WO2025073505A1 - Method, device and system for transmitting data from a terminal when changing attachment - Google Patents

Abstract

The invention relates to a method for transmitting session data, wherein the transmission is characterised by a first data file in the terminal (Term), wherein the first file is associated with a first neural network (RN_Term_STA) between the terminal and a first access device (STA) to which the terminal is attached and comprises characteristics for transmitting the session data between the terminal (Term) and the first device using first time and frequency resources (R1), wherein the method is implemented by the terminal (Term) and comprises: - attaching (Attach) the terminal to a second access device (STB) concurrently with the existing attachment of the terminal to the first access device (STA); - once the terminal (Term) has been attached to the second access device (STB), updating (MaJ) a second file in the terminal using the first file, wherein the second file is associated with a second neural network (RN_Term_STB) and comprises characteristics for transmitting the session data between the terminal (Term) and the second device (STB) using time and frequency resources (R2) distinct from the first resources (R1); - transmitting (Trans) the subsequent session data via the second access device (STB) in accordance with the characteristics of the updated second file.

Description

Method, device and system for transmitting data from a terminal during a change of attachment

The present disclosure belongs to the field of communication networks and more particularly to techniques for transferring a terminal from a first access device to a second access device of a communication network. It aims more particularly to optimize the transmission of data following the transfer of the terminal by taking advantage of the attachment of the terminal with the first access device.

Advances in Artificial Intelligence (AI) have opened many doors and enabled the application of these techniques in many fields. Among these, there is the field of digital signal processing for telecommunications.

A first approach to using artificial intelligence techniques in the telecommunications field consists of implementing physical layer functions with techniques derived from AI, such as neural networks. In this functional block approach (coding, communication channel estimation, decoding, etc.), we replace the classic algorithm used in current telecommunications networks with an element derived from AI. Thus, the same error correction and optimization algorithm can be implemented either by a standard algorithm or by means of a neural network. This allows us to benefit from the advantages of such an approach (performance, complexity, implementation flexibility) while remaining transparent from the point of view of the architecture of the signal processing chain, as well as the protocols used. In this approach, the usual functions of combating the deleterious effects of the propagation channel through equalization or error correction remain, only the way of implementing them changes, the new method being based on AI techniques.

A second, more disruptive approach abandons the objective of seamlessly integrating into a deterministic signal processing chain. It is no longer a matter of preserving the blocks as described in the first approach, but rather of considering a communication channel globally, and acting on a communication without distinguishing the different blocks. The degrees of freedom that AI then benefits from are greater, as well as, consequently, its optimization potential. In a bidirectional data exchange scheme between two learning systems, communicating through a given propagation channel, this learning characteristic allows them to gradually learn the most efficient method for communicating. By following common optimization metrics, the two systems learn to best exploit the shared propagation channel and thus define an optimized communication system. From this perspective, it is not just a matter of canceling the harmful effects of the channel to get as close as possible to a perfect channel (Gaussian channel), but rather of finding the optimal way to exploit a given channel. If we take this process to its conclusion, we arrive at what the state of the art describes as communication between two autocoders having developed their own "language".

This second approach, although promising, nevertheless faces two limitations:

It can only be implemented between two systems with similar learning capabilities, and benefiting from an initial “primary” two-way communication capability in order to enable the establishment of the feedback loop necessary for learning.

The principle of specific optimization for a given propagation channel has the direct consequence that this optimization is only relevant for this specific implementation. However, in certain contexts, such as mobile radio communications, the propagation channel is not stable, but evolves regularly, which requires a form of continuous learning.

In the context of mobile networks, these developments in the data propagation channel can take two forms:

• Soit les évolutions sont très rapides par rapport au temps nécessaire à l’apprentissage sur un canal statique. Dans ce cas, cette variabilité sera intégrée au sein d’un apprentissage, quitte à limiter les performances de la configuration qui en découlera, car mal adaptée au canal courant
• Soit les évolutions sont lentes, c’est-à-dire que le phénomène concerné peut être considéré comme statique sur le temps d’apprentissage. La survenue de telles variation va donc nécessiter une mise à jour de l’apprentissage, pour adapter le modèle retenu à la nouvelle configuration du canal. Suivant l’ampleur de cette évolution, la remise en question de l’apprentissage existant sera plus ou moins profonde.

Continuous evolution: linked to natural phenomena of the channel, such as equipment movements, or changes in the propagation environment. These evolutions are progressive and take place, for example, between the terminal and a base station. They are detected and monitored, generally through the use of reference signals that allow channel knowledge to be updated. In the case of a communication system that has learned to communicate on the channel in question, two scenarios may arise:

• Either the changes are very rapid compared to the time required for learning on a static channel. In this case, this variability will be integrated into learning, even if it means limiting the performance of the resulting configuration, because it is poorly adapted to the current channel.
• Either the changes are slow, that is to say that the phenomenon concerned can be considered static over the learning time. The occurrence of such variations will therefore require an update of the learning, to adapt the chosen model to the new configuration of the channel. Depending on the extent of this change, the questioning of the existing learning will be more or less profound.

Discontinuous evolution: Here, the evolution of the radio channel is sudden and does not present continuity with the previous situation. It is no longer an evolution of the existing channel but rather a change to a new propagation channel. In the context of a telecommunications network, such evolutions are typically caused by a change of the attached access device during communication (also called handover in English).

This development results in the resetting of the learning protocol with a fallback to the primary communication mode during handover from one access device to another access device.

In the case of a change of attached access device, restarting the learning process from scratch is inevitable since the terminal <-> access device pair has changed. The return to the primary communication mode will most often involve a significant degradation in the efficiency of the radio link and therefore in the quality of service provided by the network. In addition, the learning process requires time, which can potentially significantly alter the transmission quality during this period. Finally, this learning process involves the consumption of radio resources and therefore a temporary loss of capacity.

We therefore see that discontinuous developments can be frequent and negatively impact the efficiency of the network and the quality of the user experience since the terminal will not quickly have good quality data transmission due to the complete learning of its neural network during the change of access.

The present disclosure aims to remedy all or part of the limitations of the prior art solutions, in particular those set out above, by proposing a solution which makes it possible to optimize the transmission of data between a terminal and a new access device by using techniques relating to neural networks, by taking advantage of a transmission of data from the terminal with a previous access device, this transmission also using techniques relating to neural networks.

For this purpose, a method of transmitting data of a session is proposed, said transmission being characterized by a first data file in the terminal, said first file being associated with a first neural network between said terminal and a first access device to which the terminal is attached, and comprising characteristics of transmission of the data of the session between said terminal and said first device using first time and frequency resources, said method being implemented by said terminal and comprising:
- an attachment of said terminal to a second access device concurrently with the existing attachment of said terminal to the first access device,
- following the attachment of said terminal to the second access device, an update using the first file of a second file in the terminal, said second file being associated with a second neural network, and comprising characteristics of transmission of the session data between said terminal and said second device using time and frequency resources distinct from the first resources,
- a transmission of the following session data via the second access device in accordance with the characteristics of the second updated file.

In a context of mutation (or handover) of a terminal attached or connected successively to two access devices, and using distinct transmission resources for the transmission of data of a session with the two devices, the terminal can exploit a neural network associated with the first access device to develop a neural network associated with the second access device. More particularly, the double attachment for a period of time of the terminal with the two access devices during the handover can be exploited to obtain a neural network associated with the second access device optimized more quickly for the transmission of the data of the session via the second access device. The terminal, implementing this method, can thus transmit and receive the data of a session with good quality independently of the access device to which it is attached. In the absence of the method, and although the terminal can be simultaneously attached to both access devices, the terminal would suffer a loss of quality during the handover because the neural network associated with the second device would not be sufficiently adapted to the current session when the terminal is detached from the first access device, in particular if the period of simultaneous dual attachment is too short, which is often the case. The use of resources distinct from the resources for transmitting data via the second device implies that the terminal can manage the two neural networks, or the four neural networks if the networks are specific to the transmission and reception of data, independently. Thus, it can optimize the communication with the second device, via the training of the associated neural network(s), without impacting the communication and optimization carried out with the first device. It should be noted that the updating of the data file may include the creation of the second data file.

According to one aspect of the invention, the transmission method is characterized in that the first neural network and the second neural network constitute a single neural network.

In the case where, for example, the terminal has fewer computing resources, the terminal may comprise a single neural network for communications with one or other of the devices involved in the handover. In this case, the update of the second file associated with the second neural network is characterized by an update of the single data file of the single neural network, or of the files of the neural networks respectively associated with the transmission of data and the reception of data from an access device.

The use of a single neural network for the connection to the first device and the second device does not negatively impact the data transmission process. The terminal optimizes its neural network for communication with the second device by taking advantage of the communication with the first access device. The terminal, by gradually degrading the communication with the first device while improving the communication with the second access device, thus optimizes the neural network, or its neural networks if it has a network specific to the transmission and a neural network specific to the reception of data. This optimization is for example carried out in accordance with the evolutions of the propagation channels with the two access devices, these evolving during the handover.

According to one aspect of the invention, the transmission method is characterized by the fact that the second data file is updated according to the evolution of the transmission quality of a transmission channel established during the attachment of the terminal with the second access device.

According to one embodiment, the second file comprises information from the first file and the second file is also updated according to a transmission channel, represented by a transmission value, such as RTT (Round Trip Delay Time) data or a reception power value of a signal transmitted by the second access device. Thus, the neural network associated with the second device is progressively adapted for the data session while having saved an initial training period of the neural network by exploiting the characteristics of the first file.

According to another aspect of the invention, the transmission method is characterized by updating the second data file, this second file comprising an improvement in the characteristics of data transmission with the second access device and a corresponding degradation in the characteristics of data transmission with the first access device.

In an embodiment where the terminal does not have separate neural networks, the terminal uses information relating to the respective data transmissions with the first access device and with the second access device, to update the file associated with the single neural network, and gradually optimize it so that it becomes progressively more adapted to the session data exchanged with the second access device to the detriment of the data exchanged with the first access device, until it is completely and solely adapted to the connection with the single second access device when the terminal disconnects from the first access device at the end of the handover. Thus, despite the provision of a single neural network, or two neural networks for the transmission and reception of data, the terminal benefits from good transmission quality during the handover period.

According to another aspect of the invention, according to the transmission method, the first neural network and the second neural network constitute distinct neural networks.

According to one embodiment, the terminal having for example more internal resources (memory, performance, etc.), the terminal then uses a separate neural network specific for each access device, or two specific neural networks (one associated with the transmission and one associated with the reception of data) for each access device, which represents for example four neural networks managed by the terminal for the transmission of session data. The terminal, having a neural network optimized for the session with the first access device, can advantageously progressively optimize its (or its) neural network(s) with the second access device without impacting the neural network with the first access device. The terminal thus has differently optimized neural networks during the handover and these are independent and adapted to each access device, allowing the deactivation of the neural network with the first access device without impacting the neural network established with the first access device.

According to yet another aspect of the invention, according to the transmission method, the first data file, respectively the second data file, is specific to the reception of the session data from the first access device, respectively the second access device.

So that the neural network is more particularly adapted to the uplink (transmission of data to the access devices) and to the downlink (reception of data from the access devices), the terminal can advantageously specialize the neural networks according to the direction of data transmission and thus benefit from neural networks particularly adapted to the transmission of data in transmission and reception, the transmission characteristics being able to vary according to the direction of data transmission considered.

According to another aspect of the invention, the transmission method further comprises
an update of a fourth file associated with a fourth neural network, said fourth file comprising parameters for transmitting session data to the second device, the update being carried out according to a third data file associated with a third neural network and comprising parameters for transmitting session data to the first device.

In the case where the terminal's neural networks are specialized for the uplink and the downlink, the terminal can advantageously use the transmission parameters of the neural network associated with the first device to optimize its transmission neural network with the second device, certain transmission parameters being identical or at least very close (transmission interfaces, propagation of the transmitted data, transmission protocols used, etc.) between the terminal and the access devices. Thus, the transmission neural network, specific and adapted to the uplink, is optimized more quickly. This option is also valid for neural networks associated with the downlinks, including parameters for receiving session data.

According to another embodiment, the transmission method further comprises detaching the terminal from the first access device and deleting the first data file associated with the first neural network.

Once the handover has been completed, meaning that the terminal is only attached to the second access device, and the neural network has been optimized by benefiting from the neural network of the terminal with the first access device, this neural network with the first access device can be deleted. This has the advantage of freeing up terminal resources and simplifying the management of the neural networks. If two neural networks, transmitting and receiving, were implemented, then these two neural networks can be deleted by deleting the associated data files.

The various aspects of the transmission method just described can be implemented independently of each other or in combination with each other.

The invention also relates to a device for transmitting data of a session, said transmission being characterized by a first data file in the terminal, said first file being associated with a first neural network between said terminal and a first access device to which the terminal is attached, and comprising characteristics for transmitting the data of the session between said terminal and said first device using first time and frequency resources, said device comprising a processor and a memory coupled to the processor and further comprising:
- an attachment module, configured to attach said terminal to a second access device concurrently with the existing attachment of said terminal to the first access device,
- following the attachment of said terminal to a second access device, an update module, configured to update, using the first file, a second file in the terminal, said second file being associated with a second neural network, and comprising characteristics for transmitting session data between said terminal and said second device using time and frequency resources distinct from the first resources,
- a transmission module, configured to transmit subsequent session data via the second access device in accordance with the parameters of the updated second file.

This transmission device is capable of implementing in all its embodiments the transmission method which has just been described. This transmission device can advantageously be instantiated in a terminal, such as a smartphone, a computer, a local network gateway (box).

The invention also relates to a system for transmitting data of a session, said transmission being characterized by a first data file associated with a first neural network between said terminal and a first access device to which the terminal is attached, said first file comprising characteristics of transmission of the data of the session between said terminal and said first device using first time and frequency resources, said system comprising:
- a transmission device as described above,
- the first access device adapted to transmit and receive data from the audit session or from the terminal using first time and frequency resources,
- the second access device adapted to transmit and receive data from the audit session or from the terminal using time and frequency resources distinct from the first resources.

The invention also relates to a computer program product comprising a set of program code instructions which, when executed by at least one processor, configure said at least one processor to implement the transmission method according to any of the embodiments of the present disclosure.

The invention also relates to a computer-readable recording medium having recorded thereon a set of program code instructions which, when executed by at least one processor, configure said at least one processor to implement a transmission method according to any of the embodiments of the present disclosure.

• [  : une représentation schématique d’un réseau de communication comprenant deux dispositifs d’accès dans lequel est mis en œuvre un procédé de transmission de données d’une session, selon un mode de réalisation,
   :  une deuxième représentation schématique d’un réseau de communication comprenant deux dispositifs d’accès dans lequel est mis en œuvre un procédé de transmission de données d’une session, selon un mode de réalisation,
•  :  une représentation schématique d’un réseau de communication comprenant deux dispositifs d’accès dans lequel est mis en œuvre un procédé de transmission de données d’une session, selon un mode de réalisation,
• : une représentation d’un réseau de neurones selon un mode de réalisation,
•  :  un diagramme illustrant les principales étapes d’un procédé de transmission de données d’une session, selon un mode de réalisation,
•  : une représentation d’un dispositif de transmission de données d’une session selon un exemple.

Dans ces figures, des références identiques d’une figure à une autre désignent des éléments identiques ou analogues. Pour des raisons de clarté, les éléments représentés ne sont pas à l’échelle, sauf mention contraire.The invention will be better understood by reading the following description, given as a non-limiting example, and made with reference to the figures which represent:

• [ : a schematic representation of a communication network comprising two access devices in which a method of transmitting data of a session is implemented, according to one embodiment,
  : a second schematic representation of a communication network comprising two access devices in which a method of transmitting data from a session is implemented, according to one embodiment,
• : a schematic representation of a communication network comprising two access devices in which a method of transmitting data of a session is implemented, according to one embodiment,
• : a representation of a neural network according to one embodiment,
• : a diagram illustrating the main steps of a method of transmitting data of a session, according to one embodiment,
• : a representation of a data transmission device of a session according to an example.

In these figures, like references from one figure to another designate identical or similar elements. For reasons of clarity, the elements shown are not to scale, unless otherwise stated.

More generally, it should be noted that the methods of implementation and production considered above have been described as non-limiting examples, and that other variants are consequently conceivable.

In the remainder of the description, a communication network is presented comprising access devices (cellular network access stations, Wi-Fi access stations, for example of the Box type, or any other access station providing access to the communication network, including a satellite access station, for example of the Internet type). The stations and equipment of the network can be instantiated in the form of specific equipment or in the form of virtualized functions.

We first refer to the which presents a schematic representation of a communication network comprising two access devices in which a method of transmitting data of a session is implemented, according to one embodiment.

This presents a communications network comprising two access devices, STA and STB, providing access to a cellular type communications network to a Term terminal. The Term terminal (mobile terminal, laptop, local network access equipment, for example, box type, etc.) has a transmission neural network RN_Term_STA_Em and a reception neural network RN_Term_STA_Rec allowing respectively to optimize the transmission, respectively transmission and reception of data of a session, with a remote device (terminal, data server, machine, IoT device, etc.) not shown in the These neural networks are made up of a succession of layers representing in a more or less abstract way signal processing functions linked to the propagation channel of the session data, coding and decoding functions, hardware imperfection compensation functions, etc. Each layer is made up of a set of neurons to which learnable scalar parameters (weights and biases) are associated. The learning process makes it possible to modify these weights and biases to optimize the overall function performed by the neural network.

The terminal stores two files in memory containing the data relating to the neural networks RN_Term_STA_Em and RN_Term_STA_Rec respectively. These files describe the architecture and parameters (learnable weights and biases) of the neural networks, so that the content of a file makes it possible to reproduce the neural network in question identically. This data file can be referred to interchangeably as a standardized file format or a standard for representing a neural network.

Similarly, the STA access device has respectively a transmission neural network (RN_STA_Em) and a reception neural network (RN_STA_Rec) whose role is to optimize the data session, respectively for transmission and reception, of the terminal Term with the remote device via the STA access device. These neural networks are also transcribed into one or more data files whose data, such as the transmission characteristics, vary as the propagation channel evolves and the session is increasingly optimized, thus improving the transmission of the session data, in transmission and reception.

A data session is established between the Term terminal and a remote device (terminal, data server, machine, IoT device, etc.) not shown on the via the STA access device of the Res communication network, the session data being conveyed via the Comm_A communication between the Term terminal and the STA access device. This may be an initialization of a session or a continuation of a session initiated via another access device.

The terminal Term is attached to the cellular access device STA and exchanges session data via a Comm_A communication. The Comm_A communication allows the transmission of data between the terminal Term and the access device STA using first time and frequency resources R1. Time resources (in English Time Slot) and frequency resources (in English Resource Bloc) are in fact allocated by the operator of the Res network for each communication between two devices attached to the Res network and, without modification compared to the techniques of the prior art, the terminal Term and the STA device transmit data to each other using transmission resources allocated to them by the operator of the Res communication network.

The respective neural networks for transmitting and receiving data from the Term terminal, RN_Term_STA_Em and RN_Term_STA_Rec, and from the STA device, RN_STA_Em and RN_STA_Rec, carry out a joint learning process with respect to this session allowing proper transmission and reception of data by the Term Terminal and the STA device. In the upstream channel, according to an example, the transmitting neural network RN_Term_STA_Em is trained jointly with the receiving neural network RN_STA_Rec. Similarly, in the downstream channel, the transmitter neural network RN_STA_Em is trained jointly with the receiver neural network RN_Term_STA_Rec… Thus, the neurons representing a transmission parameter specific to the session, the propagation channel or even a hardware interface of a Term or STA device are assigned initial weights (and biases) which will be modified as the training progresses until an optimal value is reached allowing routing of session data with better quality of service. The neural networks RN_Term_STA_Em, RN_Term_STA_Rec, RN_STA_Em and RN_STA_Rec are thus possibly different and are respectively specific to the Term terminal and the STA device, to the transmission or reception of data, and are also specific to the established session. In fact, they are not directly transposable or adapted to a communication established for example between the Term Terminal and the STB access device, for the same session or a separate session.

We then refer to the which presents a second schematic representation of a communication network comprising two access devices in which a method of transmitting data from a session is implemented, according to one embodiment.

When a transfer, also called handover, of the current HDVR session occurs from the STA device to the STB device, for example following the attachment of the Term terminal with the STB access device, a Comm_B communication is established between the Term terminal and the STB device for routing the session data via the STB device.

The STB access device has respectively a transmission neural network (RN_STB_Em) and a reception neural network (RN_STB_Rec) whose role is to optimize the transmission and reception of session data to and from the Term terminal.

The Term terminal is thus attached to the cellular device STB and exchanges session data via a Comm_B communication. The Comm_B communication allows the transmission of data between the Term terminal and the access device STB using second time and frequency resources R2, allocated by the network operator Res, distinct from the first resources R1 used for the transmission of data in the Comm_A communication.

This handover is implemented while the terminal Term can be simultaneously attached to the STA and STB devices for a period, thus making it possible to prepare the handover, according to the known “Make before Break” techniques where communication with any access device is never broken and at a given moment, the terminal Term is multi-attached to both access devices, STA and STB. During this period, the terminal Term is likely to receive data transmitted by the two access devices STA and STB. In the embodiment described in the , the terminal Term has only two neural networks, RN_Term_Em, RN_Term_Rec, these neural networks specific to transmission and reception being common to both communications Comm_A and Comm_B. In another embodiment, not described in the , the terminal includes different neural networks for the two communications, Comm_A (RN_Term_STA can be structured into RN_Term_STA_Em and RN_Term_STA_Rec as described for the and represented on the ) and Comm_B (RN_Term_STB can be structured into RN_Term_STB_Em and RN_Term_STB_Rec as on the ), and can therefore optimize its neural networks in a differentiated manner depending in particular on the evolution of the propagation of the data on the two communications Comm_A and Comm_B, knowing that the transmission quality will probably be more and more adapted to the session for the data of the communication Comm_B and less and less adapted for the data Comm_A since the HDVR handover is implemented from the STA device to the STB device but the neural networks of the terminal Term, specific to the communications Comm_A and Comm_B, will make it possible to optimize the two communications Comm_A and Comm_B independently. Indeed, the evolution of the transmission channel between the terminal Term and the access device STB should gradually improve, and the second file representing the neural network associated with the device STB, and therefore with the communication Comm_B, should be regularly updated in the optimization process to take into account the evolution of the transmission quality of the data of the session to or from the access device STB.

While the Term terminal is multi-attached to both STA and STB devices, it can update its file associated with the neural network according to the parameters of the Comm_B communication. The Term terminal therefore updates or creates a second data file based on the first data file, the latter being optimized for the Comm_A communication. This second file takes over some or all of the dynamic transmission characteristics of the first file and optimizes them according to the new context, namely the fact that the terminal is attached to an STB device different from the STA device and uses R2 resources distinct from R1. Nevertheless, certain transmission characteristics specific to the Term terminal, to the data propagation context (urban, rural), to the current session remain identical or close, which makes it possible to reduce the optimization time of the neural network(s) associated with the Comm_B communication and consequently to the STB access device and to the current session.

According to one example, when the terminal Term comprises only a single neural network, or according to one example a single neural network for transmitting data and a single neural network for receiving data, the terminal is likely to manage a single data file associated with the single neural network, for the communications Comm_A and Comm_B, or a single one for transmitting and a single one for receiving. In this example, the first file and the second file are only one and the same file which is updated according to the progress or the progress of the HDVR handover. Since the single data file per direction of transmission is not specific to the two access devices STA and STB, the terminal progressively optimizes its neural networks RN_Term_Em, RN_Term_Rec in the uplink and downlink channels with the access device STB. This implies a progressive degradation of performance with the access device STA. The neural network optimization process must relate the speed of optimization of communications between the terminal and the STB access device, i.e. the improvement of parameters, including QoS, RTT, latency, etc. of data transmission with the STB device, depending on the improvement of the propagation channel between the terminal and the STB device and the degradation of the parameters of the Comm_A communication.

In the case where the terminal has specific neural networks for the uplink dedicated to the transmission of data and the downlink dedicated to the reception of data by the Term terminal, as indicated in the , then the associated files can be specific to these specific neural networks in the transmission direction. In the case where the Term terminal also has specific neural networks for data transmission with both STA and STB devices, then the terminal can manage a total of four files associated with separate neural networks during the HDVR Handover. Correspondingly, if it does not have specific neural networks for communication with both STA and STB devices, for example due to a lack of resources, then the Term terminal only manages a maximum of two neural networks in total.

We then refer to the which presents a schematic representation of a communication network comprising two access devices in which a method of transmitting data of a session is implemented, according to one embodiment. According to this , at the end of the HDVR handover, the terminal is no longer attached to the STB device and all the session data is conveyed by means of the Comm_B communication established between the Term terminal and the STB access device. This data is then conveyed on a communication channel using the second R2 resources of time and frequency distinct from the first R1 resources. The Term terminal and the STB device transmit the session data in an optimized manner by means of the respective neural networks RN_Term_STB_Em, RN_Term_STB_Rec, RN_STB_Em and RN_STB_Rec. The method of transmitting data from a session has thus enabled the Term terminal to have a neural network, represented by the data file, optimized more quickly. This file or these files (for transmission and reception) representative of one (RN_Term_STB) or two neural networks (RN_Term_STB_Em, RN_Term_STB_Rec) is updated according to the devices (Term and STB), the propagation channel and thus allows a data transmission offering criteria of better quality of service, lower latency, improved throughput in comparison with a situation where the Term terminal would have had to implement a complete process of learning its neural networks, as in the known technique. According to an alternative, the Term terminal, once it is detached from the STA access device, that is to say it no longer has a connection with this STA access device, can delete the neural networks associated with the session data routed via this STA device. This operation thus makes it possible to free up resources of the Term terminal. The STA access device can do the same.

We then refer to the which describes a representation of a neural network according to one embodiment.

The neural network is for example the neural network (RN_STA_Em) of emission of the STA access device presented in the . Binary elements (eb) emitted by an output interface of the STA device to the Term terminal are converted into a signal (sig) adapted to the propagation conditions of a channel established between the STA device and the Term terminal. This signal is optimized by progressively modifying dynamic transmission characteristics associated with the neural network RN_STA_Em and included in a data file associated with this neural network RN_STA_Em. The neural network RN_STA_Em is thus composed of several layers C1, C2....,,C7 each of the layers intervening for the transmission of data via the optimized sig signal, that is to say adapted to the propagation channel between the STA device and the terminal so that the data routed between the STA device and the terminal guarantees a good quality of service for the session and subsequently for the service associated with the data. Each layer is composed of neurons as represented for layer C1 composed of neurons n1, n2, n3. At the output of a neuron, the data generated can be represented by the value y which is here equal to a weighted sum of the neuron's inputs to which a bias b_i is added. The weighting is done by a set of weights w_i, each input being associated with a weight. Each neuron n1, n2, n3 is therefore assigned different sets of weights w_i, which will evolve during the learning process just like the biases b_i, this information w_i and b_i can be called scalar parameters. For example, the data y is calculated according to the following formula y = phi(sum w_i x_i + b_i) where x_i are input values of the neuron. In the following layers, C2 to C7, these inputs are the outputs of the neurons of the previous layers. Other equations are possible to represent the calculation carried out by a neuron. All the layers of the RN_STA_Em neural network are thus optimized for the transmission of data to the terminal, guaranteeing a good quality of service for this transmission. Furthermore, in the case of data transmission from the STA device to a terminal, the RN_STA_Em neural network is trained jointly with the RN_Term_Rec reception neural network of the terminal so that the transmission quality actually benefits from a good quality of service. The terminal can, for its part, accelerate its learning process by taking advantage, for example, of a reception neural network used for receiving data from another device. According to one embodiment, the different layers C1, C2,....., C7 and/or the scalar parameters independently represent dynamic data processing characteristics. According to another embodiment, the correspondence between the dynamic processing characteristics and the layers C1,...,C7 is not as direct and several layers of neurons or even several neurons are involved for a characteristic. The transmission method as described allows weights and/or biases or even one or more layers of a neural network, this information representing, according to particular embodiments, dynamic transmission characteristics associated with a neural network of a device, associated with a communication, for example Comm_A, to be reused by the same device, for example a terminal, for another communication established by this terminal, such as the communication Comm_B. This transfer, executed during a handover, thus makes it possible to reduce the learning process of a certain number of weights and biases or even layers of neurons in the neural network of this other device for a communication, such as the communication Comm_B by exploiting neurons, layers of neurons, weights, biases or calculation formulas of a neural network associated with another communication of the terminal, such as the communication Comm_A.

We then refer to the which presents a diagram illustrating the main steps of a method for transmitting data from a session, according to one embodiment. The STA and STB access devices are access devices of a cellular network but the steps of the method are identical for different access devices, such as satellite access devices or Wi-Fi access devices in particular. The steps described below are implemented by a Term terminal, as described in figures 1 to 3, which can be a smartphone, a PC, an Internet box, on the sole condition that this Term terminal has communication interfaces allowing it to attach to the STA and STB access devices to be able to establish a data session, for example of the http type, with a remote device (terminal, data server, machine, IoT equipment, etc.).

In a step 300, a terminal is attached to a first access device and has a session in progress with a remote device, such as another terminal or a data server, and the data of this session is routed via the first access device. This terminal then attaches itself, according to an Attach procedure, to a second access device. The attachment procedure is not modified and is described in the documents of the prior art. This attachment procedure comprises for example a procedure for authenticating the terminal and the second access device, and an allocation of configuration data of the terminal by the second device. The access technology (cellular, Wi-Fi, satellite) may be different from the technology used for attaching the terminal to the first access device. The terminal uses R1 resources for transmitting data in frequency and time with the first access device. The data of the current session previously routed via the first access device alone using the resources R1 will then also be routed via the second access device, this data being routed using resources R2 in time and frequency distinct from the resources R1. The terminal also uses transmission and reception neural networks to improve the transmission of data between the terminal and the first access device. The first access device also implements transmission and reception neural networks to improve the transmission of data with the terminal. These neural networks include neurons representative of data propagation on an access channel, such as a radio channel, processing of session data, a communication interface and processing of the signals used for data transmission. When attaching the terminal to the second access device, the terminal may use the same transmitting and receiving neural networks used for the session data exchanged with the first device or may instantiate or use one or more neural networks specific to the session data exchanged with the second access device. The choice of one or other of the options depends on the terminal's capabilities, its configuration or external constraints, for example relating to a handover deployment context (single-operator or multi-operator for example).

In a step 301, the terminal updates, during an Update procedure, a second file comprising the data associated with the neural network for routing data between the terminal and the second access device by means of resources R2 distinct from the resources R1. This update may comprise the updating of two files or sub-files, relating to the neural networks associated with the transmission and reception of data. Depending on whether the neural network(s) associated with the session data exchanged with the second device are specific or not, the update may consist of updating the file associated with the first device or of instantiating a second file, distinct from the first file associated with the neural network for the data exchanged with the first device. In the case of a single neural network for communications with both devices, the update will include the creation of a component in the file for communication with the second device, this component being able to include data transmission characteristics copied from the component associated with the first access device or the modification, creation or deletion of certain characteristics associated with the first access device but not valid for communication with the second access device. In this case, the component can be functionally identified as a second file or a sub-file. In the case where the neural networks are distinct, the second file will be instantiated or updated and will include characteristics (neurons, weights, bias, calculation formula, etc.) transferred from the first file to the second file, then modified or not depending on the new data transmission context, and in particular the attachment to a distinct access device represented by the second access device. This update operation can be carried out by a specific module in the terminal or by an existing module for managing the neural network for example. In the case where the access devices are satellite access devices, the characteristics, in particular transmission characteristics, may include satellite navigation data (speed vector, position, etc.) which data may be used to determine radio parameters (Dopler, elevation angle, MIMO configuration, etc.). These satellite characteristics and radio parameters may be used to update the second file associated with the neural network of the terminal for communication with the second STB access device.

During a step 302, the terminal optimizes in an Optim procedure its neural network (or its transmission and reception neural networks) for the data of the session with the remote equipment (terminal, server, etc.) exchanged with the second access device. It should be noted that thanks to the update of the second file, the terminal does not need to restart the learning process from scratch requiring an update of all the transmission characteristics of the neural network associated with the session and the second access device. This return to a primary communication mode is in fact costly in resources for the terminal and it leads to a transmission of data between the terminal and the second device which would not be optimized as quickly as thanks to the implementation of the data transmission method. Indeed, thanks to the update process of step 301, the terminal has a pre-optimized neural network where certain transmission characteristics, linked for example to the session, to the terminal, to the propagation context (urban, rural) are optimized or else have practically optimized values. The optimization therefore includes the modification of the values of the second file associated with the neural network which could not be directly deduced from the first file. Indeed, if a handover does indeed correspond to a new propagation channel between the terminal and the second device, it only involves a limited change in the type of propagation channel. Indeed, if the type of environment does not change, such as for example during a handover from an access device in the countryside to a neighboring access device in the countryside, the handover results in a new propagation channel but which will correspond to the same statistical model. Therefore, the learning carried out with the first device can remain at least partially adapted to the new channel between the terminal and the second device and thus the neural network associated with the second device benefits from the characteristics of the neural network of the terminal associated with the session and the first access device.

In the case where the terminal has specific neural networks per access device, it already has an optimized model with the first access device, possibly for the upstream and downstream channels. It then duplicates this model, and optimizes it in parallel during its data transmissions with the second access device. As a result, the terminal will have, for a time, different and optimized communication models with the two access devices.

In the case where the terminal does not have specific neural networks per access device, it progressively optimizes its transmission and reception neural network(s) in uplink and downlink channels with the second access device. This implies a progressive degradation of performance with the first access device. The method must relate the speed of optimization of communications between the terminal and the second access device as a function of the improvement of the propagation channel between the terminal and this second access device together with a degradation of the data transmission characteristics between the terminal and the first access device.

According to an alternative, in the case where the update fails in the optimization step 302, the terminal performs during a step 303 an Altern operation of recourse to the primary operating mode consisting of implementing a complete learning process for all of the transmission characteristics of the neural network file associated with the transmission of data with the second device. This Altern operation inevitably leads to a degradation of the handover between the two access devices since the data routed between the terminal and the second device does not quickly benefit from a transmission quality permitted by the update process of step 302 and optimization 303 comprising the inheritance of transmission characteristics between the two communications (terminal - first access device and terminal - second access device).

Once the second file has been updated, during a step 304, following step 302 or 303, the data of the current session previously routed via the first access device alone by means of the resources R1 are then also routed via the second access device, these data being routed by means of resources R2 in time and frequency distinct from the resources R1. This step 304 corresponding to the transmission Trans of the session data via the second access device is enabled by a multi-attachment of the terminal and to a routing of the data of a session by two distinct channels, in accordance with transmission characteristics present in files associated with one or more neural networks, specific to each attachment or not.

In a step 306 Detach, the terminal then detaches from the first access device, following a loss of connectivity or by a voluntary action of the terminal and/or the first access device. The data of the application session is then routed by the second access device alone, in accordance with the transmission characteristics updated in the second file relating to the neural network associated with the second access device.

According to an alternative, during a step 306, the terminal having completed the handover and no longer being attached to the first device, for example by loss of the radio signal emitted by this first device or by a disconnection procedure implemented by the terminal, it performs an operation of erasing the first file associated with the communication with the first access device. In the case where it only has a single data file associated with the single neural network, the terminal erases the component of the file comprising the transmission characteristics specific to the session data exchanged between the terminal and the first access device.

We then refer to the which describes a representation of a device for transmitting data of a session, said transmission being characterized by a first data file in the terminal, said first file being associated with a first neural network between said terminal and a first access device to which the terminal is attached, and comprising characteristics for transmitting the data of the session between said terminal and said first device using first time and frequency resources. The transmission device can advantageously be instantiated in a Term terminal which can be for example a smartphone, a PC or access equipment of a local network such as a home gateway, the transmission device of the terminal being adapted to attach to a first access device and concurrently to a second access device

For example, the transmission device 100 comprises a processing unit 130, equipped for example with a microprocessor μP, and controlled by a computer program 110, stored in a memory 120 and implementing the transmission method according to the invention. The processor of the processing unit 130 controls the operations of the device 100. Upon initialization, the code instructions of the computer program 110 are for example loaded into a RAM memory, before being executed by the processor of the processing unit 130. Such a transmission device 100 comprises
- an attachment module 101, configured to attach said terminal to a second access device concurrently with the existing attachment of said terminal to the first access device,
- following the attachment of said terminal to a second access device, an update module 102, configured to update, using the first file, a second file in the terminal, said second file being associated with a second neural network, and comprising characteristics for transmitting session data between said terminal and said second device using time and frequency resources distinct from the first resources,
- a transmission module 103, configured to transmit subsequent Donn data of the session via the second access device in accordance with the characteristics of the second updated file.

The entities described and included in the device described in relation to the can be hardware or software. The illustrates only one particular way, among several possible ones, of carrying out the method detailed above, in relation to the preceding figures. Indeed, the technique of the invention is carried out indifferently on a reprogrammable computing machine (a PC computer, a DSP processor or a microcontroller) executing a program comprising a sequence of instructions, or on a dedicated computing machine (for example a set of logic gates such as an FPGA or an ASIC, or any other hardware module).

In the case where the invention is implemented on a reprogrammable computing machine, the corresponding program (i.e. the sequence of instructions) may be stored in a removable storage medium (such as for example a USB key, a floppy disk, a CD-ROM or a DVD-ROM) or not, this storage medium being partially or totally readable by a computer or a processor.

Claims (13)

A method of transmitting session data, said transmission being characterized by a first data file in the terminal (Term), said first file being associated with a first neural network (RN_Term_STA) between said terminal and a first access device (STA) to which the terminal is attached, and comprising characteristics of transmitting session data between said terminal (Term) and said first device using first time and frequency resources (R1), said method being implemented by said terminal (Term) and comprising:
- an attachment (Attach) of said terminal to a second access device (STB) concurrently with the existing attachment of said terminal to the first access device (STA),
- following the attachment of said terminal (Term) to the second access device (STB), an update (MaJ) using the first file of a second file in the terminal, said second file being associated with a second neural network (RN_Term_STB), and comprising characteristics of transmission of the session data between said terminal (Term) and said second device (STB) using time and frequency resources (R2) distinct from the first resources (R1),
- a transmission (Trans) of the following session data via the second access device (STB) in accordance with the characteristics of the second updated file.
A transmission method according to claim 1, wherein the first neural network and the second neural network constitute a single neural network.
Transmission method, according to claim 1 or claim 2, in which the second data file is updated according to the evolution of the transmission quality of a transmission channel established during the attachment of the terminal with the second access device.
A transmission method according to one of claims 2 to 3, wherein updating the second data file comprises an improvement in the data transmission characteristics with the second access device and a corresponding degradation in the data transmission characteristics with the first access device.
Transmission method, according to claim 1, wherein the first neural network (RN_Term_STA) and the second neural network (RN_Term_STB) are distinct neural networks. Transmission method according to one of claims 1 to 5, in which the first data file, respectively the second data file, is specific to the reception of the session data from the first access device, respectively the second access device.
Transmission method, according to claim 6, further comprising an update of a fourth file associated with a fourth neural network, said fourth file comprising parameters for transmitting session data to the second device, the update being carried out according to a third data file associated with a third neural network and comprising parameters for transmitting session data to the first device.
Transmission method, according to one of claims 1 to 7, further comprising a detachment of the terminal from the first device and the deletion of the first data file associated with the first neural network.
Device (100) for transmitting data of a session, said transmission being characterized by a first data file in the terminal (Term), said first file being associated with a first neural network (RN_Term_STA) between said terminal and a first access device (STA) to which the terminal is attached, and comprising characteristics for transmitting the data of the session between said terminal and said first device using first time and frequency resources, said device comprising a processor (130) and a memory (110) coupled to the processor (130) and further comprising:
- an attachment module (101), configured to attach said terminal to a second access device concurrently with the existing attachment of said terminal to the first access device,
- following the attachment of said terminal to a second access device (STB), an update module (102), configured to update, using the first file, a second file in the terminal, said second file being associated with a second neural network (RN_Term_STB), and comprising characteristics for transmitting session data between said terminal (Term) and said second device (STB) using time and frequency resources distinct from the first resources,
- a transmission module (103), configured to transmit subsequent data (Donn) of the session via the second access device (STB) in accordance with the characteristics of the second updated file. System for transmitting session data, said transmission being characterized by a first data file associated with a first neural network between said terminal and a first access device to which the terminal is attached, said first file comprising characteristics of transmission of the session data between said terminal and said first device using first time and frequency resources, said system comprising:
- a transmission device according to claim 9
- the first access device adapted to transmit and receive data from the audit session or from the terminal using first time and frequency resources,
- the second access device adapted to transmit and receive data from the audit session or from the terminal using time and frequency resources distinct from the first resources.
Terminal comprising a transmission device according to claim 9.
Program comprising instructions which, when the program is implemented by a processor, lead to implementing the transmission method according to one of claims 1 to 8.
Recording medium readable by a transmission device on which the program according to claim 12 is recorded

Images