Multihop Uplink Communication Approach Based on Layer Clustering in LoRa Networks for Emerging IoT Applications

,


Introduction
Te Internet of things (IoT) is experiencing considerable scientifc expansion in several areas, facilitating the exchange of information between objects, hardware, and software resources [1][2][3][4][5]. Technological advancements in the feld of IoT have enabled the development of a wide variety of applications such as precision agriculture (PA), smart city, asset tracking, health care, and many more [6,7]. Most of these applications require long-range communications with characteristics such as low data rate, low deployment cost, low power consumption, and security management [8][9][10][11]. Tus, short range radio technologies are not very suitable for wide coverage applications. A wide range of data acquisition objects already participate in a broad spectrum of applications that signifcantly consider long-range communication with low power consumption to extend network life without human intervention [11]. Tus, these growing needs have led to the emergence of Low Power Wide Area Networks (LPWANs). LPWAN is a type of wireless communication that is positioned as a connectivity facilitator for IoT applications or for cellular networks. Indeed, cellular solutions have the advantage of ofering long connectivity but are energy intensive [11,12]. LPWAN technology has experienced considerable growth because it has a set of interesting features among which we can mention scalable deployment, high energy efciency, low management, and operating costs. Indeed, LPWAN technology makes it possible to build networks made up of thousands of objects with a long range (several kilometers) [13]. Figure 1 [11] presents an architecture of LPWAN technology that facilitates the increase in the number of IoT applications. You can see a panoramic view of the functionalities at diferent layers as an added value for many applications. In the literature, there are several licensed and unlicensed LPWAN technologies, and among them, the most used solutions are LoRaWAN, Sigfox, and NB-IoT [11,14].
LoRaWAN is an LPWAN-based communication solution that was designed to take advantage of a long-range, low-cost, and low-power hub-and-spoke network [15]. A LoRaWAN network consists of three essential components as follows: the terminals, the gateway, and then a link to the server, as shown in Figure 2. In addition, security (endto-end encryption in the various communications) is an important element taken into account in LoRaWAN solutions. Such a LoRaWAN network consists of objects equipped with sensors or actuators that use the LoRa physical layer to exchange messages with the [16] gateway.
Remote health care and remote patient monitoring are concepts that have captured the attention of researchers in recent years [7]. Indeed, a smart home equipped with several objects capable of remotely monitoring one or more patients appears to be an interesting solution nowadays. Specifcally, such technology can be an alternative for monitoring disabled people, quarantined people, and people with chronic illnesses. IoT provides an environment of objects connected to cloud-based applications and services, with diferent cooperation mechanisms, appropriate standardization, and advanced sensors with low-cost and low-power microprocessors [17,18]. According to Figure 3, LoRaWAN is considered as one of the best IoT solutions based on the healthcare monitoring system due to its wide communication range and perfect interoperability between IoT sensors.
In this work, we have explained a multihop communication protocol over a large-scale area with LoRa devices while ensuring a deterministic and intelligent choice of the gateway which must redirect the information submitted by a device to guarantee, for example, a real-time information treatment by the server while minimizing the amount of energy.
Tis protocol takes place in two phases. A frst phase consists of subdividing the networks into diferent layers and a second phase during which each device chooses the relay gateway for uplink communications.
We proposed a layered multihop clustering method for structuring large-scale networks. For this purpose, we supposed that the communications between two-end devices are reliable. Te proposed algorithm tries to minimize the energy consumption spent during the formation of the layers. We obtained an acceptable packet delivery rate, and the number of messages sent is very low. Moreover, the proposed protocol is energy efcient extending by the way of IoT device battery life time, and it has a linear time complexity which is an asset for real-time IoT applications.
Te rest of this paper is organized as follows: Section 2 presents a state of the art on some LoRa protocols, a mathematical formulation of the problem is presented in Section 3, our approach is presented in Section 4, the results of the experiments are presented in Section 5, and fnally, the conclusion and some future works are presented in Section 6.  [21][22][23]. Tis technique allows the reception of a multiple number of messages on channels, an orthogonal separation between the signals. Tis ofers an advantage in the management of the fow.

Review of the Literature
A characteristic of LoRa is that it ofers an improvement of criteria such as SF (spreading factor), TP (transmission power), CR (coding rate), and BW (bandwidth) with a relation given by equation (1) [23][24][25].
Tis technology is based on the ISM (industrial, scientifc, and medical) bands, whose distribution of frequencies and regulations vary according to the region of the world.  a given frequency domain and increases mean bandwidth with the aim of increasing resistance to interference, reducing energy consumption, and integrating an errorcorrecting code. Tis type of modulation, widely used for radar applications in the past, is now necessary for low-speed communications. During a transmission, the spreading factor is adjustable and imposes a preliminary search to optimize the value according to various criteria. Indeed, starting from a fxed pass-band, a high SF directly implies an increase in the range and in the transmission delay T s of a symbol expressed in seconds. Te latter can be calculated using Formula (2). On the other hand, still in the case of a high SF, the communicating system sees its bit rate decreasing but has better reception sensitivity. Formula (1) presents the relationship between the fnal fow rate and the spreading factor [26].
In LoRa, there are several types of equipment among which we can mention the following: terminals, gateways, servers, application servers, and Join Server; a layout of a LoRa architecture is given by Figure 4.
LoRaWAN is a robust solution competing with other LPWANs by taking up the strengths of LoRa technology while having a strong infuence on battery life, network load capacity, quality of service, and security. Te LoRaWAN network is organized according to a star of stars topology. Te communication proposed in this network is bidirectional while clearly favoring uplink transmissions towards the concentrators [15]. Tis organization, illustrated in Figure 4, is in fact composed of a multitude of concentrators which relay the information received from the nodes to a central server using a GSM or Ethernet protocol most of the time. Te central server makes it possible to eliminate duplicates received by the various concentrators and manages the fow rates of the nodes in order to optimize the  Mobile Information Systems capacity of the network and to extend the autonomy of the wireless devices.

Multihop Communications in LoRa Networks.
In [22], the authors propose an algorithm which addresses three main criteria which are load balance in the network, reduction of the number of hops between the root, and a terminal and connectivity problem, using a faster throughput (higher SF) for connections for insertion of each node away from the root with the Topdown Breadth First-Search (TBFS) algorithm. Tis protocol uses long range communications for subtree formations, which requires a higher spreading factor and therefore additional energy consumption.
In [24], the author proposes new multihop clustering algorithm LoRaWAN networks. Tis approach is divided into two steps. Te frst one allows to form the layers, and the second one allows to choose the gateway for the communications to the root. For the layer formation, the protocol modifes the initial structure by adding elements of layer identifcations by the gateways. Te process is managed by the root which sends a broadcast with a hop count of 0 for the identifcation of the frst layer. Te answers of the elements of this layer allow him to make another difusion for the identifcation of the following layer by passing by layer 0. Layer 1 sends an answer, and the process proceeds thus until obtaining all the layers. Te selection is made from the RSSI of each gateway except for this one at a layer lower than a node. It has been compared to the method proposed in [28]. It also proposes an approach of multihop communication in LoRa and has LoRaWAN. Tis method thus allows a good communication between terminals over a long distance. However, the layer formation mechanism is very tedious when the number of layers is very high and would consume a lot of time and energy.
In [29], a protocol that maintains connectivity and coverage with the network using multihop communications is proposed. Evaluations are made on parameters such as the spreading factor and the packet reception rate. Tis protocol shows how in a smart city, the use of LoRa technology with multihop communications can save energy, ensure network coverage, and good connectivity. However, the use of the spread spectrum factor is still very high. Tis increases the energy consumption of the nodes.
In [30], it is proposed to use a simple relay device to increase the LoRaWAN coverage area for rural areas. Te authors suggested deploying relay nodes by knowing the locations that are not covered by the gateway. Te authors proposed a simple message forwarder and a synchronization mechanism. Tey showed that the energy consumption decreases with the addition of the relay node to deliver the packets. Tis method saves energy by minimizing retransmissions and increases the network connectivity rate. However, this protocol is only valid for a maximum of 2 hops. Tis does not allow for scalability.
In [31], the authors propose a transmission protocol (concurrent transmission) which proceeds by fooding the network for the design of multihop communications. Te transmission of packets is done by transmitting identical packets in a synchronous manner, which increases the effciency of the network by solving the problem of packet collisions. Although this protocol improves the efciency of the multihop network and is robust against collisions, it results in a huge energy consumption due to the numerous identical messages broadcasted through the network.

Problem Formulation
We consider gw 1 , gw 2 , . . . , gw g the set of gateways in the network, DV � dv 1 , dv 2 , . . . , dv d the set of LoRa terminals in the network that need to be dispersed in layers, and C � c 1 , c 2 , . . . , c c the set of formed layers in the network. We seek to improve the operation of large-scale LPWANs by considering the multiobjective problem (energy-efcient formation of layers and intelligent selection of the best gateway). Te selection of the best gateway is given by où.
(1) dv i is the set of gateways determined as follows: (2) f is the objective function of each terminal in relation to a gateway.
Tis problem is formulated subject to the constraints as follows: (1) Let alpha i,j be a binary value which indicates the admission of the device dv i to the gateway g j .

Multihop Uplink Communication Approach
In this section, we present an efcient approach to multiobjective routing in a LoRa network based on layer clustering. Tis approach is divided into two steps. Te frst one is to form layers, and the second one is to select the best gateway for routing data to the root.

Assumptions
(1) Te nodes are randomly deployed on a square surface (2) Te transmission radio is the same for each equipment (3) Te gateways are deployed in a deterministic way to ensure the coverage of the deployment area 4 Mobile Information Systems

Layer Formation.
Each gateway has the following parameters for layer formation: (1) id represents the identifer of the gateway (2) layer id is the layer number to which the gateway belongs (3) gwt contains the id of its gateway to reach the root gateway We defne three types of messages for layer formation. A DIL (Discover Information Layer) and DIA (Discover Information Acceptance) which contain the felds L ID for the layer number, H NT is the number of hops from the root gateway, T NID is the identifer of the node transmitting a DIA or DIL message, and fnally, a gatewaylayer id contains the identifer of the relaying gateway towards the root. A DIN (Direct Information Notifcation) message contains the felds ID for the identifer of the transmitting gateway, gatewaylayer id contains the identifer of the relay gateway to the root, and layer contains the number of layers of the formed network.
Te root gateway initiates the broadcast of a DIA message with the LID feld set to 0 to indicate the start of the layering process. As soon as a DIS message is received, each gateway initiates a DIA message to indicate that the message has been received and initiates a DIS message again, this time with the LID number incremented by 1 and the hop number set to 1. It then sets a timer t and waits until the end of this time. If it has received another DIL message, it generates a DIL message and transmits it to its relay gateway.
Otherwise, it generates a DIL message and forwards it to its relay gateway and thus considers itself as the leaf node of the layered tree. Te process repeats itself until layer n − 1 receives DILs from layer n, transmits DILs to layer n − 2, and so on until the DILs are distributed to the root. Tis will allow the root gateway to know the number of layers formed in the network. Algorithm 1 is executed by all the gateways for the formation of the network layers.
An illustration of this approach is presented in Figure 5. Te root gateway in the center of the fgure (purple color) illustrates the formation of the layers. Te other gateways in green, red, and blue colors are Layer 1, Layer 2, and Layer 3 gateways, respectively.

Gateway Selection.
Tis step consists of intelligently choosing a gateway in its layer to be able to route information to the root gateway. Once the layer formation is complete, each gateway will issue a broadcast message to tell each device which gateways in its layer, the upper and/or perhaps lower layer, where it can access. Each node will calculate its objective function to determine which gateway should be chosen to minimize energy consumption costs. Algorithm 2 is executed by all the end devices for the selected gateway.
f � w 1 × RSSI + w 2 × SF + w 3 × Layer id, where RSSI (Received Signal Strength Indication) is the received signal strength, SF (spreading factor) and Layer id is the layer number of the gateway, and w 1 + w 2 + w 3 � 1.

Mobile Information Systems
For each gateway message received, the device evaluates the objective function f. It will therefore choose among the gateways in its layer, the gateway with the largest value of f.

Results
Te simulation is conducted on an i5 series processor with 4 GB of RAM using LoRaSim. LoRaSim is a discrete-event simulator based on SimPy for simulating collisions in LoRa networks and to analyse scalability. Teir simulator and the model energy are described in [32].
Te simulation parameters are listed in Figure 6. Figure 7 shows a comparison of the energy consumed by the RPL, Farooq, and LoRaWan protocols. We can observe that our protocol outperforms those of the others by consuming 50%, 70%, and 80% less than the Farooq protocols [24,28] and LoRaWan.
We also bring an improvement in the packet delivery rate of 2% compared to [24] in Figure 8. Figure 9 shows the number of messages sent for the layer formation process in the network. We see that compared to the Farooq protocol, the number of messages sent remains constant regardless of the number of layers formed in the network. Tis proves that the root gateway sends the same number of messages for the formation of the layers whatever the size of the network. So, it also proves that this protocol is suitable for large-scale networks. Figure 10 shows the number of messages sent for the layer formation process in the network. We see that compared to the Farooq protocol, the number of messages sent remains constant regardless of the number of layers formed in the network. Tis proves that the intermediate gateways send the same number of messages for the formation of the layers whatever the size of the network. Terefore, the     4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Our solution Farooq

Conclusion
In this article, we have proposed a low power system with LoRa technology. Tis system can be used to monitor the physiological parameters of a patient to determine his medical situation, in order to anticipate the aggravation of pathologies for patients and to reduce the time of hospitalization and cost, especially with the spread of the Covid pandemic. Our results showed very good efciency in terms of system life, with autonomy gains multiplied by 2. Work is underway to improve the study surface.

Data Availability
No underlying data was collected or produced in this study.

Ethical Approval
Tis article does not contain any studies with human participants or animals performed by any of the authors.

Consent
Informed consent was obtained from all individual participants included in the study.