Radio-based Localisation Techniques

Received Signal Strength (RSS)

RSSI values are Received Signal Strength (RSS) values averaged over a certain sampling period and are measured as a perceived power value \(P_r\) with the units of decibels. Using the physical property of attenuation, the gradual loss of intensity over a spatial propagation, RSS can be used to determine a distance d to an emitting source, as the output power value \(P_T\) is known. RSSI localisation uses trilateration with at least 3 transmitters.

\[P_r \propto P_T \frac{G_T G_R}{4 \pi d^p}\]

Attenuation is characterised by a path loss exponent \(p\) which in free space is represented by a literature value \(p = 2\), it is however heterogeneous in an indoor space. \(P_T\) is the transmitted power by the at the emitter. In theory, by calculating multiple perceived distances \(d\) from emitters to various receivers, position can be calculated via multi-lateration. This however this is subject to multi-path influences, introducing error.

Advantages:

Simple and cost-efficient due to low hardware requirements

Disadvantages:

Poor localization accuracy, especially in non-Line-of-Sight situations
Additional signal attenuation from transmission through walls, obstacles, and multipath
Potentially high fluctuation over time

Channel State Information (CSI)

Captures amplitude and phase response across different frequencies
Higher granularity than RSSI, gives more information to reduce multipath and provide more consistent measurements
Overall higher localisation accuracy than RSSI

Fingerprinting/Scene Analysis

Discrete estimation of user location based on grid
- Delicate balance between size of grid (and corresponding resolution), and the signal variation error
- Eg: impossible to estimate the correct point if the difference in RSSI values between each grid cell is less than the uncertainty range of the signal strength
Offline phase: collect RSSI/CSI measurements to form a fingerprint map
Online phase: compare real-time measurements with fingerprint map using algorithms:
- Probabilistic method: use histogram/kernel to calculate probability of receiver being in \(x\) grid of fingerprint map
- Artificial Neural Network: train a neural network (eg: Multi-Layer Perceptron) and provide real-time input with weights
- k-Nearest Neighbours: average nearest measurement patches to estimate the location
- Support Vector Machine (SVM): machine learning method applicable to indoor localisation

Angle of Arrival (AoA)

Receiver's antennae array estimates the angle of transmitted signal by the time difference of arrival at individual parts of the array
Advantages:
- Receiver location can be estimated with just 3 transmitters in a 3D environment
- Accurate estimation when transmitter-receiver distances are small
Disadvantages:
- Slight error in angle calculation translates to huge error in position determination
  - Thus not accurate at larger transmitter-receiver distances
- Line of Sight for angle calculation is hard to obtain due to indoor multipath effects

Time of Flight (ToF)

Calculate physical transmitter-receiver distance using signal propagation time and speed of light
Trilateration is used similar to RSSI to estimate position
Requires strict time synchronisation between transmitter and receiver, such that timestamps may also be included in the signal
Advantages:
- High sampling rate and large signal bandwidth increases the resolution of the position estimation
Disadvantages:
- Highly sensitive to obstacles as they deflect the signals, resulting in a longer propagation time

Time of Arrival (ToA)

This principle measures the absolute travel time of a signal from transmitter to receiver. A euclidian distance can be derived using the wave speed.

Time Difference of Arrival (TDoA)

Create hyperbolas between each pair of transmitters
Intersection point of all hyperbolas represents the position estimate
- Ie: solve the system of hyperbola equations for X,Y,Z
Difference between TDoA and ToF
- ToF uses absolute signal propagation time between each transmitter and receiver
- TDoA uses the difference between these propagation times
  - Synchronisation is thus only necessary between transmitters (receiver error is common between transmitter pairs)

Return Time of Flight (RToF)

Measures the round-trip signal propagation time from transmitter to receiver, and response signal from receiver to transmitter
Difference between RToF and ToF
- Only moderate synchronisation between transmitter and receiver is required
- All factors affecting ToF are worse in RToF due to twice the distance
- Time delay in transmitting response signal (depending on the receiver's electronics) adds to position estimation error

Phase of Arrival (PoA)

Measures phase difference from transmission to reception
Typically used in conjunction with other methods (RSSI, ToF, TDoA) to enhance localisation accuracy
High accuracy is only achievable with direct line-of-sight

Angle of Arrival (AOA)

This methodology aims to obtain the direction of an incident incoming wave signal. For this, directionally sensitive antennas are needed. By having sensors that are responsive to a specific direction in space, vector addition can be used to evaluate an angle of incidence (arrival).

Path-Loss

The ITU Model for Indoor Attenuation takes into account how radio waves propagate indoors. It models the path loss \(L\), relative to the distance \(d\) between emitter and receiver. \(\(L=20 \log f+p \log d+c(k, f)-28\)\) Here \(f\) is the radio frequency and \(c\) is an empirical floor loss penetration factor. \(k\) is the number of floors between transmitter and receiver.

Fine Timing Measurement (FTM)

Radio Frequency Identification (RFID)

A Radio Frequency Identification system consists of an antenna which reads nearby active transceivers (transmit & receive) and passive tags on the radio wave part of the EM-spectrum. These transmit a unique tag ID. To position these tags in space the Cell of Origin (CoO) principle is used. A combination of ToA and AoA is used to pinpoint a location of an RFID tag.

Active RFID sensors are equipped with batteries which interrogate active radio transceivers. They are more costly than their passive opponent but offer long detection ranges of 30m or more. Location detection accuracies on the meter level have been proven possible.

Passive receivers rely on inductive coupling of EM waves and do not require batteries. These rely on receiving enough information by a scanner in order to transmit codes back. They are less costly, but only function in ranges smaller than 2 meters as signals need to travel two ways.

Applications of passive scanners include the identification of buried pipes as well as inventory control.

Using RFID scanners for indoor locations implies either having permanent active RFID sensors casting information or scanners suitable to detect passive ones. Both require large capital cost of installation.

Comparison

Technology	Range	Advantages	Disadvantages
Wifi 802.11 n/ac/ad	35m	Widely available High accuracy Existing infrastructure	Sensitive to noise Complex processing algorithms needed
Wifi 802.11 ah (new)	1km	High coverage range Low power consumption	New technology, not yet tested for indoor environments
Ultrawide Band (UWB)	10-20m	Immune to interference High accuracy	Short Range Additional hardware High cost
Radio Frequency Indentification Device (RFID)	200m	Low power consumption High coverage range	Low accuracy
Bluetooth	100m	High throughput High coverage range Low power consumption	Low accuracy Sensitive to noise
Ultrasound	<20m	Less absorption from obstacles	High dependence on sensor placement
Acoustics	<2m	High accuracy	Niche applications Sensitive to sound pollution Additional hardware/anchor points
Visible Light (LEDs)	1.4km	Widely available High potential (with AoA technique) Multipath-free	Requires Line-of-Sight Higher power consumption