Transmission Techniques for LoRaWAN IoT Networks
Date
2025-05-15
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
0000-0002-0305-0340
Type
Thesis
Degree Level
Doctoral
Abstract
The Internet of Things (IoT) is a network of devices connected to collect data, process it, and then use it to improve automation and decision-making. The hardware of IoT includes everything from low-cost sensors in homes, industries, and farming to very advanced machines like drones and self-driving cars. The data generated by these devices is massive, diverse, and noisy. Moreover, the six Vs \textemdash volume, velocity, variety, variability, veracity, and value\textemdash \;are used to classify it. Wireless networks are crucial to forward this data to the cloud for further processing. In terms of coverage range, there are three types of IoT wireless networks: short-range, wide area and cellular, and non-terrestrial networks (NTNs) which are an extension of coverage into the third dimension.
With the fast-paced growth of IoT applications, scalability, and efficiency will be major issues for both current and future networks. To address these issues, low-power wide area network (LPWAN) technology has emerged as one of the most popular connectivity solutions for IoT networks. LPWAN provides extensive communication coverage, energy efficiency, and cost-effectiveness, making it excellent for applications that tolerate delays and have restricted device throughput, such as low-power and battery-operated devices. Among various LPWAN technologies, long-range wide area network (LoRaWAN) protocol is rapidly gaining industrial popularity throughout the globe. LoRaWAN is an open-source technology that enables autonomous network setup at a low cost, and hence, it is frequently employed for LPWAN applications. LoRaWAN has been widely used for a variety of systems and purposes. Smart homes, smart cities, smart agriculture, smart meters, and water quality monitoring are examples of general LoRaWAN-based IoT applications. Since technologies based on LoRaWAN have to be very energy efficient, LoRaWAN medium access control (MAC) tries to avoid overhead in signaling as much as possible as it can cost more energy and latency. LoRa protocol as one of LoRaWAN transmission schemes uses a very simple channel allocation strategy adopting ALOHA with additional ACK mechanisms so end-devices (EDs) do not need to peer with specific gateways (GWs). Moreover, the LoRa protocol adopts Semtech's proprietary chirp spread spectrum (CSS) modulation in the physical (PHY) layer. Recently, another transmission scheme, i.e., long-range frequency hopping spread spectrum (LR-FHSS) protocol, has been introduced by Semtech as an extension to LoRaWAN. This new transmission scheme is motivated by the emerging IoT use cases with increasingly larger and denser network deployments, including direct-to-satellite (DtS) IoT networks in which the coverage area is very large due to the satellite footprint. In the MAC layer, the LR-FHSS protocol uses frequency hopping as a method of random frequency allocation along with the ALOHA to provide immunity against co-channel interference. Also, LR-FHSS exploits Gaussian minimum shift keying (GMSK) in the PHY layer. In this thesis, we aim to improve the performance of LoRaWAN transmission techniques, particularly, LoRa and LR-FHSS protocols, via the following contributions.
LoRa protocol provides a relatively low data rate for long-range communication requirements of LPWAN. Hence, in the first contribution, we aim to improve the performance of CSS modulation in terms of spectral efficiency (SE) and error performance. LoRaWAN IoT EDs offer a long-range communication \textemdash up to tens of kilometers\textemdash\; and extended battery life of up to ten years. However, they support relatively low data rates, typically in the range of a few kbit/s. In the U.S., the highest available data rate, achieved with a spreading factor of 7 (SF7) and a $500$ kHz bandwidth, is only $21.9$ kbit/s. While such a data rate is sufficient for many applications, this limitation hinders high-data-rate scenarios such as emergency communications, multimedia IoT, video streaming, and disaster monitoring. Enhancing the SE of CSS modulation is therefore essential to meet growing IoT demands. To this end, this thesis introduces a system based on integrating multiple-input multiple-output (MIMO), CSS, and permutation matrix modulation (PMM) named MIMO-CSS-PMM. This transceiver model improves LoRaWAN's SE by encoding additional information into the CSS symbol assignment for each transmitter antenna while leveraging spatial diversity at the receiver to enhance error performance. We formulate the maximum likelihood (ML) detection for the presented system model which turns out to be of very high complexity considering the practical CSS transmission parameters. To reduce ML search space, and hence, reduce the receiver complexity, we propose two low-complexity semi-coherent detection schemes, namely, scheme I and scheme II. Both schemes exploit the principles of CSS signal detection, i.e., dechirping and taking fast Fourier transform (FFT) of the signal of each receiver antenna. By evaluating the proposed MIMO-CSS-PMM in terms of SE, it turned out that it is improved by almost $4$ times compared to the conventional CSS scheme for different settings of SF. Also, the error rate of the proposed MIMO-CSS-PMM has been improved by almost $4$ dB at a bit error rate (BER) of $10^{-4}$ compared to the conventional CSS modulation. Moreover, using computer simulations, it has been shown that going from scheme I to scheme II results in a similar error performance while decreasing the complexity of the detection by almost $95\%$.
For the terrestrial LoRa protocol setup, in which the coverage area is small compared to the DtS-IoT networks, the literature has shown that a single GW can serve almost $100$ ED/${\rm km}^2$ for the target outage probability of $10^{-2}$, for a coverage area of $0.554176$ ${\rm km}^2$ (with an ED to GW distance of $420$ m). However, regarding the emerging applications of DtS-IoT networks, since the network must support a massive number of EDs with the satellite acting as the GW providing a significantly larger coverage area, the resource allocation of LoRa protocol based on ALOHA cannot cope with the performance requirement of such networks. Based on the results provided in this thesis, LoRa protocol using ALOHA can serve EDs with density up to $6.44\times 10^{-5}$ ED/${\rm km}^2$ with an outage probability of $10^{-2}$ using the best settings in terms of outage, i.e., SF7, for a satellite coverage area of $24,839,188$ ${\rm km}^2$. As an example, for the area of Saskatchewan, CA, i.e., $651,900$ ${\rm km}^2$, this translates to supporting only $42$ EDs in the entire province for the target outage of $10^{-2}$.
As mentioned, LR-FHSS is suitable for a dense network with a high coverage area due to its random joint time-frequency allocation method, i.e., ALOHA and frequency hopping. Compared to the LoRa protocol, by exploiting the LR-FHSS, the density of users increases to $0.02$ ED/${\rm km}^2$ using DR6 and $0.06$ ED/${\rm km}^2$ using DR5, with the coverage area of $24,839,188$ ${\rm km}^2$. Considering the previous example, for the area of Saskatchewan, CA, the LR-FHSS protocol can support $13,038$ EDs for DR6 and $39,114$ EDs for DR5 at the target outage of $10^{-2}$, which is significantly higher than the LoRa protocol. Hence, LR-FHSS is considered a potential solution for LoRaWAN DtS IoT applications in which the coverage area of the GW is much higher than the terrestrial networks.
Based on the above discussions, in the second contribution, we provide a mathematical framework for the LR-FHSS protocol, where we derive a closed-form expression for the outage probability under fading, noise, and path loss. Simulation results demonstrate that LR-FHSS delivers enhancements in terms of number of IoT EDs served by a single GW for a fixed outage probability of $10^{-2}$. In particular, the obtained results show that the network can serve up to $0.02$ and $0.06$ ED/${\rm km}^2$ for $48$ bytes of information using two specified data rates (DR6 and DR5). These numbers present a huge improvement over the conventional LoRa network which completely fails in such network settings.
In the third contribution, we present an integration of a device-to-device (D2D) communication scheme with the LR-FHSS to further improve network scalability. We propose integrating D2D communication and network coding to enhance performance for massive IoT deployments. A detailed analytical framework is developed to derive a closed-form expression of the outage probability of the LR-FHSS system, incorporating real-world factors like shadowed-Rice fading, path loss, noise, and interference in both time and frequency domains. We enhance LR-FHSS by having EDs first communicate through LoRa then forward data to a LEO satellite’s IoT GW using LR-FHSS. The mathematical analysis reveals that the D2D-aided LR-FHSS approach reduces the outage probability compared to the conventional LR-FHSS transmission method. Moreover, the results show a significant boost in network performance of D2D-aided LR-FHSS in terms of the density of EDs served by a single IoT GW installed on a satellite compared to LR-FHSS protocol ($249.9\%$ and $150.1\%$ increase at a typical outage of $10^{-2}$ for DR6 and DR5, respectively).
In the fourth contribution, we aim to lower the complexity of LR-FHSS packet detection. As mentioned, the LR–FHSS transmission protocol can support very large numbers of IoT EDs through its time-frequency scheduling, and hence, millions of packets can be received by a single GW in an hour. This high traffic volume requires a low complexity detector that consumes less processing power. To address this issue, this thesis presents a symbol-by-symbol low-complexity detection scheme using the characteristics of GMSK as the underlying modulation of LR-FHSS. While significantly lowering the detection complexity compared to the existing detection methods, e.g., the Viterbi algorithm, our proposed detector suffers from only $1$ dB and $1.5$ dB error performance degradation compared to the optimal receiver in additive white Gaussian noise (AWGN) and shadowed-Rice fading conditions, respectively. This makes our proposed detection a candidate for being used in DtS IoT scenarios where a single GW will serve a massive number of IoT EDs.
Finally, in the fifth contribution, we investigate the coexistence of both LoRa and LR-FHSS. This thesis provides the first analysis of the coexistence problem of LoRa and LR-FHSS and their impact on the performance of LoRa chirp modulation. This coexistence issue is important, especially when LoRa and LR-FHSS coexist in the terrestrial and DtS-IoT networks. We investigate this problem by focusing on the decoding of the LoRa chirp signals in the presence of interference from LR-FHSS. We derive an optimal joint maximum likelihood (ML) detector for decoding the LoRa chirp in the presence of interfering LR-FHSS fragments. Due to the high complexity of joint ML detection, we further propose a low-complexity filtering-based detection scheme that can remove the LR-FHSS interference using the narrowband characteristics of the LR-FHSS packet fragments. For AWGN condition, the results show $1$ dB loss of BER of $10^{-4}$ at low signal-to-interference (SIR) of $-15$ dB compared to the case without interference. Moreover, it is observed the proposed detector can provide acceptable performance in relatively high SIR values, i.e., ${\rm SIR}=-1$ dB, for Rayleigh fading channel conditions.
Description
Keywords
CSS, IoT, LoRaWAN, LRFHSS
Citation
Degree
Doctor of Philosophy (Ph.D.)
Department
Electrical and Computer Engineering
Program
Electrical Engineering