FR3130494A1 - Method and device for dynamic management of a mesh communication network - Google Patents

Abstract

The invention relates to a device and an associated method, for dynamically defining the transmission power to be applied to the nodes of a mesh network, and which takes into account the energy limits experienced by the nodes of the network. Figure for the abstract: Fig.1

Description

Method and device for dynamic management of a mesh communication network

Field of invention

The invention lies in the technical field of communication networks, and more particularly the field of the Internet of Things (IoT) implementing mesh networks based on limited nodes in energy and capable of transmitting at different power levels.

State of the art

In the field of the Internet of Objects, networks of mesh type nodes or "mesh" according to the accepted anglicism, are made up of sensors that provide both measurement or environmental control capabilities, as well as radio communications capabilities with neighboring nodes/sensors. This type of network typically makes it possible to route measurements or commands from hop to hop in the mesh network, using some of the nodes considered as “router” nodes.

The nodes of these networks are generally "wireless" in the sense that communications with neighboring nodes are made by means of radio communication, but also in the sense that they are not connected by wire to a network either. of energy distribution.

So-called "gateway" nodes, which interconnect such mesh networks with the Internet, are typically an exception to this rule, but the other nodes - sensors and/or actuators and/or routers - of the network typically do not. exception.

This is made possible by the use of batteries in these nodes, which can be supplied with energy by an energy collection system such as "solar panels", wind turbines, etc.

A simpler mode also consists in the implementation of a maintenance process requiring the replacement of the batteries at regular intervals, these intervals being expressed in years in certain cases.

In any case, the energy consumption in the nodes must be limited, so as not to take the risk of an interruption of service due to the exhaustion of certain batteries.

Also, when designing radio communication systems, the choice of the transmission power is critical. If it is too high, the power consumed in the node is too, and if it is too low, the radio range is reduced, which leads either to multiply the number of nodes to cover a given region, or to encounter reliability issues in communications.

In practice, the transmission power is fixed during the design of the radio module and it is not likely to be modified during the operating life of a network.

In addition, all nodes in a network are typically designed to provide the same transmit power. The design of the network and the topological deployment of the nodes are then constrained by the radio range that this fixed transmission power allows.

It is nevertheless possible to design nodes for which the transmission power is variable. A continuous adjustment of the transmission power is possible. However, in order to provide optimum energy efficiency for various transmission powers, these solutions comprise, as output stage of the transmission chain, several modules each associated with a transmission power, it being understood that only one module is activated at any given time while the other modules are deactivated and unpowered, so that only the selected module consumes power during transmit periods.

With such nodes, it is then possible to control the transmission power dynamically. For example, it is possible to decide the transmission power before transmitting each data packet, depending on the context in a given node.

However, such an approach does not allow an optimized and dynamic control of the transmission power of each node, taking into account the energy constraints of each node.

In the more general context of radio communications, many solutions exist that aim to dynamically control the transmission power of transmitters. These are “automatic power control” or “transmit power control” solutions.

However, these solutions all require a certain context:
- have a device for continuous control of the transmission power; And
- have a continuous flow of data between the transmitter and the receiver.

A transmitter power control loop is based on continuous feedback from the receiver that periodically commands an increase or decrease in transmit power. There is then a data communication in the data transmitter to data receiver direction, and a command communication in the opposite direction. This is two-way communication.

To do this, a signal-to-noise ratio target can be set at the receiver. Either the measured signal-to-noise ratio is lower than this target and an increase in power is ordered, or it is higher and a decrease is ordered. This description is simplified and many variants and refinements exist and are presented in the literature.

Generally, the signal-to-noise ratio target is chosen high enough to ensure error-free transmission and low enough not to cause unnecessary power consumption or harmful interference on other radio communication links.

Such automatic transmission power control is however not directly applicable to energy-limited mesh networks, for several reasons:
- Devices for continuous control of the transmission power are typically non-optimal in terms of energy efficiency. The energy efficiency, ie the ratio between the power emitted on the antenna and the power consumed by the transmitter, is optimized only for the maximum power, and it is typically all the more bad as the power of emission is low.
- The assumption of a continuous flow of data is not an admissible assumption in the context of energy-limited mesh networks. Indeed, in this context the nodes are required to remain silent most of the time and to sporadically transmit brief data packets.

The following article presents a solution that transposes the known systems for automatic control of transmitted power to the field of mesh sensor networks: "A Transmission Power Control Mechanism for 802.15.4+RPL-Operated Wirelesss Sensor Network", Ali Qolami, Mohammad Nassiri and Hatam Abdoli, published by Betham Science, International Journal of Sensors, Wireless Communication and control, 2020, 10, pages 197_206, where the sensor networks considered correspond to the IEEE 802.15.4 standard with regard to the physical and MAC layers (using the "Beacon" mode of this standard) and the IETF 6LowPAN and RPL standards with regard to the upper layers.

This solution is equivalent to the application of the conventional mechanism of automatic control of transmitted power, but which is limited only to the short data sequences which can be exchanged between a node and its “parent”. The notion of target for the signal-to-noise ratio is played here by a target on the received power, identified by 'P _thr ' on the of the item.

The approach proposed in this article has two limitations.

A first limitation relates to the absence of a continuous flow of data. In the presence of dynamic "fading" on the radio links (i.e. dynamic variations in the attenuation of the radio signal), the initial values given to the transmission power after long periods of inactivity between a node and its parent is arbitrary . If the data communication is of short duration (which is typically the case in sensor networks), the automatic control has very little time to converge.

A second limitation is that the link between the choice of the power emitted and the energy limitations of the sensor platforms is not mentioned.

In an article by Sebastian Max and Tinghuai Wang, “Transmit Power Control in Wireless Mesh Networks Considered Harmful”, Advances in Mesh Networks, 2009, it is suggested that power control in a non-power-limited mesh network does not bring added value, the best performance being obtained in an "all or nothing" mode where the nodes are either silent or transmit at the maximum power and at the maximum rate.

Also, if one has a network of sensors not limited in energy and not limited in transmission power, the best strategy consists in forcing a star topology (with the gateway as the center of the star) in order to avoid any multi-hop communication. This supposes choosing a transmission power allowing any node to communicate directly with the gateway.

This boundary reasoning then illustrates the need not to separate the question of the dynamic selection of the transmission power from the question of the energy limits of the nodes.

But there is currently no solution that allows optimized and dynamic control of the transmission power of each node of a mesh network, taking into account the energy constraint of each node.

The present invention meets this need.

An object of the present invention is to propose a device and an associated method, making it possible to dynamically define the transmission power to be applied to the nodes of a mesh network, and taking into account the energy limits experienced by the nodes of the network.

Advantageously, the function of dynamic adjustment of the transmission power of the nodes allows greater flexibility in the operation of mesh networks, for example by reducing the constraints on the topology of the network, and by modifying the transmission power of the nodes opportunistically to respond to variations in the environment.

Indeed, according to the prior art, mesh sensor-actuator networks always assume the deployment of nodes used at constant and homogeneous power. In some cases, we can see the use of uneven power, for example to address localized coverage problems. However, we do not observe a dynamic use of the transmission power of each node in this type of so-called non-homogeneous power (or heterogeneous power) networks.

The technical difficulty to overcome did not essentially relate to the realization of a new type of nodes called "multi-power" nodes, even if such nodes did not exist, but mainly resided in the design of protocols and algorithms which make it possible to make the best use of this flexibility, i.e. the opportunistic or dynamic modification of the transmission power of the nodes.

Thus, the present invention provides a method which meets this need, in particular when the energy aspects are critical. This is the case, for example, when the nodes use energy collection systems that are very modest in size.

Advantageously, the present invention implements a hybrid architecture in which the choice of transmission power is determined by a central entity, while a distributed routing solution gives the nodes autonomy in routing decisions.

Indeed, when it comes to defining a system of dynamic configuration of a network of sensors, an important question consists in choosing if this dynamic configuration will be controlled according to a centralized or decentralized architecture.

In the first case of a centralized architecture, the intelligence is located at the level of a central entity and this entity must be provided with the means of observation enabling it to apprehend the detailed state of the entire network, and also the means of carrying the configuration orders to all the nodes of the network.

In the second case of a decentralized architecture, the network nodes are given the necessary autonomy to self-configure in collaboration with their neighbors, and a central entity is then no longer required.

To meet the aforementioned need where the question of the selection of the transmission power and the question of the routing in the mesh network cannot be separated, it is clear that the determination of the paths from hop to hop is fundamentally impacted by the range of the communications which is a function of the chosen transmission power.

Also, a centralized approach that would cover both the routing and transmission power selection aspects would not be suitable. The radio links of a mesh network are typically affected by dynamic variations in the attenuation of the radio signal (“fading”). This fading can be fast or slow, and be caused by variations in the environment (outdoor atmospheric conditions, mobility of people indoors, as well as movement of furniture, door openings and closings, etc.).

Moreover, a routing solution in a mesh network that would not be dynamic would be inefficient, unable to follow the variations of fading. If a central entity were in charge of routing, it would have to be given a very dynamic network observation capability, as well as capabilities for sending commands to the nodes, which would also be very dynamic.

However, these capacities would inevitably result in an increase in the bandwidth dedicated to the control plane on all the elementary radio links in the mesh network. This would be a paradox, insofar as one of the objectives addressed by the invention is compatibility with energy-limited nodes.

Also, the present invention proposes a hybrid architecture, where the determination of the transmission power of each node of the network is made centrally, while the routing decisions are distributed and entrusted to the nodes of the network.

The device of the present invention comprises in particular a central entity which is configured to observe the network and apply solutions to relieve the most critical nodes from the point of view of energy management.

The central observation entity, in its mission to maintain the service provided by the network, relies in one embodiment on a neural network in order to dynamically determine and select the transmission power of each node, from a list of transmit power values that the node supports.

When homogeneous sensors are available in terms of transmission power and coverage, the choice of this homogeneous transmission power is critical. Indeed, if the transmission power is too low, there is a constraint of having to "square" the coverage area by ensuring that all the nodes have in their immediate vicinity a sufficient number of neighbors to ensure a communication of end-to-end in a mesh network, which then leads to inflation in the number of router nodes to be deployed.

On the other hand, if the transmission power is too great, it emerges nodes whose energy consumption is high, and which will then require either to have heavy energy collection devices (for example a size of solar panel important), or to lead to strong maintenance constraints (for example, the need to replace the battery of all the sensors with a battery charged off site, and this every month for example).

The proposed solution is a solution with great flexibility of use, since the nodes can be used with various transmission powers depending on the context (for example: there may be obstacles, constraints on radio propagation), and their transmission power can be modified according to changes in the environment (for example: a change in the quality of the radio signal, addition or deletion of network nodes, etc.).

The fields of industrial application of the invention are mainly those of the Internet of Things, to cite a few examples: "Smart Cities", "Smart Buildings", "Smart Industry" (or "Industry 4.0") "Smart Farming". , etc.

To obtain the desired results, a method is proposed for the dynamic management of a meshed communication network implemented in an environment of the Internet of Things, the network being composed of a plurality of communicating nodes constrained in energy and configured in order to be able to transmit data at different power levels, the mesh network comprising at least one gateway node configured to transmit data packets originating from transmitter nodes to the Internet network via a central node. The method is implemented sequentially over epochs of predefined duration by a processor of the central node and comprises steps consisting of:
- for the duration of an epoch i:
(a1) receiving from said at least one gateway node, at least one data packet, said at least one data packet comprising measurement data from a transmitting node of the network and at least information relating to:
- the node-to-node path followed by said at least one packet from the sender node; And
- the residual energy 'currentEnergy' of each node of the path;
(a2) adding the information received to an observation database, the observation database being organized to store, for each node of the mesh network, the information received at each epoch, as well as a 'txPow' index representative of the transmission power and a 'rankingCost' value representative of a routing cost during the past epoch i;
- at the beginning of the following epoch i+1:
(b1) calculating on the basis of the information stored in the observation database a new configuration for the mesh network, the calculation of the new configuration consisting in determining, for at least one node of the mesh network, actions to be applied to the node as to its transmission power and its routing cost; And
(b2) broadcast to each node of the network via said at least one gateway, the new calculated configuration.

• Ledit au moins un paquet de données reçu comprend de plus des informations relatives au rapport signal à bruit plus interférence ‘SNIR’ de chaque nœud du chemin suivi, et des informations relatives au nombre de répétitions ‘ETX’ pour transmettre ledit au moins un paquet pour chaque nœud du chemin.
• L’étape de calculer une nouvelle configuration comprend des étapes consistant à :

- déterminer pour ledit au moins un nœud, si l’énergie résiduelle est inférieure à une valeur seuil ‘Tenergy’ et si l’énergie consommée par le nœud est supérieure à l’énergie produite ; et
- si oui, appliquer des actions consistant à diminuer la puissance d’émission et augmenter le coût de routage du nœud concerné ; ou
- si non, appliquer des actions consistant soit à augmenter la puissance d’émission du nœud concerné, soit à maintenir la puissance courante.

• Le procédé comprend de plus pendant la durée de chaque époque, des étapes consistant à :

- évaluer la santé du réseau maillé, la santé du réseau étant évaluée par rapport à une estimation de la qualité de transmission de données entre les nœuds du réseau maillé, et par rapport au nombre de nœuds qui sont considérés comme critiques en termes d’énergie, c’est à dire les nœuds pour lesquels l’énergie résiduelle est inférieure à une valeur seuil ;
- ajouter dans la base de données d’observation une valeur représentative de cette évaluation de la santé du réseau ;
- calculer une évolution de la santé du réseau maillé suite à des actions appliquées aux nœuds quant à la puissance d’émission et le coût de routage ; et
- ajouter dans la base de données d’observation une valeur représentative de cette évolution de la santé du réseau.

• L’étape de calculer une évolution de la santé du réseau maillé, met en œuvre des algorithmes d’apprentissage par renforcement profond, inspirés de l’apprentissage par renforcement combiné à des réseaux de neurones profonds pour générer une prédiction de l’évolution de la santé du réseau maillé.
• L’étape de prédiction de l’évolution de la santé du réseau maillé consiste à opérer de manière itérative un réseau de neurones avec des données labellisées correspondant aux valeurs de paramètres enregistrés dans la base de données d’observation pour chaque nœud pendant une époque, et aux actions qui sont appliquées sur chaque nœud au terme de cette époque, afin de déterminer l’évolution de la santé optimale.
• Le procédé comprend de plus des étapes consistant à :

- générer une nouvelle configuration pour le réseau maillé correspondant à l’évolution de santé optimale, et lui associer un nouveau numéro de version du réseau maillé ; et
- transmettre la nouvelle configuration et le nouveau numéro de version à ladite au moins une passerelle du réseau maillé, pour la diffuser à chaque nœud du réseau.

• L’étape de diffuser la nouvelle configuration calculée, à chaque nœud du réseau via ladite au moins une passerelle, consiste :

- pour un nœud, à transmettre vers un nœud récepteur un message qui contient, en plus des informations relatives au rang des nœuds, le numéro de version du réseau ;
- pour un nœud récepteur, à comparer le numéro de version reçu avec le numéro de version courante, et mettre à jour les éléments relatifs à la différence entre la nouvelle version reçue et la version courante.The invention can be implemented according to alternative or combined embodiments, where:

• Said at least one data packet received further comprises information relating to the signal to noise ratio plus interference 'SNIR' of each node of the path followed, and information relating to the number of repetitions 'ETX' to transmit said at least one packet for each node in the path.
• The step of calculating a new configuration includes steps consisting of:

- determining for said at least one node, if the residual energy is lower than a threshold value 'Tenergy' and if the energy consumed by the node is greater than the energy produced; And
- if yes, apply actions consisting in reducing the transmission power and increasing the routing cost of the node concerned; Or
- if not, apply actions consisting either in increasing the transmission power of the node concerned, or in maintaining the current power.

• The method further comprises, during the duration of each epoch, steps consisting of:

- assess the health of the mesh network, the health of the network being assessed against an estimate of the quality of data transmission between the nodes of the mesh network, and against the number of nodes which are considered energy critical , that is to say the nodes for which the residual energy is less than a threshold value;
- add in the observation database a value representative of this evaluation of the health of the network;
- calculate an evolution of the health of the mesh network following actions applied to the nodes with regard to the transmission power and the routing cost; And
- add in the observation database a value representative of this evolution of the health of the network.

• The step of calculating a health evolution of the mesh network, implements deep reinforcement learning algorithms, inspired by reinforcement learning combined with deep neural networks to generate a prediction of the evolution of the mesh network health.
• The mesh network health evolution prediction step consists of iteratively operating a neural network with labeled data corresponding to the parameter values recorded in the observation database for each node during an epoch, and the actions that are applied to each node at the end of this epoch, in order to determine the evolution of optimal health.
• The method further comprises steps of:

- generating a new configuration for the mesh network corresponding to the evolution of optimal health, and associating it with a new version number of the mesh network; And
- Transmit the new configuration and the new version number to said at least one gateway of the mesh network, to distribute it to each node of the network.

• The step of broadcasting the new calculated configuration, to each node of the network via said at least one gateway, consists of:

- for a node, to transmit to a receiving node a message which contains, in addition to information relating to the rank of the nodes, the version number of the network;
- for a receiving node, comparing the version number received with the current version number, and updating the elements relating to the difference between the new version received and the current version.

The invention also relates to a device for the dynamic management of a meshed communication network implemented in an environment of the Internet of Things, the network being composed of a plurality of communicating nodes constrained in energy and configured to be able to transmit data according to different power levels, the mesh network comprising at least one gateway node configured to transmit data packets from transmitting nodes to the Internet network via a central node, the device comprising means for implementing sequentially over epochs of duration predefined, the steps of the method of the invention.

The invention also relates to a computer program product which comprises code instructions making it possible to carry out the steps of the method of the invention, when the program is executed on a computer.

Description of figures

Other characteristics and advantages of the invention will appear with the aid of the following description and the figures of the appended drawings in which:

illustrates a mesh network environment for implementing the present invention;

schematically illustrates the structure of a multi-power node according to one embodiment of the invention;

illustrates different coverage areas depending on transmit power for a multi-power node according to the ;

illustrates a time sequencing of epochs for a central entity according to an embodiment of the invention;

illustrates a message flow diagram for broadcasting a new configuration according to one embodiment of the invention;

illustrates a hop-by-hop header of a multi-power node according to one embodiment of the invention;

symbolizes the vision that a central entity has for a network node at the end of an era;

illustrates the content of an observation database managed by the central entity according to one embodiment of the invention;

illustrates the content of an observation database managed by the central entity according to another embodiment of the invention;

illustrates the steps of the method of the invention making it possible to dynamically define the transmission power to be applied to the nodes of a mesh network, taking into account the energy limits experienced by the nodes, according to an embodiment of the invention ;

illustrates epoch time sequencing in one embodiment of the invention with network health assessment;

is a schematic representation of a neural network adapted to perform network health prediction, according to one embodiment of the invention;

illustrates the steps of the method of the invention making it possible to dynamically define the transmission power to be applied to the nodes of a mesh network, taking into account the energy limits experienced by the nodes, according to another embodiment of the invention.

Claims (10)

Method for dynamic management of a mesh communication network implemented in an Internet of Things environment, the network being composed of a plurality of communicating nodes constrained in energy and configured to be able to transmit data according to different power levels , the mesh network comprising at least one gateway node configured to transmit data packets from transmitter nodes to the Internet network via a central node, the method being implemented sequentially over epochs of predefined duration by a processor of the central node and comprising steps of:
- for the duration of an epoch i:
(a1) receiving from said at least one gateway node, at least one data packet (600), said at least one data packet comprising measurement data from a transmitter node of the network and at least information (602, 604 ) relative:
- the node-to-node path followed by said at least one packet from the sender node; And
- the residual energy of each node of the path;
(a2) adding the information received to an observation database (800), the observation database being organized to store, for each node of the mesh network, the information received at each epoch, as well as an index representative of the transmission power and a value representative of a routing cost during the past epoch i;
- at the beginning (902) of the following epoch i+1:
(b1) calculating on the basis of the information stored in the observation database a new configuration for the mesh network, the calculation of the new configuration consisting in determining, for at least one node of the mesh network (904), actions to applying to the node as to its transmit power and routing cost (908, 916); And
(b2) broadcasting (502) to each node of the network via said at least one gateway, the new calculated configuration. The method of claim 1 wherein said at least one data packet received further includes information relating to the signal-to-noise-plus-interference ratio of each node of the path followed, and information relating to the number of repetitions to transmit said at least one packet for each node in the path. The method of claim 1 or 2 wherein the step of calculating a new configuration includes steps of:
- determining for said at least one node, if the residual energy is below a threshold value and if the energy consumed by the node is greater than the energy produced; And
- if yes, apply actions consisting in reducing the transmission power and increasing the routing cost of the node concerned; Or
- if not, apply actions consisting either in increasing the transmission power of the node concerned, or in maintaining the current power. The method according to any one of claims 1 to 3 further comprising during the duration of each epoch, the steps of:
- evaluating a health parameter of the mesh network, the health of the network being evaluated with respect to an estimate of the quality of data transmission between the nodes of the mesh network, and with respect to the number of nodes which are considered to be critical in terms of energy, ie the nodes for which the residual energy is lower than a threshold value;
- add in the observation database a value representative of this evaluation of the health of the network;
- calculate an evolution of the health of the mesh network following actions applied to the nodes with regard to the transmission power and the routing cost; And
- add in the observation database a value representative of this evolution of the health of the network. The method according to claim 4 wherein the step of calculating an evolution of the health of the mesh network, implements deep reinforcement learning algorithms, combining reinforcement learning with deep neural networks to generate a mesh network health evolution prediction. The method according to claim 5 wherein the step of predicting the evolution of the health of the mesh network consists of iteratively operating a neural network with labeled data corresponding to the values of parameters recorded in the database of observation for each node during an epoch, and the actions that are applied on each node at the end of this epoch, in order to determine the evolution of the optimal health. The method according to claim 6 further comprising the steps of:
- generating a new configuration for the mesh network corresponding to the evolution of optimal health, and associating it with a new version number of the mesh network; And
- Transmit the new configuration and the new version number to said at least one gateway of the mesh network, to distribute it to each node of the network. The method according to any one of claims 1 to 7 in which the step of broadcasting the new calculated configuration, to each node of the network via the said at least one gateway, consists of:
- for a node, to transmit to a receiving node a message which contains, in addition to information relating to the rank of the nodes, the version number of the network;
- for the receiving node, to compare the version number received with the current version number, and update the elements relating to the difference between the new version received and the current version. A computer program product, said computer program comprising code instructions for carrying out the steps of the method according to any one of claims 1 to 8, when said program is executed on a computer. A device for dynamic management of a mesh communication network implemented in an Internet of Things environment, the network being composed of a plurality of communicating nodes constrained in energy and configured to be able to transmit data according to different levels of power, the mesh network comprising at least one gateway node configured to transmit data packets from transmitter nodes to the Internet network via a central node, the device comprising means for implementing sequentially over periods of predefined duration, the steps of the method according to any one of claims 1 to 8.