# Received Signal Strength Fingerprinting-Based Indoor Location Estimation Employing Machine Learning

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Related Work

#### 2.1. RSS-Based Fingerprinting Improvement

#### 2.2. Fusion of RSS and Other Information

## 3. Indoor Measurement Campaign

#### 3.1. Remote and Self-Positioning

#### 3.2. Experimental Setup

#### 3.3. Indoor Environment

#### 3.4. Measurement Scenario

#### 3.5. Dataset

- The raw measurement, collected as the received signal levels at anchors in the remote positioning mode, contain, for each anchor, four channel numbers and four corresponding RSS levels, as measured by four BLE modules (onboard each anchor). At this point, it is not assured that RSS levels at all three advertisement channels are recorded and that all values are valid. In the self-positioning mode, one measurement contains the same set of values, although taken in the opposite direction.
- All values where RSS level of −110 dBm was recorded are considered to be missing measurements as at this level the receiver fails to measure the actual received power level.
- In order to ensure there are no missing values, records from two consecutive measurements are taken as one sample. For each anchor, the first valid measurement is recorded in the sample and the remaining measurements are discarded (e.g., measurements from further antennas). This allows to clean the data set in such a way that no missing values appear in the data.

## 4. One-Dimensional Positioning

#### 4.1. kNN Classifier

- The entire training set needs to be stored, which can lead to large memory requirements for bigger datasets.
- The metric to measure the distance from neighbors needs to be selected. A common choice is the Minkowski metric, defined for two points $X,Y$ in the n-dimensional space as:$$D(X,Y)={\left(\sum _{i=1}^{n}{|{x}_{i}-{y}_{i}|}^{p}\right)}^{\frac{1}{p}},$$
- The number of neighbors to be considered is a design choice for the algorithm and depends on the characteristics of the dataset. In general, higher values of k lead to smoother decision boundaries, while smaller numbers of k capture data variations more faithfully.

#### 4.2. Support Vector Machines

#### 4.3. Random Forest

#### 4.4. Multi-Layer Perceptron

## 5. Two-Dimensional Positioning

#### 5.1. kNN Classifier

#### 5.2. SVM Classifier

#### 5.3. Random Forest Classifier

#### 5.4. Multi-Layer Perceptron

## 6. Summary and Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## References

- Zafari, F.; Gkelias, A.; Leung, K.K. A Survey of Indoor Localization Systems and Technologies. IEEE Commun. Surv. Tutor.
**2019**, 21, 2568–2599. [Google Scholar] [CrossRef] [Green Version] - Al-Ammar, M.A.; Alhadhrami, S.; Al-Salman, A.; Alarifi, A.; Al-Khalifa, H.S.; Alnafessah, A.; Alsaleh, M. Comparative Survey of Indoor Positioning Technologies, Techniques, and Algorithms. In Proceedings of the 2014 International Conference on Cyberworlds, Santander, Spain, 6–8 October 2014; pp. 245–252. [Google Scholar] [CrossRef]
- Stavrou, V.; Bardaki, C.; Papakyriakopoulos, D.; Pramatari, K. An Ensemble Filter for Indoor Positioning in a Retail Store Using Bluetooth Low Energy Beacons. Sensors
**2019**, 19, 4550. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Xiao, J.; Zhou, Z.; Yi, Y.; Ni, L.M. A survey on wireless indoor localization from the device perspective. ACM Comput. Surv. CSUR
**2016**, 49, 1–31. [Google Scholar] [CrossRef] - Yassin, A.; Nasser, Y.; Awad, M.; Al-Dubai, A.; Liu, R.; Yuen, C.; Raulefs, R.; Aboutanios, E. Recent Advances in Indoor Localization: A Survey on Theoretical Approaches and Applications. IEEE Commun. Surv. Tutor.
**2017**, 19, 1327–1346. [Google Scholar] [CrossRef] [Green Version] - Baert, M.; Rossey, J.; Shahid, A.; Hoebeke, J. The Bluetooth mesh standard: An overview and experimental evaluation. Sensors
**2018**, 18, 2409. [Google Scholar] [CrossRef] [Green Version] - Yang, J.; Poellabauer, C.; Mitra, P.; Neubecker, C. Beyond beaconing: Emerging applications and challenges of BLE. Ad Hoc Netw.
**2020**, 97, 1–12. [Google Scholar] [CrossRef] [Green Version] - Kriz, P.; Maly, F.; Kozel, T. Improving indoor localization using Bluetooth low energy beacons. Mob. Inf. Syst.
**2016**, 2016, 1–11. [Google Scholar] [CrossRef] [Green Version] - Neburka, J.; Tlamsa, Z.; Benes, V.; Polak, L.; Kaller, O.; Bolecek, L.; Sebesta, J.; Kratochvil, T. Study of the performance of RSSI based Bluetooth Smart indoor positioning. In Proceedings of the 2016 26th International Conference Radioelektronika (RADIOELEKTRONIKA), Kosice, Slovakia, 19–20 April 2016; pp. 121–125. [Google Scholar] [CrossRef]
- Pelant, J.; Tlamsa, Z.; Benes, V.; Polak, L.; Kaller, O.; Bolecek, L.; Kufa, J.; Sebesta, J.; Kratochvil, T. BLE device indoor localization based on RSS fingerprinting mapped by propagation modes. In Proceedings of the 2017 27th International Conference Radioelektronika (RADIOELEKTRONIKA), Brno, Czech Republic, 19–20 April 2017; pp. 1–5. [Google Scholar] [CrossRef]
- Rozum, S.; Sebesta, J. SIMO RSS measurement in Bluetooth low power indoor positioning system. In Proceedings of the 2018 28th International Conference Radioelektronika, Prague, Czech Republic, 19–20 April 2018; pp. 1–5. [Google Scholar] [CrossRef]
- Rozum, S.; Kufa, J.; Polak, L. Bluetooth Low Power Portable Indoor Positioning System Using SIMO Approach. In Proceedings of the 2019 42nd International Conference on Telecommunications and Signal Processing (TSP), Budapest, Hungary, 1–3 July 2019; pp. 228–231. [Google Scholar] [CrossRef]
- Li, G.; Geng, E.; Ye, Z.; Xu, Y.; Lin, J.; Pang, Y. Indoor positioning algorithm based on the improved RSSI distance model. Sensors
**2018**, 18, 2820. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Giuliano, R.; Cardarilli, G.C.; Cesarini, C.; Di Nunzio, L.; Fallucchi, F.; Fazzolari, R.; Mazzenga, F.; Re, M.; Vizzarri, A. Indoor localization system based on Bluetooth low energy for museum applications. Electronics
**2020**, 9, 1055. [Google Scholar] [CrossRef] - Atashi, M.; Malekzadeh, P.; Salimibeni, M.; Hajiakhondi-Meybodi, Z.; Plataniotis, K.N.; Mohammadi, A. Orientation-matched multiple modeling for RSSI-based indoor localization via BLE sensors. In Proceedings of the 2020 28th European Signal Processing Conference (EUSIPCO), Amsterdam, The Netherlands, 18–21 January 2021; pp. 1702–1706. [Google Scholar] [CrossRef]
- Ng, P.C.; She, J.; Rong, R. Compressive RF Fingerprint Acquisition and Broadcasting for Dense BLE Networks. IEEE Trans. Mob. Comput.
**2020**. [Google Scholar] [CrossRef] - Heyn, R.; Kuhn, M.; Schulten, H.; Dumphart, G.; Zwyssig, J.; Trosch, F.; Wittneben, A. User Tracking for Access Control with Bluetooth Low Energy. In Proceedings of the 2019 IEEE 89th Vehicular Technology Conference (VTC2019-Spring), Kuala Lumpur, Malaysia, 28 April–1 May 2019; pp. 1–7. [Google Scholar] [CrossRef]
- Wisanmongkol, J.; Klinkusoom, L.; Sanpechuda, T.; Kovavisaruch, L.; Kaemarungsi, K. Multipath Mitigation for RSSI-Based Bluetooth Low Energy Localization. In Proceedings of the 2019 19th International Symposium on Communications and Information Technologies (ISCIT), Ho Chi Minh City, Vietnam, 25–27 September 2019; pp. 47–51. [Google Scholar] [CrossRef]
- Sun, X.; Ai, H.; Tao, J.; Hu, T.; Cheng, Y. BERT-ADLOC: A secure crowdsourced indoor localization system based on BLE fingerprints. Appl. Soft Comput.
**2021**, 104, 1–10. [Google Scholar] [CrossRef] - Abed, A.; Abdel-Qader, I. RSS-Fingerprint Dimensionality Reduction for Multiple Service Set Identifier-Based Indoor Positioning Systems. Appl. Sci.
**2019**, 9, 3137. [Google Scholar] [CrossRef] [Green Version] - Sthapit, P.; Gang, H.S.; Pyun, J.Y. Bluetooth Based Indoor Positioning Using Machine Learning Algorithms. In Proceedings of the 2018 IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia), JeJu, Korea, 24–26 June 2018; pp. 206–212. [Google Scholar] [CrossRef]
- Duong, N.S.; Dinh, T.M. Indoor Localization with lightweight RSS Fingerprint using BLE iBeacon on iOS platform. In Proceedings of the 2019 19th International Symposium on Communications and Information Technologies (ISCIT), Ho Chi Minh City, Vietnam, 25–27 September 2019; pp. 91–95. [Google Scholar] [CrossRef]
- Kajioka, S.; Mori, T.; Uchiya, T.; Takumi, I.; Matsuo, H. Experiment of indoor position presumption based on RSSI of Bluetooth LE beacon. In Proceedings of the 2014 IEEE 3rd Global Conference on Consumer Electronics (GCCE), Tokyo, Japan, 7–10 October 2014; pp. 337–339. [Google Scholar] [CrossRef]
- Campaña, F.; Pinargote, A.; Domínguez, F.; Peláez, E. Towards an indoor navigation system using Bluetooth Low Energy Beacons. In Proceedings of the 2017 IEEE Second Ecuador Technical Chapters Meeting (ETCM), Salinas, Ecuador, 16–20 October 2017; pp. 1–6. [Google Scholar] [CrossRef]
- Iqbal, Z.; Luo, D.; Henry, P.; Kazemifar, S.; Rozario, T.; Yan, Y.; Westover, K.; Lu, W.; Nguyen, D.; Long, T.; et al. Accurate real time localization tracking in a clinical environment using Bluetooth Low Energy and deep learning. PLoS ONE
**2018**, 13, e0205392. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Lovon-Melgarejo, J.; Castillo-Cara, M.; Orozco-Barbosa, L.; García-Varea, I. Supervised learning algorithms for indoor localization fingerprinting using BLE4.0 beacons. In Proceedings of the 2017 IEEE Latin American Conference on Computational Intelligence (LA-CCI), Arequipa, Peru, 8–10 November 2017; pp. 1–6. [Google Scholar] [CrossRef]
- Jondhale, S.R.; Deshpande, R.S. GRNN and KF framework based real time target tracking using PSOC BLE and smartphone. Ad Hoc Netw.
**2019**, 84, 19–28. [Google Scholar] [CrossRef] - Takayama, T.; Umezawa, T.; Komuro, N.; Osawa, N. An Indoor Positioning Method Based on Regression Models with Compound Location Fingerprints. In Proceedings of the 2018 Ubiquitous Positioning, Indoor Navigation and Location-Based Services (UPINLBS), Wuhan, China, 22–23 March 2018; pp. 1–7. [Google Scholar] [CrossRef]
- Sou, S.I.; Lin, W.H.; Lan, K.C.; Lin, C.S. Indoor Location Learning Over Wireless Fingerprinting System with Particle Markov Chain Model. IEEE Access
**2019**, 7, 8713–8725. [Google Scholar] [CrossRef] - Zhuang, Y.; Yang, J.; Li, Y.; Qi, L.; El-Sheimy, N. Smartphone-Based Indoor Localization with Bluetooth Low Energy Beacons. Sensors
**2016**, 16, 596. [Google Scholar] [CrossRef] [Green Version] - Theodoridis, S.; Pikrakis, A.; Koutroumbas, K.; Cavouras, D. Introduction to Pattern Recognition: A MATLAB Approach; Academic Press: Cambridge, MA, USA, 2010. [Google Scholar]
- Duong, S.N.; Trinh, A.V.T.; Dinh, T.M. Bluetooth Low Energy Based Indoor Positioning on iOS Platform. In Proceedings of the 2018 IEEE 12th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC), Hanoi, Vietnam, 12–14 September 2018; pp. 57–63. [Google Scholar] [CrossRef]
- Liu, H.; Darabi, H.; Banerjee, P.; Liu, J. Survey of Wireless Indoor Positioning Techniques and Systems. IEEE Trans. Syst. Man Cybern. Part C Appl. Rev.
**2007**, 37, 1067–1080. [Google Scholar] [CrossRef] - Telegies. ETRX35x ZigBee Modules; TG-ETRX35X-PM-010-103; Telegies: Austin, TX, USA, 2010. [Google Scholar]
- Laird Connectivity. BL652 Series Bluetooth v5; BL652-SA and BL652-SC, Version 2.2; Laird Connectivity: Akron, OH, USA, 2016. [Google Scholar]
- STMicroelectronics. ARM-Based Microcontroller; STM32F091xB and STM32F091x; STM32F091xB and STM32F091x; Rev. 3; STMicroelectronics: Geneva, Switzerland, 2015. [Google Scholar]
- PulseLarsen Electronics. Wireless External Antenna for 2.4 GHz Applications; W1030; Version 1.1; Pulse Electronics: San Diego, CA, USA, 2007. [Google Scholar]
- Jung, Y. Multiple predicting K-fold cross-validation for model selection. J. Nonparametr. Stat.
**2017**, 30, 197–215. [Google Scholar] [CrossRef] - Fong-Mata, M.B.; García-Guerrero, E.E.; Mejía-Medina, D.A.; López-Bonilla, O.R.; Villarreal-Gómez, L.J.; Zamora-Arellano, F.; López-Mancilla, D.; Inzunza-González, E. An Artificial Neural Network Approach and a Data Augmentation Algorithm to Systematize the Diagnosis of Deep-Vein Thrombosis by Using Wells’ Criteria. Electronics
**2020**, 9, 1810. [Google Scholar] [CrossRef] - Tian, X.; Wang, M.; Li, W.; Jiang, B.; Xu, D.; Wang, X.; Xu, J. Improve accuracy of fingerprinting localization with temporal correlation of the RSS. IEEE Trans. Mob. Comput.
**2017**, 17, 113–126. [Google Scholar] [CrossRef] - Hoang, M.T.; Zhu, Y.; Yuen, B.; Reese, T.; Dong, X.; Lu, T.; Westendorp, R.; Xie, M. A soft range limited K-nearest neighbors algorithm for indoor localization enhancement. IEEE Sens. J.
**2018**, 18, 10208–10216. [Google Scholar] [CrossRef] [Green Version] - Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V. Scikit-learn: Machine Learning in Python. J. Mach. Learn. Res.
**2011**, 12, 2825–2830. [Google Scholar]

**Figure 1.**Timing of advertisement transmission and reception. The BLE module sends three identical consecutive advertising packets on different advertisement channels on every advertising event. The scanner receives only one report depending on the listening channel. It can actively listen for the whole Scan Interval or only a part of it. Time marks are relative to the Advertiser or Scanner and as such they are intentionally misaligned to illustrate the real-world situations, where the BLE communication is not synchronized.

**Figure 2.**Block diagram of the BLE-based localization system using SIMO approach (based on [12]).

**Figure 3.**Measurement scenario and floor plan. Green boxes denote the anchors. The height of the anchors is 2.2 m above the floor level, the height of tags is 1.0 m. Scale of X axis and Y axis is different. Room dimensions are approximate.

**Figure 5.**A visualisation of RSS values received from three anchors in a three-dimensional space, channel 37 only. The color of the points represents the location distance. For simplicity, only 20% of data has been used in the plot.

**Figure 6.**Mean values of received signal strengths at four anchor location when the transmitter is the moving tag. For each location, the mean of all received measurement samples is displayed. Peaks of RSS are visible for all advertisement channels for distances from 0 to 35 m.

**Figure 7.**Flowchart of the model training and testing procedure. There are two phases of the training. At first, we look for the best model parameters on the training data (with no knowledge of the test data) by splitting the training data into n folds, then training on $n-1$ folds and validating on the rest. This way, for example, we find the number of neighbors for the best performance in the kNN model (e.g., we find that $k=1$). Then, when we are ready to select the best model parameters—we create the final model on the whole training set with the prerequisite that $k=1$. Only this final training is then used to evaluate the performance on the test set.

**Figure 8.**Scatter plot showing the predicted versus the real distances (test set) in a single dimension for the kNN classifier with k = 1. The color of the points represents the number of points at a given location.

**Figure 9.**Mean accuracy of prediction achieved on the validation set for a varying number of neigbors in the kNN classifier.

**Figure 10.**Scatter plot showing the predicted versus the real distances in a single dimension for the SVM classifier with $C=20$.

**Figure 11.**Accuracy of prediction achieved for a varying regularization parameter C in the SVM classifier.

**Figure 12.**Scatter plot showing the predicted versus the real distances in a single dimension with the Random Forest classifier with the number of trees equal to 160.

**Figure 13.**A sample structure of a Multi-Layer Perceptron with five inputs, two hidden layers and three outputs.

**Figure 14.**Scatter plot showing the predicted versus the real distances in a single dimension using the Multi Layer Perceptron with the two hidden layers containing 640 and 320 neurons.

**Figure 15.**Mean values of received signal strengths (RSSs) at four anchor location when the transmitter is the moving tag in a two-dimensional position grid. All values (see legends) are in units dBm. For each location, the mean of all received measurement samples is displayed, Channel 37.

**Figure 16.**Scatter plot diagram of the real and predicted distance along the X (

**left**) and Y (

**right**) axes for the kNN classifier applied to two-dimensional localization. The colour of the points corresponds to normalized count of samples at a given position of the plot.

**Figure 17.**Scatter plot diagram of the real and predicted distance along the X (

**left**) and Y (

**right**) axes for the SVM classifier applied to two-dimensional localization.

**Figure 18.**Scatter plot diagram of the real and predicted distance along the X (

**left**) and Y (

**right**) axes for the Random Forest classifier applied to two-dimensional localization.

**Figure 19.**Scatter plot diagram of the real and predicted distance along the X (

**left**) and Y (

**right**) axes for the Multi-Layer Perceptron classifier applied to two-dimensional localization.

**Table 1.**Summary of literature considering Machine Learning for localization algorithm improvement and data fusion in the context of RSS localization.

Classifier or Regessor | RSS Fingerprint. Improvement | Fusion of RSS and Other Information |
---|---|---|

Support Vector Machine | [21,26] | [28] |

Random Forest | [24] | [28] |

Logistic Regression | [21] | |

k Nearest Neighbors | [26,29,32] | [22,29] |

Sum of Squared Differences | [23] | |

Particle Markov Model | [29] | |

Polynomial Regression | [30] | |

Naive Bayes | [24] | |

Artificial Neural Network | [25,27] |

Classifier | Test Set Accuracy | Max. Error | Mean Error |
---|---|---|---|

kNN (1 neighbor) | 0.995813 | 4.5 m | 0.00697 m |

SVM | 0.996046 | 5.0 m | 0.00837 m |

Rand. forest | 0.999534 | 0.5 m | 0.00023 m |

MLP | 0.970697 | 22.0 m | 0.04162 m |

Classifier | Test Set Accuracy | Max. Error | Mean Error |
---|---|---|---|

kNN | 0.992174 | 10.51 m | 0.01872 m |

SVM | 0.991553 | 14.50 m | 0.02176 m |

Rand. forest | 0.999503 | 7.01 m | 0.00188 m |

MLP | 0.975965 | 29.54 m | 0.08150 m |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Polak, L.; Rozum, S.; Slanina, M.; Bravenec, T.; Fryza, T.; Pikrakis, A.
Received Signal Strength Fingerprinting-Based Indoor Location Estimation Employing Machine Learning. *Sensors* **2021**, *21*, 4605.
https://doi.org/10.3390/s21134605

**AMA Style**

Polak L, Rozum S, Slanina M, Bravenec T, Fryza T, Pikrakis A.
Received Signal Strength Fingerprinting-Based Indoor Location Estimation Employing Machine Learning. *Sensors*. 2021; 21(13):4605.
https://doi.org/10.3390/s21134605

**Chicago/Turabian Style**

Polak, Ladislav, Stanislav Rozum, Martin Slanina, Tomas Bravenec, Tomas Fryza, and Aggelos Pikrakis.
2021. "Received Signal Strength Fingerprinting-Based Indoor Location Estimation Employing Machine Learning" *Sensors* 21, no. 13: 4605.
https://doi.org/10.3390/s21134605