METHOD FOR THE CONJOINT COMMUNICATION AND SENSING OF THE ENVIRONMENT OF A NETWORK NODE

Abstract

The present invention relates to a conjoint communication and sensing method for detecting the movement of an obstacle in the environment of a node in a wireless telecommunication network and, where applicable, localising this obstacle. The node divides its environment into angular sectors and measures, at a plurality of observation instants, a quantity characteristic of the signal-to-noise-plus-sector-interference ratio in each of these sectors (410-420) to form a sensing matrix (Γτ,n(t)). A blind separation of sources is implemented to extract from this matrix a matrix of sector contributions (Γτ,n(•)(t)) of the interfering signals (430). The movement of any obstacle is detected from the history of the sector contributions (440), which makes it possible, where applicable, to predict a blockage situation (450) and to modify the allocation of the radio resources accordingly (460).

Description

TECHNICAL FIELD

The present invention relates to the field of environment sensing in a telecommunication network, in particular for radio resource management (RRM). One particular embodiment also relates to the field of distributed learning.

PRIOR ART

With the deployment of 5th-generation (5G) cellular networks, radio resource management techniques have had to evolve in order to take into account new cases of use having recourse to services that are highly heterogeneous in terms of quality of service (QOS). Thus RRM techniques now use more and more complex strategies combining power allocation, beam formation, allocation of transmission resources (time intervals, frequencies, codes), etc to be able to satisfy such heterogeneity. In such a context, it is important to be able to sense the environment of the various nodes in the network (UEs and base stations) and to predict changes therein, knowledge of this environment making it possible to finely manage the radio resources of the network in terms of time and space. Thus, for example, in a 5G cellular network where the communications in the millimetric band may be subject to strong attenuation in the case of blockage, predicting the change in the environment makes it possible to anticipate the loss of QoS due to weakening of a radio link in this band and to take steps to remedy this, in particular by initiating a handover procedure in good time. More widely, it may also make it possible to optimise, according to circumstances, the quality of service in certain communications, to reduce the frequency of handover operations, and to minimise the energy consumption of certain nodes or of certain types of node.

The environment of the nodes of a network can be sensed by equipping these nodes with a radar device, so that they have a dual functionality of communication and radar sensing. In this case, a node can be equipped with distinct communication and radar devices or even an integrated communication and radar sensing (C&S) system, using the same radio resource or distinct radio resources for the radar function and the radio communication function. A description of the various possible variant embodiments of nodes with dual functionality can be found in the article by J. A. Zhang et al. entitled “Perceptive mobile network: a cellular network with radio vision via joint communication and radio sensing”, published in IEEE Vehicular Technology Magazine, vol. 16, No 2, pp. 20-30, June 2021.

In general terms, radar sensing of the environment by a node with dual functionality is based on conventional techniques of measuring angle of arrival (AoA), time of flight or ToF and received signal strength or RSS to locate the radio-propagation obstacles. However, it does not exploit (or does so only little) the measurements coming from the radio communication and supposes to significantly modify the architecture of the nodes, which makes retrofitting complex and expensive.

The object of the present invention is to propose a method for the conjoint communication and sensing of the environment of a network node relying only on the measurements coming from the radio system, without adding a radar device and without complex modification to the architecture of the node.

DESCRIPTION OF THE INVENTION

The present invention is defined by a conjoint communication and sensing method for sensing the movement of an object in the environment of a node, referred to as the reference node, of a wireless communication network, said reference node having previously established a radio link with a sending node of the network, said method being original in that the reference node:

• • determines a plurality of non-overlapping angular sectors in said environment and the vertex of which is the reference node, as well as a plurality of observation instants;
  • makes, in each of the observation instants, a measurement of a quantity characteristic of a signal-to-noise-plus-sector-interference ratio on said radio link, for each of the angular sectors, and constructs a sensing matrix (Γ_τ,m(t), Γ_τ,n^j0(t)) the elements of which are the measurements of said characteristic quantity;
  • implements a blind separation of sources using said sensing matrix and extracts therefrom a matrix of sector contributions (Γ_τ,n^(•)(t), Γ_τ,n^j0(•)((t)) of interfering signals in the various angular sectors;
  • detects the movement of any obstacle in the environment of the reference node from at least one history of said sector contributions.

Advantageously, a situation of blockage of said link is predicted from characteristics of the movement of the obstacle in the environment of the reference node.

When such a situation of blockage of said link is predicted, a modification of the radio resources in the network is implemented.

Typically, the blind separation of sources is implemented by means of a singular value decomposition of the sensing matrix or a principal components analysis.

According to a second embodiment, the reference node determines the nodes of its neighbourhood and makes a simultaneous mapping and localisation on said environment using the matrix of sector contributions that it has extracted as well as matrices of sector contributions that it has received, in the form of messages, from the nodes of its neighbourhood.

Advantageously, a mapping dictionary is constructed and updated, each entry in the dictionary corresponding to the intersection of a plurality of angular sectors the vertices of which are the reference node or nodes of its neighbourhood, each entry containing a signature consisting of the sector contributions of the angular sectors involved in the intersection associated with said entry.

In the event of modification of at least one signature in the mapping dictionary between two observation instants, a change is detected and a reallocation of radio resources of the network is implemented.

Alternatively, a Kullback-Leibler divergence is calculated between the distribution of the signatures at a current observation instant and that of a previous observation instant, over the whole of the mapping dictionary or a part thereof, and a reallocation of radio resources of the network is implemented when the Kullback-Leibler divergence exceeds a predetermined threshold.

The neighbourhood of the reference node can be obtained by making a preselection of nodes of the network by means of a random sampling or a K-nearest neighbours method, then a calculation, for each node thus preselected, of a score from its matrix of sector contributions and that of the reference node, the neighbourhood of this node being determined by the preselected nodes the score of which exceeds a predetermined threshold or by the M nodes having the highest scores, where M is a non-zero integer.

The score of a preselected node can be calculated by means of a cosine similarity metric, a Kullback-Leibler divergence or an attention mechanism or by a neural network implementing a classification operation.

BRIEF DESCRIPTION OF THE DRAWINGS

Other features and advantages of the invention will emerge from the reading of a preferential embodiment of the invention, described with reference to the accompanying figures, among which:

FIG. 1 illustrates schematically an interference situation in a telecommunication network;

FIG. 2 shows a model illustrating the calculation of a signal-to-noise-plus-sector-interference ratio;

FIG. 3 shows an example of sensing of an obstacle in the environment of a node using a matrix representing the signal-to-noise-plus-sector-interference ratio over time;

FIG. 4 shows, in the form of a flow chart, the principle of a conjoint communication and environment-sensing method, according to a first embodiment of the invention;

FIG. 5 illustrates an example of cooperative sensing of the position of an obstacle in the environment of a reference node using the measurement matrices obtained by various nodes in the neighbourhood of this reference node;

FIG. 6 shows, in the form of a flow chart, the principle of a conjoint communication and environment-sensing method, according to a second embodiment of the invention;

FIG. 7 details the step of determining the neighbourhood of a node in the flow diagram of FIG. 6.

DETAILED DESCRIPTION OF PARTICULAR EMBODIMENTS

Hereinafter, a network of wireless communication nodes, not necessarily cellular, will be considered, wherein a first node has established a communication with a second node by means of a link, said link being interfered with by signals sent by third-party nodes, where applicable reflected by the environment or even by reflective surfaces (Reflective Intelligent Surfaces).

Without loss of generality, and for reasons of simplification of the presentation of the invention, we shall suppose that this telecommunication network is cellular, that the first node is an item of user equipment, UE, typically a mobile terminal, and that the second node is a base station, BS, serving this node. The present invention can in particular be applied to the 5G cellular network, in particular for communications in the millimetric domain or in the THz domain.

We shall consider any node in the network, for example an item of user equipment UE, and shall take it as the reference node. It is clear however that this node can alternatively be a base station, BS, of the network.

The idea at the basis of the present invention is to benefit from the spatial interference affecting the reference node and from the variation in this interference over time to deduce therefrom information with regard to the presence of any moving obstacles present in the neighbourhood of this node. More precisely, the reception of a communication by the reference node (via a downlink in the case of a UE and an uplink in the case of a BS) can be affected by interferences due to the reception of signals sent by the nodes located in the neighbourhood of this node.

This situation is illustrated in the example shown in FIG. 1. The reception diagram of the antenna of the reference node, 100, also has a main lobe pointing towards the base station, secondary lobes able to receive the signals sent by neighbouring nodes such as 110. If an obstacle, 120, moves in the neighbourhood of the reference node, the interference situation changes over time. In general terms, spatial interferences due to communications of the neighbouring nodes and fluctuations in these interferences over time can provide information on the presence and movement of obstacles in the neighbourhood of the reference node. Whilst the phenomenon of spatial interference degrades the quality of the radio communications and is combated for this reason, it is here usefully exploited to detect an activity in the environment of the reference node.

The reference node estimates a quantity characteristic of the signal-to-noise-plus-sector-interference ratio, for a plurality of sectors covering at least part of the neighbourhood of the reference node. Sector interference is defined here as the sum of the powers of the interfering signals received in a given angular sector, for example in the angular sector indicated at 150.

It is supposed, to illustrate the invention, that the nodes are distributed in a plane in accordance with a Poisson stochastic point process and that the channel between two nodes follows a Friis model, namely that the power PRx of the signal received by a receiving node is expressed using the power PTx sent from a sending node by means of:

$\begin{array}{cc}\left[\mathrm{Math}.1\right]& \\ \text{?}=h\left(t\right)\chi \left(t\right)P^{\mathrm{Tx}}{G}^{\mathrm{Tx}}\text{?}{\mathrm{Cd}}^{-\eta }& \left(1\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

where h(t) is a coefficient is the fading coefficient of the channel, χ(t) is a shadowing coefficient, G^Tx, G^Rx are the respective antenna gains of the sender and of the receiver in the propagation direction, C is an attenuation constant, d is the distance between the sending and receiving nodes and η is a transfer exponent on the propagation path.

It is clear for a person skilled in the art that other channel models can be used without departing from the scope of the present invention.

FIG. 2 shows a model illustrating the calculation of a signal-to-noise-plus-sector-interference ratio.

In this model, the base stations are referenced by the index i and the user equipment by the index j.

The reference node, here an item of user equipment, 210, bearing the index j₀ receives a communication signal from a sending node, here a base station BS, 220, bearing the index i₀. The communication is established on a downlink between the base station i₀ and the user equipment j₀.

It is supposed that the main sending lobe of the base station and the main reception lobe of the UA are aligned, i.e. the formation of the beam is optimum. In this configuration, the antenna gains in sending and receiving are respectively denoted G_m^Tx and G_m^Rx and the power of the communication signal received by the reference node is none other than:

$\begin{array}{cc}\left[\mathrm{Math}.2\right]& \\ \text{?}=P^{\mathrm{Tx}}{h}_0\left(t\right){\chi }_0\left(t\right)G_m^{\mathrm{Rx}}{G}_m^{\mathrm{Tx}}{\mathrm{Cd}}_0^{-\eta }& \left(2\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

with the same notations as before and where the index 0 indicates that the quantities relate to the nodes i₀ and j₀. The distance do is the one separating the nodes i₀ and j₀.

The base stations other than the one serving the reference node are now considered, i.e. such that i≠i₀. Each of these base stations is liable to cause spatial interference on the link between the nodes i₀ and j₀. The power of the interfering signal generated by the link between the node i (base station) and the node j (UE) on the link between the nodes i₀ and j₀ can be expressed in the following form:

$\begin{array}{cc}\left[\mathrm{Math}.3\right]& \\ \text{?}=P_i^{\mathrm{Tx}}{h}_i\left(t\right){\chi }_i\left(t\right)G^{\mathrm{Rx}}\left(\theta ,\text{?}\right)G^{\mathrm{Tx}}\left(v,{\phi }_{i,j}\right){\mathrm{Cd}}_i^{-\eta }& \left(3\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

where d_i is the distance between the nodes i and j₀, G^Tx(ν,φ_i,j) is the gain of the sending antenna of the base station i in the direction given by the angle φ_i,j, where ν is the angular width of the main lobe of the sending beam and φ_i,j is the relative angle of departure (AoD) of the interfering signal with respect to the direction of this main lobe, G^Rx(θ, ψ_i) is the gain of the receiving antenna of the user equipment where θ is the angular width of the main lobe of the reception beam and ψ_i is the relative angle of arrival (AoA) of the interfering signal with respect to the direction of this main lobe.

The total interfering signal on the downlink between the nodes i₀ and j₀, coming from a direction ψ, at the instant t, can then be written:

[Math.4]

$\begin{array}{cc}I\left(\psi ,t\right)=C\text{?}{P}_i^{\mathrm{Tx}}{h}_i\left(t\right){\chi }_i\left(t\right)G^{\mathrm{Rx}}\left(\theta ,\text{?}\right)G^{\mathrm{Tx}}\left(v,\text{?}\right)\text{?}\delta \left(\psi -{\psi }_i\right)& \left(4\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

where δ(.) is the Dirac symbol. This spatial interference can be added on each sector. In other words, the sector interference I_s(ϕ,α)(t), i.e. that generated by the nodes i∈s(ϕ, α) where s(ϕ, α) is the sector of angular width 2α and of absolute angular orientation ϕ, can be obtained by:

$\begin{array}{cc}\left[\mathrm{Math}.5\right]& \\ \text{?}\left(t\right)=\text{?}I\left(\psi ,t\right)d\psi & \left(5\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

This sector interference, i.e. created by the nodes of the network belonging to a given angular sector and having the reference node as the vertex, makes it possible to define a quantity characteristic of the signal-to-noise-plus-sector-interference ratio:

$\begin{array}{cc}\left[\mathrm{Math}.6\right]& \\ \text{?}\left(t\right)=\frac{\text{?}}{\text{?}\left(t\right)+N\text{?}{B}_0}& \left(6\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

where N is the noise spectral density and B₀ is the bandwidth used for the communication (downlink) between the nodes i₀ and j₀.

It should be noted that the sector interference I_s(ϕ,α)(t) in a given sector depends on the time. If it is supposed that the sending beams of the various base stations are static in a given time range, the time variation of a sector interference results from a variation in configuration of the environment, for example the movement of an obstacle over time, as illustrated in FIG. 2 (movement of a robot, 230, through the sector s₀).

The neighbourhood of the reference node can be the subject of a division into continuous and non-overlapping angular sectors, not necessarily having identical angular widths. For reasons of simplification, we shall however suppose hereinafter that this division is implemented by means of n+1 sectors all having the same angular width

$2\alpha =\frac{2\pi }{n+1}$

where n is a non-zero integer. The neighbourhood of the reference node is then partitioned into sectors s_k, k=0, . . . , n, the sector s₀ being defined as the one the orientation of which, ϕ₀, corresponds to the link between the nodes i₀ and j₀, the other sectors shaving respective orientations ϕ_k=ϕ₀−2α·k.

For an observation time window ending at the time t and of duration τ, the values taken by said quantity characteristic of the signal-to-noise-plus-sector-interference ratio at the instants

$t-\tau +p\frac{\tau }{m},$ $p=0,..,$

m are collected, where τ/m is a detection period where m is a non-zero integer, and this for the various sectors covering the neighbourhood of the reference node, s_k, k=0, . . . , n.

In this way a sensing matrix of the environment of the reference node is obtained:

$\begin{array}{cc}\left[\mathrm{Math}.7\right]& \\ \text{?}\left(t\right)=\left(\begin{array}{ccc}\text{?}\left(t-\tau \right)& \dots & \text{?}\left(t-\tau \right)\\ ⋮& \ddots & ⋮\\ \text{?}\left(t\right)& \dots & \text{?}\left(t\right)\end{array}\right)& \left(7\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

When an obstacle is located in the environment of the reference node, this results in a modification of the signal-to-noise-plus-sector-interference ratio, this modification depending on the secondary lobe or lobes shadowed and on the shadowing coefficient. The resolution with which this obstacle can be sensed depends in particular on the angular width of the sectors and the density of the interfering nodes (i≠i₀). It will be understood in fact that, the greater the density of the interfering nodes and the smaller the angular width, the better the spatial resolution.

In practice, the reception of the interfering signals and the integration of the sector interference in the various sectors take place in a manner multiplexed in time (TDD), by scanning in turn the various angular sectors. The refresh frequency can be adaptive according to the activity in the various angular sectors. In all cases, the refresh frequency in the various sectors can be selected so that the matrix Γ_τ,n(t) faithfully represents the signal-to-noise-plus-sector-interference ratio of the various sectors at the same observation instants.

According to a first embodiment, the sensing matrix can make it possible to sense the movement of an obstacle in the environment of the reference node. For this purpose, the measurements of the signal-to-noise-plus-sector-interference over time, in other words the row vectors of the sensing matrix, can allow a blind source separation. This is because a sensing matrix can be decomposed into three distinct components:

$\begin{array}{cc}\left[\mathrm{Math}.8\right]& \\ \text{?}=\text{?}\left(t\right)+\text{?}\left(t\right)+\text{?}\left(t\right)& \left(8\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

The first component, Γ_τ,n⁽⁰⁾(t), corresponds to the contribution relating to static shadowing, the second component, Γ_τ,n^(•)(t), corresponds to the contribution of moving obstacles in the environment of the reference node, and the third component, Γ_τ,n^(noise)(t), corresponds to the other contributions of noise such as random scattering and scattering.

The blind separation of sources makes it possible to isolate the various contributions to the sensing matrix, in particular that due to the moving obstacles. Various source-separation methods can be applied, for example singular value decomposition (SVD) of the sensing matrix or principal components analysis (PCA), in a manner known per se. For example, the sensing matrix, of size (m+1)×(n+1), can be the subject of an SVD decomposition:

$\begin{array}{cc}\left[\mathrm{Math}.9\right]& \\ \text{?}\left(t\right)=\text{?}& \left(9\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

where U is a semi-unitary matrix of size (m+1)×r, V is a semi-unitary matrix of size (n+1)×r, and E is a diagonal square matrix of size r×r with r=rank(Γ_τ,n(t))≤min(m+1, n+1), the diagonal elements of E being the singular values σ₁≥σ₂≥ . . . ≥σ_r.

Equation (9) can also be written:

$\begin{array}{cc}\left[\mathrm{Math}.10\right]& \\ \text{?}\left(t\right)=\text{?}=\text{?}+\text{?}+\text{?}& \left(10\right)\end{array}$ $\text{?}\text{indicates text missing or illegible when filed}$

wherein the three adding terms appearing on the right correspond respectively to the components Γ_τ,n⁽⁰⁾(t), Γ_τ,n^(•)(t), and Γ_τ,n^(noise)(t) of the sensing matrix. This is because the matrix Γ_τ,n⁽⁰⁾(t) corresponds essentially to an attenuation on the direct propagation path or the multi-paths, the second to the signals of the interferers in the neighbourhood and the third to a random noise (random scattering, thermal noise).

The ordered list of the singular values shows three different decrease regimes with changes of slope between successive regimes, which makes it possible to determine the indices ₀, and ₁, and therefore makes it possible to isolate the matrix Γ_τ,n^(•)(t)=.

The determination of the indices ₀, and ₁ from the sensing matrix has been illustrated in FIG. 3. The left-hand part of the figure, 310, shows (the amplitude) of the elements of the matrix Γ_τ,n(t), the sector index appearing on the Y-axis and the index of the instant of detection in the time window appearing on the X-axis. The right-hand part of the figure shows the result of the singular value decomposition. More precisely, the diagram 320 gives the singular values of the matrix, ordered by decreasing values, in which the aforementioned three decrease regimes can be distinguished. The elements of the matrix Γ_τ,n^(•)(t) corresponding to the eigenvalues to are shown at 330. The indices of the sectors have been rearranged in ordinates to centre the diagram on the sector s₀. A blockage of the link corresponding to the appearance of an obstacle between the nodes i₀ and j₀ is noted at 340. This blockage can be anticipated since the obstacle first of all masks certain interference sources and therefore increases the signal-to-noise-plus-sector-interference ratio in certain sectors before the blockage in question. The chronological sequence of the sectors in which this ratio increases, and more generally the history of the signal-to-noise-plus-sector-interference ratio in the various angular sectors, makes it possible to estimate the trajectory (direction, angular velocity of the movement along the trajectory) and, where applicable, the apparent size of the obstacle in question.

The estimation of the trajectory and, where applicable, of the apparent size of the object makes it possible to anticipate the situation of blocking of the link and to manage the RRM resources accordingly, for example to prepare a handover for the reference node or to modify an allocation of resources.

FIG. 4 shows the principle of a conjoint communication and environment-sensing method, according to a first embodiment of the invention.

In this embodiment, a reference node of a wireless telecommunication network makes the measurements of a quantity characteristic of the noise-to-sector-interference ratio in a plurality of non-overlapping angular sectors covering the neighbourhood of this node and detects the movement of any object in this neighbourhood. Optionally, this detection enables the network to manage its radio resources, for example by preparing a handover or by modifying an allocation of resources.

At step 410, the reference node, j₀, having previously established a communication with a sending node, i₀, defines a plurality of non-overlapping angular sectors of which it is the origin, and a plurality of observation instants, for example by determining a sector angular width, α, or a given number of sectors, n+1, as well as a number m+1 of observation instants in a time window of duration τ.

At step 420, the reference node, at each observation instant and for each angular sector, makes a measurement of a quantity characteristic of the signal-to-noise-plus-sector-interference ratio. From these measurements it constructs a matrix Γ_τ,n(t) of size (n+1)×(m+1).

At step 430, the reference node implements a blind separation of sources using the measurements made. This separation can be implemented by means of a principal components analysis or by means of a singular value decomposition of the sensing matrix. At step 430, the reference node determines, among the sources determined at the previous step, those that correspond to interferences due to signals sent by nodes belonging to the neighbourhood of the reference node. For example, when the separation of sources is implemented by means of an SVD decomposition, this determination can be made by identifying breaks in slope in the series of singular values according to the observation instant. The matrix Γ_τ,n^(•)(t) corresponding to the contribution of the interferences generated by the nodes in the neighbourhood is deduced therefrom, to the matrix at Γ_τ,n(t). More precisely, the matrix Γ_τ,n^(•)(t) supplies, at each observation instant, the sector contributions of the nodes in the neighbourhood to the sensing matrix.

At step 440, the reference node senses, from the history of the sector contributions, the movement of any obstacle in the neighbourhood of the reference node. At step 450, if a movement is sensed at the previous step, the reference node makes a prediction of occurrence of a blocking situation from the characteristics of this movement.

At step 460, if a blocking situation is predicted by the reference node, the network implements a modification of the allocation of the radio resources, example implements a handover, a sending beam switching.

Steps 450 and 460 are optional since the reference node can, in some applications, limit itself to sensing the movement of an obstacle in its environment.

In a second embodiment of the invention, the reference node, by virtue of the cooperation of its neighbouring points, senses the position of an obstacle in its environment.

FIG. 5 illustrates an example of cooperative sensing of the position of an obstacle in the environment of a reference node using the sensing matrices obtained by various nodes in the neighbourhood of this reference node.

In this embodiment, the reference node, j₀, makes measurements of a quantity characteristic of the signal-to-noise-plus-sector-interference ratio in a plurality of angular sectors and a plurality of observation instants in a time window. It deduces therefrom a sensing matrix, denoted Γ_τ,n^j0(t), as in the first embodiment. Unlike the first embodiment, however, the nodes of a neighbourhood of j₀, within the meaning defined below, also each make measurements of said quantity characteristic of the signal-to-noise-plus-sector-interference ratio in a plurality of angular sectors and deduce therefrom their own sensing matrices. In the case illustrated, the nodes j₁ and j₂ belong to the neighbourhood of the reference node and construct their respective sensing matrices Γ_τ,n^j1(t) and Γ_τ,n^j2(t).

The reference node and the nodes belonging to the neighbourhood of this node implement a blind separation of sources as described previously, to obtain matrices of sector contributions denoted Γ_τ,n^j0(•)(t), Γ_τ,n^j1(•)(t), Γ_τ,n^j2(•)(t). The nodes j₁ and j₂ respectively transmit, in message form, the matrices Γ_τ,n^j1(•)(t) and Γ_τ,n^j2(•)(t) to the reference node. Alternatively, the reference node j₀ and the nodes belonging to its neighbourhood each transmits its matrix of sector contributions to an administrator node. From the chronological sequences of the sector contributions coming from the matrices Γ_τ,n^j0(•)(t), Γ_τ,n^j1(•)(t), Γ_τ,n^j2(•)(t), the reference node or, where applicable, the administrator node, implements simultaneous localisation and mapping (SLAM). A description of a simultaneous localisation and mapping method can be found in the article by H. Durrant-Whyte et al. entitled “Simultaneous Localization and Mapping: part I” published in IEEE Robotics & Automation Magazine, vol. 13, No 2, pp. 99-110. Being able to have available measurements coming from various “angles of view”, or more precisely from various distinct angular sectors, the vertices of which are located at distinct points, enables the node responsible for the calculation to determine and localise any moving obstacle. The size of the intersections of the various angular sectors gives an estimation of the errors in localising an obstacle in the environment of the reference node (right-hand part of the figure).

FIG. 6 details, in the form of a flow chart, a method for communicating and sensing the environment of a reference node, according to a second embodiment of the invention.

Sensing the environment of the reference node here aims to cooperatively estimate the position of a moving obstacle in the environment of the node in question.

As in the first embodiment, the reference node j₀ determines at 610 a plurality of angular sectors of which it is the vertex as well as a plurality of observation instants, and then measures, at 620, at each observation instant, a quantity characteristic of the signal-to-noise-plus-sector-interference ratio for each of said angular sectors, so as to construct the sensing matrix, Γ_τ,n^j0(t). At 630, it makes a blind separation of sources, for example by means of a singular value decomposition, to obtain a matrix of sector contributions Γ_τ,n^j0(•)(t).

In parallel, the reference node selects at 615, from the nodes in the network, those that belong to its neighbourhood, V_j0(t), in accordance with a selection algorithm described below, on the basis of the matrices of sector contributions Γ_τ,n^j(•)(t) received in the form of messages from the set U of nodes in the network or even from a subset of these nodes.

At step 625, it takes into account only the matrices of sector contributions respectively obtained from the nodes in its neighbourhood, or in other words Γ_τ,n^j(•)(t) with j∈V_j0(t).

At 640, the reference node (or the administrator node) executes a simultaneous localisation and mapping algorithm, SLAM, using all the sector contribution matrices {Γ_τ,n^j0(•)(t)}∪{Γ_τ,n^j(•)(t)|j∈V_j0(t)}, to determine the presence of any object in the environment of the reference node. The presence of an object is sensed by a signature formed by the respective sector contributions of angular sectors coming from the nodes {j₀}∪V_j0(t) such that their intersection is non-void. The reference node (or the administrator node) produces and updates, at each sensing instant, a mapping dictionary, 670, the entries of which correspond to these intersections, each entry containing a signature, a signature consisting of the sector contributions of the angular sectors involved in the intersection associated with said entry.

At 650, the reference node (or the administrator node) determines, for each of the entries in the dictionary in question, whether a change of signature has occurred with respect to the previous detection instant. Alternatively, the determination of a change can be made over the whole of the dictionary or over only part, associated with a region in space, using the calculation of a Kullback-Leibler divergence between the distribution of signatures relating to the current observation instant and the distribution of signatures relating to the previous detection instant.

If such a change is detected, the network at 660 adapts its allocation of radio resources (RRM), for example makes a handover or a switching of a sending beam.

FIG. 7 illustrates schematically the way in which the reference node determines the nodes that belong to its neighbourhood, V_j0(t), for each observation time window.

The neighbourhood of the node V_j0(t) can be determined by selecting (or by subsampling) from the set U of nodes in the network, or even a subset V thereof obtained previously in a network discovery phase, those the score of which is above a certain threshold.

At step 710, the reference node j₀ makes a preselection of the nodes in the network, for example by means of a random sampling. This preselection may alternatively have been obtained by a clustering method in the set U or V, for example a K-nearest neighbours method.

At step 720, a score is calculated for each of these nodes thus preselected.

The score of a node j is based on its local observation, o_j, this local observation being understood as the matrix of the sector contributions Γ_τ,n^j(•)(t) coming from its sensing matrix Γ_τ,n^j(t), as well as the local observation o_j0 of the reference node, i.e. the matrix of the sector contributions Γ_τ,n^j0(•)(t). The score of the node reflects the importance of this node in predicting the movement of an obstacle in the environment of the reference node j₀.

The score can be obtained from a cosine similarity metric, a Kullback-Leibler divergence, an attention mechanism, etc.

Alternatively, the score of a node can be given by a classification process, implemented by a neural network, supplying for each node the probability of the observation o_j being in the same class as o_j0. This classification process can be trained online in a centralised and supervised manner.

The reference node next makes, at 730, a discrimination of the preselected nodes on the basis of their respective scores. This discrimination can for example be made by retaining only the M nodes (M being a non-zero integer) that obtained the best scores or those that achieve a score above a certain threshold.

The nodes thus retained form the neighbourhood V_j0(t) of the reference node.

In the description of the invention, we have supposed that the network was deployed in a 2D space and that the angular sectors were defined in one plane. It is however clear for a person skilled in the art that the network can be deployed in a 3D space (in particular when it uses drones) and that the angular sectors can be defined by their angular extents in azimuth and elevation. Furthermore, the conjoint communication and sensing method of the present invention is not limited to sensing the movement of a single obstacle in the environment of the reference node, and that the movement of a plurality of objects can be sensed in the first embodiment, the localisation of the obstacles in addition being obtained in the second embodiment. Finally, we have supposed, in the description of the invention, that the interfering signals were sent by nodes in the network. When the spatial density of the nodes is insufficient, use can be made of intelligent reflecting surfaces (IRS) to artificially increase the density of these nodes by creating additional apparent nodes.

Claims

1. Conjoint communication and sensing method for sensing the movement of an object in the environment of a node, referred to as the reference node, of a wireless communication network, said reference node having previously established a radio link with a sending node of the network, said method comprising: determining a plurality of non-overlapping angular sectors in said environment and the vertex of which is the reference node, as well as a plurality of observation instants;measuring, in each of the observation instants, a measurement of a quantity characteristic of a signal-to-noise-plus-sector-interference ratio on said radio link, for each of the angular sectors, and constructs a sensing matrix Γτ,n(t), Γτ,nj0(t) the elements of which are the measurements of said characteristic quantity;implementing a blind separation of sources using said sensing matrix and extracts therefrom a matrix of sector contributions Γτ,n(•)(t), Γτ,nj0(•)(t)) of interfering signals in the various angular sectors; anddetecting the movement of any obstacle in the environment of the reference node from at least one history of said sector contributions.

2. Conjoint communication and sensing method according to claim 1, wherein a situation of blockage of said link is predicted from characteristics of the movement of the obstacle in the environment of the reference node.

3. Conjoint communication and sensing method according to claim 1, wherein, if such a situation of blockage of said link is predicted, a modification of the radio resources in the network is implemented.

4. Conjoint communication and sensing method according to claim 1, wherein the blind separation of sources is implemented by means of a singular value decomposition of the sensing matrix or a principal components analysis.

5. Conjoint communication and sensing method according to claim 1, wherein the reference node determines the nodes of its neighbourhood and makes a simultaneous mapping and localisation on said environment using the matrix of sector contributions that it has extracted as well as matrices of sector contributions that it has received, in the form of messages, from the nodes of its neighbourhood.

6. Conjoint communication and sensing method according to claim 5, wherein a mapping dictionary is constructed and updated, each entry in the dictionary corresponding to the intersection of a plurality of angular sectors the vertices of which are the reference node or nodes of its neighbourhood, each entry containing a signature consisting of the sector contributions of the angular sectors involved in the intersection associated with said entry.

7. Conjoint communication and sensing method according to claim 6, wherein, in the event of modification of at least one signature in the mapping dictionary between two observation instants, a change is detected and a reallocation of radio resources of the network is implemented.

8. Conjoint communication and sensing method according to claim 6, wherein a Kullback-Leibler divergence is calculated between the distribution of the signatures at a current observation instant and that of a previous observation instant, over the whole of the mapping dictionary or a part thereof, and in that a reallocation of radio resources of the network is implemented when the Kullback-Leibler divergence exceeds a predetermined threshold.

9. Conjoint communication and sensing method according to claim 5, wherein the neighbourhood of the reference node is obtained by making a preselection of nodes of the network by means of a random sampling or a K-nearest neighbours method, then calculates, for each node thus preselected, a score from its matrix of sector contributions and that of the reference node, the neighbourhood of this node being determined by the preselected nodes the score of which exceeds a predetermined threshold or by the M nodes having the highest scores, where M is a non-zero integer.

10. Conjoint communication and sensing method according to claim 9, wherein the score of a preselected node is calculated by means of a cosine similarity metric, a Kullback-Leibler divergence or an attention mechanism or by a neural network implementing a classification operation.