Low Power Wide Area Network Analysis: Can LoRa Scale

ABSTRACT

Low Power Wide Area (LPWA) networks are making spectacular progress from design, standardisation, to commercialisation. At this time of fast-paced adoption, it is of utmost importance to analyse how well these technologies will scale as the number of devices connected to the Internet of Things (IoT) inevitably grows. In this letter, we provide a stochastic geometry framework for modelling the performance of a single gateway LoRa network, a leading LPWA technology. Our analysis formulates unique peculiarities of LoRa, including its chirp spread-spectrum modulation technique, regulatory limitations on radio duty cycle, and use of ALOHA protocol on top, all of which are not as common in today’s commercial cellular networks. We show that the coverage probability drops exponentially as the number of end-devices grows due to interfering signals using the same spreading sequence. We conclude that this fundamental limiting factor is perhaps more significant towards LoRa scalability than for instance spectrum restrictions. Our derivations for co-spreading factor interference found in LoRa networks enables rigorous scalability analysis of such networks.

INTRODUCTION

The last years have seen much interest in Low Power Wide Area (LPWA) technologies, which are gaining unprecedented momentum and commercial interest towards the realisation of the Internet of Things (IoT). There are many candidates that have taken the research community by surprise, actively pursuing standardisation, adoption, and commercial deployments in parallel. Most LPWA networks operate in the unlicensed ISM bands at 169, 433, 868/915 MHz, and 2.4 GHz depending on the region of operation. Some of the most pronounced LPWA candidates are SigFox, LoRa, Weightless, and Ingenu [1].

In this paper, we focus on LoRa (Long Range), one of the most promising wide-area IoT technologies proposed by Semtech and further promoted by the LoRa Alliance [2]. At the heart of LoRa’s success is its adaptive data rate chirp modulation technology allowing for flexible long-range communication with low power consumption and low cost design. Essentially, this is achieved via spread spectrum multiple access techniques accommodating multiple users in one channel. LoRa Alliance has defined the higher layers and network architecture on top the LoRa physical layers and termed them LoRaWAN. Together, these features make LoRa attractive to developers who can build complete system solutions on top of it for both geographical and residential/industrial types of IoT networks, thus fast-tracking its market adoption.

Despite this success, LoRa has not yet attracted similar levels of attention from the academic and research community with only very few peer-reviewed studies published to date [1], [3]–[7]. In fact, most of these studies are either review articles [1], [3], [4], or measurement reports [5], [6]. Significantly, it has consistently been assumed in peer-reviewed and industry published white papers that the different spreading sequences employed by LoRa are orthogonal and provide interference immunity at the receiver end. While this technology does indeed create an extra set of “virtual” channels thus increasing the capacity of the gateway, transmissions of similar spread are susceptible to a new type of interference, unique to LoRa networks, which we term co-spreading factor interference. Since LoRaWAN’s MAC protocol is essentially an ALOHA variant with no collision avoidance provisions [3], in very dense deployment scenarios, LoRa networks will inevitably become interference-limited, rather than noise-limited, thus necessitating for new interference-related performance metrics which capture the interference peculiarities of LoRa networks.

It is the purpose of this letter to apply state-of-the-art mathematical tools to model the uplink coverage of single gateway LoRa networks and further understand its unique PHY and MAC features. Namely, we leverage tools from Stochastic Geometry [8] to address two independent linkoutage conditions, one concerned with SNR, and the other with co-spreading factor interference. We show that LoRa is susceptible to the latter, and that end-device coverage probability decays exponentially with increasing number of end-devices, despite the ‘cushioning’ provided by the low duty cycling and chirp orthogonality.

CONCLUSION

We have investigated the effects of interference in a single gateway LoRa network, a LPWA technology with promising IoT applications. Unlike other wireless networks, LoRa employs an adaptive CSS modulation scheme, thus extending the communication range in the absence of any interference. Interference is however present when signals simultaneously collide in time, frequency, and spreading factor. Leveraging tools from stochastic geometry, we have formulated and solved two link-outage conditions, one based on SNR, and the other on co-spreading sequence interference. Each displays interesting behaviours, unique to LoRa, with the latter causing performance to decay exponentially with the number of end-devices, despite various interference mitigation measures available to LoRa, thus limiting its scalability. It is interesting that LoRa networks appear to be impervious to cumulative interference effects (typically modelled as shot-noise [8]). If this assumption is invalid, then our qualitative results are simply optimistic upper bounds towards network scalability. Going beyond this first foray into the modelling of LoRa, it would be interesting to understand the effets of multiple gateways [6], and spatially inhomogeneous deployments. Finally, we point towards recently developed packet-level simulators [13] which can further shed light into the performance of LoRa networks