Journal:Wireless positioning in IoT: A look at current and future trends

Abstract

Connectivity solutions for the internet of things (IoT) aim to support the needs imposed by several applications or use cases across multiple sectors, such as logistics, agriculture, asset management, or smart lighting. Each of these applications has its own challenges to solve, such as dealing with large or massive networks, low and ultra-low latency requirements, long battery life requirements (i.e., more than ten years operation on battery), continuously monitoring of the location of certain nodes, security, and authentication. Hence, a part of picking a connectivity solution for a certain application depends on how well its features solve the specific needs of the end application. One key feature that we see as a need for future IoT networks is the ability to provide location-based information for large-scale IoT applications. The goal of this paper is to highlight the importance of positioning features for IoT applications and to provide means of comparing and evaluating different connectivity protocols in terms of their positioning capabilities. Our compact and unified analysis ends with several case studies, both simulation-based and measurement-based, which show that high positioning accuracy on low-cost low-power devices is feasible if one designs the system properly.

Keywords: internet of things (IoT), wireless positioning, indoor location

Introduction

Nowadays, the amount of connected wireless devices is growing, e.g., smart watches, smart light bulbs, smart toothbrushes, smart coffee mugs, etc. The trend in the information technology industry is towards connecting and extracting analytics from a variety of inter-connected wireless devices.

While many IoT applications have so far focused on the consumer realm, more and more industrial applications are also appearing, such as utilities measurement (e.g., water, electricity, and gas), industrial lighting, logistics, and smart agriculture. Enabling such industrial applications means that IoT networks need to support large amounts of devices, multiple years of operation on battery, different latency requirements, and low costs per unit.

We believe that, on top of the communications and reliability requirements of a wireless link, many IoT applications will require or benefit from knowing the location of certain devices. Such location information will be needed seamlessly, both indoors and outdoors, and without the battery-draining Global Navigation Satellite Systems (GNSS) chipsets. The need for localization and tracking appears not only from the network management point of view, but also from a business perspective, driving new business models and new business avenues.

Nevertheless, enabling or creating a positioning system with an IoT network is not a trivial task. The reason behind this is that industrial applications seek a low per unit cost of their IoT devices, which results in devices with very limited hardware components, such as CPU, memory, and battery. The limited hardware has an impact on the number of devices that a single device can serve and how fast it can process network and application requests. However, while CPU and memory will have an important impact on the scale of the network, the biggest challenge for enabling a positioning system lies on the proper management of the devices’ radio.

The need for proper radio management becomes evident as there are devices with known coordinates which will broadcast specific payloads on a regular basis and other devices whose locations are to be determined, which will need to scan the spectrum frequently. Hence, too frequent broadcasting will lead to spectrum congestion and increased packet collision, whereas frequent scanning leads to high battery consumption, which is particularly problematic for battery-operated devices.

Overall, the biggest challenge to tackle for an IoT positioning network is to balance the power consumption against the performance of the system. A very reactive system will have to rely on frequent scanning and broadcasting of its members, which means that devices will need to draw large quantities of power. A low reactive system will draw less power with devices scanning very seldom.

The goal of this paper is to provide an insight on positioning capabilities of the current IoT technologies and other relevant IoT-enabling wireless technologies, such as WiFi. The paper starts by classifying three domains of positioning and discussing the main shortcomings of each of these domains for IoT devices. It then classifies the different IoT solutions according to six classification criteria, and it provides a discussion on the main system parameters relevant to positioning and tracking purposes. This discussion acts as a basis for comparison between the different IoT wireless solutions. To further complement this discussion, we present positioning results based on simulation-based scenarios and field experiments with a platform built on top of the Wirepas Mesh connectivity solution. In the end, we provide a short summary and conclusions of our findings.

Related work

At this moment, to the authors’ best knowledge, there are no comprehensive comparisons in the literature between different IoT protocols in terms of their positioning capabilities. There are, however, other studies that compare specific IoT technologies and which look at IoT from the communications point of view, as well as studies focusing on positioning with a particular technology, such as narrow-band IoT (NB-IoT) or BLE. In this section, we highlight the related work from literature studies.

A survey of localization methods for 5G, containing a short section also on IoT positioning, has been recently published as a white paper by the European Cooperation in Science & Technology (COST). It has also been emphasized in this paper that localization will become a key component of future 5G systems, though accurate future localization solutions in 5G should exploit the multipath and non-line-of-sight information and should put more emphasis on heterogeneous data fusion mechanisms. However, such advanced solutions would also increase the power consumption on the devices and are not well-suited for the majority of IoT systems. Distinct from COST's work, our paper focuses mostly on low-cost low-power consumption IoT solutions.

Lin et al. focus on the Long-Term Evolution (LTE) Machine type communications (LTE-M) and Narrow Band Internet of Things (NB-IoT) protocols and their positioning capabilities. The authors demonstrate that at 46 dBm power of the transmit AN, positioning accuracy goes to around 10 m and that NB-IoT protocol supports better positioning accuracy than LTE-M protocol. A similar study focuses on indoor localization via improved received signal strength (RSS) fingerprinting in generic IoT devices. The results are based on 802.11b/g/n signals where location errors below 5 m are achieved in more than 50 percent of the studied cases.

del Peral-Rosado et al. investigate a time-domain based positioning with additional frequency hopping for the NB-IoT system. The obtained positioning accuracy is down to 30–50 m under strong signal-to-noise ratio conditions, and it deteriorates quickly for medium and low signal-to-noise ratios.

Chen et al. released a study complementary to our work, looking at IoT positioning from the perspective of security, privacy, and robustness of the localization technology. No positioning results were reported in the study. Another complementary study by Singh and Kapoor focuses on existing and emerging software and hardware platforms for IoT applications, but positioning was not part of that study. IoT positioning has recently been considered by Zhang et al. from the point of view of spoofing resistance in time of arrival (TOA) ultra-wideband (UWB) for IoT systems.

Other complementary comprehensive studies, focusing solely on the communication aspects of IoT, are authored by Al-Sarawi et al. and Raza et al.

==Designing an IoT positioning system At its core, a positioning system translates a set of measurements from well-known reference points into a coordinate pair. The reference points—known as anchors in localization terminology or Access Nodes (AN) in IoT terminology—act as a means for the device of interest, a mobile or an IoT tag, to be in a local or global reference frame. Depending on who takes the measurements, the positioning is considered to be network-centric (i.e., when the anchors make the positioning-related measurements) or device-centric (i.e., when the IoT end nodes or tags perform the positioning-related measurements).

These two types of positioning have very different implications on security and privacy, which should always be carefully considered regarding the final application. For example, privacy-preserving positioning solutions are easier to be achieved in a device-centric approach than in a network-centric approach as the device would not need to disclose its position to the network.

Positioning domains

In terms of measurements, there are multiple domains from which they can be extracted from, as long as there are means to do so in the devices. For that reason, we briefly present three of the main domains we consider of interest for an IoT positioning system:

• power or signal strength-based
• time-based
• space-based

Other domains, such as natural or artificial fields, e.g., geo-magnetic field, light, sounds, or smell are out of the scope of our study, but they could also serve as relevant sources of information for future IoT positioning systems.

The following subsections provide a short summary of main challenges in each of these three positioning domains and their system-wide impacts.

Power domain

Signal strength measurements are derived from the protocol operation, which most of the times results in a measurement of no additional cost to the device and battery consumption. However, positioning solutions in the power domain must tackle several challenges, in particular those related to the fast fluctuations of the Received Signal Strength (RSS) or of the backscattered power (BP), due to fading and shadowing caused by the surrounding environment. One key factor to model the RSS measurements relies on the possibility of understanding, with a given degree of accuracy, how the signal power changes in its surrounding environments. The signal power models as a function of the distance between the transmitter and the receiver are known as path-loss models. A typical empirical Log distance model is the single-slope path loss model:

$P_{r}\left(d\right)=P_{r}\left(d_{0}\right)-10\eta \log _{10}\;\left({\frac {d}{d_{0}}}\right)+w$,

where $P_{r}\left(\right)<0$ is the received signal power in logarithmic scale dependent on distance d, $d_{0}$ is a reference distance (usually 1 m), $\eta >0$ is the path-loss exponent, and $w\sim log\left(N\left(0,\sigma ^{2}\right)\right)$ is a log-normally distributed random variable that models the slow fading phenomenon and possible RSS measurements errors (e.g., due to quantization). Both $\eta$ and w are dependent on the propagation environment and are typically dependent on the device type and environment type. In addition, w can depend on factors such as device orientation and the amount of people present in the measurement area at the time of acquisition.

In terms of an IoT positioning system, the fact that one can extrapolate this information directly from the communication’s signal, which means that there is no additional cost for the device. In terms of battery, the cost will depend on the amount of positioning location requests demanded per second. Ideally, if the requirement is to have an opportunistic location, based on the sporadic communication of the device, acquiring the RSS-based positioning will have no impact on the battery life. However, if the device or the infrastructure will have to listen periodically for a specific pilot signal, acquiring the RSS-based positioning will cause further demands in terms of battery consumption. One limitation of RSS-based approaches is that some current IoT standards support only a coarse RSS measurement (e.g., in steps of 6 dB), which can adversely impact the positioning accuracy, as the noise variance $\sigma ^{2}$ will increase.

Another interesting aspect of the RSS measurements is that, based on simulations, RSS-based positioning errors are shown to be frequency independent (as shown later in Figure 1). However, one would expect different levels of location-based service at different frequency ranges. The frequency ranges can be coarsely divided into three categories: sub-GHz (i.e., carrier frequencies less than 1 GHz), GHz (1 to 30 GHz) and mmWave (above 30 GHz). The scattering becomes more prominent as frequency increases, thus one would expect different target positioning accuracy according to the frequency range. In addition, as the operating frequency increases, the antenna’s effective area is smaller, and the signal coverage decreases. This is possible to see in Figure 2 where the ideal signal propagation in drawn over a 100 by 100 square area, based on the Friis equation and assuming zero system gains, G,

$P_{r}\left(d\right)=P_{t}+G+20\log _{10}\;\left({\frac {c}{f4\pi }}\right)-20\log _{10}\;\left(d\right)$,

where $P_{t}$ is the transmission power, f the operating frequency and c the speed of light.

Figure 1. Comparative analysis of RSS-based estimates at various carrier frequencies and various AN densities

Figure 2. Ideal radio signal propagation at 0.5, 2.4, 30 and 60 GHz

Based on the signal’s behavior, it is easy to understand that a sparser infrastructure at higher frequencies will likely result in a degradation of the positioning performance (as shown in Figure 1).

Time domain

Positioning estimation based on timing information is based on estimating the time-of-arrival (TOA) or the time-difference-of-arrival (TDOA) from three or more fixed access nodes and then converting those timing estimates into distances. For example, 3D location based on TOA is possible with three synchronized measurements from three known devices. The goal is to solve the following set of equations and find out the node’s location, $\xi _{n}=\left(x_{n},y_{n},z_{n}\right)$, assuming that $\xi _{n}=\left(x_{a},y_{a},z_{a}\right)$ are known coordinates:

$\mathbf {L} _{n}={\sqrt {\left(\xi _{a}-\xi _{n}\right)^{2}}}$.

For TDOA, the range is now a difference of ranges, based on the TOA at the measurement device. Hence, the TDOA from a node n to a measurement device m would be written as

$\mathbf {L} _{nm}=\mathbf {L} _{n}-\mathbf {L} _{m}$.

Due to this difference, the range measurement is free of errors imposed by the measurement device’s clock, since it cancels out when subtracting the two TOA measurements.

Overall, the time measurements require synchronized clocks, either at the receiver or at the transmitter side, leading to a significant burden on device cost. This does not play well for IoT applications, which are driven by the need of having low-cost devices.

It is also important to keep in mind the relationship between bandwidth and accuracy for TOA measurements. This is illustrated in Figure 3, where the positioning error is plotted against the available channel bandwidth at different Signal-to-Noise Ratio (SNR) values. Clearly, sub-m positioning accuracy with time-based approaches is achievable only with high bandwidths (of the order of 100 MHz), but it is very challenging for narrowband and ultra-narrowband systems even at very high SNR.

Figure 3. Comparative analysis of TOA-based position estimates at various bandwidths

Space domain

In the space domain, the ranges are estimated by measuring the angle (or direction) of arrival (AoA) for the signal of interest. Often, this is done by the means of an antenna array or a sectorized antenna. For a given device n, it is possible to describe its measurement at m as

$\left(x_{n},y_{n}\right)=r_{m}cos\left(\theta _{m}\right)+r_{m}sin\left(\theta _{m}\right)$,

where $r_{m}$ is the distance m to n and $\theta$ the angle of arrival determined at m. Hence, by solving for the unknown coordinates, one can obtain a range estimate.

In summary, AoA is particularly interesting for IoT, as the major constraint for achieving angle measurements relies only on the antenna design. However, its major drawback is that the error increases with the distance to the transmitter, which means that a small deviation in the angle results in a large error for the devices at the service edge.

Notes

This presentation is faithful to the original, with only a few minor changes to presentation. In some cases important information was missing from the references, and that information was added. Some grammar, punctuation, and minor wording issues have been corrected.