Accuracy of Bluetooth Low Energy for Indoor Positioning Applications,” in 27th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+ 2014), Sep 2014 [27] J Rezazadeh, R Subramanian, K Sandrasegaran, X Kong, M Moradi and F Khodamoradi, “Novel iBeacon Placement for Indoor Positioning in IoT”, IEEE Sensors Journal, vol 18, no 24, pp 10240-10247, 2018 38 [28] Apple Developer (2019) What’s New in Core WWDC 2013 - Videos - Apple Developer [Online] https://developer.apple.com/videos/play/wwdc2013/307 Location Available at: [29] S Subedi, H Gang, N Ko, S Hwang and J Pyun, “Improving Indoor Fingerprinting Positioning With Affinity Propagation Clustering and Weighted Centroid Fingerprint”, IEEE Access, vol 7, pp 31738-31750, 2019 [30] Weisstein, Eric W “Least Squares Fitting–Logarithmic.” From MathWorld–A Wolfram Web Resource http://mathworld.wolfram.com/LeastSquaresFittingLogarithmic.html [31] Weinberg, H Using the ADXL202 in Pedometer and Personal Navigation Applications In Application Notes American Devices; Analog Devices, Inc.: Norwood, MA, USA, 2002 [32] Yassin, A., Nasser, Y., Awad, M., Al-Dubai, A., Liu, R., Yuen, C., Aboutanios, E “Recent advances in indoor localization: a survey on theorectical approaches and applications”, IEEE Communications Surveys & Tutorials, vol 19, no 2, pp 13271346, 2017 [33] A Zanella, “Best Practice in RSS Measurements and Ranging”, IEEE Communications Surveys & Tutorials, vol 18, no 4, pp 2662-2686, 2016 [34] D Feng, C Wang, C He, Y Zhuang and X Xia, “Kalman-Filter-Based Integration of IMU and UWB for High-Accuracy Indoor Positioning and Navigation”, IEEE Internet of Things Journal, vol 7, no 4, pp 3133-3146, 2020 [35] B Hofmann-Wellenhof, James Collins and Herbert Lichtenegger “Global Positioning System: Theory and Practice”, Springer-Verlag GmbH 2000 [36] S He and K Shin, “Geomagnetism for Smartphone-Based Indoor Localization”, ACM Computing Surveys, vol 50, no 6, pp 1-37, 2018 [37] J Jiao, F Li, Z Deng and W Ma, “A Smartphone Camera-Based Indoor Positioning Algorithm of Crowded Scenarios with the Assistance of Deep CNN”, Sensors, vol 17, no 4, p 704, 2017 [38] M Arulampalam, S Maskell, N Gordon and T Clapp, “A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking”, IEEE Transactions on Signal Processing, vol 50, no 2, pp 174-188, 2002 [39] I Guvenc, C Abdallah, R Jordan, and O Dedoglu, “Enhancements to RSS-based indoor tracking systems using Kalman filters,” in Proc Int Signal Processing Conf., 2003 [40] Greg Welch and Gary Bishop “An Introduction to the Kalman Filter” Technical Report University of North Carolina, Chapel Hill, USA, 1995 [41] S Sadowski and P Spachos, “RSSI-Based Indoor Localization With the Internet of Things”, IEEE Access, vol 6, pp 30149-30161, 2018 [42] L Kanaris, A Kokkinis, A Liotta and S Stavrou, “Fusing Bluetooth Beacon Data with Wi-Fi Radiomaps for Improved Indoor Localization”, Sensors, vol 17, no 4, p 812, 2017 39 [43] Thai-Mai Thi Dinh, Ngoc-Son Duong, Kumbesan Sandrasegaran, “Smartphone-based Indoor Positioning Using BLE iBeacon and Reliable Lightweight Fingerprint Map”, IEEE Sensors Journal, 2020 In press https://doi.org/10.1109/JSEN.2020.2989411 [44] A Goldsmith, Wireless Communications New York, NY, USA: Cambridge University Press, 2005 [45] Yan, C Tiberius, G Janssen, P Teunissen and G Bellusci, “Review of range-based positioning algorithms”, IEEE Aerospace and Electronic Systems Magazine, vol 28, no 8, pp 2-27, 2013 [46] M D Yacoub, “The α-µ distribution: A physical fading model for the stacy distribution,” IEEE Transactions on Vehicular Technology, vol 56, no 1, pp 27 –34, 2007 [47] Yunye Jin, Hong-Song Toh, Wee-Seng Soh, and Wai-Choong Wong, “A Robust DeadReckoning Pedestrian Tracking System with Low Cost Sensors”, in IEEE International Conference on Pervasive Computing and Communications (PerCom), Seattle, WA, USA, 2011 [48] Weisstein, Eric W “Newton’s Method.” From MathWorld-A Wolfram Web Resource [Online] Available: https://mathworld.wolfram.com/NewtonsMethod.html 40 ... Wi-fi and BLE A detailed comparison between Wi-fi and BLE is shown in Tab 1.1 Table 1.1: Comparison between Wi-Fi or BLE Beacons for indoor location Accuracy Compatible with Android? BLE Beacon... database using a matching algorithm Due to the instability of the BLE signal, indoor localization using only BLE beacon result in large errors Thus, many studies have combined BLE beacon with other... Improved Indoor Localization? ??, Sensors, vol 17, no 4, p 812, 2017 39 [43] Thai-Mai Thi Dinh, Ngoc-Son Duong, Kumbesan Sandrasegaran, ? ?Smartphone- based Indoor Positioning Using BLE iBeacon and Reliable