Radio-based identification systems were originally designed to monitor and identify objects, animals and people based on the proximity of a transponder to a reading device. After simple modifications, the same system architecture can be regarded as a tool for estimating the position of objects or people bearing transponders, since they are radio transmitting devices, when they are found within the operation range of the system. The location of the agent is then determined relative to a reference location (Bouet & dos Santos, 2008). RFID localization systems are more commonly found indoors, in buildings where the quality of satellite positioning service is dramatically weakened. Examples of applications are in construction sites and storerooms.
RFID positioning implementations face problems similar to the problems of other radiowave-based positioning systems, such as multipath, shadowing, interference. Especially for signal-strength ranging techniques, these systems are even more prone to the effects of the environment, however, a strong correlation between the strength of the signal emitted by a base station and its distance does exist. This correlation can be exploited for location estimation (Locher, Wattenhofer, & Zollinger, 2005).
The methods used for RFID positioning make use of the design and specifications of the technology. For indoor location systems (but can also apply to outdoors) that use range-based distance measurements, the methods can be classified into two categories: received signal strength (RSS) methods and time-of-fly (ToF) methods. The methods of first category function under the law of signal decay by distance, while those of the second category are based on the signal propagation physics. Methods such as time-of-arrival (ToA), time-difference-of-arrival (TDoA), angle-of-arrival (AoA), and phase-of-arrival (PoA) are examples of ToF methods. ToF methods are also used in satellite positioning systems, such as GPS (López, Gómez, Álvarez, & Andrés, 2011).
Each one of the methods mentioned above comes with its advantages and disadvantages. The general opinion expressed in the relevant literature is that RSS methods are not as accurate and reliable as the ToF methods due to the fact that distance is only one of the numerous parameters that affect the RF signal strength. However, RSS methods are simple in their setup and do not require expensive equipment.
The RFID technology
Radio frequency identification (RFID) systems consist of transponders (tags) and readers that communicate using electromagnetic energy at the radio spectrum, hence their name. RFID systems can be classified according to their energy source into three types: passive, active and semi-active. In a passive system, the tag draws all the energy it needs to function wirelessly from the reader. On the other hand, in the active type the sole purpose of the reader is to communicate with the tag and not to provide power, which is provided by another source, usually a non-replaceable internal battery. A semi-active RFID system uses tags with an internal battery which solely powers the internal circuitry for the tags internal functions, while the energy provided by the reader is only used for the activation (wake up) of the tag and the communication (transmission of data).
The technology behind identification using radio waves is far from new, as it dates back to the beginning of the century and especially in the time before the Second World War, when a simple implementation called IFF, acronym for Identification, Friend or Foe, was used for identifying approaching aircrafts and avoiding friendly fire. A custom antenna was mounted on the friendly aircrafts; it was designed to answer to interrogation electromagnetic signals emitted by ground stations. Precursor to the modern passive RFID systems, however, is the “Thing”, aka the “Great Seal Bug”; a small eavesdropping device invented by Léon Theremin that was used by the Soviet intelligence agencies to spy on the Americans. The device had a simple microphone connected to an antenna. The device was powered wirelessly by a remote source in a manner similar to the modern passive RFID systems.
In the industry, RFID is widely regarded as a replacement for barcodes. When compared to previous Auto-ID technologies, RFID has significant benefits and some drawbacks. Amongst the main advantages is the ability to communicate and track items without physical contact and out of line of sight, the increased automation in reading/writing and the fast and accurate data transfers, the larger amount of programmable data they can store, and the flexibility and robustness of the entire system. When used in a positioning application, these advantages are of particular importance. However, some disadvantages do exist, the most important of which are the potential incompatibilities between devices coming from different manufacturers, the increased implementation costs and the cost of the equipment, and the fact that the propagation of radio waves is affected by the environment and even blocked by some materials such as water and metallic surfaces (Erande, 2008).
Standardization of RFID was proposed and has been developed by the International Organization for Standardization (ISO) and by the Auto-ID Center with their Electronic Product Code technologies.
RFID system architecture
In its simplest form, a fully functional RFID system consists of two main components: the transponder (tag) and the reader (a.k.a. RFID scanner) (Finkenzeller, 2003, p. 7). The transponder is a device generally simple and cheap that is attached to the object or person being tracked/identified; the reader is a larger, more complex and more expensive unit that can power, read, and write the transponder without physical contact and when the latter is placed within the reader’s read range. The image below shows a typical RFID setup used in logistics. This setup is the basis for a positioning system, where the distance between the antenna and the objects bearing the tags is to be estimated.
The transponders used in RFID systems typically consist of three sections: a radio-frequency front end, an analog section, and a digital section. Not all sections are present in all kinds of tags; for example, the very simple 1-bit tag usually found in antitheft systems does not have a digital section. The role of each section is outlined below.
RF front end:
- Energy harvesting from the electromagnetic field
- Demodulation of the received signals
- Transmission of the outgoing signal
- Clock of the digital subsystem
- Powers the rest subsystems/components of the tag chip
- Stabilizes the voltage of the front end
- ‘Power on reset signal’
- Manages the power
- Data recovery
- Executes the protocol operations
The Reader consists the first communication layer between the tag and the rest of the RFID system. They consist of subsystems that enable the transmission of energy and data to and from the tag and the forwarding of the data to the next communication layer in the system; these subsystems are typically a RF unit, an external (or internal/embedded) antenna, an electronic control unit, and a communication interface. Readers communicate with the middleware, or another control device, through the communication interface (e.g. RS 232, RS 485, USB, UART).
Readers of passive tags:
- High-power (up to 4W) RF transmission for the activation of the passive tags
- High power consumption, in the order of Watts
- Maximum distance of communication in the orders of centimeters or meters (up to 20 m)
- Reading capacity of ~100 tags in a few seconds
Readers of active tags:
- Low-power RF transmission (10–20 mW)
- Power consumption requirements significantly lower (~mW), which facilitates the design of portable readers
- Long reading range
- Reading capacity of hundreds of active tags in a few milliseconds
- Small antenna
The interface that connects the reader with the other units down the communication chain is referred to as “middleware”. Middleware units perform the following tasks:
- Filtering of the data incoming from the reader, so that the system is not overwhelmed
- Routing of the filtered data towards the proper software application
- Data logging
- Management of technical parameters, such as reading frequency, transmission energy levels, et al.
The most meaningful categorization of RFID systems is based on the power requirements of the tags. Three categories can be recognized” passive, active and battery assisted passive, aka “semi-passive”. Passive tags are powered entirely by the reader and constitute the simplest design of all three; active tags have an internal power source, usually a small, non-replaceable battery which provides energy not only for the transmission of the signal, but also for internal circuitry functions; battery-assisted tags embed a small battery which is used only for the internal functions of the tag, while communication with the reader requires passive-like energy harvesting. The capabilities of a tag are greatly affected by the energy source it uses. For example, the maximum reading distance of an active tag is significantly greater than the range of an entirely passive system, which can be from a few centimeters (NFC devices) and can reach 10–15 meters maximum for UHF systems. For comparison, the maximum reading distance of an active tag can exceed 100 m. In general, semi-passive tags are the middle ground between entirely active and entirely passive systems.
The number of RFID system types available in the market today is extremely large because of the different parameters and characteristics of the architecture of the technology. As with the personal computer market, where the buyer has to take into account the characteristics of a number of subsystems that identify a specific model like the CPU, the RAM the HDD etc., a number of ‘selection criteria’ (Finkenzeller, 2003, p. 25) are present in a RFID system. These criteria are:
- Operating frequency
The criteria are not entirely independent; systems that operate at higher frequencies tend to have a higher range. For example, microwave systems operating at frequencies in the vicinity of 2.4 GHz can typically achieve read distances of tens of meters, while systems in the LF bands have range of only a few centimeters. The architecture of the systems is comprised of a number of properties that together define the system itself.
Of significant importance to a RF-based positioning system is the maximum distance of communication between the transponders and the agent. The range of RFID systems is dependent on its design and spans from a few centimeters to over 100 m. According to (Finkenzeller, 2003, p. 26), the key factors of an application that will define the range are the ‘positional accuracy of the transponder’, the presence and the number of transponders
The operating frequency of a RFID system dictates may other parameters, such as the range and the environment they can be used in mainly due to different permeability. The most established bands in the industry are four, but more can be defined. The four bands are: LF, HF, UHF, SHG or microwave.
|SHF||30 GHz – 3 GHz||2,45 GHz; 5,8 GHz; 24,125 GHz|
|UHF||3 GHz – 300 MHz||433,920 MHz; 869 MHz; 915 MHz|
|VHF||300 MHz – 30 MHz||40,680 MHz|
|HF||30 MHz – 3 MHz||6,78 MHz; 13,56 MHz; 27,125 MHz|
|MF||3 MHz – 300 kHz|
|LF||300 kHz – 30 kHz||9–135 kHz|
|VLF||30 kHz – 3 kHz||9–135 kHz|
For an entirely passive RFID-based positioning system, the best choice seems to be a system operating in the UHF range, where maximum read distance can be up to a few meters. For battery- assisted or active systems, one may opt for a microwave system offering significantly higher range in the orders of tens of meters. Passive LF systems were amongst the first that were used and became widely popular in animal tagging applications and in high-accuracy timing systems used in sports. There systems are low-power, the reading range is a few centimeters and data rates are very low, usually under 8 kbps. LF requires large antennas because the coupling is inductive and reading range is within centimeters. Similarly, HF systems are found in book tracking in libraries and smart cards, and have a reading range of up to 1 m, which is insufficient for passive positioning.
The demand for small size and low-cost RFID tags has an impact on the size of the memory they can have. In general, cheap and low-capacity memories store identification data only, while advanced (and more expensive) “smart” tags can afford higher capacity memory and circuity. Typical memory technologies used in RFID tags are:
|Type||Volatility||Energy cons.||Size||Write cycles||Speed|
|EEPROM||Non-volatile||High||16 bytes – 8 kb||104 – 106||Slow|
|SRAM||Volatile||High||256 bytes – 64kB||∞||Fast|
Tag memory can be used for storing location data, such as coordinates, or semantic position data, such as cell ID. This can be useful in autonomous systems that do not have access to a spatial database from where to retrieve location data.
In sensitive RFID applications, for example where transaction of personal identification data takes place, security needs to be included.
Passive communication between reader and tag
Data transfer between readers and transponders in an RFID system presupposes the establishment of a communication channel between the two devices. Two methods are described: inductive coupling, which is used for systems operating at LF and HF bands, and modulated backscatter coupling for UHF and higher bands.
In resonant inductive coupling, the reader antenna has a coil which is powered by alternating current generated by an internal oscillator. As the electric current passes through the coil, it generates an alternating magnetic field that serves as a power source for the tag. The latter’s antenna coil is energized by the electromagnetic field which subsequently charges a nearby capacitor and activates the tag’s integrated circuit. Data transfer takes place through the electromagnetic energy exchange, which occurs in pulses translating to data.
Backscatter coupling is used in UHF tags and requires a dipole antenna. The reader generates high- power electromagnetic signals that the tags modulate and reflect back to the reader. Some readers can sense the power levels of the reflected signal, which is the basis for return-signal-strength positioning.
Sequential, Half-duplex, Full-duplex
RFID tags that could have any use in a positioning system must be able to harvest energy and transmit data that is stored in their small chip. Both of these actions occur through the only antenna system that the tag has, it is therefore necessary to define the timings for energizing, receiving data, and transmitting data. Three alternative procedures are used: sequential, full-duplex and half-duplex. All three procedures use three lanes of operation that “pass through” the single hardware interface, i.e. the antenna. The first lane is the energy transfer, the second lane is the downlink (data transfer from reader to tag), and the third lane is the uplink (from tag to reader).
The main characteristic of the sequential procedure is the intermittent energy transfer from the reader to the tag. Energy and data are transferred simultaneously in the first time slot, which is followed by uplink-only activity in the second time slot, followed by a simultaneous energy transfer/downlink activity, and the cycle continues until it is terminated with an uplink activity that is not followed by energy transfer.
Half- and full-duplex procedures utilize a constant energy transfer and non-constant data transfer. Of the two lanes for data transfer (uplink and downlink), only one is active at any given time slot in half- duplex procedures, while full-duplex procedures have them operating in parallel.
RFID positioning approaches
Positioning approaches that implement the lateration technique on RFID systems include Phase of Arrival (PoA) and Phase Difference of Arrival (PDoA), where the phase of the signals are used for estimation of distances between the tags and the readers (Povalač & Šebesta, 2010) and are found in systems operating at the UHF range.
Proximity-based (cell-of-origin) RFID methods
The extent of the interrogation zone of an RFID system defines the granularity of the proximity-based location methods, where the approximate location of the agent equipped with a tag is determined as they move into the zone. Proximity-based positioning might better be described by the term “tracking” (Song, Haas, & Caldas, 2007) rather than positioning because as a method, it does not necessarily disclose the geographical position, absolute or relative, of the tag, but it has a semantic meaning, e.g. “tag is found at gate B17”. The position of “gate B17” is therefore the position of the tag, to a certainty characterized by the granularity of the system.
When this technique is used for mobile phone tracking using the cell tower they are connected to, it is generally referred to as Cell-of-Origin (CoO), but this term could also be used in RFID.
Proximity-based methods are meant to be used in occasions where no information on the distance between the agent and a key location (in this case, a properly placed transponder of known coordinates) is available, such as strength of returned RF signal, propagation time for the signal, angle of arrival, etc. These methods offer the following benefits:
- Reduced cost of equipment. RFID readers and systems with the ability to return or extract some kind of information regarding the propagation parameters that can later be used for positioning estimation cost significantly more. For the present thesis, the purchased RFID reader model without the return signal strength data output would have cost approximately 30% less. The cost increases in more complicated systems based on time-of-fly or angle.
- Higher certainty in precision. Precision in PBPs is tightly linked to the interrogation range of the tag or the overlapping tags, and can be adjusted from the settings panel of the RFID system.
The drawbacks of PBP systems are the following:
- Number of tags. For a system to be able to perform as mentioned above, the precision of a PBP system is linked to the read range
For a 2D implementation, the unknown Cartesian coordinates of an agent that has interacted with a transponder of known coordinates will always be found within the radian interrogation range of the transponder/reader system. In other words, assuming that the signal is uniformly homogenous and unobstructed, the following expression holds:
In this example, as the agent moves in the interrogation range of the tag, i.e. as the distance between the agent and the tag becomes , the reading process is activated and the location of the agent is registered at high accuracy and at precision equal to . Higher precision can be achieved by reducing the transmission power of the RFID reader, thus reducing the ; noted that the effective positioning range for that specific tag will also decrease.
Higher precision can be achieved by placing several tags at close proximity with overlapping interrogation ranges. Estimating the location of the tag comes down to selecting a point from the intersecting region, which should be smaller than the entire range of a single reader. However, this implementation must be using anti-collision and anti-interference systems otherwise quality readings will not be possible.
RSS-based RFID methods
RF signals can be represented by the following units of measurement, which can be more or less converted from one another for measurements that are not close to the extremes and with a level of accuracy that varies (Bardwell, 2002):
- mW (milliwatts), where
- dBm (db-milliwatt)
- RSSI (Receive Signal Strength Indicator)
- % (percent)
Milliwatts (mW) and db-milliwatts (dBm) are measurements of the emitted RF energy but mW is linear, while dBm is logarithmic. Conversion between the measurements is possible by calculating the common logarithm (base 10) and multiplying by 10, as shown in the example that follows:
|Measurement in mW||Conversion||Measurement in dBm|
|100 mW||10 ∙ (log100) = 10 ∙ 2 = 20||20 dBm|
|50 mW||10 ∙ (log50) = 10 ∙ 1,69897 ≈ 17||17 dBm|
Doubling the emitted RF energy increases the dBm values by approximately 3 dBm. In addition, while it is possible to have negative dBm values, this is not the case with the mW since the latter refers to emitted RF energy which does not make sense to be negative.
RSSI is usually expressed in the form of a single byte character with value range 0-255, althoughmany vendors of consumer products generally report it in a ‘percentage’ scale of 101 values, where0 refers to minimum signal strength and 100 to maximum. This is mostly found in devices that are not related to RFID, such as wireless network (WiFi) devices etc.
- is the RSS at 1 m distance
- is the distance
- is the signal propagation exponent
Attenuation is expressed as ratio of change, in dB (decibels). For two power states (measured) and (reference), where , attenuation is calculated as (Couch, 1999, p. 212):
At this point, it should be pointed out that RSSI and distance are reversely related, i.e. higher distance will give lower RSSI returns.
Applications of RSS
Signal-strength-based ranging systems have been used in robotics for adding depth (distance) data to the pixels of tagged objects as they are seen by cameras attached to the robot (Deyle, Nguyen, Reynolds, & Kemp, 2009), in locating systems used for tracking of people in hospitals, livestock, workers in construction sites (Choi, Lee, Elmasri, & Engels, 2009). In fact, RSS-based positioning in construction sites has been the focus in quite a few projects, as the special requirements of these environments require study of interference and filtering of the signals prior to the actual location finding algorithms (Ibrahim & Moselhi, 2015).
RSSI value fluctuations
In practice, the strength of the returned signal is not stable, experiencing variations that are caused by several factors. These factors have been found to be physical distances between the transponders, obstacles, orientation of the antennas, and interference (Chapre, Mohapatra, Jha, & Seneviratne, 2013). These fluctuations, however, could be used for building of spectral maps in positioning using the fingerprinting method, were each cell must be assigned to a unique set of attributes in its spectral signature.Fingerprinting in RFID
With fingerprinting, location determination takes place over an area that has analyzed beforehand. It is a two-step setup that requires a preparatory phase (calibration) and the actualization phase, where real-time localization occurs. In the first phase, the entire area is divided into cells and RSS value samples are taken in each cell, with coordinates (xi,yj). The data of the most representative values is stored in a database (lookup table, location fingerprint map or radio map) and remains available for the next phase. By “most representative values”, it is meant that many RSS values are measured but only the average values are kept (Kaemarungsi & Krishnamurthy, 2004). The survey area can be monitored by several RFID readers (Ting, Kwok, Tsang, & Ho, 2011). The level of granularity in fingerprinting is characterized by the cell size which is defined by the distance between each set of coordinates (x1,y1), (x2,y1), …
In the actualization phase, the reader scans the environment for tags and receives RSSI values for each tag, which are communicated to a pattern-matching algorithm. Location is determined by querying the database for a matching cell record (xi,yj). However, it might not be possible to match the measured RSSI values to an exact cell, therefore an algorithm of Nearest Neighbors is used (Guvenc, Abdallah, Jordan, & Dedeoglu, 2003) which matches the unmatched measurements to the database records where the Euclidian distance is minimum. It should be noted that it is important that the conditions during the mapping phase and the positioning phase are exactly the same in order to limit the variations in the signal strength measurements.
As the signal leaves the source, it propagates through the medium (or the free space) surrounding the source, which is generally filled with background noise, towards the receiver. Based on the reaction of the receiving unit with regards to the signal, three zones can be defined:
- Transmission zone, where communication between the transmitter and the receiver is possible without any errors (or with an insignificant error rate)
- Detection zone, where the signal is detected but the error rate is too high for actual communication, and
- Interference zone, which is the space past detection range, where background noise “covers” the signal rendering practically undetectable by the receiver.
Multipath particularly affects UHF tags because of the fact that communicate via backscatter, i.e. by reflecting the reader’s interrogation signals. In addition, RFID is popular in tagging applications, meaning that it is expected to find them in confined places, such as indoors, and in places with many obstacles, such as storerooms. Multipath can be filtered out using a statistical profile for each ID (Wang & Katabi, 2013).
Collision of tag communication is caused by the presence of many tags in a confined space. In such cases, the reader is prevented from communicating properly and signals cannot be registered; a problem that appears in logistics but can also appear in a GIS application. Each communication path opened between the reader and one tag acts as interference to the path between the (same) reader and the nearby tags. If less dense placement of the tags is not feasible, then anti-collision systems (protocols) address these mishaps using the following techniques:
- by forcing the reader to time-manage the readings, lock communication channel with one tag at a time and block all the others,
- Block all others and open alternative communication channel
- Use time-sharing so that each tag has a time slot to communicate
- Install to tags a subsystem that sorts them in the reading queue according to their distance from the reader
- (for moving tags) install to tags a subsystem that sorts them in the reading queue according to their relative speed in relation to the reader
Anti-collision protocols in RFID is a field of ongoing study and new techniques emerge at times.
When designing a model for a system that uses RFID for positioning, the following two aspects need to be taken into account:
- False-negative readings, meaning that a tag is not detected, even though it lies within the reader’s read range. According to (Hähnel, Burgard, Fox, Fishkin, & Philipose, 2004), false- negative readings are frequent in these RFID model scenarios. In positioning applications, false negatives will affect the accuracy of the system by denying the third required distance measurement (for 2D systems). The effects of false negatives can be mitigated if the agent is moving in space covered by the read ranges of multiple tags by implementing correction algorithms (probabilistic distance-aware models) which function under the principle that there is a minimum travel time between the nodes and a minimum number of tags a traveler must encounter, effectively assuming the presence of a node if it’s been too long since the last tag was encountered (Baba, Lu, Pedersen, & Xie, 2013).
- False-positive readings, where the reader detects a tag located at a distance greater than its maximum read distance, as specified by its manufacturer.
The parameters that affect the aforementioned false-readings are the orientation and positioning of the tag with respect to the reader(s), the material environment and the presence of objects foreign to the RFID reading process. In particular, the orientation of the antennas affects the minimum required RF energy for a tag to be read correctly, i.e., more energy has to be radiated by the reader and absorbed by the tag in the case of non-ideal alignment of the antennas. Regarding the material environment parameter, a reader’s read range field is shaped by the materials used and their positions. The material of the surface (or underneath) where the tag is attached can absorb part of the RF energy thus reducing the maximum read distance for the said tag; an example of such a material is metal. False-positive readings can be caused by objects that reflect the RF waves in such a way that communication between readers and tags that seemingly should not have been read may occur (a.k.a. multipath contributions).
Data obtained by a wireless positioning device such as a GPS or, as in this case, the RFID reader, contains noise that needs to be removed prior to feeding the data to the model. This task is often carried out algorithmically by the Kalman filter (Kálmán, 1960), a mathematical tool that is used in cases where input data contains statistically independent noise that needs to be removed. Kalman filters are widely used in signal processing (Welch & Bishop, 2001) and function under a “predict- correct” algorithm. In the predict phase, the state of the system and its error covariance are projected one step ahead, and in the correct phase, the estimation corrector is applied on the real-time measurement. As a result, the data feed is smoothed as shown in the diagram below:
Designing the RSS-based system
As previously explained, the beneficiary’s location is estimated by the positioning system in relation to a reference position and later extrapolated to a common system of coordinates. In the case of RFID, two components of the system can play either role—that of the reference position, or the one of the unknown position. These two components are the tags/transponders and the readers; either the position of the tag is estimated relative to the position of the reader or vice versa. Each of these implementations have strengths and weaknesses. The final choice will be based on the following considerations:
- Physical properties of the beneficiary of the positioning service. For example, if the objective is to increase the accuracy of GPS in driverless cars moving in dense urban environments, then it is possible for the cars to carry a somewhat heavy and bulky long-range UHF reader and perform ranging to fixed passive tags. On the other hand, for tracking of livestock in a farm, it would make more sense to install the readers on fixed positions and tag the animals with the much lighter passive tags.
- Power. Readers need to be connected to a sufficient power source, while tags are passive. As an example, a typical 9V battery can power a long-range UHF RFID reader like the one used in the present thesis for up to one hour of continuous operation.
- Cost. Passive tags are inexpensive (indicative price from USD 0,20 per piece), while long- range readers are significantly more expensive (indicative price USD 220 per unit). Installation of readers incurs additional costs as well.
- Reader-to-reader interference, reader-to-tag interference, and multipath effects caused by overlapping interrogation ranges in multiple-reader implementations (Bekkali, Zou, Kadri, Crisp, & Penty, 2014).
Read part 2 here.