Method, device and system for transferring attachment of a terminal from a first access device to a second access device

Abstract

The invention relates to a method for transferring attachment of a terminal from a first access device (STA) to a second access device (STB), wherein the terminal (Term) has previously established a data session (Sess) via the first access device (STA), and wherein the session data is routed between the terminal and the first access device (STA) using time and frequency resources (R1), wherein the method is implemented in the second access device (STB) and comprises receiving, from the first access device (STA), at least one transmission characteristic (Caract) of a first data file associated with a neural network (RN_STA), attaching the terminal to the second access device, updating a second data file on the basis of the at least one received characteristic, wherein the second file is associated with a neural network (RN_STB) comprising a plurality of characteristics for transmitting the session data specific to the session (Sess) and to the second access device (STB), and routing all subsequent session data (Sess) via only the second access device (STB) in accordance with the updated second data file, following detachment (404) of the terminal from the first access device.

Description

Method, device and system for transferring an attachment of a terminal from a first access device to a second access device

[0001] 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.

[0002] 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.

[0003] 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 for 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.

[0004] 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 cancelling out 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 approach to its conclusion, we end up with what the state of the art calls communication between two autocoders that have developed their own "language".

[0005] This second approach, although promising, nevertheless faces two limitations:

[0006] 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.

[0007] 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 that of mobile radio communications, the propagation channel is not stable, but evolves regularly, which requires a form of continuous learning.

[0008] In the context of mobile networks, these developments in the data propagation channel can take two forms:

[0009] 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 which allow the knowledge of the channel to be updated. In the case of a communication system having learned to communicate on the channel in question, two scenarios can arise:Either the evolutions are very rapid compared to the time required for learning on a static channel. In this case, this variability will be integrated into a learning process, even if it means limiting the performance of the resulting configuration, because it is poorly adapted to the current channel.Or the evolutions 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 evolution, the questioning of the existing learning will be more or less profound.

[0010] 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).

[0011] 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.

[0012] In the case of a change of the attached access device, restarting the learning process is inevitable since the terminal <-> access device pair has changed.

[0013] Returning 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, not guaranteeing good transmission quality during this period. Finally, this learning process involves the consumption of radio resources and therefore a momentary loss of capacity.

[0014] We therefore see that discontinuous developments can be frequent and negatively impact the efficiency of the network and the quality of the user experience.

[0015] The present disclosure aims to remedy all or part of the limitations of the solutions of the prior art, 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.

[0016] For this purpose, a method is proposed for transferring an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the method being implemented in the second access device and comprising: - reception from the first access device of at least one transmission characteristic of a first data file associated with a neural network, said file comprising a plurality of transmission characteristics of the session data specific to the session and to the first access device, - reception of an attachment request from the terminal,- an update of a second data file from the at least one received characteristic, said second file comprising a plurality of characteristics for transmitting the session data specific to the session and to the second access device, said plurality being associated with the routing of the session data via the second device using said time and frequency resources,- a routing of all the following data of the session via the second access device alone in accordance with the updated second data file, following the detachment of the terminal with the first access device.,

[0017] The transfer method is implemented between two access devices, these two devices being involved in the routing of data of a session established by a terminal performing a handover between these two access devices, the session data being transmitted using the same time and frequency resources, regardless of the access device routing the session data. The transfer method is advantageous since it allows the second access device to update more quickly its neural network, used to optimize the routing of data sent to the terminal or received from the terminal, during or after the handover.The transfer of the terminal's access between the two access devices includes a phase during which the terminal is attached or connected simultaneously to both access devices and communicates with both access devices using unique transmission means, namely time and frequency resources common to the terminal's communications with both of the access devices.By transmitting one or more transmission characteristics of a data file, specific to the session and to the first device, to the second access device prior to the attachment of the terminal to the second access device, this second access device can update or initialize its file associated with its own neural network, specific to itself and to the current session, by reusing part of the transmission characteristics of the first file, thus saving it a complete initialization of the neural network learning process, which can slow down on the one hand the increase in quality of the data transmission via the second access device as well as the completeness of the handover when the terminal is detached from the first access device.When a neural network is indicated, it may be understood that the neural network may comprise a neural network and an associated file for transmitting data to the terminal as well as a neural network and an associated file for receiving data from the terminal. The transfer method comprises, in this case, the transmission of a characteristic of a file per neural network, for transmission and reception, or of a common file for the two transmitting and receiving neural networks. The second device then updates two neural networks via two or one file.

[0018] According to one aspect of the invention, the transfer method further comprises, prior to detaching the terminal from the first access device, routing the session data received from the terminal, in substitution for the first access device, in the case where the processing of the data received by the second access device is greater than a reference value.

[0019] During the handover, the two communications between the terminal and the two respective access devices use the same communication resources, in time and frequencies, and the two devices therefore receive the same signals, in separate channels, from the terminal having for its part a single transmitting neural network. In the absence of combination of the signals, each device takes charge of the received signal and processes it, for example by demodulating it. It is therefore appropriate that one of the two signals processed by each of the access devices is actually taken into account and that the data resulting from this processing is routed to the destination.The embodiment, based on a reference value, for example data having zero errors or a transmission quality greater than a threshold, makes it possible to select one of the two access devices, for the selection of the device whose data will be routed to the remote equipment with which the terminal has established the data session. Knowing that the handover is implemented from a first access device to a second access device, this second access device should more likely be selected as the signal received by the terminal from the second device improves.

[0020] According to another aspect, the transfer method further comprises combining the session data received from the terminal with the session data received by the first access device, prior to detaching the terminal from the first access device.

[0021] According to another embodiment, the second access device has the ability to combine the session data received from the terminal with the data received from the terminal by the first access device while the terminal is simultaneously attached to both access devices. Thus, the second device can reconstruct the session data by combining the data received by both devices, from the terminal, the latter having a single neural network for transmitting session data to both access devices.

[0022] According to a particular embodiment, the transfer method further comprises a modification of a transmission characteristic specific to the transmission of data to the terminal, included in the second file, as a function of the evolution of a capacity of a propagation channel used for routing data between the terminal and the second access device.

[0023] In the downlink, that is to say for the transmission of session data to the terminal, the terminal receives the data on two propagation channels, respectively between the terminal and the first access device on the one hand and the terminal and the second access device on the other hand. The second access device, to which the terminal is transferred during the handover, updates its transmission characteristics in the file associated with the neural network or the transmission neural network, so that the weight of the signal, and therefore of the data transmitted by the second device, has a preponderance over the data transmitted by the first device, thus indicating to the terminal which recombines the received signals that the data from the second device are to be taken into account with increasing priority in the recombination of the signal for the data received from the two access devices.

[0024] According to another particular embodiment, the transfer method is characterized by the updating of the second file, the second file comprising a creation of a second data file, the second file comprising characteristics of transmission of the session data specific to the session and to the second access device, said characteristics being derived from a complete learning of a neural network with which the new file is associated, in the case where the at least one transmission characteristic from the first device is not exploitable.

[0025] In the unfavorable case where the transmission of the characteristic between the two access devices is not possible or the received characteristic cannot be used by the second access device due to a transmission problem, a version problem, a file problem, then the second access device performs a complete reset of its file associated with its neural network (or neural networks if it has a transmission neural network and a reception neural network), requiring learning of all the transmission characteristics of the file.

[0026] The various aspects of the transfer process just described can be implemented independently of each other or in combination with each other.

[0027] The invention also relates to a method for mutating an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the method being implemented in the terminal and comprising:- a transmission of an attachment request from the terminal to the second access device,- a transmission of the session data via the first access device and the second access device, using the time and frequency resources common for communications with the two access devices, as a function of at least one transmission characteristic of a data file associated with a neural network of the terminal,common for communications with both access devices,- a detachment of the terminal with the first access device and then the transmission of subsequent data of the session via the second access device only, depending on the at least one transmission characteristic of the data file updated according to the detachment.,

[0028] The terminal, having a single neural network (this neural network may comprise a neural network for transmitting data and a neural network for receiving data) uses the same time and frequency resources to communicate with the two access devices. When the handover occurs and the terminal is further attached to the second device and then only attached to it, it must gradually adapt its data file associated with its neural network, for example by varying the scalar parameters, such as the weights of each access device as the handover occurs until the weights associated with the first device become zero, the data session then being of the unicast type, i.e. between the terminal and the second device.

[0029] According to one aspect, the mutation method further comprises combining the session data received from the first access device and the second access device, based on at least one transmission characteristic, specific to the reception of the session data, the at least one characteristic being included in the data file, prior to detachment of the terminal with the first access device.

[0030] Prior to the completion of the handover, corresponding to the situation where the terminal is no longer attached to the second access device only, the terminal is multi-attached to both devices and receives the same data from both access devices, possibly with different transmission qualities. The terminal can advantageously modify the file associated with its neural network to increase the weights of the second access device in the neural network, as the quality of the signal between the terminal and the second access device, and as a corollary, to decrease the weight of the data of the first device in its file, thus giving increasing priority to the data received from the second device, in the combination of data, as the handover progresses.

[0031] According to another aspect, the mutation method is characterized by the fact that the terminal comprises two data files associated with two common neural networks for the two access devices in the terminal, one for receiving the session data from the access devices and the other for transmitting the data to the access devices.

[0032] The terminal can advantageously comprise two neural networks, respectively for the transmission and reception of session data, so as to optimize the neural network for each data transmission channel, and thus improve the quality of transmission, respectively of reception of data, for the terminal, each neural network being specifically adapted to the constraints of each transmission direction.

[0033] The various aspects of the mutation process just described can be implemented independently of each other or in combination with each other.

[0034] The invention also relates to a device for transferring an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the device comprising a transmitter, a receiver, a processor and a memory coupled to the processor with instructions intended to be executed by the processor to:

[0035] - receiving from the first access device at least one transmission characteristic of a first data file associated with a neural network, said file comprising a plurality of transmission characteristics of the session data specific to the session and to the first access device,

[0036] - receive an attachment request from the terminal,

[0037] - updating a second data file from the at least one received characteristic, said second file comprising a plurality of session data transmission characteristics specific to the session and to the second access device, said plurality being associated with the routing of the session data via the second device using said time and frequency resources, - routing all of the following session data via the second access device alone in accordance with the second data file, following the detachment of the terminal with the first access device.

[0038] This transfer device is capable of implementing in all its embodiments the transfer method which has just been described.

[0039] The invention also relates to a device for mutating an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the mutation device comprising a transmitter, a receiver, a processor and a memory coupled to the processor with instructions intended to be executed by the processor to:- transmit to the second access device a request for attachment of the terminal,- transmit session data via the first access device and the second access device, using the time and frequency resources common for communications with the two access devices,based on at least one transmission characteristic of a data file associated with a neural network of the terminal, common for communications with the two access devices,

[0040] - transmitting subsequent session data via the second access device only, following detachment of the terminal with the first access device, based on the at least one transmission characteristic of the data file, updated based on the detachment of the terminal with the first access device.

[0041] This mutation device is capable of implementing in all its embodiments the mutation method which has just been described.

[0042] The invention also relates to a system for transferring an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the system comprising:- a transfer device as described above,- a mutation device as described above,- a change device comprising a transmitter, a receiver, a processor and a memory coupled to the processor with instructions intended to be executed by the processor to transmit to the second access device the at least one transmission characteristic, and to ensure the detachment of the terminal with the first access device.

[0043] The invention also relates to a terminal comprising a mutation device as described previously.

[0044] The terminal can be a smartphone, a PC or a local network access device such as a home gateway (box).

[0045] The invention also relates to computer program products comprising a set of program code instructions which, when executed by at least one processor, configure said at least one processor to implement the respective transfer and mutation methods according to any of the embodiments of the present disclosure.

[0046] 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 transfer method or a mutation method according to any of the embodiments of the present disclosure.

[0047] The invention will be better understood upon reading the following description, given by way of 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 the methods of transferring an attachment of a terminal from a first access device to a second access and mutation device are implemented, according to one embodiment, : a schematic representation of a communication network comprising two access devices in which the methods of transferring an attachment of a terminal from a first access device to a second access and mutation device are implemented, according to one embodiment,: a representation of a neural network according to one embodiment: a diagram illustrating the main steps of a method for transferring an attachment of a terminal from a first access device to a second access device, according to one embodiment, : a representation of a device for transferring an attachment of a terminal from a first access device to a second access device, according to one example, : a representation of a device for mutating an attachment of a terminal from a first access device to a second access device, according to one example.,

[0048] 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 indicated.

[0049] More generally, it should be noted that the methods of implementation and embodiment considered above have been described as non-limiting examples, and that other variants are consequently conceivable.

[0050] 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.

[0051] We first refer to the which presents a schematic representation of a communication network comprising two access devices STA and STB in which a method of transferring an attachment of a terminal from a first access device to a second access device is implemented, according to one embodiment.

[0052] This presents a communications network comprising two access devices, STA and STB, providing access to a cellular type communications network, to a Term terminal.

[0053] A data session Sess is established between the terminal Term and a device Disp via the access device STA of the communication network Res, the session data being carried via the communication Comm_A between the terminal Term and the access device STA. This can be an initiation of a session or a continuation of a session initiated via another access device of the network Res.

[0054] 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 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 network Res for each communication between two devices and, without modification compared to the techniques of the state of the art, the terminal Term and the device STA transmit data to each other using transmission resources allocated to them by the operator of the communication network Res.

[0055] The Term terminal (mobile terminal, laptop, local network access equipment, for example, such as a box, etc.) has a neural network RN_Term that can optimize transmission, including the transmission and reception of data from a Sess session with a remote device Disp. The Term terminal includes a file representing the data relating to the RN_Term neural network. The file thus includes data relating to the neurons as well as their respective weights. The different layers of the neural networks thus represent hardware descriptions of the terminal (processor, interfaces, etc.), signal processing functions linked to the data propagation channel of the Sess session, coding and decoding functions, etc. The different neurons in each layer also include weights indicating their importance in the neural network.All this information is transcribed into a file transcribing the RN_Term neural network. This data file can be referred to as a standardized file format or a standard for representing a neural network. This RN_Term neural network is made up of a succession of layers representing, in a more or less abstract way, signal processing functions linked to the session data propagation channel, 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 allows these weights and biases to be modified to optimize the overall function performed by the neural network.

[0056] The Term terminal stores a file containing data related to the RN_Term neural network. This file describes the architecture and parameters (learnable weights and biases) of the neural network, so that the contents of a file can be used to reproduce the neural network in question identically.

[0057] Correspondingly, the STA access device has a neural network RN_STA whose role is to optimize the data session by the STA station both with regard to the transmission of session data to the Term terminal and with regard to the reception of session data from the Term terminal. This neural network is also transcribed into a data file whose data, such as the transmission characteristics, vary as the propagation channel between the STA access device and the Term terminal evolves and as the data session is increasingly optimized, thus improving the transmission of session data, in transmission and reception.

[0058] The respective neural networks of the terminal Term, RN_Term and the access device STA, RN_STA carry out a joint learning process with respect to this session allowing a good transmission and reception of data by the Terminal Term and the device STA. Thus, the neurons representative of a dynamic transmission characteristic specific to the session, to the propagation channel or even to a hardware interface of a device (Term or access device STA) are assigned initial weights (and biases) which will be modified as the learning progresses until an optimal value allowing a routing of the session data with a better quality of service.These neural networks RN_Term and RN_STA are thus possibly different and are respectively specific to the Term terminal and the STA access device, and include data relating to the transmission and 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.

[0059] Reference is then made to which presents a schematic representation of a communication network comprising two access devices in which the methods of transferring an attachment of a terminal from a first access device to a second access device and mutation are implemented, according to one embodiment.

[0060] When a transfer, also called handover, of the current HDVR session occurs from the STA access device to the STB access device, a Comm_B communication is established between the Term terminal and the STB device for routing the Sess session data with the Disp device via the STB device.

[0061] The STB access device has a neural network (RN_STB) whose role is to optimize the transmission of data, emission and reception, of the session of the Term terminal with the remote device Disp by the STB device.

[0062] 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 the same R1 time and frequency resources as those used for the transmission of data in the Comm_A communication.

[0063] This HDVR handover is implemented while the Term terminal can be simultaneously attached to the STA and STB devices for a period, thus making it possible to prepare the handover, the communication with an access device never being broken and at a given moment, the Term terminal is multi-attached to both access devices, STA and STB. During this period, the Term terminal is likely to receive data from the Sess session transmitted by the two access devices STA and STB. The STA and STB access devices receive the same signals from the terminal, since they are transmitted on the same R1 resources, except for the propagation channel specific to the STA and STB access devices. These signals come from a single RN_Term neural network present in the Term terminal.

[0064] The method for transferring an attachment of a terminal from a first access device STA to a second access device STB thus allows the STA device to transmit to the STB device, successively or prior to the attachment of the terminal Term to the STB device, one or more transmission characteristics of the session, included in a file associated with the neural network RN_STA. The STA device can thus transmit to the STB device, the file representative of its neural network RN_STA or a part thereof, such as for example one or more transmission characteristics of the file. Thus, the access device STB will have a set of reusable characteristics for its own data file associated with its neural network RN_STB, after the attachment or at the time of the attachment of the terminal Term and will be able to avoid complete training of its neural network for the established communication Comm_B.Indeed, certain transmission characteristics of the file, relating to the session, to the STA device, to routing conditions, may be reused for the RN_STB neural network. The optimization time of the RN_STB neural network will be reduced and the transmission characteristics specific to the session via the STB station will be adapted more quickly to this communication Comm_B than if a complete learning process had had to be carried out. According to an example, the STA device may decide to transmit the entire file representing the RN_STA neural network or transmit only a part of the file, such as, for example, certain transmission characteristics that will need little or no modification for optimization.The transmitted transmission characteristics can advantageously represent a scalar parameter such as a weight or a bias, or even information representing a neural network, such as a neuron or a layer of neurons.

[0065] The transmission of the characteristic Caract makes it possible, according to one example, to transmit the types and configuration of neural networks used for the realization of the different functions of the signal processing chain for the communication Comm_A, in order to constitute a possible starting point for a new learning (in English, transfer learning and finetuning method). To accelerate the learning of the STB device, the access device STA can indicate all the layers of the neural network or those that need to be optimized for the communication Comm_B. From this model, the access device STB can start to optimize the uplink for the data of the session Sess received from the terminal Term.

[0066] Once it has received the file or the characteristic Caract from the STA device, the STB access device updates or creates a second file, representative of its neural network RN_STB for the Sess session, and modifies the neurons and / or the weights of the neurons to adapt it to the Comm_B communication. This update of the file received from the STA device is carried out according to a parameter of the data Sess session established via the second STB access device. Thus, the characteristics specific to the STB access device or to the conditions for routing the data between the terminal and the STB device can be updated because this type of characteristic, specific to the STB device or to the propagation channel between the terminal and the STB device for example, is not present in the data file or the characteristics Caract received.The modification of the data transmission parameters to the terminal may thus evolve according to the propagation channel between the STB device and the terminal and require the updating of the data file associated with the neural network of the STB device.

[0067] Alternatively, the STA and STB access devices do not recombine the data received from the Term terminal. Each STA and STB access device performs a local demodulation and if at least one of the two devices obtains, for example, an error-free packet, corresponding, for example, to a reference value, then this device is selected by the network. The optimization of the upstream path (data transmission to the Term Terminal) must be done progressively with the STB device (neural network of the Term terminal and neural network of the STB device). In this situation, even if the terminal transmits its data to both STA and STB devices, using the same transmission resources R1, only one of the two devices and most likely the STB device to which the HDVR handover occurs routes the data received from the Term terminal to the remote device Disp.

[0068] Alternatively, the STA and STB access devices constructively recombine the Sess session data received from the Term terminal. This recombination can occur in one of the two STA or STB access devices, or in another device in the Res network, connected directly or indirectly to the STA and STB access devices, responsible for this recombination. For example, a BBU (BaseBand Unit) type device responsible for the STA and STB devices can ensure this recombination in the context of a cellular network.

[0069] According to an example, if the transfer process fails and the creation of the second file based on the first updated file cannot be carried out, then a complete training process of the neural network of the STB device for the session must be initiated. This complete training process does not allow for optimized training and faster implementation of good quality Comm_B communication but it offers a backup solution or fallback in the case where the feature Caract or the file cannot be correctly transmitted or cannot be used by the STB device.

[0070] Alternatively, in the downstream channel, when the Term terminal is multi-attached to both access devices STA and STB, the data from the STA and STB access devices coming from the Disp device and going to the Term terminal are constructively recombined to facilitate their demodulation by the Term terminal. The Term terminal then has a single neural network RN_Term in reception which it optimizes to demodulate the common signal received from the two STA and STB devices. The STA and STB devices optimize their respective neural networks RN_STA and RN_STB for the downstream channel (transmission of data to the Term terminal) by taking into account the recombination of the signals at the Term terminal, thus forming a jointly optimized distributed system.As the propagation channel between the Term terminal and the STB device improves, the weight of the STB device transmission becomes increasingly strong in the recombined signal while the contribution of the STA device decreases. The neural networks within the terminal and the base station B, respectively RN_TERM and RN_STB, are progressively optimized to ensure improved communication for the transmission of the Sess session data between the Term terminal and the STB access device. The contribution of the STA device therefore gradually becomes a minority, until the link with the STA device is cut and the HDVR handover is finalized.

[0071] Indeed, when the HDVR handover is finished, and the Term terminal no longer receives a signal from the STA device or for some reason, the attachment of the Term terminal to the STA device is broken, all the data of the Sess session with the Disp device are routed via the STB device, in accordance with the transmission characteristics of the second file associated with the RN_STB neural network. This file associated with the RN_STB neural network has been trained more quickly, thanks to obtaining the characteristics of the first file to be improved more quickly. This improvement thus allows routing of the session data with quality of service, latency, transport delay characteristics adapted to the propagation environment and more quickly corrected, according to the intrinsic characteristics of the Comm_B communication, to allow good quality of service for the transmission of the session data.

[0072] We then refer to which describes a representation of a neural network according to one embodiment.

[0073] The neural network is for example the neural network (RN_STA) of the STA access device presented in the. Binary elements (eb) 15 emitted by an output interface of the STA device to the terminal Term are converted into a signal (sig) adapted to the propagation conditions of a channel established between the STA device and the terminal Term. This signal is optimized by progressively modifying dynamic transmission characteristics associated with the RN_STA neural network and included in a data file associated with this RN_STA neural network. The RN_STA neural network is thus composed of several layers C1, C2....,,C7 each of the layers involved in the transmission or reception of data via the optimized sig signal, i.e. 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 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 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 neural network are thus optimized for the transmission and reception of data to or from the terminal, guaranteeing good quality of service for this transmission.Furthermore, in the case of data transmission from the STA device to a terminal, the neural network RN_STA, or more precisely a neural network for transmitting data, is trained jointly with the receiving neural network RN_Term of the terminal, or more precisely with a neural network for receiving data, 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 receiving 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 adaptation 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, to be transmitted and reused by another device, thus making 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.

[0074] Reference is then made to which presents a diagram illustrating the main steps of a method for transferring an attachment of a terminal from a first access device to a second access device, according to one embodiment.

[0075] During a step 400, a data session Sess is established between a terminal and a remote device, this remote device being able to be a terminal, a server, a machine adapted to communicate with the terminal or any type of equipment adapted for establishing a data session with the terminal. The data of this session are routed within a communication network via a first access device, such as an eNodeB type device or a gateway of a cellular network or an access device to a satellite network. The terminal and the first access device each comprise a neural network whose information corresponding to transmission characteristics (neurons, weights, bias) of the session data are transcribed into a data file.The terminal's neural network includes transmission characteristics related to the terminal (interface, processing capabilities, communication protocols), to the session (protocols, latency, quality of service, coding, decoding, etc.) and to the first access device with which communication is established for routing the session data. The neural network of the first access device also includes the type of characteristics indicated above, but these characteristics are specific to the first access device. The values ​​of the transmission characteristics of these neural networks have evolved as these neural networks are trained, depending on the Sess session data transported, the Sess session data transport characteristics, the signal propagation characteristics between the terminal and the first device, and the evolution of the communication channel between the terminal and the first device.Once the learning process is complete, the terminal and the first device respectively have neural networks allowing routing of the session data with the best possible quality of service given the session establishment conditions (interfaces, propagation, context of session establishment, etc.). The Sess session data are routed using time and frequency resources allocated by the network operator Res.

[0076] In a step 401, the terminal performs a transfer, also called a handover, to a second access device, the communication between the terminal and the second device using the same technology or a different technology than the communication between the terminal and the second access device, the second access device being managed by the same operator as the first access device or by a separate operator. This handover may be caused by a movement of the terminal, or by a lower quality of reception of the signal emitted by the first access device, or by a detection of the second access device. The terminal transmits an attachment request to the second access device and establishes, during an Etab procedure, the session with the remote Disp device via the second access device while maintaining the session established via the first access device.The objective is to maintain the Sess session established during the handover at all times. The time and frequency resources used for data exchanges between the terminal and the second access device are the same as those used for session data exchanges between the terminal and the first access device. At this time, the session data between the terminal and the remote device are routed on two separate paths, one via the first access device and the other via the second access device, while using the same time and frequency resources for both terminal-first device and terminal-second device communications.

[0077] In a step 402, the second access device receives, during a Recept operation from the first access device, a transmission characteristic of the data file associated with the neural network of the first device, or the complete file. This transmission characteristic is specific to the first device and to the session but can thus be used by the second device so that the communication between the terminal and the second access device for the transport of the data of the Sess session is optimized more quickly and more efficiently.Indeed, certain transmission characteristics, linked to the session, to the signal propagation context, of the terminal are identical or close for the two communications, between on the one hand the terminal and the first device and on the other hand between the terminal and the second device, and the transfer of one (or more) transmission characteristics is likely to facilitate the learning of the neural network of the second access device. The communication between the terminal and the second device can thus be optimized more quickly and the data of the Sess session routed between the terminal and the second device can be routed more quickly with good quality of service, the learning process of the neural network of the second device being trained more quickly thanks to the transmitted characteristics.This also has the advantage of being able to improve the handover time by being able to switch the routing of the session data more quickly via the second access device alone. The first device can select from the transmission characteristics present in its data file, those that it transfers to the second access device in order to limit the size of the information or of the file transmitted and also to limit itself to the characteristics that the second device can use as is or that the second device will have to modify in the optimized learning process that it will have to implement for the communication established with the terminal for the routing of the session data. In the case where the access devices are satellite access devices, the characteristics, in particular transmission characteristics, can include satellite navigation data (speed vector, position, etc.)) this data can be used to determine radio parameters (dopler, elevation angle, MIMO configuration, etc.). These satellite characteristics and radio parameters can be used to update the second file associated with the terminal's neural network for communication with the second access device.

[0078] In a step 403, the second access device updates, during an Update procedure, its file associated with its neural network based on the characteristics received from the first device. This is then the learning phase of its own neural network for routing the session data by proposing the best routing quality in view of the signal propagation conditions. The parameters of the data session are taken into account for the update.Knowing that certain transmission characteristics transmitted by the first access device can be reused without modification, the update concerns only the characteristics which vary with respect to the characteristics received or which have not been transmitted, such as for example the characteristics linked specifically to the second access device, such as the type of interface, the capacities of this interface, the protocols or versions of protocols available, the processing capacities of the second device. The propagation conditions of the communication channel between the terminal and the second device can also vary with respect to the corresponding characteristics of the first device, even if these variations are probably limited since the handover occurs in a space probably offering equivalent data routing conditions from one access device to the other.

[0079] In a step 404, the terminal detaches during a Detach procedure from the first access device. This detachment may be sudden, for example if the terminal no longer receives the radio signal emitted by the first access device, or voluntary on the part of the terminal or the first access device. The terminal may decide to detach once the update step is completed and communicated by the second access device or once the joint learning is completed. This detachment may then be faster because the neural network of the second device will have reached a stable state more quickly thanks to the accelerated learning enabled by the reception of the transmission characteristics.The first access device may also initiate the detachment of the terminal once it has transmitted one or more transmission characteristics, or even its data file, or after a period of time following the sending of the file to allow time for the second device to carry out the accelerated learning process of its neural network, for example in conjunction with the learning of the terminal's neural network.

[0080] In a step 405, once the terminal is detached from the first access device, the second access device routes in an Achem process all the data of the session Sess by always using the same time and frequency resources, also previously used for the communication between the terminal and the first access device, and by exploiting its optimized neural network, comprising modified transmission characteristic values ​​adapted to the propagation channel between the terminal and the second device, thus allowing a good quality transfer of the session data to or from the terminal.

[0081] In the method for transferring an attachment of a terminal from a first access device to a second access device, and as described in the, knowing that on the one hand the terminal uses the same time and frequency resources for both paths when it is multi-attached to the two access devices and that on the other hand, the terminal comprises only one neural network, the terminal is adapted to demodulate the common data received from the two access devices. Similarly, in this method, the access devices are able to recombine in a concerted manner the data received from the terminal, or, according to another example, the two devices individually process the received data by demodulating them, and obtaining better demodulation by one of the two devices qualifies it as a device selected for the transmission of the demodulated data to the remote device.

[0082] Alternatively, in a step 406, the first access device deletes its own data file associated with its neural network for transmitting session data between the terminal and the remote device, in order to free up resources in the first access device.

[0083] In another embodiment, the terminal and / or one or both access devices comprise a neural network associated with the transmission of session data and a neural network associated with the reception of session data. The neural network specific to the session and to a device (terminal, access device) as presented above is then structured into two neural networks, respectively for the transmission and reception of data. According to this mode, the first device transmits characteristics present in the files associated with the two transmission and reception neural networks, or included in a single file comprising the transmission characteristics specific to the transmission and reception of session data.The second access device then updates two data files, associated respectively with its transmitting and receiving neural networks, based on the transmitting and receiving transmission characteristics transmitted by the first access device or a single file if the characteristics of the transmitting and receiving neural networks are included in a single file.

[0084] Reference is then made to which presents a device for transferring an attachment of a terminal from a first access device to a second access device, implementing the transfer method, according to one aspect of the invention.

[0085] The transfer device 100 can be implemented in an access station of a cellular network, a Wi-Fi network or a satellite network. It can for example be, without limitation, a NodeB, eNodeB, NR type device or even a Wi-Fi gateway or access equipment of a local network, for example of the residential type (Box).

[0086] For example, the transfer device 100 comprises a transmitter 101, a receiver 102, 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 transfer method according to the invention. At 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.

[0087] Such a transmitter (101), such a receiver (102), such a processor (130) and such a memory (120) are configured to:

[0088] - receiving from the first access device (STA) at least one transmission characteristic (Caract) of a first data file associated with a neural network, said file comprising a plurality of transmission characteristics of the session data specific to the session (Sess) and to the first access device (STA),

[0089] - receive an Attach request from the terminal,

[0090] - updating a second data file from the at least one received characteristic, said second file comprising a plurality of session data transmission characteristics specific to the session and to the second access device (STB), said plurality being associated with the routing of the session data via the second device (STB) using the time and frequency resources (R1),- routing all of the following session data via the second access device only in accordance with the updated second data file, following the detachment of the terminal (Term) with the first access device (STA).

[0091] We then refer to the which presents a device for transferring an attachment of a terminal from a first access device (STA) to a second access device (STB).

[0092] The mutation device 200 can be implemented in a terminal, which can more specifically be a 3G, 4G, 5G…xG compatible smartphone, a PC, an IoT (Internet of Things) device, a box or an access gateway of a residential network.

[0093] For example, the mutation device 200 comprises a transmitter 201, a receiver 202, a processing unit 230, equipped for example with a microprocessor μP, and controlled by a computer program 210, stored in a memory 220 and implementing the mutation method according to the invention. At initialization, the code instructions of the computer program 210 are for example loaded into a RAM memory, before being executed by the processor of the processing unit 230. 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 method for determining a programmable device according to the invention. 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.

[0094] Such a transmitter (101), such a receiver (102), such a processor (130) and such a memory (120) are configured to:

[0095] - transmitting to the second access device (STB) a request for attachment of the terminal,- transmitting session data via the first access device and the second access device, using the time and frequency resources common to the two access devices, according to at least one transmission characteristic of a data file associated with a neural network of the terminal, common to the two access devices,

[0096] - transmitting subsequent session data via the second access device only, following detachment of the terminal with the first access device, based on the at least one transmission characteristic of the data file updated based on the detachment.

[0097] The entities described and included in the devices described in relation to Figures 5 and 6 may be hardware or software. Figures 5 and 6 illustrate only one particular way, among several possible ones, of carrying out the methods 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).

[0098] 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

Method for transferring an attachment of a terminal (Term) from a first access device (STA) to a second access device (STB), the terminal (Term) having previously established a data session (Sess) via the first access device (STA) and said session data being routed between the terminal and the first access device (STA) using time and frequency resources (R1), the method being implemented in the second access device (STB) and comprising: - a reception (402) from the first access device (STA) of at least one transmission characteristic (Caract) of a first data file associated with a neural network (RN_STA), said file comprising a plurality of transmission characteristics of the session data specific to the session (Sess) and to the first access device (STA), - a reception (401) of an attachment request from the terminal (Term),- an update (403) of a second data file from the at least one received characteristic, said second file associated with a neural network (RN_STB), comprising a plurality of characteristics for transmitting the session data specific to the session (Sess) and to the second access device (STB), said plurality being associated with the routing of the session data (Sess) via the second device using said time and frequency resources (R1),- a routing (405) of all the following data of the session (Sess) via the second access device (STB) only in accordance with the updated second data file, following the detachment (404) of the terminal with the first access device., Transfer method, according to claim 1, further comprising, prior to detaching the terminal from the first access device, routing the session data received from the terminal, in substitution for the first access device, in the case where the processing of the data received by the second access device is greater than a reference value. The transfer method of claim 1, further comprising combining the session data received from the terminal with the session data received by the first access device, prior to detaching the terminal from the first access device. Transfer method, according to one of claims 1 to 3, comprising a modification of a transmission characteristic specific to the transmission of data to the terminal, included in the second file, as a function of the evolution of a capacity of a propagation channel used for routing data between the terminal and the second access device. Transfer method, according to one of claims 1 to 4, according to which the updating of the second file comprises a creation of a second data file, the second file comprising characteristics of transmission of the session data specific to the session (Sess) and to the second access device (STB), said characteristics being derived from a complete learning of a neural network (RN_STB) with which the new file is associated, in the case where the at least one transmission characteristic from the first device is not exploitable. Method for mutating an attachment of a terminal (Term) from a first access device (STA) to a second access device (STB), the terminal having previously established a data session (Sess) via the first access device (STA) and said session data (Sess) being routed between the terminal (Term) and the first access device (STA) using time and frequency resources (R1), the method being implemented in the terminal (Term) and comprising:- a transmission of an attachment request from the terminal to the second access device (STB),- a transmission of the session data via the first access device (STA) and the second access device (STB), using the time and frequency resources (R1) common for communications with the two access devices (STA, STB),depending on at least one characteristic (Caract) of transmission of a data file associated with a neural network (RN_Term) of the terminal (Term), common for communications with the two access devices (STA, STB), - a detachment of the terminal with the first access device (STA) then the transmission of following data of the session (Sess) via the only second access device (STB), depending on the at least one characteristic (Caract) of transmission of the data file updated according to the detachment., A mutation method, according to claim 6, comprising a combination of the session data received from the first access device and the second access device, as a function of at least one transmission characteristic, specific to the reception of the session data, the at least one characteristic being included in the data file, prior to detachment of the terminal with the first access device. A mutation method according to claim 6 or claim 7, wherein the terminal comprises two data files associated with two common neural networks for the two access devices in the terminal, one for receiving session data from the access devices and the other for transmitting data to the access devices. A device (100) for transferring an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the device comprising a transmitter (101), a receiver (102), a processor (130) and a memory (120) coupled to the processor with instructions intended to be executed by the processor for:- receiving from the first access device at least one transmission characteristic of a first data file associated with a neural network, said file comprising a plurality of transmission characteristics of the session data specific to the session and to the first access device,- receiving an attachment request from the terminal,- updating a second data file from the at least one received characteristic, said second file associated with a neural network (RN_STB), comprising a plurality of session data transmission characteristics specific to the session and to the second access device, said plurality being associated with the routing of the session data via the second device using said time and frequency resources,- routing all of the following session data via the second access device only in accordance with the updated second data file, following the detachment of the terminal with the first access device., Device (200) for mutating an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the mutation device comprising a transmitter (201), a receiver (202), a processor (230) and a memory (220) coupled to the processor with instructions intended to be executed by the processor to:- transmit to the second access device a request for attachment of the terminal,- transmit session data via the first access device and the second access device, using the time and frequency resources common to the two access devices,depending on at least one transmission characteristic of a data file associated with a neural network of the terminal, common for the two access devices, - transmitting subsequent data of the session via the second access device only, following the detachment of the terminal with the first access device, depending on the at least one transmission characteristic of the data file, updated depending on the detachment of the terminal with the first access device., A system for transferring an attachment of a terminal from a first access device to a second access device, the terminal having previously established a data session via the first access device and said session data being routed between the terminal and the first access device using time and frequency resources, the system comprising:- a transfer device according to claim 9,- a mutation device according to claim 10,- a change device, included in the first access device (STA), comprising a transmitter, a receiver, a processor and a memory coupled to the processor with instructions intended to be executed by the processor to transmit to the second access device the at least one transmission characteristic, and to ensure the detachment of the terminal with the first access device. Terminal comprising a mutation device according to claim 10. Program comprising instructions which, when the program is implemented by a processor, lead to implementing the transfer method according to one of claims 1 to 5. Program comprising instructions which, when the program is implemented by a processor, lead to implementing the transfer method according to one of claims 6 to 8. Information medium readable by an access station (STA) and comprising instructions of a computer program according to claim 13. Information medium readable by a terminal (Term) and comprising instructions of a computer program according to claim 14.

Images