In a system where several IoT devices are connected to each other to exchange data with each other and with the remote server, different communication, application and security protocols are used. In this exchange between objects, the connected devices do not necessarily have the same characteristics in terms of communication capacity, memory, energy, etc. This is known as heterogeneity. This is known as IoT device heterogeneity.

Figure 1 illustrates a solution to heterogeneity and the use of the gateway that enables interconnection between IoT sensors and the Internet.

The algorithm, as defined in [12] , plays a key role in the gateway. It is thanks to the algorithm that the gateway becomes intelligent in order to resolve problems of heterogeneity and even range.

The solutions for interconnecting connected objects are based on the algorithms implemented in the gateways [13] .

An IoT gateway establishes a bridge between IoT devices (sensors) and the cloud to enable data to be exchanged. It systematically connects “the field”, made up mainly of sensors, with the remote server or the cloud. However, the heterogeneous nature of sensors and networks means that gateways have to support different types of interfaces (e.g. Power Line Communication (PLC), Bluetooth, cellular network or WiFi) and network protocols [1] . Depending on the nature of the sensors or networks, the configuration of an IoT gateway and its functions must also change to ensure interconnection.

The algorithms used have several functions: optimum management of congestion in the network, optimum management of data and communication security, optimum management of application memory resources, etc. So, IoT gateways perform well if and only if they use appropriate algorithms. We present some of the algorithms implemented in gateways.

1) Network selection algorithm for heterogeneous wireless networks

In [14] , an efficient mechanism for network selection over heterogeneous wireless networks is proposed. In this work, the authors put special emphasis on user preferences towards the use of heterogeneous wireless networks in order to improve system performance and support reliable connectivity. User preferences are summarised in three parameters: latency, cost and energy consumption. In this system, the user is at the centre of the gateway’s decision-making processes.

2) SIA (Service-to-Interface Assignment) algorithm

In [15] , Abdellatif et al. presented a mechanism called SIA (Service-to-Interface Assignment). As its name suggests, this is a mechanism for assigning one or more services to a given interface. In this work, they considered an IoT device with multiple interfaces [16] to highlight the problem of efficiently allocating resources in a network. Since an IoT device is called upon to provide a wide variety of services with different requirements, the mechanism proposed by Vangelis Angelakis et al. uses two different algorithms to solve the interface assignment problem in an optimal way. The first algorithm takes into account the most demanding services, and then immediately allocates interfaces to them. The second algorithm first calculates the shares of existing resources and then allocates them to services by random selection.

3) ARTPoS algorithm (Adaptive Radio and Transmission Power Selection system)

In [17] , Mu et al. proposed an adaptive radio and transmission power selection algorithm that makes several wireless technologies available. This system supports several wireless protocols, such as WiFi, Zigbee and Bluetooth. The mechanism consists of selecting the appropriate network for the traffic and network conditions while minimising energy consumption. It is known as ARTPoS (Adaptive Radio and Transmission Power Selection system). In its architecture, ARTPoS incorporates a selection module that chooses the most optimal radio interface. This choice is made on the basis of bit rate and residual network capacity. ARTPoS also includes a second module whose role is to monitor the radio interfaces selected by the first module. In this module, energy consumption is calculated for the selected interfaces. The results of these calculations are analysed by the first module to make a final decision. At the end of their work, the authors obtained an algorithm enabling a considerable reduction in energy consumption, while maintaining the desired link reliability.

4) Real-time low-power M2M control algorithm

In [18] , Elghoul et al. developed a control mechanism for real-time, low-energy M2M communication. In a monitoring application, for example, sensors transmit sensed data to an IoT gateway. This transmission must take place in real time. To maintain this exchange for a long time, the IoT nodes and gateways must, above all, have sufficient autonomy.

In Table 1 , the algorithm of Di Mu, Yunpeng Ge, M. Sha, S. Paul, N. Ravichandra & S. Chowdhury is our concern considering our objective, which is to extend its functionalities. Initially, this algorithm includes only three wireless protocols: WiFi, Zigbee and Bluetooth.

Our proposed solution described above is represented by the diagram in Figure 3 .

In Figure 3 , we have an IoT gateway with multiple interfaces at the centre of the system. This gateway enables the interconnection of four different networks such as the cellular network, WiFi, Bluetooth and the LoRa network. It also facilitates the interconnection of several heterogeneous IoT devices. What’s more, these devices are more or less distant from each other. On the LoRa network, we have a sensor that not only collects and sends data over a long distance, but also has long-lasting autonomy. This figure highlights the following elements:

We will be interconnecting four different networks: Zigbee, LoRa, WiFi and Bluetooth. To do this, we will use an IoT gateway, in particular the Raspberry Pi. Interconnecting these four networks involves connecting each of them to the gateway.

In order to ensure communication between the LoRa network and the IoT gateway, we rely on the solution presented in section 1.2 (connection between the LoRa gateway and the application server). A LoRa gateway whose role is to send all the data collected to an application server. The application server is responsible for processing, analyzing and displaying the data. In the same way, the application server is connected to an IoT gateway, as shown in Figure 4 .

This architecture makes it possible to connect the LoRa network to the IoT gateway. The gateway now acts as a bridge between the LoRa network and the application server. When data is received by the gateway, it will redirect it to the application server for analysis. At the same time, the gateway can decide to perform an action (start the fan, start the air conditioner, start the heating or send a notification to a user in our previous example). The complete architecture of this system is shown in Figure 5 .

At the same time, the gateway is responsible for interconnecting the different networks with heterogeneous devices at their ends and deciding what action to take when data is received. To perform all these tasks efficiently, the gateway must be equipped with an algorithm. When a task requires significantly more resources, it calls on the application server for intelligent processing and then receives the response. The GitHub link below contains the code written in Python that can be run in the gateway for these four points ( https://github.com/saidou-supptic/haman ):

In addition to this, LoRa offers the following advantages: