Machine Learning-Based Approaches to IoT Localization

Version	Summary	Created by	Modification	Content Size	Created at	Operation
1		Reza Shahbazian	--	2907	2023-04-19 23:07:03	\|
2	layout & references	Sirius Huang	Meta information modification	2907	2023-04-20 10:27:52	\|

This entry is adapted from the peer-reviewed paper 10.3390/s23073551

The widespread use of the Internet and the exponential growth in small hardware diversity enable the development of Internet of things (IoT)-based localization systems. Because of their high prediction accuracy, machine learning methods are being used to solve localization problems.

machine learning localization Internet of things fingerprinting Industry 4.0

1. Introduction

The term Internet of things (IoT) is used to indicate a network of physical objects, including machinery, furniture, and construction materials, equipped with sensors, software, and connectivity to gather and exchange data ^[1]. The Internet of things is important because it allows individuals and companies to better comprehend and exert control over their environment, enhancing efficiency, safety, and convenience. The Fourth Industrial Revolution (Industry 4.0) ^[2] is the term used to describe the present trend of automation and data interchange in manufacturing technology, including the use of the IoT. By utilizing cutting-edge technologies such as artificial intelligence, machine learning, and the Internet of things, Industry 4.0 is crucial because it enables organizations to boost productivity, reduce costs, and improve goods and services. In general, the integration of the IoT and Industry 4.0 is promoting the growth of a society that is more interconnected, intelligent, and productive ^[3]^[4].

Localization is the process of pinpointing the precise location of an IoT device. The problem is known as IoT network localization if the goal is to localize all the devices. The utilization of numerous technologies, including GPS, wireless communication signals, and sensor data, can be used to do this ^[5]. In the context of IoT localization, machine learning can be used to analyze data from many sources ^[3] to enhance the precision and dependability of position predictions. For instance, trends in data gathered from IoT devices could be examined by machine learning algorithms in order to identify elements that influence the precision of the localization. This may involve the strength of the wireless signal, the presence of objects that obstruct or skew the signals, or the existence of other interference sources. By learning to recognize and account for these characteristics, machine learning algorithms can increase the precision of the localization.

Many IoT localization systems use learning algorithms, which allow the system to adapt and improve its performance over time. IoT localization can be categorized in several ways, such as based on measurement methods, range-based localization, and range-free localization, which can be carried out in both indoor and outdoor settings. Range-based IoT localization refers to methods that leverage an IoT device’s distance from one or more reference points to pinpoint its location ^[6]. They could be specialized equipment deployed particularly for localization, or they could be other devices fixed in recognized locations such as cell towers or WiFi routers. Range-based IoT localization methods include triangulation, trilateration, and multilateration. Techniques for locating an IoT device without using distance measurements are referred to as range-free IoT localization. Instead, these methods rely on other kinds of data, such as the intensity of the wireless signal, the existence of certain landmarks or other objects, or patterns in the data from the device ^[7].

The ideal IoT localization method relies on the particular needs of the application. Both range-based and range-free strategies have benefits and drawbacks. Although range-based methods are typically more precise, they could be limited by the lack of reference locations or the precision of distance measurements. Range-free methods may be less precise, but in some circumstances, they are more flexible and simpler to use.

Indoor IoT localization is the process of locating an IoT device inside a structure or another enclosed place. Because the gadget could not have access to GPS or other external signals for localization, this can be a challenging process. Instead, indoor localization systems usually rely on other kinds of information, like the intensity of the wireless signal, the existence of particular landmarks or other objects, or patterns in the data that the device collects. Outdoor IoT localization is the procedure used to pinpoint an IoT device’s position in an outdoor setting. Because the gadget can commonly use GPS or other exterior signals for localization, this is typically simpler than indoor localization. On the other hand, obstructions or other sources of interference may pose problems for outdoor localization systems, affecting the precision of the position estimates ^[8]^[9].

In IoT localization, a variety of wireless radiofrequency techniques could be used. Wireless technologies are critical in an IoT ecosystem because they enable device-to-device communication and data transfer. WiFi, for example, communicates between devices using radio waves, and signal strength can be measured using the received signal strength indicator (RSSI) and channel state information (CSI), which can be used for indoor localization. Bluetooth Low Energy (BLE) is a low-power wireless technology that is used for short-range communication between devices and is also used in IoT localization via beacons. RFID uses radio waves to transfer data between a reader and a tag attached to an object, allowing it to be identified and located. Ultrawideband (UWB) technology transmits data over a wide bandwidth using high-frequency radio pulses, providing fast and accurate location information. Ultrasonic technology, which uses high-frequency sound waves to measure distance and detect objects, is also widely used in IoT localization applications. Some of the most common technologies referred in the covered publications are as follows:

The received signal strength (RSS) is a measurement of the strength of a wireless signal at a specific location. It is commonly used in IoT localization systems to determine the distance between a device and a reference point, such as a WiFi router or cell tower. By measuring the RSS at multiple reference points, the device’s location can be triangulated. The RSS is typically expressed in decibels relative to one milliwatt. It is a measurement of the strength of a wireless signal at a specific location that can be used to determine the distance between the device and the reference point. The stronger the signal, the shorter the distance. The relationship between the RSS and distance, on the other hand, is not always straightforward, as it can be influenced by a variety of factors such as the presence of obstacles, interference from other signals, and antenna characteristics ^[10]^[11].
The channel state information (CSI) is a measure of the characteristics of a wireless signal, such as its phase and amplitude at different frequencies. It is frequently used in WiFi localization systems to improve the accuracy of location estimates. The CSI can be used to collect data about the environment in which the signal is being transmitted, such as the presence of reflections or specific objects. The CSI is typically expressed in complex numbers that represent the phase and amplitude of the signal at each frequency. The distance between the device and the reference point, the presence of obstacles or other sources of interference, and the antenna characteristics of the device can all influence the signal’s phase and amplitude ^[12].
Bluetooth Low Energy (BLE) is a wireless communication technology popular in IoT devices. Because of its low power consumption and short range, it is well-suited for use in location-based services. BLE can be used in both range-based and range-free localization systems, depending on how it is implemented ^[13].
Other radiofrequency techniques used in IoT localization include radiofrequency identification (RFID) and ultrawideband (UWB). RFID employs passive tags that are attached to objects and read by a reader, whereas UWB employs radio pulses that are extremely brief in duration and can be used for high-precision localization ^[14].

There are numerous machine learning algorithms that can be used in IoT localization, some of the common algorithms are presented in the following:

k-Nearest neighbors (KNN): A simple but effective machine learning algorithm that can be used for both classification and regression. In the context of IoT localization, KNN could be used to predict the location of an IoT device based on the locations of other nearby devices.
Decision trees: A machine learning algorithm that produces a treelike model of decisions and their potential outcomes. It can forecast an IoT device’s location based on the values of various features, such as the wireless signal strength or the presence of specific landmarks.
Support vector machine (SVM): SVMs are powerful machine learning algorithms that can be used for classification, regression, and a variety of other tasks. In the context of IoT localization, SVMs could be used to predict the location of an IoT device based on patterns in data collected from it.
Neural networks are a type of machine learning algorithm that is inspired by the structure and function of the human brain. Neural networks can perform a variety of tasks, including classification, regression, and pattern recognition. In the context of IoT localization, neural networks could be used to predict the location of an IoT device based on patterns in data collected from it.
Deep learning (DL): DL is a type of machine learning that makes use of deep neural networks with multiple layers of processing. It is capable of detecting complex patterns in data and predicting outcomes based on those patterns. In the context of IoT localization, deep learning could be used to predict the location of an IoT device based on patterns in data collected from it. Deep learning is a subset of machine learning that makes use of deep neural networks with multiple processing layers. Deep learning can be used in IoT localization to analyze data from various sources, such as GPS, wireless communication signals, and sensor data, in order to improve the accuracy and reliability of location estimates. The ability of deep learning to handle large and complex datasets is a significant advantage for IoT localization. Deep learning algorithms can learn to see patterns and relationships in data that humans cannot, allowing them to make more accurate predictions. Deep learning algorithms can also learn and improve their performance as they are exposed to more data.
Reinforcement learning is a subset of machine learning in which an agent learns how to interact with its surroundings in order to maximize a reward signal. Reinforcement learning could be used in the context of IoT localization to optimize the behavior of an IoT device in order to improve the accuracy of its location estimates.

Figure 1 depicts a high-level overview of the IoT localization classification.

Figure 1. A high-level overview of the IoT localization classification.

2. Applications of Learning in IoT Localization

This section discusses the use of machine learning (ML) algorithms in IoT localization. There are some well-known IoT localization applications. Asset tracking is one of the most common applications of IoT localization. For example, a company may use IoT devices to track the movement and location of its vehicle fleet or commodity inventory ^[15]. This can help the company improve the effectiveness of its field service operations as well as streamline its supply chain and logistics procedures. Another industry that benefits from IoT localization is public safety ^[16]. Emergency responders, for example, can use IoT devices to track and locate first responders in real time to better coordinate their efforts and ensure everyone’s safety ^[17]. The localization of IoT devices is also applicable in smart city and smart home settings ^[18]. A smart home system, for example, could use IoT localization to monitor occupants’ movements throughout the house and make necessary adjustments to lighting, temperature, and other settings. The movement of residents and visitors can also be tracked using IoT localization, allowing a smart city to maximize resource utilization and improve the overall quality of life for its residents.

Learning algorithms can be used in several ways to improve the accuracy and reliability of IoT localization systems. Some of these applications are summarized as follows:

Calibration: Learning algorithms can be used to calibrate sensors and other IoT device components, ensuring that they function correctly and provide accurate data. This can help to improve the overall accuracy of the location estimates ^[19].
Noise reduction: Machine learning algorithms can be used to remove noise and other sources of error from data from IoT devices. This can aid in improving the accuracy of location estimates by reducing the impact of errors and other factors that can distort the data. Machine learning algorithms can be used to identify the most relevant features in data collected from IoT devices, thereby improving location estimation accuracy by focusing on the most important factors ^[20].
Model selection: machine learning algorithms can be used to identify the best model or combination of models for a given application, improving location estimation accuracy by selecting the best model that fits the data.
Improving accuracy: large quantities of data can be analyzed by machine learning algorithms to provide more accurate estimates of device location.
Automating the localization process: machine learning has the potential to automate the process of determining device location, removing the need for manual input.
Adapting to changes in the environment: to provide more accurate location estimates, machine learning algorithms can adapt to changes in the environment, such as new obstacles or changes in signal strength.

3. Challenges of Learning-Based IoT Localization

Several challenges must be overcome before learning algorithms for IoT localization can be used effectively. One challenge is the need for a large quantity of high-quality data to train the algorithms. Another issue is the computational complexity of deep learning algorithms, which can necessitate a significant investment in training and implementation. Finally, there are concerns about the interpretability of deep learning models because it may be difficult to understand how they arrived at their predictions. An IoT device, for example, could be trained to move in a specific way to improve the quality of the data it collects (for instance, by following a given path or turning in a specific direction). Data collection that yields precise location estimates may result in rewards for the device, whereas inaccurate estimates may result in penalties. The device would gradually learn how to change its behavior in order to maximize its rewards. One potential issue with using reinforcement learning for IoT localization is the need for a well-defined incentive signal that accurately reflects the location prediction accuracy. Furthermore, it may be necessary to carefully balance the demands of gathering high-quality data and protecting device resources (such as battery life).

Data quality: The quality of data collected from IoT devices heavily influences the accuracy and dependability of location estimates. If the data are noisy, incomplete, or corrupted, the learning algorithms may struggle to locate the devices accurately.
Computational complexity: Deep learning algorithms, for example, can require significant computational resources to train and deploy. Because IoT devices may have limited processing power and storage capacity, this can be difficult.
Limited data: In some cases, the data collected by IoT devices may be limited in quantity or quality, making it difficult to train accurate learning models. This is especially challenging in applications where devices are deployed in unusual or rare environments, or where data are highly variable.
Concerns about privacy: The use of learning algorithms in IoT applications can raise concerns about the privacy of device data. The need for accurate location estimates must be carefully balanced in some cases with the need to protect users’ privacy.
Security risks: Because learning algorithms are vulnerable to hacking and other forms of tampering, their use in IoT applications can pose security risks. Appropriate security measures must be implemented to protect the data and the integrity of the learning algorithms.
Model selection: Choosing the best machine learning algorithm for a particular localization task can be difficult. Different algorithms have different strengths and weaknesses, making it difficult to select the best one for the job.
Model interpretability: Deep learning neural networks, for example, are notoriously difficult to interpret. This can make it difficult to understand why a particular prediction was made and to improve the model.

There are numerous challenges, but there are also numerous solutions. To improve the quality of the training data, techniques such as data cleaning and normalization can be used. These methods are commonly known as preprocessing techniques. For example, model compression can be used to reduce the computational resources needed to run machine learning algorithms. Different machine learning algorithms can be compared and evaluated using metrics such as accuracy and the computational resources required to choose the best machine learning algorithm for a specific task. To improve interpretability, techniques such as feature importance analysis and model visualization can be used. Transfer learning techniques can also be used to leverage pretrained models on large datasets, allowing knowledge to be transferred to smaller datasets. Federated learning can also be used to train models while maintaining data privacy because the data are kept on the IoT devices and only model parameters are exchanged. Furthermore, data augmentation techniques such as random cropping, rotation, or translation can be used to generate synthetic data to address the limited quantity of data. Unsupervised learning techniques such as clustering can be used in some cases to identify patterns in data where labels are not available. Edge computing can also be used to reduce the quantity of data that must be transmitted to the cloud, saving network bandwidth and reducing the computational burden on the cloud. Encryption and secure communication protocols can be used to protect device data and learning algorithms, addressing privacy and security concerns. Machine learning techniques that preserve privacy, such as differential privacy, can be used to provide a mathematically rigorous way of ensuring the privacy of device data. To summarize, we can overcome these challenges and develop accurate and reliable machine learning models for IoT localization by carefully considering the limitations of machine learning and IoT devices and leveraging preprocessing techniques, transfer learning, federated learning, data augmentation, unsupervised learning, and privacy-preserving techniques.

References

Patel, K.K.; Patel, S.M.; Scholar, P. Internet of things-IOT: Definition, characteristics, architecture, enabling technologies, application & future challenges. Int. J. Eng. Sci. Comput. 2016, 6, 6122–6131.
Devezas, T.; Sarygulov, A. Industry 4.0; Springer: Berlin/Heidelberg, Germany, 2017.
Bécue, A.; Praça, I.; Gama, J. Artificial intelligence, cyber-threats and Industry 4.0: Challenges and opportunities. Artif. Intell. Rev. 2021, 54, 3849–3886.
Mirtaheri, S.L.; Shahbazian, R. Machine Learning: Theory to Applications; CRC Press: Boca Raton, FL, USA, 2022.
Shit, R.C.; Sharma, S.; Puthal, D.; Zomaya, A.Y. Location of Things (LoT): A Review and Taxonomy of Sensors Localization in IoT Infrastructure. IEEE Commun. Surv. Tutor. 2018, 20, 2028–2061.
Alanezi, M.A.; Bouchekara, H.R.; Javaid, M.S. Range-Based Localization of a Wireless Sensor Network for Internet of Things Using Received Signal Strength Indicator and the Most Valuable Player Algorithm. Technologies 2021, 9, 42.
Barshandeh, S.; Masdari, M.; Dhiman, G.; Hosseini, V.; Singh, K.K. A range-free localization algorithm for IoT networks. Int. J. Intell. Syst. 2021, 37, 10336–10379.
Nabati, M.; Ghorashi, S.A.; Shahbazian, R. Joint coordinate optimization in fingerprint-based indoor positioning. IEEE Commun. Lett. 2020, 25, 1192–1195.
Asaad, S.M.; Maghdid, H.S. A Comprehensive Review of Indoor/Outdoor Localization Solutions in IoT era: Research Challenges and Future Perspectives. Comput. Netw. 2022, 212, 109041.
Tomic, S.; Beko, M.; Dinis, R. RSS-AoA-Based Target Localization and Tracking in Wireless Sensor Networks; CRC Press: Boca Raton, FL, USA, 2022.
Shahbazian, R.; Guerriero, F. Optimized distributed large-scale analytics over decentralized data sources with imperfect communication. J. Supercomput. 2020, 76, 8510–8528.
Fard Moshiri, P.; Shahbazian, R.; Nabati, M.; Ghorashi, S.A. A CSI-based human activity recognition using deep learning. Sensors 2021, 21, 7225.
Nouali, O.; Moussaoui, S.; Derder, A. A BLE-based data collection system for IoT. In Proceedings of the 2015 First International Conference on New Technologies of Information and Communication (NTIC), Mila, Algeria, 8–9 November 2015; IEEE: Piscataway, NJ, USA, 2015; pp. 1–5.
Mohanty, S.; Tripathy, A.; Das, B. An overview of a low energy UWB localization in IoT based system. In Proceedings of the 2021 International Symposium of Asian Control Association on Intelligent Robotics and Industrial Automation (IRIA), Goa, India, 20–22 September 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 293–296.
Wang, S.; Hou, Y.; Gao, F.; Ji, X. A novel IoT access architecture for vehicle monitoring system. In Proceedings of the 2016 IEEE 3rd World Forum on Internet of Things (WF-IoT), Reston, VA, USA, 12–14 December 2016; IEEE: Piscataway, NJ, USA, 2016; pp. 639–642.
Papadakis, N.; Koukoulas, N.; Christakis, I.; Stavrakas, I.; Kandris, D. An IoT-based participatory antitheft system for public safety enhancement in smart cities. Smart Cities 2021, 4, 919–937.
Rodriguez-Sanchez, M.C.; Fernández-Jiménez, L.; Jiménez, A.R.; Vaquero, J.; Borromeo, S.; Lázaro-Galilea, J.L. Helpresponder—System for the security of first responder interventions. Sensors 2021, 21, 2614.
Mpeis, P.; Roussel, T.; Kumar, M.; Costa, C.; LaoudiasDenis, C.; Capot-Ray, D.; Zeinalipour-Yazti, D. The anyplace 4.0 iot localization architecture. In Proceedings of the 2020 21st IEEE International Conference on Mobile Data Management (MDM), Versailles, France, 30 June–3 July 2020; IEEE: Piscataway, NJ, USA, 2020; pp. 218–225.
Zheng, L.; Hu, B.J.; Qiu, J.; Cui, M. A deep-learning-based self-calibration time-reversal fingerprinting localization approach on Wi-Fi platform. IEEE Internet Things J. 2020, 7, 7072–7083.
Li, Y.; Zhuang, Y.; Hu, X.; Gao, Z.; Hu, J.; Chen, L.; He, Z.; Pei, L.; Chen, K.; Wang, M.; et al. Toward location-enabled IoT (LE-IoT): IoT positioning techniques, error sources, and error mitigation. IEEE Internet Things J. 2020, 8, 4035–4062.

© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.

Upload a video for this entry

Information

Subjects: Computer Science, Artificial Intelligence

Contributors MDPI registered users' name will be linked to their SciProfiles pages. To register with us, please refer to https://encyclopedia.pub/register :

Reza Shahbazian

Giusy Macrina

Edoardo Scalzo

Francesca Guerriero

View Times: 321

Update Date: 20 Apr 2023

Table of Contents

Video Upload Options

Confirm

1. Introduction

2. Applications of Learning in IoT Localization

3. Challenges of Learning-Based IoT Localization

References