Overview of Indoor Positioning System Technologies

Transcription

1 Overview of Indoor Positioning System Technologies Luka Batistić *, Mladen Tomić * * University of Rijeka, Faculty of Engineering/Department of Computer Engineering, Rijeka, Croatia lbatistic@riteh.hr; mtomic@riteh.hr Abstract - Constant changes in the field of indoor positioning systems (IPS) dictate that we keep up with new and developing trends and technologies. With improving accuracy, IPS found widespread application in commercial environments, such as asset or personnel tracking. However, it is yet to be established in the everyday personal usage applications. In this paper, we present a technological overview and classification of the most relevant IPS technologies. We compare technologies based on their cost, accuracy, performance and complexity. Hybrid approaches are also investigated as a way to overcome limitations of specific IPS technologies. The paper is mostly focused on IPS technologies that can be used in personal applications, using pocket-sized devices (e.g. smartphone). Keywords Indoor positioning systems; Location based services; Navigation I. INTRODUCTION Outdoor location based services, such as global navigation satellite systems (GNSS), are widely used in everyday life. In recent years, location awareness in indoor environments became a popular topic. Whether the application is asset tracking, location detection of personnel or simply finding a way around an unfamiliar building, there is a great need for indoor positioning systems (IPS). Unlike outdoor positioning systems where 1-3 m accuracies can be achieved using satellites, IPS cannot effectively use GNSS for location detection. The main reason is signal degradation that occurs when radio signals used by GNSS hit obstacles (such as walls and roofs). This results in signal attenuation and scatter, so existing infrastructure like Global Positioning System (GPS) cannot provide sufficient accuracy [1]. In the last two decades, a great effort has been put into improvement of indoor location sensing. This resulted in commercial availability of several IPSs as well as a number of systems currently being tested and improved. In this overview, we will focus on most relevant IPS technologies and their corresponding measuring principles, methods and algorithms, that can be implemented on pocket-size devices (e.g. smartphone). We will compare the technologies based on their cost, accuracy, performance (latency, advantages and disadvantages) and complexity (installation and usage). Since IPS technologies often have drawbacks, mostly in accuracy, we will also investigate combinations of different technologies (hybrid technologies), which help to overcome disadvantages of a single IPS. II. MEASURING PRINCIPLES AND METHODS Location detection can be achieved using different techniques and algorithms, which provide data in the form of signal strength or range. From the acquired data, a position estimation of an object can be calculated. In this section, we will explain measuring principles, methods and algorithms that are used in the technologies described in Section III. Note that there are other techniques which are out of scope of this overview, and are, therefore, not explained. A. Triangulation One of the most common techniques for location detection is triangulation. Triangulation uses geometric properties of triangles to determine distance or orientation. It can be done in two ways: lateration or angulation. Time of arrival (TOA), time difference of arrival (TDOA) and angle of arrival (AOA) are most common techniques of triangulation, but received signal strength (RSS) and return time of flight (RTOF) can sometimes also be used. With all triangulation techniques, line of sight (LOS) between transmitters and receivers is important. If LOS is not established, signal attenuation, scatter and multipath can have a great impact on accuracy. 1) Lateration TOA uses the time t, that a signal takes to travel from a transmitter to a receiver, in order to calculate the distance between the two [2]. After obtaining the time, the distance is calculated using a simple linear motion equation: where c is the speed of light. d = c t, (1) In 2-D spaces this leads to the following circle equation: d = (x ref x) 2 + (y ref y) 2, (2) where (x ref, y ref) are coordinates of a known reference point and (x, y) are coordinates of a target. Having measurements from several devices, it is possible to calculate the position of a target. Say we have a target T and three reference points R i, as shown in Fig. 1. If 2-D coordinates of each reference point R i are known, and the distance between the target T and each of the reference points R i is obtained, the 2-D position of the target T can be calculated. At least three reference points are needed to 522 MIPRO 2018/CTI

2 determine 2-D location unambiguously. When using TOA, it is a requirement that the internal clock of each device is synchronized with other devices and that timestamps of transmissions are embedded in the signal being transmitted. TDOA is similar to TOA, since they both use the time a signal takes to travel a distance between a transmitter and a receiver. TDOA is a technique used to determine the difference in time at which the same signal arrives at multiple receivers [3]. This is in contrast to measuring the absolute time that a signal takes to travel between a single transmitter and a single receiver in TOA. Once the signal has been received at multiple reference points, a difference in distances between two reference points Δd can be calculated by multiplying the speed of light c and a difference in the time of arrival - Δt: Δd = c Δt. (3) In 2-D spaces, this leads to the following hyperbolic equation: Δd = (x i x) 2 (y i y) 2 (x j x) 2 (y j y) 2, (4) where (x i, y i) and (x j, y j) are known coordinates of two reference points I and J, and (x, y) are coordinates of the target T. Once at least two hyperbolas are obtained, the target s location can be estimated, and when at least three hyperbolas are obtained, the target can be pin-pointed. Since absolute time of arrival is not used in this calculation, there is also no need for the time of transmission to be known, hence, transmitter s clock does not need to be synchronized with the receiver s clock (receiver s clock still needs to be synchronized to other receivers). 2) Angulation Figure 1. TOA measurements AOA technique, sometimes called direction of arrival (DOA), determines the target s location by estimating the intersection of directions (angles) at which signals arrive at receiving sensors. Once the intersection is estimated, a distance between the target and the reference points can be calculated. Only two angles are needed to determine a 2-D location. Technologies implementing AOA usually use directional antennae. B. Fingerprinting One of the most widespread and successful techniques in indoor positioning is radio fingerprinting. It is based on a scene analysis mapped area is analyzed and signal properties in different locations are collected and stored to a database. When a target demands position estimation, signal properties of the target s position are compared to the database and the closest match is returned as a position estimate. A received signal strength (RSS) is commonly collected during scene analysis. There are different algorithms that can be used for fingerprinting, such as neural networks, k-nearest-neighbors, support vector machine or probabilistic methods, as discussed in [4]. For this paper, the k-nearest-neighbors (knn) is particularly interesting. The knn uses RSS from multiple reference points, collected at the target s position, and determines the k closest reference points by comparing collected values to the database values. While it is advantageous that fingerprinting based technologies can often use existing infrastructure (such as WLAN access points), often there is a need for upgrading that infrastructure to achieve satisfactory accuracy. A disadvantage of RSS, and fingerprinting in general, is susceptibility to errors caused by radio signal scatter, multipath and attenuation. C. Signal propagation modeling Signal propagation modeling, also called radio propagation modeling, is a technique that provides an alternative to the empirical method of fingerprinting. With this technique, mathematical signal propagation models are made, and theoretically-computed signal strength data set is created. User s location can then be estimated by matching the signal strength measured in real-time with the theoretically-computed signal strengths. Most common models are Ricean and Rayleig fading models. Ricean fading model is used in cases where dominant multipath component can be determined, and Rayleig fading model is used in cases where dominant multipath component cannot be determined. D. Dead reckoning Dead reckoning is one of the oldest positioning techniques. It relies upon estimation of the current position based on a previously known position and estimated movement distance and direction (heading). Distance and direction are usually estimated using inertial sensor (gyroscope, accelerometer, and compass) measurements. Dead reckoning techniques are rarely used as the only technique in an IPS, since small errors in direction estimation can result in larger error in position estimation. To improve accuracy and reduce errors, particle filtering (PF) was introduced [5]. PF uses impassable obstacles (e.g. walls and furniture) to eliminate positions in which a target could not have possibly been located. PF is a numerical approximation to a Bayesian filter [6]. Bayesian filter uses probability distribution for the location estimate, and PF represents the probability (Bel(x t)) using a set of weighted samples ({x t i,w t i }): Bel(x t ) = {x t i, w t i }, i = 1 n, (5) MIPRO 2018/CTI 523

3 where (x ti ) is a discrete hypothesis of object s location, and (w t i ) is a weight representing importance factor which sums up to one. In each iteration, PF updates each weighted sample using motion model, weighs all samples by sensor s likelihood model (for the current measurement), and resamples (using importance factor). E. Proximity Proximity location technique provides information on the location of a target with respect to a known position in an area. It doesn t necessarily provide absolute location of an object, but a relative approximate of the location. When the target is detected at a reference point, it is considered to be in proximity of that reference point. The Approximation accuracy of the target s location depends on number and sensitivity of detectors. F. Cooperative positioning Cooperative positioning is an approach which allows users of a positioning system to be connected in a peer-topeer (P2P) network where they can share their position information and receive position information from their peers. Using information acquired from peers in their vicinity, users can estimate, or improve accuracy and reliability of their position. Social-Loc is a middleware that can be installed to improve accuracy of a system by using cooperative positioning. It has been implemented and tested with Wi-Fi fingerprinting and dead reckoning systems (separately) [9], both on Android smartphone. Authors report improved accuracy by at least 22% (for Wi-Fi) and 37% (for dead reckoning). System works in such a way that users, upon encountering each other, scan their respective positions (via RFID), compare them and do the corrections if necessary. III. INDOOR POSITIONING TECHNOLOGIES A. Wireless local area network Wireless local area network (WLAN) technology is used for wireless exchange of data in computer networks, in a limited distance of m from access points. WLANs are based on IEEE standards (Wi-Fi brand name). They operate in 2.4 or 5 GHz frequency bands and can achieve throughput in excess of 1Gbit/s. Indoor positioning systems relying on WLAN technology use multiple WLAN access points to determine the target s position in an area. In order to improve accuracy, additional access points must often be added to the existing network. Most common techniques for IPS implemented with WLANs are fingerprinting and signal propagation modeling. One of the first WLAN indoor positioning systems was RADAR [8], which used two techniques to determine location: fingerprinting, i.e. measurement of signal strength for each access point, and signal propagation modeling. To improve the system s accuracy, wall attenuation and floor attenuation factors are used (in contrast to Rayleigh fading model and Rician distribution model which are used in outdoor positioning systems). RADAR managed to achieve accuracy of around 2-3 m and could be used by multiple tracked devices at the same time. Ekahau positioning system [9] is one of the most popular commercial systems that uses existing WLAN infrastructure (APs). The system consists of APs, tags and mapping and calibration software (site survey). Tags are worn by users or attached to objects and they transmit Wi- Fi signals periodically, when the tracked object moves, or when the specific button is pressed. When the signal is detected on different APs, received signal strength indicator (RSSI) values are used to triangulate target s 2-D location. The reported accuracy is 1-3 m. During mapping and calibration, the site survey software detects network coverage area, overlapping of WLAN APs, signal strength and SNR. Ekahau system is a low-cost system and it is simple to use with multiple tags. B. Radio-frequency identification Radio-frequency identification (RFID) is a technology that uses radio waves to store and retrieve data between a reader and a tag. RFID readers are able to read data emitted from RFID tags. RFID is a cheap and flexible wireless technology that already found widespread usage in identification of objects (inventory, personnel, animals etc.), but it can also be used to determine position of an object in a mapped area. There are two types of RFID technologies. One is the passive RFID where the tracked tag is a receiver. The working range of passive RFID systems is short (1-2 m). They work in several frequency bands (usually LF: khz, HF and NFC: MHz, UHF: MHz) and their price is low. The tag modulates and reflects a signal back to the transmitter and does not require a power source for its operation. The other RFID type is active RFID, where tracked device is a transmitter. Active RFID systems can cover longer distances ( m) and usually operate in the 433 MHz or 915 MHz frequency bands. Their price is higher than the price of passive RFIDs. The device acts as a beacon and periodically transmits signals containing ID, or some other data, to readers. RFID usually uses proximity techniques to detect weather a target is in range of a reference point. LANDMARC (Indoor Location Sensing Using Active RFID) [10] was a pioneering system that used RFID to determine location indoors. The system used active RFID tags, attached to the tracked objects, and multiple readers placed in fixed locations. Except the tags for tracked objects, the system also introduced reference tags in order to improve calibration accuracy. Signal strength from each tag was measured at readers and knn method was used to calculate the location of target s RFID. In a 40 m 2 room, with 4 receivers and 16 reference tags, accuracy of the system was 1 m for 50th percentile and maximum error distances were less than 2 m. Battery life of a tag was 3-5 years. Downside of the system was that it took a long time to estimate target s location (7.5 s in a system of 500 tags) because additional processing of signal was needed to calculate signal strength - RFID receivers only reported if a tag is detectable or not detectable and did not give any information on signal strength. C. Ultrasonic Ultrasonic positioning systems use sound frequencies above the upper audible limit of human hearing (usually around 40 khz). Such systems determine target s position by measuring the time needed for a signal to travel from a 524 MIPRO 2018/CTI

4 Figure 2. Active Bat Tag transmitter to receivers. Audible sound IPSs work on similar principles but ultrasonic systems are more convenient since they are not intrusive to people who are using (or are surrounded by) the system. A disadvantage of the technology is that new infrastructure, in form of sensors and transmitters, is needed in every room where the system is used. Active Bat positioning system [11] gives 3D position of the tracked tag, Fig. 2. Tag broadcasts ultrasonic signals which are received by receivers (microphones) mounted on a room ceiling. A position is calculated using triangulation technique by measuring signal s TOA. To calculate the 3D position, at least 3 receivers need to receive target s signal. In an area of 1000 m 2 with 750 receivers, accuracy is 3 cm for 95th percentile and each tag can be located 50 times per second. Tags have battery life of about 15 months. Lok8 [12] is an indoor positioning system for smartphones that uses off-the-shelf smartphone speaker to produce ultrasound signals (around 22 khz). The system uses TDOA, hence no need for synchronization of clocks on system devices is needed, making it even more simple to use. It was tested in 49 m 2 room with 4 receivers (one in each corner of the room) and accuracy is 10 cm on average. D. Bluetooth Bluetooth technology is used to transfer data over short distances (10-100m, with Bluetooth 2.0). It is defined in IEEE as a form of a Wireless personal area network (WPAN) and works in 2.4 GHz frequency band with data transfer rates of 1-3 Mbit/s. The main purpose of the Bluetooth technology is to create a simple wireless communication that doesn t require LOS. Because of Bluetooth s low price and complexity, today, this technology is present in most smartphones and wireless computer peripherals. Topaz local positioning system [13] utilizes Bluetooth technology to locate targets indoors. The initial system could only provide 2-D location, with accuracy around 2 m for 95th percentile, but later it was enhanced with usage of IR technology to improve accuracy. Locating delay is s. System consists of a positioning server, wireless access points placed every 2-15 m, and tags worn by targets. ibeacon by Apple [14] is a technology standard that uses Bluetooth Low Energy (BLE) for proximity measuring. One of ibeacon s purposes is to enable users to locate products in stores by telling them proximity to the wanted item or a shelf. The system can also be used for indoor positioning by distributing beacons on predetermined fixed points in an area and then triangulating position with usage of multiple proximity measurements. Beacons transmit signals, while an application installed on portable devices determine proximity of each beacon: immediate (less than 50cm), near (between 50 cm and 3 m) and far (3-30 m). ibeacon applications are available on both Android and ios. E. Inertial Inertial sensors, such as accelerometers and gyroscopes, in combination with other sensors (e.g. compass) can be used to determine a target s speed and direction. When speed and direction are obtained, having knowledge of initial position, we can estimate the future position dead reckoning technique. An advantage of this technology is that most modern smartphones have all the sensors necessary for dead reckoning. A disadvantage is the accumulation of errors in cases where small deviations in calculation of direction can cause large errors in estimation of position as long distances are traveled. To reduce the error accumulation, a lot of dead reckoning based IPSs also use particle filtering (PF) algorithms. Inertial technology IPSs are usually used in combination with some other technologies (i.e. WLAN, Bluetooth, RF). Beauregard et al. [15] proposed a pedestrian dead reckoning (PDR) system with inertial technology and backtracking particle filter (BPF). BPF introduces improvement in comparison to regular PF in sense that it enables deletion of previous impossible particles and trajectories when target meets an impassable barrier, which can be seen in Fig. 3. Beauregard et al. report accuracy of 2.5 m when only an external map of walls is available and 0.74 m when a detailed map of walls is available. In their experiments, a motion sensor was attached to one shoe of a subject who walked in a 60m by 60m multi-story building. Advantage of this system is that it can be used to record movements and trajectories of a target even if the map is not available at the moment of recording. This feature is very useful for first responders (e.g. firefighters) as their movement can be recorded during intervention at an unmapped location, and the map can be added afterwards. Figure 3. BPF example [15] MIPRO 2018/CTI 525

5 Table I. Han Zou et al. experiment accuracies [19] Figure 4. Ubisense Tags ( F. Ultra-wideband Ultra-wideband (UWB) is a technology where electromagnetic waves are emitted in short pulses (less than 1 ns) and in a wide frequency band (>500 MHz), which makes it possible to filter reflected signals from the original signal, hence minimizing the multipath problem and improving the accuracy of TOA. Unlike other RF based technologies, UWB transmits signals in multiple frequency bands (from 3.1 to 10.6 GHz) simultaneously. Besides high accuracy, low power requirements made this technology commercially successful in environments with pervasive usage of IPSs (e.g. warehouses). A disadvantage is a costly equipment and installation. Ubisense company made UWB indoor positioning system [16], which consists of sensors distributed across a mapped area and tags (Fig 4.) attached to the objects being tracked. The tags have both UWB transmitter (for location) and radio transceiver for communication with network. The system uses TDOA and AOA of the UWB signal to calculate target s location. Tag s signal needs to be received by two sensors in order to calculate a 3D location of a tag. The accuracy is about 15 cm. Disadvantages of the system are that it is costly and, when used to locate people, users location is known to the system at any time. G. Hybrid technologies Recently, most promising results in indoor positioning were achieved with hybrid IPSs, which use multiple technologies to overcome the disadvantages of any individual technology. With many combinations possible, in this article we are going to cover only some promising combinations that can be implemented in pocket-sized and affordable devices. Leppäkoski et al. [17] proposed a hybrid system with inertial sensors, particle filtering and WLAN RSS. Inertial sensors composed of a gyro sensor and a three-axis accelerometer attached to users back, along with WLAN signal strengths, were scanned every 2.3 s with a handset WLAN device, during a 17-minute walk in a 1000 m 2 area. They reported accuracy of 3 m in distance (and 12 in heading), in contrast to inertial only PDR system where accuracy was 5 m (and 24 in heading). The advantage of the system is that it is cheap and it can be implemented with a smartphone. Cricket [18] is an ultrasound and RF indoor positioning system which works similar to the Active Bat system, but in this case, emitters are mounted to the fixed points, and a receiver is carried by the target. Cricket also uses TOA and triangulation technique to locate a target. Emitters transmit RF messages that are used to synchronize TOA measurements and give out approximate location (of the emitter) in cases when there is not enough emitters for triangulation. Reported accuracy is around 10 cm. Advantages of the Cricket system is that the receiver being carried by the target provides more privacy to the user (target) and the system is low cost ($10 per receiver). On the other hand, since Cricket uses both ultrasound and RF, it is power consuming. Han Zou et al. [19] built a smartphone system that utilized inertial sensors, WiFi signal strength and ibeacon (Bluetooth technology). They made tests and compared accuracies of different technologies and their combinations. Results are shown in Table I. The tests were performed in a 600 m 2 multifunctional office using Google Nexus 6 smartphone. During the tests with ibeacons, six beacons were installed in the same office room. The advantage of the system is that it works on a smartphone. IV. Approach TECHNOLOGIES COMPARISON Comparison of technologies we described can be seen in Table II. In the Cost column, we describe the cost of installation (which includes equipment and labor, if required) and the cost of each user unit. The performance is described in columns Advantages and Disadvantages. The Complexity describes total complexity of installation and usage of the system. V. CONCLUSION Dead Reckoning (DR) WiFi fingerprinting DR + WiFi DR + WiFi + ibeacon Mean Error m m m m In this paper, we provide an overview and compare the most relevant indoor positioning technologies that had been developed over the recent years and can be incorporated in pocket-size devices. We describe the underlying technologies and techniques, and give leading examples in each one of them. Each IPS architecture and working principle is discussed, investigating their advantages and disadvantages, complexity, accuracy, and cost. For deployments, most interesting parameters are accuracy and complexity. Sub-meter accuracies had been achieved in multiple systems using ultrasound, ultra-wideband, Bluetooth and hybrid technologies, but not all of them are fit for everyday usage by an average person since most of them are too expensive and too complex to be implemented in most indoor areas. We show that each system has its limitations and strengths. Some of the described systems could be 526 MIPRO 2018/CTI

6 Table II Comparison of IPS technologies IPS Technology Accuracy Cost (installation/ unit) Advantages Disadvantages Complexity RADAR WiFi 2-3 m L/L Low price, existing Low accuracy, complex system Medium infrastructure Ekahau WiFi 1-3 m H/L Existing Expensive mapping software Low infrastructure, good mapping software LANDMARC RFID 2 m H/L Very cheap user Locating delay 7.5 s Medium units Active Bat Ultrasound 3 cm H/L Cheap user units, Requires a lot of beacons, High very precise medium battery life Lok8 Ultrasound 10 cm L/L Smartphone user Requires new infrastructure in Medium units, precise every room Topaz Bluetooth 2 m L/L Low price Locating delay s Medium ibeacon Bluetooth m H/L Smartphone user units, ease of access Beauregard et al. Inertial m L/L Cheap, map can be added post-hoc Ubisense Ultrawideband 15 cm H/H Very precise, very robust Leppäkoski et al. Inertial +WiFi 3 m L/L Cheap, could work on smartphone Cricket Ultrasound + 10 cm H/L Cheap user units, RF provides privacy, Han Zou et al. Inertial +WiFi +Bluetooth Requires a lot of beacons for better precision Requires a detailed map for better precision Expensive installation and units Low Medium Low accuracy, time consuming Medium installation complex installation, low High battery life precise 0.59 m L/L Cheap, precise Requires detailed mapping Medium High combined into hybrid systems, to improve accuracy and overcome their limitations. Smartphones are also a promising land for hybrid systems, since modern smartphones already include most of the necessary sensors that can be used for location sensing. REFERENCES [1] G. Lachapelle, GNSS Indoor Location Technologies, GNSS 2004, The 2004 International Symposium on GNSS/GPS, [2] D. Munoz, F.B. Lara, C. Vargas, and R. Enriquez-Caldera, Position Location Techniques and Applications, 1st edition, Academic Press, [3] R. Roberts, TDOA localization techniques, Harris Corporation, IEEE P Working Group for Wireless Personal Area Networks (WPANs), [4] H. Liu, Survey of Wireless Indoor Positioning Techniques and Systems, IEEE Transactions on Systems, Man, and Cybernetics Part C: Applications and Reviews, vol. 37, no. 6, [5] F. Gustafsson, F. Gunnarsson, N.Begman, U. Forssell, J. Jansson, R. Karlsson, and PJ. Nordlund, Particle filters for positioning, navigation, and tracking, IEEE Transactions on Signal Processing, vol. 50, issue 2, Pages , [6] J. Hightower, G. Borriello, Particle Filters for Location Estimation in Ubiquitous Computing: A Case Study, International Conference on Ubiquitous Computing UbiComp 2004: Ubiquitous Computing pp , [7] J. Jun et al., Social-Loc: Improving indoor localization with social sensing, Proceedings of the 11th ACM Conference on Embeded Networked Sensor Systems, [8] P. Bahl, and V.N. Padmanabhan, Radar: An in-building RF-based user location and tracking system, Ninteenth Annual Joint Conference of the IEEE Computer and Communications Societies, [9] Z. Li Z. Deng, W. Liu, and L. Xu, A Novel Three-Dimensional Indoor Localization Algorithm Based on Multi-Sensors, China Satellite Navigation Conference, [10] L. M. Ni, Y. Liu, Y. C. Lau, and P. Patil, LANDMARC: Indoor location sensing using active RFID, Wireless Networks, vol. 10, no. 6, pp , [11] A. Ward, A. Jones, and A. Hopper, A New Location Technique for The Active Office, IEEE Personal Communications, vol. 4, issue 5, [12] V. Filonenko, C. Cullen, and J. Carswell, Indoor Positioning for Smartphones Using Asynchronous Ultrasound Trilateration, ISPRS International Journal of Geo-Information, vol. 2, pages , [13] Topaz, 2018, [14] ibeacon, 2018, [15] S. Beauregard, Widyawan, and M. Klepal, Indoor PDR Performance Enhancement using Minimal Map Information and Particle Filters, Position, Location and Navigation Symposium, [16] Ubisense, 2018, [17] H. Leppäkoski, J. Collin, and J. Takala, Pedestrian Navigation Based on Inertial Sensors, Indoor Map, And WLAN Signals, IEEE Inernational Conference on Acoustics, Speech and Signal Processing (ICASSP), [18] N. B. Priyantha, A. Chakraborty, and H. Balakrishnan, The Cricket Location-Support System, 6th ACM International Conference on Mobile Computing and Networking, Boston, MA, [19] H. Zou, Z. Chen, H. Jiang, L. Xie, and C. Spanos, Accurate Indoor Localization and Tracking Using Mobile Phone Inertial Sensors, WiFi and ibeacon, IEEE International Symposium on Inertial Sensors and Systems, MIPRO 2018/CTI 527