Experiences from Using LoRa and IEEE 802.15.4 for IoT-enabled Classrooms
Lidia Pocero, Stelios Tsampas, Georgios Mylonas, Dimitrios Amaxilatis
EExperiences from Using LoRa and IEEE802.15.4 for IoT-enabled Classrooms (cid:63)
Lidia Pocero , Stelios Tsampas , Georgios Mylonas , and DimitriosAmaxilatis Computer Technology Institute and Press “Diophantus”, Rio, Patras, Greece { pocero, tsampas, mylonasg, amaxilat } @cti.gr Abstract.
Several networking technologies targeting the IoT applica-tion space currently compete within the smart city domain, both in out-door and indoor deployments. However, up till now, there is no clear win-ner, and results from real-world deployments have only recently startedto surface. In this paper, we present a comparative study of 2 popu-lar IoT networking technologies, LoRa and IEEE 802.15.4, within thecontext of a research-oriented IoT deployment inside school buildings inEurope, targeting energy efficiency in education. We evaluate the actualperformance of these two technologies in real-world settings, presenting acomparative study on the effect of parameters like the built environment,network quality, or data rate. Our results indicate that both technologieshave their advantages, and while in certain cases both are perfectly ade-quate, in our use case LoRa exhibits a more robust behavior. Moreover,LoRa’s characteristics make it a very good choice for indoor IoT deploy-ments such as in educational buildings, and especially in cases wherethere are low bandwidth requirements.
Keywords:
IoT · LoRa · IEEE 802.15.4 · Educational buildings · Real-world deployment · LPWAN · Evaluation.
The smart cities and the Internet of Things (IoT) domains are currently amongthe most active research areas, having gradually progressed from being merebuzzwords to having actual large-scale installations deployed and applications (cid:63)
This work has been supported by the EU research project “European Extreme Per-forming Big Data Stacks” (E2Data), funded by the European Commission underH2020 and contract number 780245, and by the “Green Awareness In Action”(GAIA) research project, funded by the European Commission and the EASMEunder H2020 and contract number 696029. This document reflects only the authors’views and the EC and EASME are not responsible for any use that may be made ofthe information it contains.
Preprint version of the paper submitted to 2019European Conference on Ambient Intelligence, 13-15 November 2019,Rome, Italy. AmI 2019. Lecture Notes in Computer Science, vol 11912.Springer, Cham. https://doi.org/10.1007/978-3-030-34255-5 13 a r X i v : . [ c s . N I] F e b developed. In this context, a number of competing wireless networking technolo-gies have surfaced in recent years, aiming to appeal to the communities thatengage within these two domains. Advancements in wireless communicationstechnology have enabled a multitude of different approaches to the trade-offbetween power consumption, communication range and bandwidth, in order toanswer to all the various types of application use-case requirements. In this con-text, recent technologies like LoRaWAN and NB-IoT have surfaced, aiming toclaim a place in the area originally covered by technologies like ZigBee.As part of our research activity, we have developed an IoT platform thatcombines sensing, web-based tools and gamification elements, in order to ad-dress the educational community. Within the context of a research project, itsaim is to increase awareness about energy consumption and sustainability, basedon real-world sensor data produced by the school buildings where students andteachers live and work, while also lead towards behavior change in terms ofenergy efficiency. This real-world IoT deployment developed through the afore-mentioned project provides real-time monitoring of 25 school buildings spreadin 3 European countries.Due to the multi-year development phase of the project, a number of con-ditions, like limited availability of certain networking components and appear-ance on the market of new ones, have led us to follow a heterogeneous approachwith several networking technologies utilized in different buildings of our deploy-ments. During the previous development phases, we have used almost exclusivelyIEEE 802.15.4-based 2.4GHz modules. However, school buildings have certaincharacteristics that in practice lead to less than optimal results in terms of relia-bility and connectivity. For this reason, we decided to shift towards LoRa-basedmodules for our deployments in some specific school buildings. LoRa is also well-suited to application use-cases where devices mostly transmit data to the cloud ora nearby gateway (uplink), versus downlink, which also reflects better in the de-sign of other, higher-level, protocols used in IoT. ZigBee and other technologiesare better suited for use-cases with more symmetric bandwidth requirements.The frequencies used in LoRa aim for longer range, while they also help to pro-vide a higher degree of wall penetration than other protocols, although 802.15.4modules are also available in similar frequencies (i.e., apart from 2.4GHz) buttheir availability is limited and not guaranteed.In this paper, we present a comparison between LoRa and IEEE 802.15.4 asa networking backbone of an IoT deployment inside a number of school buildingsin Europe. We relay our experiences from using both technologies in practice,to develop IoT real-world, reliable and well-performing deployments as a foun-dation for pervasive computing applications. We present an overview of the twotechnologies and how we used them in our use-case, along with an analysis ofthe effect of changing parameters like network density, application data rate anddistance between nodes. Our results indicate that in our use-case and under thedesign constraints that we had, LoRa works in a more reliable manner while alsosatisfying our data rate requirements. xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 3 Regarding recent comparisons between protocols used for low power wide areanetworks (LPWAN) in IoT, [15] and [18] discuss aspects related to LoRa, NB-IoT and ZigBee. All of these technologies are being used especially in smartcity applications, and currently there is a lot of interest in understanding theparameters related to the their performance in the real world. This aspect isdiscussed in [16], where a smart city deployment using LoRa and IEEE 802.15.4is evaluated using mostly simulation methods and limited real-world studies.[11] provides a survey of LoRaWAN for IoT and recent examples of relatedapplications, along with a discussion on its advantages and shortcomings.Although most related works describing aspects like the ones mentionedabove are limited to simulation, there are some recent ones performing measure-ments in real-world settings. LoRa performance is explored to a certain degree in[9], with a discussion on possibilities and limitations. In [8], a performance eval-uation of LoRaWAN and its integration in IoT devices is discussed, while [13]explores its scalability in the context of large-Scale sensor networks. [14] presentsan evaluation of LoRaWAN using a permanent outdoor deployment, while [17]provides an experimental study on LPWANs for mobile IoT applications. [10]provided a study of LoRa in long-range use-cases and produced certain radiopropagation models to be used when designing LoRa-based solutions. Their workconfirmed coverage of up to 8 kilometers in urban and 45 kilometers in urbanareas (in line-of-sight conditions). [19] provided a simulation-based comparativestudy between LoRa and NB-IoT, describing the advantages of each technologyin specific areas and use-cases.However, so far most works are either mostly based on simulation, or they donot attempt a straight apples-to-apples comparison between different networkingtechnologies in specific use-cases e.g., for IoT, pervasive computing or smartcities. Our work here contributes to the discussion over which technology is bettersuited for real-world application in a representative use-case; school buildingare a characteristic and ubiquitous example of public building. Our applicationrequirements in terms of data sampling and quality of service (QoS) are alsosimilar to other related application scenarios (e.g., office building monitoringand automation).
In this section, we present a brief comparison between the IEEE 802.15.4 andLoRa networking, in order to give a context for the sections that follow anddiscuss their performance in more detail.
The IEEE 802.15.4 is a standard for wireless communication. It specifies theuse of Direct Sequence Spread Spectrum (DSSS) and an Offset Quadrature
Phase Shift Keying (O-QPSK). The IEEE 802.15.4 protocol specification in-cludes both a Physical and a MAC layer definition. The physical layer definedthe frequency (possible frequencies are 868
M Hz , 915
M Hz and 2 . GHz ) andthe number of channels. The MAC layer defines the device types (physical ad-dress) and channel access. The 802.15.4 physical layer defines the possibility of 16channels in ISM band from 5
M Hz channel spacing, beginning at 2405
M Hz andending at 2480
M Hz . The carrier-sense multiple access with collision avoidance(CSMA/CA) protocol is implemented as part of the MAC layer by using a CCA(clear channel assessment) technique to determine if the channel is availablebefore to transmitting a packet [2].Moreover, the European Telecommunication Standards Institute (ETSI) reg-ulates the maximum transmitted RF power in wireless networking modules viathe ETSI EN 300 328 standard. Two clauses are the most important: the maxi-mum transmit power, which limits power to 100 mW , and the maximum EIRPspectral density, which is limited to 10 mW/Hz [2]. The ETSI standard sets asafe limit for RF output power around 12 dBm [5]. Furthermore, in the 2 . GHz band, a maximum over-the-air data rate of 250 kbps is specified, but due to theoverhead of the protocol, the actual theoretical maximum data rate is approxi-mately half of that [4].For our network implementation, for the 802.15.4 part we have chosen to useXBee network modules; in the rest of the text, XBee refers to 802.15.4 aspects.We set every XBee module at the 802.15.4 MAC mode with ACKs acknowl-edgment protocol. The RF module operates in a unicast mode that supportsretries. The receiving modules send an ACK of RF packets to confirm receptionto the transmitter. If the transmitting module does not receive the ACK, it willresend the packet up to three times, or until the ACK packet is received. Thetransmission happens directly without any delays. The modules are configuredto operate with a peer-to-peer network topology with no master/slave relation-ship and each module of the network shares both roles master and salve. TheNetwork ID and Channel must be identical across all the modules in the net-work. Each RF packet contains a maximum of 100 characters (100 bytes ). In ournetwork, the payload of the RF packet will be variable but always smaller thanthe 100 character limit, which means all messages are transmitted within onepacket.
Long Range (LoRa) was originally conceived as a long-range wireless communica-tion technology that operates on the sub-GHz license free ISM bands (868
M Hz in Europe and 915
M Hz in the U.S.). This means that, in contrast to other re-lated technologies like NB-IoT, it operates in frequencies that are free to use andanybody can potentially operate a LoRa network without requiring a license forit. Regarding features of LoRa that are examined in this work, the over-the-airLoRa modulation technique can be understood as a MFSK modulation on topof a Chirp Spread Spectrum (CSS) method. Each bit is spread by a chippingfactor, with the number of chips per bit called Spread Factor (SF). Chirps are xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 5 used to encode data in LoRa networks for transmission, while inverse chirps areused on the receiver side for signal decoding. The modulation across the channelis weeping so that the transmission signal occupies the chosen bandwidth (BW).SFs specifically set the data transfer rate relative to the range, by essentially in-dicating how many chirps are used per second, and define bit rates, per symbolradiated power, and achievable range. The possible values of SF are between 6and 12. The data rate depends on the selected SF, e.g., SF SF M Hz to 870
M Hz and is regu-lated for the European zone[12]. The rules are based on two restrictions: a) themaximum power transmission that can be used on a channel at the communi-cation is 25 mW (equivalent of 14 dB ); b) the duty cycle that is defined as theratio of maximum time-on-air (ToA) per hour and is limited to 1%, which inpractice restricts the communication of each LoRa device with other nodes to36 seconds per hour. The MAC layer of LoRa does not implement any listen-before-talk (LBT) or CSMA to avoid collisions. Instead it implements a pureAloha protocol, sending data whenever available, thus the number of collisionsincreases together with transmission rate or network node density. Overall, the deployed devices provide 1250 sensing points organized in four cate-gories: (1) classroom environmental sensors; (2) atmospheric sensors (outdoors);(3) weather stations (on rooftops); and (4) power consumption meters (attachedto electricity distribution panels). Given the diverse building characteristics andusage requirements, deployments vary between schools (e.g., number of sensors,manufacturer, networking, etc.). The IoT devices (Fig. 1) used are either open-design IoT nodes, or off-the-shelf products from IoT device manufacturers. In-door devices use IEEE 802.15.4 or LoRa wireless networks. These devices areconnected to cloud services via IoT gateway devices, which coordinate commu-nication with the rest of the platform, while outdoor nodes use wired networkingor WiFi.
The IEEE 802.15.4 communication between the IoT nodes is provided by XBeemodules connected to each IoT node operated by the Arduino XBee [1] andXBeeRadio [6] software libraries. Node-to-node communication includes the check-sum of the payload, which is validated at the network level for each node todetermine erroneous or invalid messages, which are discarded.All IoT nodes form ad-hoc networks and report their measurements throughthe designated IoT gateways. Because IEEE 802.15.4 is a short-range commu-nication technology and end-to-end communication is not possible due to power
Fig. 1.
Examples of the IoT infrastructure located inside school buildings in Greece(a-b) IoT nodes based on Arduino and Raspberry Pi, c) actual node inside a classroom,d) power meter installed on a electricity distribution panel. on the right part of thefigure, the latest hardware revision of the actual environmental nodes used in our IoTinfrastructure, utilizing a LoRa communication module. limitations and propagation obstacles, all indoor IoT nodes form an ad-hoc over-laying multi-hop bidirectional tree network. The gateway is the root of this treeand the orchestrator of the network. New nodes can join the network at anytime either directly below the gateway, or as a child of the node that is closer tothe gateway and has a received signal strength indication (RSSI) lower than aspecific threshold ( in our case 90 db ). The resulting routing tree allows for bidi-rectional communication between the IoT nodes and the gateway. The routinglibrary developed for the Arduino and XBee devices is also available on GitHub.An example of a formed network can be seen in Fig. 2(a). Once the networkhas been established, each node collects environmental or other sensor data andemits a data packer (e.g., an Environmental Data Packet (EDP) ) to the GWevery 10 seconds. The payload size varies depending on the sensing activity, butin our case it is always lower than the limit of a 100 character payload to fit ina single packet. In addition, each environmental node checks its motion (PIR)sensor every 2 seconds and emits a
PIR Data Packet (PDP) independently eachtime motion is detected.
Our IoT nodes based on LoRa use a single-hop topology to cover the neces-sary distance (tested with up to 3-floor concrete-built buildings), thanks to thecommunication range and signal penetration characteristics. An example of aformed network showing the difference with the IEEE 802.15.4 network can beseen in Fig. 2(b). A network, in which the IoT nodes communicate directly with xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 7
Fig. 2.
Examples of data collection routes in an IEEE 802.15.4 (left) and a LoRa (right)IoT deployment, using a tree/multihop and a star network topology, respectively. the IoT GW has been installed inside multiple school buildings. We use LoRa tobuild our own wireless LoRa Personal Area Network (PAN) with a star topology.The IoT nodes communicate using the Grove LoRa 868
M Hz [4] modules witha LG01-N Single Channel LoRa Dragino gateway [3], suitable for small-scaleLoRa networks. The communication device for both the Dragino and the Grovemodules is based on the RF95 SX1276 LoRa module [7].The GW coordinates the communication with every node to guarantee thatthe nodes do not occupy the medium at the same time. This implementation isnecessary to avoid the interference due to ALOHA MAC protocol and guaranteeno interference between the nodes at the same network. The network is createdby the GW announcing itself through broadcast messages. The new nodes replyto the broadcast with a connection request which, if it is accepted by the GW, isacknowledged by a confirmation message. The ALOHA protocol with a random-ized delay is used by the nodes to answer to the GW network announcement.Once the network is setup, the GW requests data periodically from each nodein a sequential fashion with a
Data Request Packet (DRP) . The nodes reply witha a
Data Packet (DP) consisting of the sensor measurements. The request rateof the GW is configurable to adjust to the requirements of ToA EU regulations.In this case, the node samples the PIR sensor between GW requests and includesthe motion sensing information in the DP to avoid creating overhead. If a replyis not received or the reply is corrupted, the GW can repeat the DRP up tothree times for each node. We implemented our own
CRC (Cycle RedundancyCheck) method at network level to detect message corruption instead of usingthe LoRa module functionality at MAC level. The network is refreshed every 15minutes. On each refresh, new nodes can be attached while unreachable nodesare not removed to speedup future reconnects.
The maximum
ToA is restricted in Europe, thus limiting the packet rate foreach network device. The ToA of each LoRa packet depends on the spreadingfactor (SF), coding rate (CR), signal bandwidth (BW) and the packet payload (PL). The LoRa packet duration is the sum of the duration of the preamble andthe transmitted packet. The data-sheet of the SX1276 module [7] describes theformula to calculate the number of payload symbols and the preamble length.We use this to determine the ToA of each packet of our network in milliseconds,and thus we can calculate the maximum legal packet rate to accommodate the36 seconds ToA per node limit under different network configurations. The max-imum PL of each DP is 60 bytes and the PL of each DRP is fixed at 4 bytes . Acomparison for each type of packet between different network configuration isdescribed in Table 1 showing the corresponding PL , ToA for a single packet inmilliseconds and the minimum
Period between transmissions in seconds.
Table 1.
Packet ToA and Period per deviceNode (Data Packets) Gateway (Data Request Packets)SF BW PL
Packet ToA Min. Period PL Packet ToA Min. Period [kHz] [Bytes] [ms] [s] [Bytes] [ms] [s]7 125 60 112.896 11.289 4 30.976 3.0977 250 60 56.448 5.644 4 15.488 1.5487 500 60 25.088 2.8224 4 7.744 0.7749 125 60 319.488 36.966 4 123.904 12.3909 250 60 159.744 18.483 4 61.952 6.1959 500 60 79.872 9.241 4 30.976 3.097
As described, the GW requests data from each node periodically by sending a
DRP with a 4 bytes payload. As a consequence, in our case the maximum packetrate (minimum period between
DPs ) per node is limited by the maximum packetrate (minimum period between
DRPs ) of the GW, which depends on the totalnumber of nodes in the network. Table 2 presents the theoretical minimum period per node and the maximum packets per 15 minutes in our network as influencedby the restrictions of the
ToA of the GW under different network configurationsfor two schools (
LoRa School A , LoRa School B ) of our installation.The maximum packet rate per node is achieved with SF BW kHz ,which we implemented in our final network installation due to our priority ofmaximizing the sensing rate in the school and achieving a better sampling ofthe environmental reality in the public buildings. It is noteworthy that higherspreading factors allows for longer range at the expense of lower data rate, andvice versa.We aim to compare the quality of the network under two extreme configura-tions. Configuration A provides higher rate (SF 7, BW 500 kHz ) and Configura-tion B provides longer range (SF 9, BW 125 kHz ). In order to study the network xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 9 Table 2.
Theoretical minimum Period and maximum Packets per 15 min. n representsthe number of nodes in the networkn Nodes LoRa School A (6 nodes) LoRa School B (7 nodes)SF BW Min. Period Max. Packets Min. Period Max. Packets Min. Period Max. Packets[kHz] [s] [ n n n n n n n n n n n n behaviour, we collected the following measurements per node: the number of DRPs from the GW, the number of received
DPs , and the number of packetsreceived with
CRC errors over a period of 15 minutes.The maximum number of packets received under Configuration 1 is limited bythe
ToA imposed by communications regulations. Thus we have to set the GWrequest rate accordingly to implement this restriction. As such, Configuration 1 islimited to 12 packets, per 15 minutes, per node. On the other hand, Configuration2 is limited by the regulation at 193.69 (Table 2) packets, per 15 minutes, pernode. Effectively, Configuration 2 is restricted by the node design constraints.The GW requests data from each node after a 50 ms delay to guarantee thecorrect communication between the LoRa module and the micro-controller. Inaddition, the Environmental Nodes consume time to communicate through I2Cwith their digital sensors to collect the data for each request, limiting the finalrate of the node. Due to these factors, in School Building A’s installation everynode can achieve a maximum of 174 packets (Table 3) per 15 minutes, which islower but close to the theoretical limit. On both configurations, the average ofthe delivered DP rate (Table 3) is higher for the nodes near the GW (Nodes 1,5 and 6). Table 3.
Number of delivered
DPs per node under different configurations in a 15minute period node1 node2 node3 node4 node5 node6
Configuration 1
Avg. 11.17 11.17 11.15 11.16 11.16 11.16SF 9, BW 125 kHz Min. 5.00 5.00 0.00 0.00 5.00 5.00Max. 12.00 12.00 12.00 12.00 12.00 12.00
Configuration 2
Avg. 155.24 149.77 145.79 150.87 155.23 155.22SF 7, BW 500 kHz Min. 77.00 0.00 0.00 1.00 77.00 77.00Max. 174.00 174.00 174.00 174.00 174.00 174.000
As an indicator of the quality of the network, we define
CRC Error Ratio and
Re-transmission Ratio . CRC Error Ratio is the ratio of
DRPs from the GWwhich resulted in a corrupted DP being received. Re-transmission Ratio is theratio of
DRPs required to be repeated, either because of CRC errors, malformed
DRP or due to not receiving a reply. We are also interested in the connectivitybetween the GW and every node in our LoRa network. We quantify the qualityof each link by calculating the
Packet Delivery Ratio (PDR) for every node. ThePDR of the link between node A and GW can be measured as the ratio betweenthe number of
DPs received by the GW from node A, and the number of
DRPs sent from the GW to the node A. The GW makes one
DRP and a maximum of3 re-transmissions of the
DRP per node. In addition, we study the variation ofthe
Received Signal Strength Indicator (RSSI) per node in the network for bothconfigurations.
Table 4.
Packet Delivery Ratio (PDR) per node under different configurationsnode1 node2 node3 node4 node5 node6
Configuration 1
Avg. 0.96 0.96 0.96 0.95 0.95 0.93SF 9, BW 125 kHz SD 0.03 0.03 0.04 0.05 0.03 0.03Min. 0.72 0.80 0.00 0.00 0.86 0.80Max. 1.00 1.00 1.00 1.00 1.00 1.00
Configuration 2
Avg. 0.99 0.93 0.86 0.89 0.98 0.97SF 7, BW 500 kHz SD 0.01 0.22 0.27 0.19 0.01 0.02Min. 0.91 0.00 0.00 0.00 0.93 0.88Max. 1.00 1.00 1.00 1.00 1.00 1.00
Table 5.
Received Signal Strength Indicator (RSSI) per node under different configu-rations node1 node2 node3 node4 node5 node6
Configuration 1
Avg. -45.06 -50.55 -87.78 -87.08 -53.52 -52.40Sf 9, BW 125 kHz SD 0.41 0.98 2.59 1.73 1.35 1.29Min. -46.13 -54.25 -95.47 -95.29 -59.40 -55.77Max. -43.74 -48.75 -82.12 -84.12 -51.73 -50.17
Configuration 2
Avg. -42.04 -46.29 -82.23 -86.86 -49.88 -44.78Sf 7, BW 500 kHz SD 3.17 10.84 16.92 4.66 1.68 3.25Min. -55.33 -56.98 -89.00 -88.44 -57.70 -60.79Max. -37.40 0.00 0.00 0.00 -48.17 -41.27
We expected to observe a worse quality network under Configuration 2, asa consequence of selecting parameter values that achieve a higher packet rate.We can observe that the median value for the
CRC Error Ratio distribution xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 11
Fig. 3.
Per node statistics for configurations SF 7, BW 500kHz and SF 9, BW125kHz . of each node is higher with similar standard deviation with the exception ofthe closest node (Node 1, CRC Error Ratio graph in Fig. 3). The degradation ofnetwork quality is evident at the farthest nodes (3 and 4) from the GW regardingthe number of Re-transmissions (Re-transmission Ratio graph in Fig. 3), whichexhibits higher standard deviation and more frequent and distant upper outliers.In addition, the PDR (Packet Delivery Ratio in Fig. 3) of these nodes is worsethan in Configuration 1, with higher standard deviation and more frequent anddistant lower outliers. On the other hand, the nodes closest to the GW achievebetter network behavior regarding Packet Delivery and Re-transmission Ratios.Furthermore, the RSSI distribution (RSSI in Figure 3) exhibits greater standarddeviation, entailing less stable signal strength.In conclusion, we observed the expected cost in network quality, only inthe further nodes, while in nearby nodes we observed an increase in the link’sefficiency. This combined with the increase in the per node packet rate, resultedin a significant increase in sensor measurements across the whole network. Every node in the XBee network tries to deliver an
Environmental Data Packet(EDP) to the GW every 10 seconds while emitting an extra
PIR Data Packet(PDP) each time motion is detected. The network overhead due to the extra PDPs can saturate the medium and provoke a decrease of
EDPs delivered pernode, thus decreasing the EDP rate. The data-set considered to analyze thespecific behaviour in XBee School C is composed of the total number of packetsdelivered in the network in 5 minute periods (
EDPs and
PDPs from every nodes).To quantify the effects of the independent
PDPs , we use their ratio against theobserved maximum of the aggregation.Fig. 4 shows clearly how during school hours the
PDPs can cause an observ-able decrease in the number of
EDPs , due to the saturation of the network atpeak of
PDP
Ratio. The maximum number of
EDPs is observed when the num-ber of
PDPs is zero (Table 6). In addition, when the number of
PDPs exhibits amaximum, the number of
EDPs decreases below their average. The decision toinclude real-time motion detection to the network, can potentially be a hindranceto the stability of the
EDP rate of our network.
Table 6.
XBee Network behaviour. Number of
Aggregated Packets , PDPs and
EDPs delivered at the time of maximum and minimum
Aggregated Packet Ratio , EDP Ratio and
PDP Ratio respectively.Average Aggregated Packets PIR Packets Env. PacketsMax Min Max Min Max MinAggregated Packets [
We aim to compare the quality of our LoRa and IEEE 802.15.4 IoT networksby comparing our observations from two real school buildings. LoRa School Aconsists of a LoRa Network with 6 nodes where the Node 3 and 4 are located atthe farthest positions and the Node 1 and 5 at the nearest in relation with theGW. XBee School C has an XBee network consisting of 6 nodes, the farthestis node 6 and the nearest is the node 5. Due to the significant differences inradio and network architectures, to compare them we quantify the quality ofthese networks by the
Network Delivery Ratio (NDR) . The
NDR is defined asthe ratio between the measured Delivered Packets and the potential maximumnumber of Packets that could be delivered by each node in the network in atime period, which in our case is 15 minutes. Every node in the XBee networkis scheduled to attempt to send data to the GW every 10 seconds, resulting ina maximum of 90 packets in a 15 minute period. In the LoRa network usingConfiguration 2 and with 6 nodes, the network can achieve the delivery of amaximum of 174 packets. The data-set consists of
NDR measurements collectedduring a period of both business and weekend days for LoRa School A and XBeeSchool C can be seen in Table 7 and Figure 5. xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 13
Fig. 4.
Aggregated Packet Ratio , EDP Ratio and
PDP Ratio in XBee School C over 48hours (Sunday and Monday)
We observe that the best network quality in terms of
NDR is observed inthe LoRa School A under Configuration 1 which is the one with lowest ratewhere the maximum number of Data Packets per node in a 15 minute period is12 packets. The network in XBee School C exhibits a better
NDR than LoRaSchool A under Configuration 2 regarding the averages for every node with theexception of the farthest one that achieve a delivery of 20% of the generatedpackets. On the other hand, the network in LoRa School A under Configuration2 achieves a more stable
NDR across the installation, including the farthestnodes, with successful packet deliveries between 86% and 89% for every nodeand a significantly higher delivery rate.The tree topology necessary for the XBee network to achieve comparablerange to LoRa, influences negatively the packet rate of the nodes placed at theextremes of the tree. In comparison, LoRa’s star network topology offered bettercoverage with a more stable data rate on all nodes.
Our work in recent years has resulted in the deployment of a large-scale IoTinfrastructure inside a number of school buildings in Europe. In this context,we have opted to use different wireless networking technologies in order to testin practice their performance. With this work, we wanted to relay our practicalexperiences from using both IEEE 802.15.4 and LoRa for our specific application Table 7.
Network Delivery Ratio (NDR) per node for LoRa and IEEE 802.15.4 Net-works in different school buildingsNetwrok node1 node2 node3 node4 node5 node6
LoRa School A
Avg. 0.94 0.94 0.94 0.94 0.94 0.94Conf. 1 SD 0.03 0.03 0.04 0.04 0.03 0.03SF 9, BW 125 Min. 0.41 0.41 0.00 0.00 0.41 0.41Max. 1.00 1.00 1.00 1.00 1.00 1.00
LoRa School A
Avg. 0.89 0.86 0.84 0.87 0.89 0.89Conf. 2 SD 0.12 0.22 0.25 0.17 0.12 0.12SF 7, BW 500 Min. 0.44 0.00 0.00 0.01 0.44 0.44Max. 1.00 1.00 1.00 1.00 1.00 1.00
XBee School C
Avg. 0.85 0.91 0.92 0.92 0.92 0.20SD 0.04 0.05 0.05 0.04 0.05 0.05Min. 0.28 0.32 0.31 0.31 0.32 0.09Max. 0.91 0.98 0.98 0.98 0.99 0.38
Fig. 5.
Network Delivery Ratio (NDR) per node in LoRa School A (left) under Con-figuration 2 and XBee School C (right)xperiences from Using LoRa & IEEE 802.15.4 for IoT-enabled Classrooms 15 use-case and provide some practical examples and guidelines for IoT deploymentsthat are similar to ours.We have studied the behavior of both networks, in the scenario of changingthe number of nodes in the network, varying the sampling rate of the sensors andthe required data rate, or changing the distance between the IoT nodes insidethe building. As an example of the results from the comparisons we made, LoRadecreases its delivery rate when increasing the number of nodes because of ToAEuropean regulations which restricts the number of GW data requests in ournetwork design. In 802.15.4 we expect an increased number of collisions whenadding nodes due to CSMA. In the case of increasing the distance between nodes,LoRa achieves longer range with a stable rate, while 802.15.4 will need hop nodesin the middle, leading to increased number of collisions and an unstable rate inextreme nodes as a side effect.Overall, our results show that in the use-case scenario and environmentalsettings of school buildings in Greece, LoRa-based wireless communication canhave an advantage against competing technologies, in terms of reliability andcomplexity of networking. Regarding our future work, we plan to conduct a morethorough performance evaluation and explore in additional dimensions practicalaspects like networking performance and reliability.
References PP , 1–1 (03 2019). https://doi.org/10.1109/JIOT.2019.2906838611. Haxhibeqiri, J., De Poorter, E., Moerman, I., Hoebeke, J.: A survey oflorawan for iot: From technology to application. Sensors (11) (2018).https://doi.org/10.3390/s1811399512. Lavric, A., Popa, V.: A lorawan: Long range wide area networks study. 2017 In-ternational Conference on Electromechanical and Power Systems (SIELMEN) pp.417–420 (2017)13. Lavric, A., Popa, V.: Performance evaluation of lorawan communication scalabil-ity in large-scale wireless sensor networks. Wireless Communications and MobileComputing , 1–9 (06 2018). https://doi.org/10.1155/2018/673071914. Marais, J., Malekian, R., Abu-Mahfouz, A.: Evaluating the lorawan protocol us-ing a permanent outdoor testbed. IEEE Sensors Journal PP , 1–1 (02 2019).https://doi.org/10.1109/JSEN.2019.290073515. Mekki, K., Bajic, E., Chaxel, F., Meyer, F.: A comparative study of lp-wan technologies for large-scale iot deployment. ICT Express (1), 1 – 7(2019). https://doi.org/https://doi.org/10.1016/j.icte.2017.12.005,
16. Pasolini, G., Buratti, C., Feltrin, L., Zabini, F., De Castro, C., Verdone, R., An-drisano, O.: Smart city pilot projects using lora and ieee802.15.4 technologies.Sensors , 1118 (04 2018). https://doi.org/10.3390/s1804111817. Patel, D., Won, M.: Experimental study on low power wide area net-works (lpwan) for mobile internet of things. In: 2017 IEEE 85th Ve-hicular Technology Conference (VTC Spring). pp. 1–5 (June 2017).https://doi.org/10.1109/VTCSpring.2017.810850118. Sinha, R.S., Wei, Y., Hwang, S.H.: A survey on lpwa tech-nology: Lora and nb-iot. ICT Express (1), 14 – 21 (2017).https://doi.org/https://doi.org/10.1016/j.icte.2017.03.004,
19. Sinha, R.S., Wei, Y., Hwang, S.H.: A survey on lpwa tech-nology: Lora and nb-iot. ICT Express (1), 14 – 21 (2017).https://doi.org/https://doi.org/10.1016/j.icte.2017.03.004,(1), 14 – 21 (2017).https://doi.org/https://doi.org/10.1016/j.icte.2017.03.004,