Recent Advancements in Indoor Positioning and Localization

1. Introduction

The current era celebrates the rise of mobile devices, most of which are mobile phones. Penetrating into different fields of life, the explosive growth of mobile phones led to the inception of several innovative applications that use the current position of the phone to meet users’ needs. Such applications such as location-based services (LBSs), mobile applications, etc., rely heavily on accurate position information of the mobile user, both outdoors and indoors. The global positioning system (GPS) has been regarded as, de facto technology, a suitable technology accurate enough to provide the position of a mobile device in the outdoor environment to meet the standards for such applications. For the indoor environment, on the other hand, GPS faces several challenges, and its accuracy is affected by phenomenona such as signal loss, multipath, shadowing, etc., and is not suitable for indoor positioning. To overcome such issues and provide high accuracy indoor positioning and localization, recent years have seen the introduction of many new technologies.

Recently, wireless local area networks (WLANs), Bluetooth, magnetic-field-based positioning systems, and pedestrian dead reckoning have been investigated as potential candidates for indoor positioning. However, such technology has shortcomings in providing the desired levels of accuracy for indoor positioning and localization. Some of these are affected by their inherent limitations, while others are limited by real-time changes in the environment. To meet accuracy requirements, advanced algorithms, custom models, hybrid solutions, and novel fusion frameworks are needed that can provide unparalleled accuracy and prove their efficacy for indoor positioning and localization. In this regard, the role of different sensors for data collection, and machine learning techniques are potentially important. Additionally, the launch of 5G has opened up new dimensions for cellular-based indoor positioning, where small cells can be utilized to obtain fine time resolution, which helps to obtain higher positioning accuracy. The current Special Issue was designed to analyze novel and recent sensors, techniques, models, and frameworks that can help to realize the demand for accurate indoor positioning and localization.

2. Current Special Issue

This Special Issue published five papers that focus on the application of sensors, the use of machine learning and deep learning models, and an overview of different techniques and technologies for indoor positioning and localization. Of the published papers, one paper discusses the use of acoustic signals from multiple microphone arrays to perform indoor positioning. One paper deals with the visible light-based channel modeling for underground coal mine areas. One paper focuses on energy optimization and designs an energy-saving strategy. One paper presents a novel indoor positioning algorithm based on an improved K nearest neighbor model. Finally, the last paper elaborates on the use of smartphone sensors, their applications, and the pros and cons of indoor positioning.

Table 1. Summary of the works published in the Special Issue.

Reference	Application	Sensor	Approach
[1]	Positioning	Acoustic	Time difference of arrival
[2]	Channel modeling	Visible light	Bimodal Gaussian distribution in line of sight and non-line of sight
[3]	Energy optimization	-	Macro model for energy optimization
[4]	Positioning	Wireless	Improved K nearest neighbor
[5]	Positioning	Smartphone sensors	Analytical review

provides the details of these papers, while the discussions are given in the following paragraphs.

The study in [1] utilizes sound signals for indoor localization, which is one of the most widely experimented with and deployed positioning and localization technologies. Based on the direction of the sound and the traveling time of the sound, it can provide highly accurate positioning. The focus of the paper is to use two microphone linear arrays for locating the sound source. Each linear array is equipped with two microphones. Microphone arrays are configured to receive channels in the right and left terminals. In addition, the setup contains an audio amplifier, TDoA, and voice detection. On reception, the A/D converter converts the digital signal. TDoA works when sound is received and angles are calculated for left and right channels. Delay in the sound reception is resolved by using the time difference of arrival (TDoA). Location is determined by generalized cross-correlations and trigonometric functions. Experiments are performed using the European Telecommunications Standards Institute (ETSI) channel environment and carried out using different frequency ranges, including 1 K, 5 K, 10 K, 12 K, and 15 K, with a fixed signal-to-noise ratio of 35. The results reveal that the x–y distance error is high at lower frequencies, and the same is true for angle estimation, which yields a higher error when the frequency is low. These errors are more stable when the operating frequency is 10 K. With higher frequency, calculated angles are more stable, leading to higher positioning accuracy.

The authors investigate the characteristics of visible light channels for the underground hazardous environment of coal mines in [2]. Communicating with and locating the people working in the high-risk environment of coal mines is very important, and traditional wired communication channels are susceptible to damage, fading, and complete failures in case of gaseous explosions, and other potential alternatives should be explored to locate the workers. It requires an in-depth analysis and investigation of channel models with respect to the influence of coal dust particles, scattering, and complex non-line-of-sight (NLOS) scenarios. In addition, the shadowing effect of a large number of pillars in the coal area, the impact of mining machinery, and attenuation from dust clouds must be analyzed for visible light communication. The study employs bimodal Gaussian distribution for the coal area, which is divided into galleries and sub-galleries. The experimental setup includes the use of an array of light-emitting diodes (LEDs) on ceilings within a 4 m distance, and a Lambertian radiation pattern is considered for LEDs. The results indicate that coal dust concentration has an inverse relationship with the visibility, which is reduced to 2 m when suspended particles are 10 mg/m³ or higher. Dust attenuation also increases with coal dust intensity. The size of the coal particle is reported to have a much lesser impact on the scattering coefficient. However, a high change is reported for extinction coefficient with particle size exceeding 500 nm. Regarding the channel impulse response, the arrival time is high for higher-order reflections. Sub-galleries have a higher delay due to larger dimensions. The path loss is exponential for regions close to the transmitting source; however, it becomes linear while moving away. The study reports that the difference between the NLOS and line-of-sight (LOS) link is high. The reflected surface absorbs, as well as scatters, the light, thus affecting the signal power. Higher delay spread is reported for the mine gallery environment as compared to the mine sub-gallery environment. Similarly, the probability of error is increased with an increase in the distance between the source and target.

The study [3] focuses on the energy optimization for 3D printing due to its wide and diverse applications. The authors devise a macro model for energy optimization and compare its performance with commercial printers. The experimental results show that the power has a direct relationship with the printing time. Higher printing speed is assumed to reduce the printing time; however, the study finds that the relationship between the speed and time is not linear and that the melting point should also be incorporated. Similarly, the angle of rotation for printing is reported to have a small influence on energy. The study reports that the major element for higher energy is the heating process, which needs to be optimized for energy optimization.

Given the importance of K nearest neighbor (KNN) for indoor positioning and localization, ref. [4] proposes a modified weighted KNN (WKNN) approach to improve the positioning accuracy. Predominantly, the Euclidean distance, Manhattan distance, etc., are adopted for reference point (RP) selection, where received signal strength (RSS) is considered to estimate the distance. However, due to the exponential relationship between the RSS and distance, RPs selection may be inaccurate. The current study considers the spatial distance and physical distance for the proposed WKNN approach and thus refines the position estimation using two distances, where the former has a linear relationship while the latter has an exponential relationship with the RSS value. The performance of the proposed approach is measured using mean position error, root mean squared error, standard deviation, and position estimation error in comparison to KNN, Euclidean-distance-based WKNN, and physical-distance-based WKNN. The results indicate superior results as compared to state-of-the-art WKNN, and the proposed WKNN approach offers higher positioning accuracy.

Finally, a state-of-the-art review is provided in [5] regarding the use of microelectromechanical systems (MEMS) sensors embedded in smartphones. With a brief introduction of each sensor, its working mechanism, and its use for indoor positioning and localization, a discussion of different approaches is presented. The use of a smartphone Wi-Fi sensor is discussed regarding positioning based on fingerprinting and multilateration. The multilateration, or trilateration, is further discussed within the context of the time of arrival, TDoA, and angle of arrival. Given the wide use of fingerprinting approaches, several types of fingerprinting approaches are elaborated, such as RSS, hyperbolic location fingerprinting, ordered sequence of receive signal strength indicator, etc. The study also points out the challenges of Wi-Fi-based fingerprinting and asserts that such approaches are limited by data collection labor, dependence on indoor infrastructure, the influence of mobility, and changes in the indoor settings, all of which necessitate new fingerprint collection. In addition, noise in data due to sensors’ quality and diversity of hardware poses real challenges for Wi-Fi fingerprint solutions. Pedestrian dead reckoning (PDR) is discussed as a solution to provide short-term relevant positions for indoor environments by using multiple smartphone sensors such as accelerometers, gyroscopes, magnetometers, etc. PDR is further categorized into inertial navigation systems and step and heading systems. The challenges of PDR are investigated within the context of smartphone sensors, where the error is accumulated over time and re-calibration is required to reduce the threshold of errors.

The study also discusses the magnetometer sensor of the smartphone as a separate solution for indoor positioning and localization where the magnetic anomalies of indoor buildings are used as the fingerprint. Often caused by ferromagnetic materials, such anomalies are reported to be more stable than Wi-Fi fingerprints. The study provides a brief, yet precise, account of indoor positioning approaches which are based on magnetic field data and include fingerprinting, pattern matching, data-intensive solutions, and multi-sensor fusion techniques, where the data from Wi-Fi, PDR, and Bluetooth low energy (BLE) is combined for increased accuracy. The use of a smartphone camera is discussed regarding positioning as a stand-alone unit, as well as complementing technology for the magnetic field, BLE, and Wi-Fi-based solutions. In addition, the use of LEDs for indoor positioning is discussed. The study points out that the use of the camera for positioning is expensive regarding battery consumption and image processing tasks that require a dedicated server for processing feature extraction, training, and image matching. BLE is discussed as a potential alternative to Wi-Fi both as a separate positioning module and complementary technology for Wi-Fi, PDR, and magnetic-field-based positioning. With short communication distance, it offers higher accuracy than Wi-Fi; however, it requires installing BLE beacons where a higher number of beacons are needed to obtain higher accuracy. The use of a lux meter is beneficial when the positioning environment contains indoor to outdoor transitions, or vice versa, or has semi-outdoor or semi-indoor environments. Similarly, the use of a barometer is elaborated within the context of multi-floor buildings where the change in atmospheric pressure can be measured by the barometer and used to detect the change in the floor level of the user. The study thus provides a comprehensive review of the indoor approaches that utilize one or more sensors embedded in modern smartphones.

3. Conclusions

The papers published in this Special Issue show several novel trends and can be pivotal to determining future trends

First, the studies emphasize different aspects of indoor positioning and localization approaches which show that the accuracy does not solely depend on the positioning algorithm. For example, the quality of the channel model, selection of appropriate RP selection for indoor positioning, the computational complexity of the positioning algorithm, and the appropriate energy optimization approach are all important.

Second, given the complex indoor environments of modern multi-floor buildings which may contain small openings (outdoor-like environments) or uneven surfaces, using only one technology such as Wi-Fi or PDR is not enough to obtain an accurate position. Instead, the use of multiple complementary technologies is reported to obtain higher accuracy. The fusion of such technologies is a matter of sensor data availability, desired accuracy, available computation power, and the complexity of indoor infrastructure.

Third, the wide use of deep learning and machine learning models is reported; however, this requires the need for dedicated servers and a communication link. Carrying out computations on the smartphone is only possible for sensors such as Wi-Fi, gyroscope, magnetometer, BLE, etc., yet image-processing approaches still need servers. Similarly, deploying machine learning and deep learning models is not possible even on modern smartphones. In addition, the use of crowdsourcing is reported for indoor positioning and localization, which is expected to ease the labor of data collection. With increased data sharing and connected smartphones through device-to-device (D2D) communication, better solutions can be designed to obtain highly accurate positioning.

Finally, with the reports of the inclusion of new sensors such as ultra-wideband and LiDAR in the upcoming models of iPhones and Samsung smartphones, novel solutions need to be devised. Another important aspect to consider is the privacy concerns that arise out of the higher level of D2D connectivity, as with the increased position sharing comes several threats.