In this study, the concepts of positioning and location sensing are explored with a focus on RFID technology as applied in urban environments. RFID positioning is a simple and cheap way to identify the location of an object or a person, indoors and outdoors, alone or in combination with different positioning methods, such as satellite systems. The implementations are numerous and an asset to location-based services and GIS.
Radio Frequency Identification, generally abbreviated as RFID, finds its origin in military applications decades ago, when backscatter radiation was used to identify hostile aircrafts during World War II. The technology has evolved since then, enabling more complicated implementations, such as supply chain monitoring, product labeling, and item tracking. Currently deployed RFID systems are generally focused on the real-time tracing, tracking and monitoring, with numerous implementations in logistics, inventory management, supply chain and access control, to name a few. The key features of the technology have helped to find its place in industrial and urban settings: being contactless, no line of sight communication, and automation that does not require any human intervention (Goshey, 2008). RFID uses wireless radio frequency (RF) communication technology to interlink remote devices, usually with a unique ID (UID), and transfer data among them for further processing.
The following four components are found in a basic RFID system:
RFID tags are found in the following three types: passive, active and semi-active. This categorization refers to the energy requirements and the method of energy consumption for the tag. Active tags are powered entirely by a battery, while passive tags are powered exclusively by the electromagnetic waves emitted by the reader. The third category of semi-active tags includes transponders that use an internal power source (battery) only for their internal functions like storing data to memory and microprocessing; the communication between tag and reader is done by consuming energy provided by the reader similarly to the entirely passive tags. Passive tags can be very small and cheap, however the maximum read distance is shorter than the active type. On the other hard, active tags can transmit to the reader from a longer distance, but the battery has a limited life (Shikada, Shiraishi, & Takeuchi, 2012).
The passive communication method generally involves three main phases of operation, as described below:
Tags and readers communicate with each other mainly by either inductive coupling or by electromagnetic backscatter coupling (electromagnetic waves). Inductive coupling is used for RFID systems operating at frequency bands LF and HF (Table 1), while electromagnetic backscatter is used for higher frequencies.
The frequencies that RFID systems use to communicate range from the Low Frequency (LF) part of the spectrum (<135 kHz) to microwaves (2.45–5.8 GHz). Water and other nonconductive substances significantly absorb frequencies around 1 GHz, so this frequency range is not used (Finkenzeller, 2010). Several properties are dependent upon the operation frequency, such as the read range, the penetration of the signal and the data transfer rate. Table 1 gives an overview of different frequency band characteristics.
|Band||Frequency range||Applications||Read range|
|micro-wave||2,4 GHz – 5,8 GHz||Toll roads, train wagon positioning||up to 100 m|
|VHF, UHF||30 MHz – 3 GHz||Luggage management, pallet-level tracking in logistics||2 – 5 m|
|HF||3 MHz – 30 MHz||Libraries, smart cards, item-level tracking||up to 1 m|
|LF||30 kHz – 300 kHz||Farm animal tagging, car immobilizers||~10 cm|
Table 1: Common RFID frequency bands and application examples
Higher operation frequencies are generally related to higher data transfer rates. Applications of tracking objects moving at a high speed, such as a speeding car or a train wagon, require relatively high data rates, otherwise the tag will not be at a reading distance from the transceiver long enough. However, data rate is not the only parameter affecting the maximum tracking speed, since there are other factors such as the necessary charging time for the passive transmission methods. In practice, the communication can take nine times longer that the theoretical limit of the data transfer according to (Chon, Jun, Jung, & An, 2004), who showed that a reader moving at a speed of 165 km/h can read 128 bits from a road tag as long as the read range is greater or equal to 81 cm.
Of particular importance to a GIS-based implementation of an RFID system are the characteristics regarding memory and data transfer of the communication path tag-reader. RFID tags come in a variety of memory sizes, from 16 bytes to 64 kB, using ferroelectric random access memory (FRAM) (Fujitsu, 2014).
In the context of the present project and geography in general, positioning refers to the act of a localization system and its relation to a location-based service (LBS), which is a product of combination of contemporary informatics, the Internet, and geography (Figueiras & Frattasi, 2010). An LBS is an informatics service that serves its user in a manner relevant to their location. A number of definitions have been given to the LBS, all relating to a, chiefly mobile, user and their interaction with technologies of informatics.
Examples of LBS are (Kolodziej & Hjelm, 2006):
In all of the examples above, the key input for the LBS is of course the location of the users, which is determined by various means. Of particular importance is the meaning of the term “location”, which generally refers to a physical place but it can also be associated with non-tangible concepts. Physical locations, relevant to LBS implementations, are subcategorized by (Küpper, 2005) as follows:
A distinction is made between physical and virtual locations, the latter being locations in a virtual system, like a video game, a chat room, an instant messaging (IM) application (Küpper, 2005). However, the aforementioned examples generally still point to a specific network location: a website is stored on a server with a specific IP address, translated by a domain name system (DNS) to a human-friendly Uniform Resource Locator (URL).
The term refers to cell identification. In this case, the position is identified according to the location of the cell the object of interest is ‘connected’ to. This approach is generally used in combination with others, for example in an RFID positioning system, the reader’s ID number would be the Cell ID whereas in a WiFi network, the MAC address of the wireless module would be the Cell ID. In theory, this method is used for object tracking with an RFID system in symbolic space, where no coordinate system is used per se, but tracking takes place in the form of true-false logical functions, for example “if the tag $T$ is in the area of coverage of reader $R$ at time $t$, then the function $F$ is true, otherwise (else) false” (Kang, Kim, & Li, 2010)
Radio signals propagate in space but due to fading caused by destructive propagation effects, such as signal attenuation, shadowing, scattering and diffraction, they lose strength. The loss of power is a function of distance between the transmitter and the receiver, it is therefore possible to calculate how far the tracked object is from the transmitter based on the Received Signal Strength (RSS), calculated in dB, or the Received Signal Strength Index (RSSI). However, signal-strength-based positioning can be very inaccurate compared to the techniques below because it is very difficult to be calibrated and because the signal strength is affected by many factors, distance being only one of them.
The principle of this method is the fact that electromagnetic radiation propagates in space at a finite rate, equal to the speed of light c. Very accurate clocks must be used, which set timestamps when a signal is received at the tracked object. By knowing the propagation speed and the time it took the signal to reach the destination (receiver), it is possible to calculate the distance. In this method, however, the concept of clock granularity is delved, which generates the need for special mechanisms to correct errors due to the non-contiguous function of the clock, and therefore the timestamp operation, which runs with clock ticks at specific frequency.
The position of the tracked object is estimated by triangulation the measurements of distance through the technique of lateration. In a two-dimensional space, three measurements of distance are required, while in three dimensions, four measurements are required. The location of the receiver is at the point where the three (2D) or four (3D) circles intersect, as illustrated in Image 1.
The angle at which the signal leaves the transmitter and reaches the receiver is the key to this method. However, such a system requires high cost directional antennas and is generally more complicated than the Time of Arrival method. The measurements of this method are used in the process of angulation, a kind of triangulation that calculates the position of the tracked object based on the angles of at least two signals transmitted from two different sources.
The principle of this method is based on the difference in the phase of the signal emitted by the reader and the signal backscattered by the tag that is read, which is related to the distance of the backscatter device. This distance estimation technique is often used in combination with other techniques, such as AOA. It is possible to design a system with multiple frequency pairs that will be able to estimate the location with a higher accuracy (Zhang, Li, & Amin, 2010).
The term positioning refers to a technique that can determine and share the location of living and inanimate objects continuously and in real-time. According to (Esri, 2006), positioning can be either static (determining a position on the earth by averaging the readings taken by a stationary antenna over a period of time) or kinematic (determining the position of an antenna on a moving object).
An LBS can deliver services and exchange data among static and moving users. Advanced location sensing mechanisms need to be deployed, as indexing of moving objects that use techniques to “exploit the volatility of the data values being indexed” (Jensen, Lin, & Ooi, 2008), which is a characteristic of objects that move into and out of the area covered by the LBS. In the example of a $B^x$–Tree indexing of moving objects, a positioning system needs to be able to capture a position vector $x ⃗$, a velocity vector $y ⃗$ and a specific time value when these inputs are valid $t_u$.
The most known positioning technique is the one used by satellite navigation systems, where the satellite positions are known and the time the signal needs to propagate to a target object is measured. Assuming that the electromagnetic waves travel at the speed of light, the distance between the satellite and the target object is calculated by multiplying propagation time by speed of light. In reality, however, electromagnetic radiation experiences an atmospheric delay, which is defined as the reduction of their propagation speed when they pass through the ionosphere and the troposphere.
The technologies dominating outdoor positioning are mainly based on satellite systems, such as the Global Navigation Satellite Systems (GNSS). Such systems have now reached maturity and have a number of advantages: high precision, continuity, able to function regardless of the weather conditions, near-real-time observation, and increased reliability. With regard to their advantages, GNSS are widely used in applications other than positioning and navigation as well, such as remote sensing (Shuanggen, Estel, & Feiqin, 2013). The most widely used satellite systems for global navigation and positioning are:
Global Positioning System (GPS) finds it roots in the early 1960s, when a number of U.S. governmental organizations joined forces to develop a system that would provide navigation and positioning services primarily for military and secondarily for civilian use. At present, the American positioning system consists of a constellation of 24 operational satellites at an orbital radius of approximately 26 600 km, and has been used not only for positioning and navigation, but also for delivering precise timing.
Former USSR had developed their own satellite positioning system, known as Global Navigation Satellite System (GLONASS). The constellation of its satellites was completed in 1995, four years after the collapse of the Soviet Union in 1991. The project’s operation was suspended for the rest of the 1990s but is now operational with 24 satellites (Russian Federal Space Agency, 2016).
The European Space Agency’s (ESA) ongoing GNSS project Galileo, named after the Italian astronomer Galileo Galilei, is a €5 billion project intended for civilian and commercial use. The system will become fully operational until 2020, but the initial services will be made available in late 2016. In full deployment, the system will consist of 24 operational satellites at 23 222 km altitude; it will provide basic, relatively low-accuracy services to everyone and advanced, high-precision services to paying customers (European Space Agency, 2015).
The fourth main GNSS is the Chinese BeiDou/COMPASS, scheduled to deploy by 2020. The system’s constellation will have 35 satellites, five of which in geostationary orbit, and will be offering basic services for civilian use and higher accuracy services for special uses, similarly to the systems mentioned earlier (BeiDou Navigation Satellite System, 2015).
Satellite technologies have well met the needs for determining the location in outdoor environments. However, they are not well suited for indoor areas because of the poor reception of satellite signals, which are greatly degraded inside of buildings. The task of positioning in such environments has been addressed by several techniques based on wireless technologies. Examples of such systems include sonars (audio waves), radio signal triangulation and beacons (electromagnetic waves) (Fernandes, Filipea, Costa, & Barroso, 2014), as well as infrared and physical contact methods (Youssef, 2008).
Indoor positioning faces significant challenges that place it in a less widespread position, when compared to the extensive use of ‘conventional’ GNSSs. These challenges are (Gubi, et al., 2010):
Indoor positioning implementations can be found in office buildings, warehouses, factories. Examples of indoor location sensing technologies include infrared positioning, indoor GPS-based systems, Ultra Wide Band (UWB), Wireless Local Area Network (WLAN), and of course RFID positioning. The dominant RF-based methods, however, are RFID and WLAN. In an indoor environment, localization of an object is possible through tracking in two or three dimensions. If tracking takes place in a 2-dimension space but on multiple planes, like tracking of people in a building of many floors, then the term 2.5-dimension can be used. The processing of the input data and the determination of the location can be implemented on a mobile device on the object or person being traced (client device), or on a central unit (server), the location of which can be, theoretically, anywhere in the world. The position can be reported either symbolically or in coordinates, which in turn can be absolute (e.g. longitude, altitude etc.) or relative (e.g. distance from a nearby point of reference).
Hybrid positioning systems usually combine satellite location technology with ground-based systems in order to either increase the accuracy served by the satellite system or to provide location information for those areas where satellites cannot reach, such as indoors. The assisting technologies usually are mobile phone cell tower signals (GSM, LTE), WiFi, WiMAX, Bluetooth and others (AlterGeo, 2015).
RFID systems can be used to obtain indoors location information in such a way, that when combined with GNSS can create a system of seamless positioning. It is not mandatory for areas not covered by GNSS to be indoor areas; outdoors obstacles such as canopies can negatively affect satellite signal quality as well (Shikada, Shiraishi, & Takeuchi, 2012). An example of such a seamless system output is shown in Image 4. The light green points are obtained from the passive RFID system, while the red points from GPS. In (Shikada, Shiraishi, & Takeuchi, 2012), where this image was taken from, it shows a problem of overlapping data for the positions where both the RFID system and the GPS system feed the system with location data.
Hybrid positioning techniques can also be combining RFID and another technology (Bai, Wu, Wu, & Zhang, 2012), not necessarily aimed at indoor implementations only (Wen, 2010). Some examples of non-GNSS technologies that have been used to enhance the capabilities of RFID-based systems include ZigBee (a protocol used for Wireless Sensor Networks – WSN), Wi-Fi/WLAN, ultra wide band (UWB), infrared (IR), and ultrasonic. Such combinations can provide the user with increased accuracy and reliability.
There are generally two kinds of implementing RFID-based location sensing: fixed tags–moving reader, which implies that the object or person whose position is to be estimated carries a reader, and fixed reader(s)–moving tags, where the setup is the opposite. In addition, that there are systems where readers and reference tags are stationary and a non-fixed tag is being tracked.
A two dimensional space can be interpreted as a Cartesian coordinate system, where each point that belongs to the said plane can be specified by a pair of numbers, usually (x, y). The implementation described below mainly consists of a grid of passive RFID tags equally distributed over the area where the positioning takes place and a portable RFID reader which is carried by the person/object being tracked. A system like the one illustrated below can achieve positioning accuracy of 50 cm if tag placement is 50 cm, the antenna is linearly polarized and the the RFID transmission power is 18dBm (Shiraishi, Komuro, Ueda, Kasai, & Tsuboi, 2008).
Each one of the installed tags has a unique ID number and its spatial location is known with a high accuracy. The spatial location and ID data are stored in a database; therefore, the memory requirement for the tags is low, as it only stores the ID number. The reader moves along the plane of tags at a distance, along the z axis.
As illustrated in image 5, the reader moves over the $(x,y)$ plane at a distance $d$. The interrogation field of the reader reaches the tags on the floor. As it moves, several tags are read. Experimental applications have verified that the closer the reader to the tag plane is, the better the location estimation is. In addition, the reader detects significantly fewer tags if it is placed closer to the plane than the wavelength of the RFID system. For example, for tags operating at 950 MHZ, the wavelength, and therefore the minimum recommended distance from the tag plane, is about 0,30 m (Shiraishi, Komuro, Ueda, Kasai, & Tsuboi, 2008).
After receiving the tag IDs that have been detected, the position of the reader is estimated computationally. The easiest approach is to calculate the center of gravity of all the tags detected, but this method might not be the most appropriate because the reader might detect tags that are way off. Thus, a clustering approach can be more suitable. This method can locate clusters of tags and disregard outliers. For this purpose, GIS software that can perform spatial statistics operations can be used. In particular, the Hot Spot Analysis (Getis-Ord Gi*) can locate the clustering (Esri). There should be no reason to use global statistics first, which shall identify whether or not clustering exists, since it is taken for granted that there is a cluster of detected tags.
The reader in an implementation of a fixed grid of passive RFID tags can be small enough to be portable by blind people. An example is a prototype called SmartVision (Fernandes, Filipea, Costa, & Barroso, 2014), aimed at providing location-based services for the blind. The prototype is based on the same principle as the example above. There is a base layer of RFID tags which marks the area the people can move and several “layers” of points of interest (POI). Of course, the tags used still only store their ID information; whether a specific tag belongs to layers with POIs is associated in the GIS database. This way, people can navigate (or be informed about their position) on many levels, e.g. on a position marked as ‘safe’ (not wet, for example, since the database can be updated in real-time) and marked as ‘room 5’.
It is possible, however, to use RFID-based positioning system as above but for linear location sensing; along a path, for example. Active RFID systems operating at high frequencies (microwave) are already being used for electronic toll collection on motorways. Though they can be used (and are used) for tracking of vehicles and road traffic monitoring, they do not provide ‘position’ information in the sense of exact location of a vehicle on the road. Contrary to indoor positioning, the main challenge here is the moving speed of the tracked object (the car) and the selection of an RFID system with transponders able to communicate whilst moving at high speeds. In addition, the reader and the middleware installed on the car must be able to handle high data rate, given that many tags are read in a short time frame. Even if the tags only communicate their ID codes, the data adds up. In the experimentation setting of (Chon, Jun, Jung, & An, 2004), the reader moving at 165 km/h needs 81 cm of travel distance to read 128 bits of data from a stationary tag, therefore, its read range has to be at least 81 cm.
The implementation opposite of §3.1 is to set readers to a fixed and known position and estimate the position of RFID tags within their interrogation field. The techniques used for this purpose are the ones presented in §1.3.
Implementing an RFID-based system for positioning, especially indoor positioning, can have some advantages over competitive systems. RFID generally is a simple, flexible, portable and low-cost system that can provide identification, in a supply chain for example, and location information at the same time. However, the communication is one-way when dealing with passive tags and undesirable multipath effects are observed (Bai, Wu, Wu, & Zhang, 2012).
While the positioning techniques mentioned above are mainly focused on either 2D or 2.5D, the possibilities offered by a 3D positioning system are significant, especially in work sites. An example of such a location sensing system is by (Ko, 2013) and provides users with an alternative to the RFID-based symbolic tracking methods, such as the ones in a supply chain.