CN118784507A - A model monitoring processing method and device - Google Patents

Abstract

The present invention provides a model monitoring processing method and device, which relate to the field of communication technology. The method of the present invention: a first node receives model monitoring related information sent by a second node; the first node performs at least one of the following according to the model monitoring related information: model monitoring; distribution mode of decision model; training mode of decision model. The solution of the present invention achieves the purpose of improving the performance of AI/ML models in node applications.

Description

A model monitoring processing method and device

Technical Field

The present invention relates to the field of communication technology, and in particular to a model monitoring processing method and device.

Background Art

With the development of technology, artificial intelligence (AI) and machine learning (ML) have been applied to the communication field to improve the performance of communication. Among them, AI/ML models can be distributed in various network nodes.

However, how to ensure the performance of AI/ML models in node applications has become a technical problem that needs to be solved urgently.

Summary of the invention

The purpose of the present invention is to provide a model monitoring processing method and device to achieve the purpose of improving the performance of AI/ML models in node applications.

In order to achieve the above object, an embodiment of the present invention provides a model monitoring processing method, including:

The first node receives the model monitoring related information sent by the second node;

The first node performs at least one of the following according to the model monitoring related information:

Model monitoring;

How decision models are distributed;

How to train the decision model.

Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, it also includes:

Before receiving the first information, the first node sends first configuration information to the second node, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, it also includes:

Before receiving the first information, the first node receives a third indication sent by the second node, where the third indication is used to notify the first node that the second node has acquired the first information.

Optionally, it also includes:

Before receiving the second information, the first node receives second configuration information sent by the second node, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, it also includes:

Before receiving the second information, the first node sends a sixth indication to the second node, where the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, it also includes:

Before receiving the second information, the first node sends third configuration information to the second node, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, it also includes:

Before receiving the second information, the first node receives a ninth indication sent by the second node, where the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, it also includes:

Before receiving the second information, the first node sends the first information to the second node.

Optionally, the first information includes at least one of the following:

Input of the channel state information CSI generation part;

Output of the CSI reconstruction part;

Compressed CSI;

Beam assist information.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Original channel;

Precoding matrix indication PMI;

Eigenvector;

Precoder matrix.

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the distribution mode of the decision model includes at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

Optionally, the training method of the decision model includes centralized or distributed training of the decision model.

In order to achieve the above object, an embodiment of the present invention further provides a model monitoring processing method, including:

The second node sends model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, it also includes:

Before sending the first information, the second node receives first configuration information sent by the first node, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, it also includes:

Before sending the first information, the second node sends a third indication to the first node, where the third indication is used to notify the first node that the second node has acquired the first information.

Optionally, it also includes:

Before sending the second information, the second node sends second configuration information to the first node, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, it also includes:

Before sending the second information, the second node receives a sixth indication sent by the first node, where the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, it also includes:

Before sending the second information, the second node receives third configuration information sent by the first node, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, it also includes:

Before sending the second information, the second node sends a ninth indication to the first node, where the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, it also includes:

Before sending the second information, the second node receives the first information sent by the first node.

In order to achieve the above-mentioned purpose, an embodiment of the present invention further provides a model monitoring processing device, comprising: a memory, a transceiver, and a processor: the memory is used to store program instructions; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the program instructions in the memory;

The transceiver performs the following operations: receiving model monitoring related information sent by the second node;

The processor performs the following operations: performing at least one of the following according to the model monitoring related information:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In order to achieve the above object, an embodiment of the present invention further provides a model monitoring processing device, including:

A first receiving module, used to receive model monitoring related information sent by the second node;

A processing module is configured to perform at least one of the following according to the model monitoring related information:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In order to achieve the above-mentioned purpose, an embodiment of the present invention further provides a model monitoring processing device, comprising: a memory, a transceiver, and a processor: the memory is used to store program instructions; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the program instructions in the memory, and the transceiver is used to perform the following operations:

Sending model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In order to achieve the above object, an embodiment of the present invention further provides a model monitoring processing device, including:

A first sending module, used to send model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In order to achieve the above objective, an embodiment of the present invention further provides a processor-readable storage medium, wherein the processor-readable storage medium stores program instructions, and the program instructions are used to enable the processor to execute the method described above.

The above technical solution of the present invention has at least the following beneficial effects:

In the above-mentioned technical scheme of the embodiment of the present invention, after receiving the model monitoring related information sent by the second node, the first node can perform at least one of model monitoring, distribution method of the decision model, and training method of the decision model based on the model monitoring related information to ensure that the model used by each second node has good performance and the model utilization rate is relatively high, so that the model distribution achieves the overall optimality among generalization performance, training complexity, and reasoning performance.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a flow chart of a method according to an embodiment of the present invention;

FIG2 is a second schematic flow chart of the method according to an embodiment of the present invention;

FIG3 is a structural block diagram of a device according to an embodiment of the present invention;

FIG4 is a schematic diagram of a module of a device according to an embodiment of the present invention;

FIG5 is a second structural block diagram of the device according to an embodiment of the present invention;

FIG. 6 is a second schematic diagram of modules of the device according to the embodiment of the present invention.

DETAILED DESCRIPTION

In the embodiments of the present invention, the term "and/or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and/or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character "/" generally indicates that the associated objects before and after are in an "or" relationship.

In the embodiments of the present application, the term "plurality" refers to two or more than two, and other quantifiers are similar.

The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

In addition, in this embodiment, "instruct" may also be understood as "request", and "obtain" may also be understood as "receive" or "generate".

The embodiment of the present invention provides a model monitoring processing method and device. The method and device are based on the same application concept. Since the method and device solve the problem in a similar principle, the implementation of the device and the method can refer to each other, and the repeated parts will not be repeated.

As shown in FIG1 , a model monitoring processing method provided by an embodiment of the present invention includes:

Step 101, a first node receives model monitoring related information sent by a second node;

Step 102: The first node performs at least one of the following according to the model monitoring related information:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In this way, the first node follows the above steps and, after receiving the model monitoring related information sent by the second node, can perform at least one of model monitoring, distribution method of the decision model, and training method of the decision model based on the model monitoring related information to ensure that the model used by each second node has good performance and a relatively high model utilization rate, so that the model distribution achieves an overall optimum among generalization performance, training complexity, and inference performance.

The model may be an AI/ML model.

Optionally, in this embodiment, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

That is, the first node can complete model monitoring based on the received first information; the first node can decide the distribution mode and/or training mode of the model based on the received second information. Of course, when the first node decides the distribution mode and/or training mode of the model, it can combine the second information with other information, or it can not use the second information.

For the implementation mode of the first node performing model monitoring, optionally, the method further includes:

Before receiving the first information, the first node sends first configuration information to the second node, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

That is, the first node will send the first configuration information to the second node, so that the second node can obtain and/or send the first information after receiving the first configuration information, so as to perform model monitoring. Of course, if the first node has stored/can obtain all the information for model monitoring, the second node does not need to obtain and/or send the first information, and the first node may not send the first configuration information.

Optionally, if the first configuration information includes the first indication, the first configuration information may also include the sending configuration of the first information, such as the sending resources used, the sending period, etc.

In this embodiment, optionally, the second node obtains the first information, which may be by obtaining the first information from the terminal through the air interface, such as receiving the first information sent by the terminal via the air interface; or by obtaining auxiliary information required to generate the first information from the first node or the terminal, and the second node generates the first information based on the auxiliary information.

Optionally, the auxiliary information for generating the first information may be compressed channel state information (Channel State Information, CSI), input of a CSI generation part (input of CSI generation part), output of a CSI reconstruction part (output of CSI reconstruction part), etc.

Optionally, in this embodiment, the method further includes:

Before receiving the first information, the first node receives a third indication sent by the second node, where the third indication is used to notify the first node that the second node has acquired the first information.

That is, before sending the first information, the second node will send the third indication to the first node to inform the first node that it has obtained the first information so that the first node can obtain it when necessary. Specifically, after the second node obtains the first information based on the second indication and/or the third information, it sends the third indication to the first node.

Here, after receiving the third indication, the first node may instruct the second node to send the first information, for example, the first node sends the first indication to the second node. That is, the second node will dynamically send the first information in response to the needs of the first node. The first node may also instruct the second node to send the first information through an independent indication rather than an indication in the configuration information. Of course, the second node may also send the first information to the first node directly (automatically) or periodically (for example, a preset period or the first indication is a periodic sending indication) after obtaining the first information.

In this way, after receiving the first information, the first node can complete the model monitoring and obtain the model monitoring result. If the distribution mode and/or training mode of the decision model is required, the first node can complete it based on the model monitoring result.

For the implementation method of the second node performing model monitoring, optionally, the first node can decide the distribution mode and/or training mode of the model. Optionally, the method further includes:

Before receiving the second information, the first node receives second configuration information sent by the second node, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

That is, the second node will send the second configuration information to the first node, so that the first node can obtain and/or send the first information after receiving the second configuration information, so that the second node can receive the first information sent by the first node and perform model monitoring based on the first information.

Optionally, if the first configuration information includes the fourth indication, the second configuration information may further include the sending configuration of the first information, such as the sending resources used, the sending cycle, etc.

In this embodiment, optionally, the first node obtains the first information, similar to the second node, and may obtain the first information from the terminal through the air interface, such as receiving the first information sent by the terminal via the air interface; or may obtain auxiliary information required to generate the first information from the second node or the terminal, and the first node generates the first information based on the auxiliary information.

Optionally, in this embodiment, the method further includes:

Before receiving the second information, the first node sends a sixth indication to the second node, where the sixth indication is used to notify the second node that the first node has acquired the first information.

That is, before sending the first information, the first node sends the sixth indication to the second node to inform the second node that it has obtained the first information so that the second node can obtain it when necessary. Specifically, after the first node obtains the first information based on the fifth indication and/or the fourth information, it sends the sixth indication to the second node.

Here, optionally, after receiving the sixth indication, the second node may instruct the first node to send the first information, for example, the second node sends the fourth indication to the first node. That is, the first node will dynamically send the first information in response to the needs of the second node. The second node may also instruct the first node to send the first information through an independent indication rather than an indication in the configuration information. Of course, the first node may also send the first information to the second node directly (automatically) or periodically (for example, a preset period or the fourth indication is a periodic sending indication) after obtaining the first information.

In this way, after receiving the first information, the second node can complete the model monitoring and obtain the model monitoring result. Optionally, the second node can also send the second information to the first node, and the first node decides the distribution mode and/or training mode of the model.

Optionally, in this embodiment, the method further includes:

Before receiving the second information, the first node sends third configuration information to the second node, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

That is, the first node will send the third configuration information to the second node, so that the second node can obtain and/or send the second information after receiving the third configuration information. Of course, if the first node has stored/can obtain all the information for model monitoring, the second node does not need to obtain and/or send the second information, and the first node may not send the third configuration information.

Optionally, if the third configuration information includes the seventh indication, the third configuration information may also include the sending configuration of the second information, such as the sending resources used, the sending cycle, etc.

In this embodiment, optionally, the second node performs model monitoring based on the first information to obtain the model monitoring result in the same manner as the first node performs model monitoring, which is not described in detail here. After the model monitoring, the second node obtains the second information.

Optionally, in this embodiment, the method further includes:

Before receiving the second information, the first node receives a ninth indication sent by the second node, where the ninth indication is used to notify the first node that the second node has acquired the second information.

That is, before sending the second information, the second node will send the ninth indication to the first node to inform the first node that it has obtained the second information so that the first node can obtain it when necessary. Specifically, after the second node obtains the second information based on the eighth indication, it sends the ninth indication to the first node.

Here, optionally, after receiving the ninth indication, the first node may instruct the second node to send the second information, for example, the first node sends the seventh indication to the second node. That is, the second node will dynamically send the second information in response to the needs of the first node. The first node may also instruct the second node to send the second information through an independent indication rather than an indication in the configuration information. Of course, the second node may also send the second information to the first node directly (automatically) or periodically (for example, a preset period or the seventh indication is a periodic sending indication) after obtaining the second information.

In this way, after receiving the second information, the first node can decide the distribution method and/or training method of the model.

Optionally, in this embodiment, the method further includes:

Before receiving the second information, the first node sends the first information to the second node.

That is, if the second node performs model monitoring, the first node will decide the distribution method and/or training method of the model based on the second information. The first node will send the first information to the second node before receiving the second information so that the second node can perform model monitoring.

Optionally, in this embodiment, the first information includes at least one of the following:

Channel State Information (CSI) generation part input (input ofCSI generation part);

Output of CSI reconstruction part;

Compressed CSI;

Beam auxiliary information.

Optionally, the compressed CSI is CSI compressed using an AI/ML model encoder.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Raw channel;

Pre-coding Matrix Indicator (PMI);

eigenvectors;

Precoder matrix (precoders matrix).

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the beam width may be a 3dB beam width.

Optionally, the content of the first information may be determined based on different use cases, for example, a spatial-frequency domain CSI compression sub use case, and the first information may be one or more of the following:

Input of CSI generation part (can be one or more of raw channel, PMI, eigenvectors, precoders matrix);

output of CSI reconstruction part (can be one or more of raw channel, PMI, eigenvectors, precoders matrix);

CSI (CSI compressed using AI/ML model encoder).

For example, in a beam prediction use case, the first information may include beam auxiliary information (which may be one or more of beam shape-related information, radiation pattern, 3dB beam bandwidth, gNB coding index, receiving beam angle, terminal location, and base station transmitting antenna angle).

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the model monitoring result information may be information indicating the performance of the model (such as the output value of the performance monitoring algorithm, etc.). The enumeration type ENUMERATED{Good performance, Poor performance} may be used to define good performance or poor performance. The output value of the specific performance monitoring algorithm depends on different use cases and different monitoring algorithms.

Optionally, the monitoring algorithm may be squared generalized cosine similarity (SGCS), system throughput, or the like.

Optionally, the model identification information may be: model index/identification (model id), function index/identification (functionality id), configuration index/identification (configuration id), model parameter, model input/output, etc.

Optionally, the functional identification information may be: functionality id, model id, configuration id, model parameters, model input/output, etc.

Optionally, the time information may be the length of time that the monitoring result is stable or continuous.

It should be noted that model monitoring is to determine whether the performance of the current model (AI/ML model) is good or meets the requirements. There are many ways to judge the performance of AI/ML models, which may be based on different usage scenarios and functions. For model performance monitoring of AI/ML models on both sides, it is necessary to consider AI/ML model monitoring on both sides. In this embodiment, the model monitoring methods may include: monitoring based on inference accuracy; monitoring based on system performance; monitoring based on data distribution; monitoring based on applicable conditions.

Optionally, in this embodiment, the distribution mode of the decision model includes at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

That is, the first node decides the model distribution mode based on the second information, that is, whether the model distribution adopts centralized, fully distributed or semi-distributed. If the model distribution adopts semi-distributed, the first node can also decide the model allocation based on the second information, such as which models are used on which second node.

It should be known that: 1. Centralized: It can be understood that the first node is a centralized node, the second node is a plurality of distributed nodes under the first node, and each second node uses the same AI/ML model;

2. Fully distributed: It can be understood that the first node is a centralized node, and the second node is multiple distributed nodes under the first node, and each second node uses a different AI/ML model;

3. Semi-distributed: It can be understood that the first node is a centralized node, and the second node is multiple distributed nodes under the first node, where some of the second nodes can use the same AI/ML model, and some of the second nodes use different AI/ML models.

Specifically, taking the CU-DU separation scenario as an example: if the first node determines that one or more of the following conditions (not limited to the following conditions) are met based on the second information, the first node needs to re-distribute the model, that is, optimize the model distribution (the optimization here refers to changing the model distribution method):

1. Currently using the "centralized approach", but one or more of the following problems occur:

1) The model monitoring results show that the number of second nodes with poor model performance is higher than a certain threshold, or the model monitoring results show that the number of second nodes with good performance is lower than a certain threshold, such as more than 50% of the second nodes have poor model performance, or less than 30% of the second nodes have good model performance;

2) Model monitoring results for a specific model(s) are poor or do not work;

3) There are sufficient models available, such as the number of models is greater than the number of second nodes;

2. Currently using "Fully Distributed", but one or more of the following problems occur:

1) The model training cost is too high, such as the training cost exceeds the preset value;

2) The model storage space is limited, such as the storage space is less than the preset value;

3) A specific model has good generalization performance, such as a generalization performance greater than a preset value;

4) Model management is difficult;

5) Exceeded the maximum number of models;

3. Currently using "semi-distributed", but one or more of the following problems occur:

1) Model training costs are too high;

2) Model storage space is limited;

3) A specific model has good generalization performance;

4) Model management is difficult;

5) The maximum number of models has been exceeded.

In addition, in this embodiment, the model allocation in the semi-distributed manner may be determined based on one or more of the following information:

1. Historical model usage of each second node;

2. Model monitoring result information of the historical model used by each second node;

3. Current model usage of each second node;

4. Model monitoring result information of the model currently used by each second node;

5. Generalization ability information of each model;

6. Model priority information (if any);

7. Priority information of the second node (if any);

8. The scenarios/business information applicable to each model;

9. Current scenario/business information of each second node.

Optionally, the priority information of the model is used to indicate whether the model is assigned/used first or whether the model is assigned/used first in certain specific usage scenarios/specific second nodes.

Optionally, the priority information of the second node is used to indicate whether to preferentially allocate model information to the second node.

Optionally, in this embodiment, the training method of the decision model includes centralized or distributed training of the decision model.

That is, the first node decides whether to adopt centralized or distributed model training for (re)training based on the second information, so as to further improve the performance and obtain the desired model.

Optionally, when the first node finds that there is no effective or better model that can meet the application requirements, such as the first node expects to use a "centralized approach" based on some considerations (such as storage considerations), but currently no model has the generalization performance that can meet the requirements, then the first node can make the following decisions:

1. Conduct "centralized training", that is, train a model whose generalization performance can meet the requirements. This "centralized training" can obtain the model training data set information collected by each second node through the first node, and use the received model training data set information of each second node to perform model training, so that the trained model has better generalization ability and can be applied to each second node; this "centralized training" can also collect model training data set information from each second node and perform their own model training respectively, and send their trained models to the first node. The first node trains a unified overall model based on the models trained by each second node. This overall model has better generalization ability and can be applied to each second node;

2. Perform "distributed training", that is, each second node trains the model separately, so that each model can achieve better performance on this second node. This "distributed training" can collect model training data set information from each second node and perform their own model training separately, and apply the trained model to this second node.

It should also be noted that, in this embodiment, optionally, the first node and the second node may be one or more of the following combinations:

The first node is the central unit (CU), and the second node is the distributed unit (DU);

The first node is the control plane of the CU (CU-CP), and the second node is the user plane of the CU (CU-UP);

The first node is the core network (including the Location Management Function (LMF)), and the second node is the base station;

The first node is a server/operations administration and maintenance (OAM), and the second node is a core network/base station.

Optionally, the combination of the first node and the second node is not limited to the above combinations, that is, the first node may be a DU and the second node may be a CU, which is not limited here.

Optionally, there may be one or more second nodes under the first node, and the first node may obtain the first information and/or the second information from the multiple second nodes.

In this embodiment, information can be obtained through the air interface, and air interface signaling can be used, such as Radio Resource Control (RRC) signaling, Media Access Control Element (MAC CE) signaling, Downlink Control Information (DCI), etc. Information can be obtained through the interface, and interface signaling can be used, such as F1 interface signaling, E1 interface signaling, Xn interface signaling, X2 interface signaling, other interface signaling, etc.

Below, the first node is taken as node A, and the second node is taken as node B, and the application of the method of the embodiment of the present invention is described in combination with a specific scenario.

Scenario 1: Model monitoring at node A

Step 1: (optional), node A sends first configuration information to node B, where the first configuration information includes one or more of the following information:

a first indication, where the first indication is used to instruct the Node B to send the first information;

a second indication, where the second indication is used to instruct the Node B to obtain the first information;

The third information is used to assist the Node B in acquiring the first information.

The first information may be the input of CSI generation part or CSI (CSI compressed by an AI/ML model encoder) or the output of CSI reconstruction part.

Step 2: (optional) Node B obtains the first information, and the obtaining method includes one or more of the following:

Method 1: obtaining first information from a terminal (UE) through an air interface;

Method 2: Generate the first information based on information obtained from node A and/or UE.

Note 1: Step 2 is optional mainly because node A has stored/can obtain all the information for model monitoring, and node B does not need to obtain and/or send the first information.

Step 3: (optional) After acquiring the first information, node B sends indication information to node A to notify node A that node B stores the first information.

Step 4: (optional) Node A sends a request message to node B, requesting node B to send the first information to node A.

Step 5: (Optional) Node B sends first information to node A.

Note 2: Step 5 is optional mainly because node A has stored/can obtain all the information for model monitoring, and node B does not need to send the first information to node A.

Step 6: Node A performs model monitoring and obtains the model monitoring results of each node B.

Step 7: Node A makes a decision based on the model monitoring results of each node B.

Note 3: The air interface signaling used in the above steps can be RRC signaling, MAC CE signaling, DCI, etc.

The interface signaling used in the above steps may be F1 interface signaling, E1 interface signaling, Xn interface signaling, X2 interface signaling, other interface signaling, and the like.

Note 4: The above-mentioned node A can be CU, CU-CP, core network (including LMF), server/OAM, and the above-mentioned DU can be DU, CU-UP, base station, core network/base station, etc.

Note 5: The relationship between node A and node B can be one-to-one or one-to-many. It can be understood that A is a centralized node and B is a distributed node.

Scenario 2: Model monitoring at node B

Step 1: (optional) Node B sends second configuration information to node A. The specific content of the second configuration information and the first information is similar to step 1 in scenario 1 and will not be repeated here.

Step 2: (optional) Node A obtains the first information. Node A may obtain the first information in one or more of the following ways:

Method 1: obtaining first information from UE through an air interface;

Method 2: Generate the first information based on information obtained from the Node B and/or the UE.

Note 1: Step 2 is optional mainly because node B already stores/can obtain all the information for model monitoring, and node A does not need to obtain and/or send the first information.

Step 3: (optional) After node A obtains the first information, it sends indication information to node B to notify node B that node A stores the first information.

Step 4: (optional) Node B sends a request message to node A, requesting node A to send first information to node B.

Step 5: (Optional) Node A sends first information to node B.

Note 2: Step 5 is optional mainly because node B has stored/can obtain all the information for model monitoring, and node A does not need to send the first information to node B.

Step 6: (Optional) Before any of the above steps, node A may send third configuration information to node B, where the third configuration information includes one or more of the following information:

a seventh indication, where the seventh indication is used to instruct the Node B to send the second information;

An eighth indication, where the eighth indication is used to instruct the Node B to obtain the second information.

Step 7: Node B performs model monitoring to obtain second information of the node B.

Step 8: (Optional) After acquiring the second information, Node B sends indication information to Node A to notify Node A that Node B stores the second information.

Step 9: (optional) Node A sends a request message to node B, requesting node B to send second information to node A.

Step 10: Node B sends second information to node A.

Step 11: Node A makes a decision based on the second information sent by each node B.

Note 3: The air interface signaling used in the above steps can be RRC signaling, MAC CE signaling, DCI, etc.

The interface signaling used in the above steps may be F1 interface signaling, E1 interface signaling, Xn interface signaling, X2 interface signaling, other interface signaling, and the like.

Note 4: The above-mentioned node A can be CU, CU-CP, core network (including LMF), server/OAM, and the above-mentioned DU can be DU, CU-UP, base station, core network/base station, etc.

Note 5: The relationship between node A and node B can be one-to-one or one-to-many. It can be understood that A is a centralized node and B is a distributed node.

Scenario 3: Distribution and/or training method of node A’s decision model

Based on the above scenario 1 and scenario 2, when node A obtains the model monitoring result information/second information of each node B, node A makes a decision, which may include one or more of the following:

(1) The distribution mode of the decision model (e.g., centralized, fully distributed, semi-distributed);

Centralized: It can be understood that node A is a centralized node, node B is multiple distributed nodes under node A, and each node B uses the same AI/ML model;

Fully distributed: It can be understood that node A is a centralized node, and node B is multiple distributed nodes under node A. Each node B uses a different AI/ML model.

Semi-distributed: Node A can be understood as a centralized node, and Node B is a number of distributed nodes under Node A, some of which can use the same AI/ML model, and some use different AI/ML models.

Specifically, taking the CU-DU separation scenario as an example: if node A determines that one or more of the following conditions (not limited to the following conditions) are met based on the result information/second information of model monitoring, node A needs to re-distribute the model, that is, optimize the model distribution (the optimization here refers to changing the distribution mode of the model):

1) Currently using the "centralized" approach, but one or more of the following problems occur:

The model monitoring results of the AI/ML models of each Node B are poor performance and above a certain threshold, or the model monitoring results are good performance and below a certain threshold, such as more than 50% of the Node Bs have poor performance, or less than 30% of the Node Bs have good performance;

The model monitoring results of one or some models are very poor or do not work;

There are enough models available;

2) Currently using "Fully Distributed", but one or more of the following problems occur:

Model training is too expensive;

Limited model storage space;

Some models have good generalization performance;

Model management is difficult;

Exceeded the maximum number of models;

3) Currently using "semi-distributed", but one or more of the following problems occur:

Model training is too expensive;

Limited model storage space;

A model has good generalization performance;

Model management is difficult;

Exceeded the maximum number of models;

(2) decide on the model allocation within the semi-distributed approach (which models are suitable for use in which DUs);

Specifically, the model allocation in the semi-distributed manner may be determined based on one or more of the following information:

Historical model usage for each DU;

Model monitoring result information of the historical model used by each DU;

Current model usage of each DU;

Model monitoring result information of the model currently used by each DU;

Information about the generalization capabilities of each model;

Model priority information (if any);

DU priority information (if any);

·The scenarios/business information to which each model is applicable;

Current scenario/service information of each DU;

(3) Decide whether to use a centralized training method or a distributed training method to retrain the model with better performance or the required model;

Specifically, when the network finds that there is no effective or better model that can meet the application requirements, such as the network expects to use "centralized" based on some considerations (such as storage considerations), but no model currently has the generalization performance that can meet the requirements, the network can make the following decisions:

· Carry out a "centralized training method", that is, train a model whose generalization performance can meet the requirements. This "centralized training method" can obtain the model training data set information collected by each node B through node A, and use the received model training data set information of each node B to perform model training, so that the trained model has better generalization ability and can be applied to each node B; this "centralized training method" can also collect model training data set information from each node B and perform their own model training respectively, and send their trained models to node A. Node A trains a unified overall model based on the models trained by each node B. This overall model has better generalization ability and can be applied to each node B;

Conduct "distributed training", that is, training models per node B, so that each model can achieve better performance on this node B. This "distributed training method" can collect model training data set information from each node B and conduct their own model training separately, and then apply the trained model to this node B.

To sum up, in the method of the embodiment of the present invention, after the first node receives the model monitoring related information sent by the second node, it can perform at least one of model monitoring, distribution method of the decision model, and training method of the decision model based on the model monitoring related information to ensure that the model used by each second node has good performance and relatively high model utilization, so that the model distribution achieves the overall optimum among generalization performance, training complexity, and inference performance.

As shown in FIG2 , a model monitoring processing method provided by an embodiment of the present invention includes:

Step 201, the second node sends model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In this way, the second node sends model monitoring related information to the first node, so that after receiving the model monitoring related information sent by the second node, the first node can perform model monitoring, distribution method of decision models, and training method of decision models based on the model monitoring related information to ensure that the model used by each second node has good performance and relatively high model utilization, so that the model distribution achieves the overall optimality among generalization performance, training complexity, and reasoning performance.

Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, it also includes:

Before sending the first information, the second node receives first configuration information sent by the first node, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, it also includes:

Before sending the first information, the second node sends a third indication to the first node, where the third indication is used to notify the first node that the second node has acquired the first information.

Optionally, it also includes:

Before sending the second information, the second node sends second configuration information to the first node, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, it also includes:

Before sending the second information, the second node receives a sixth indication sent by the first node, where the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, it also includes:

Before sending the second information, the second node receives third configuration information sent by the first node, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, it also includes:

Before sending the second information, the second node sends a ninth indication to the first node, where the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, it also includes:

Before sending the second information, the second node receives the first information sent by the first node.

Optionally, the first information includes at least one of the following:

Input of the channel state information CSI generation part;

Output of the CSI reconstruction part;

Compressed CSI;

Beam assist information.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Original channel;

Precoding matrix indication PMI;

Eigenvector;

Precoder matrix.

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the distribution mode of the decision model includes at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

Optionally, the training method of the decision model includes centralized or distributed training of the decision model.

It should be noted that the method of this embodiment is applied to the second node and is implemented in conjunction with the above-mentioned model monitoring and processing method applied to the first node. The implementation method of the above-mentioned method embodiment is applicable to this method and can achieve the same technical effect.

As shown in FIG3 , an embodiment of the present invention further provides a model monitoring processing device, including: a memory 320, a transceiver 310, and a processor 300: the memory 320 is used to store program instructions; the transceiver 310 is used to send and receive data under the control of the processor 300; the processor 300 is used to read the program instructions in the memory 320 and perform the following operations:

Receiving model monitoring related information sent by the second node;

According to the model monitoring related information, perform at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In FIG. 3 , the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 300 and various circuits of memory represented by memory 320 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 310 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium includes a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 300 is responsible for managing the bus architecture and general processing, and the memory 320 may store data used by the processor 300 when performing operations.

The processor 300 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, the processor is further configured to perform the following operations:

Before receiving the first information, first configuration information is sent to the second node, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, the processor is further used to perform the following operations: before receiving the first information, receive a third indication sent by the second node, where the third indication is used to notify the first node that the second node has acquired the first information.

Optionally, the processor is further configured to perform the following operations: before receiving the second information, receive second configuration information sent by the second node, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, the processor is further used to perform the following operations: before receiving the second information, send a sixth indication to the second node, where the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, the processor is further configured to perform the following operations: before receiving the second information, send third configuration information to the second node, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, the processor is further used to perform the following operations: before receiving the second information, receive a ninth indication sent by the second node, where the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, the processor is further configured to perform the following operations: before receiving the second information, sending the first information to the second node.

Optionally, the first information includes at least one of the following:

Input of the channel state information CSI generation part;

Output of the CSI reconstruction part;

Compressed CSI;

Beam assist information.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Original channel;

Precoding matrix indication PMI;

Eigenvector;

Precoder matrix.

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the distribution mode of the decision model includes at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

Optionally, the training method of the decision model includes centralized or distributed training of the decision model.

The device in the embodiment of the present invention, after receiving the model monitoring related information sent by the second node, can perform at least one of model monitoring, distribution method of decision model, and training method of decision model based on the model monitoring related information to ensure that the model used by each second node has good performance and relatively high model utilization, so that the model distribution achieves the overall optimality among generalization performance, training complexity, and reasoning performance.

It should be noted here that the above-mentioned device provided in the embodiment of the present invention can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

As shown in FIG4 , the present invention also provides a model monitoring processing device, including:

A first receiving module 410 is used to receive model monitoring related information sent by the second node;

The processing module 420 is configured to perform at least one of the following according to the model monitoring related information:

Model monitoring;

How decision models are distributed;

The training method of the decision model. Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, it also includes:

The second sending module is configured to send first configuration information to the second node before receiving the first information, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, it also includes:

The second receiving module is used to receive a third indication sent by the second node before receiving the first information, wherein the third indication is used to notify the first node that the second node has acquired the first information.

Optionally, it also includes:

The third receiving module is configured to receive second configuration information sent by the second node before receiving the second information, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, it also includes:

The third sending module is used to send a sixth indication to the second node before receiving the second information, where the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, it also includes:

A fourth sending module is configured to send third configuration information to the second node before receiving the second information, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, it also includes:

The fourth receiving module is used to receive a ninth indication sent by the second node before receiving the second information, wherein the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, it also includes:

A fifth sending module is used to send the first information to the second node before receiving the second information.

Optionally, the first information includes at least one of the following:

Input of the channel state information CSI generation part;

Output of the CSI reconstruction part;

Compressed CSI;

Beam assist information.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Original channel;

Precoding matrix indication PMI;

Eigenvector;

Precoder matrix.

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the processing module is further used for at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

Optionally, the processing module is also used to decide whether the model training is centralized or distributed.

After receiving the model monitoring related information sent by the second node, the device of the embodiment of the present invention can perform at least one of model monitoring, distribution method of the decision model, and training method of the decision model based on the model monitoring related information to ensure that the model used by each second node has good performance and relatively high model utilization, so that the model distribution achieves the overall optimality among generalization performance, training complexity, and reasoning performance.

It should be noted here that the above-mentioned device provided in the embodiment of the present invention can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

In some embodiments of the present invention, a processor-readable storage medium is further provided, wherein the processor-readable storage medium stores program instructions, and the program instructions are used to enable the processor to execute the following steps:

Receiving model monitoring related information sent by the second node;

According to the model monitoring related information, perform at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

When the program instructions are executed by the processor, all implementations of the method embodiment applied to the first node side as shown in FIG. 1 can be implemented. To avoid repetition, they are not described here.

As shown in FIG5 , the present invention also provides a model monitoring processing device, including: a memory 520, a transceiver 510, and a processor 500: the memory 520 is used to store program instructions; the transceiver 510 is used to send and receive data under the control of the processor 500; the processor 500 is used to read the program instructions in the memory 520 and perform the following operations:

Sending model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

In FIG. 5 , the bus architecture may include any number of interconnected buses and bridges, specifically various circuits of one or more processors represented by processor 500 and memory represented by memory 520 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 510 may be a plurality of components, including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, including transmission media such as wireless channels, wired channels, and optical cables.

The processor 500 is responsible for managing the bus architecture and general processing, and the memory 520 can store data used by the processor 500 when performing operations.

Optionally, the processor 500 may be a CPU (central processing unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array) or a CPLD (Complex Programmable Logic Device), and the processor 500 may also adopt a multi-core architecture.

The processor 500 calls the program instructions stored in the memory to execute any of the methods provided in the embodiments of the present application according to the obtained executable instructions. The processor 500 and the memory 520 can also be arranged physically separately.

Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, the processor is further configured to perform the following operations:

Before sending the first information, first configuration information is received from the first node, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, the processor is further used to perform the following operations: before sending the first information, send a third indication to the first node, where the third indication is used to notify the first node that the second node has acquired the first information.

Optionally, the processor is further configured to perform the following operations: before sending the second information, send second configuration information to the first node, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, the processor is further used to perform the following operations: before sending the second information, receive a sixth indication sent by the first node, where the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, the processor is further configured to perform the following operations: before sending the second information, receive third configuration information sent by the first node, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, the processor is further used to perform the following operations: before sending the second information, send a ninth indication to the first node, where the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, the processor is further configured to perform the following operations: before sending the second information, receiving first information sent by the first node.

Optionally, the first information includes at least one of the following:

Input of the channel state information CSI generation part;

Output of the CSI reconstruction part;

Compressed CSI;

Beam assist information.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Original channel;

Precoding matrix indication PMI;

Eigenvector;

Precoder matrix.

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the distribution mode of the decision model includes at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

Optionally, the training method of the decision model includes centralized or distributed training of the decision model.

The device of the embodiment of the present invention sends model monitoring related information to the first node, so that after receiving the model monitoring related information sent by the second node, the first node can perform model monitoring, distribution method of the decision model, and at least one of the training method of the decision model based on the model monitoring related information, so as to ensure that the model used by each second node has good performance and relatively high model utilization, so that the model distribution achieves the overall optimum among generalization performance, training complexity, and reasoning performance.

It should be noted here that the above-mentioned device provided in the embodiment of the present invention can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

As shown in FIG6 , the present invention also provides a model monitoring processing device, including:

A first sending module 610 is used to send model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

Optionally, the model monitoring related information includes at least one of the following:

first information, wherein the first information is used to assist model monitoring;

Second information, wherein the second information is used to assist in determining the distribution method and/or training method of the decision model.

Optionally, the device further comprises:

A fifth receiving module is configured to receive first configuration information sent by the first node before sending the first information, where the first configuration information includes at least one of the following:

a first indication, where the first indication is used to instruct the second node to send the first information;

a second indication, where the second indication is used to instruct the second node to obtain the first information;

The third information is used to assist the second node in acquiring the first information.

Optionally, the device further comprises:

The sixth sending module is used to send a third indication to the first node before sending the first information, and the third indication is used to notify the first node that the second node has obtained the first information.

Optionally, the device further comprises:

A seventh sending module is configured to send second configuration information to the first node before sending the second information, where the second configuration information includes at least one of the following:

a fourth indication, where the fourth indication is used to instruct the first node to send the first information;

a fifth indication, where the fifth indication is used to instruct the first node to obtain the first information;

Fourth information, where the fourth information is used to assist the first node in acquiring the first information.

Optionally, the device further comprises:

A sixth receiving module is used to receive a sixth indication sent by the first node before sending the second information, and the sixth indication is used to notify the second node that the first node has acquired the first information.

Optionally, the device further comprises:

A seventh receiving module is configured to receive third configuration information sent by the first node before sending the second information, where the third configuration information includes at least one of the following:

a seventh indication, where the seventh indication is used to instruct the second node to send the second information;

An eighth indication, where the eighth indication is used to instruct the second node to obtain the second information.

Optionally, the device further comprises:

An eighth sending module is used to send a ninth indication to the first node before sending the second information, wherein the ninth indication is used to notify the first node that the second node has acquired the second information.

Optionally, the device further comprises:

An eighth receiving module is used to receive first information sent by the first node before sending the second information.

Optionally, the first information includes at least one of the following:

Input of the channel state information CSI generation part;

Output of the CSI reconstruction part;

Compressed CSI;

Beam assist information.

Optionally, the input of the CSI generation part or the output of the CSI reconstruction part includes at least one of the following:

Original channel;

Precoding matrix indication PMI;

Eigenvector;

Precoder matrix.

Optionally, the beam assist information includes at least one of the following:

Information related to beam shape;

Radiation pattern;

Beam width;

Base station code index;

Receive beam angle;

Terminal location;

Base station transmit antenna angle.

Optionally, the second information includes at least one of the following:

Model monitoring result information;

Monitoring algorithms;

Model identification information;

Function identification information;

Time information.

Optionally, the distribution mode of the decision model includes at least one of the following:

The decision model distribution adopts centralized, fully distributed or semi-distributed;

Model allocation within decision semi-distribution.

Optionally, the training method of the decision model includes centralized or distributed training of the decision model.

The device of the embodiment of the present invention sends model monitoring related information to the first node, so that after receiving the model monitoring related information sent by the second node, the first node can perform model monitoring, distribution method of the decision model, and at least one of the training method of the decision model based on the model monitoring related information, so as to ensure that the model used by each second node has good performance and relatively high model utilization, so that the model distribution achieves the overall optimum among generalization performance, training complexity, and reasoning performance.

It should be noted here that the above-mentioned device provided in the embodiment of the present invention can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

In some embodiments of the present invention, a processor-readable storage medium is further provided, wherein the processor-readable storage medium stores program instructions, and the program instructions are used to enable the processor to execute the following steps:

Sending model monitoring related information to the first node;

The model monitoring related information is used for at least one of the following:

Model monitoring;

How decision models are distributed;

How to train the decision model.

When the program instructions are executed by the processor, all implementations of the method embodiment applied to the second node side as shown in FIG. 2 can be implemented. To avoid repetition, they are not described here.

It should be noted that the division of units in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

The technical solution provided in the embodiment of the present application can be applicable to a variety of systems, especially 5G systems. For example, applicable systems can be global system of mobile communication (Global System of Mobile communication, GSM) system, code division multiple access (Code Division Multiple Access, CDMA) system, wideband code division multiple access (Wideband Code Division Multiple Access, WCDMA) general packet radio service (General Packet Radio Service, GPRS) system, long-term evolution (Long Term Evolution, LTE) system, LTE frequency division duplex (Frequency Division Duplex, FDD) system, LTE time division duplex (Time Division Duplex, TDD) system, advanced long-term evolution (Long Term Evolution Advanced, LTE-A) system, universal mobile system (Universal Mobile Telecommunication System, UMTS), global interconnection microwave access (Worldwide interoperability for Microwave Access, WiMAX) system, 5G new air interface (New Radio, NR) system, etc. These various systems include terminal equipment and network equipment. The system may also include core network parts, such as the Evolved Packet System (EPS), 5G System (5GS), etc.

The terminal involved in the embodiment of the present application may be a device that provides voice and/or data connectivity to a user, a handheld device with a wireless connection function, or other processing devices connected to a wireless modem. In different systems, the name of the terminal device may also be different. For example, in a 5G system, the terminal device may be called a user equipment (UE). A wireless terminal device can communicate with one or more core networks (CN) via a radio access network (RAN). The wireless terminal device may be a mobile terminal device, such as a mobile phone (or a "cellular" phone) and a computer with a mobile terminal device. For example, it may be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device that exchanges language and/or data with a wireless access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs) and other devices. The wireless terminal device may also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, an access point, a remote terminal device, an access terminal device, a user terminal device, a user agent, and a user device, but is not limited thereto in the embodiments of the present application.

The network device involved in the embodiment of the present application may be a base station, which may include multiple cells providing services for the terminal. Depending on the specific application scenario, the base station may also be called an access point, or may be a device in the access network that communicates with the wireless terminal device through one or more sectors on the air interface, or other names. The network device may be used to interchange received air frames with Internet Protocol (IP) packets, and serve as a router between the wireless terminal device and the rest of the access network, wherein the rest of the access network may include an Internet Protocol (IP) communication network. The network device may also coordinate the attribute management of the air interface. For example, the network device involved in the embodiments of the present application may be a network device (Base Transceiver Station, BTS) in the Global System for Mobile communications (Global System for Mobile communications, GSM) or Code Division Multiple Access (Code Division Multiple Access, CDMA), or a network device (NodeB) in Wide-band Code Division Multiple Access (WCDMA), or an evolutionary network device (evolutional Node B, eNB or e-NodeB) in the Long Term Evolution (Long Term Evolution, LTE) system, a 5G base station (gNB) in the 5G network architecture (next generation system), or a home evolved Node B (Home evolved Node B, HeNB), a relay node, a home base station (femto), a pico base station (pico), etc., which is not limited in the embodiments of the present application. In some network structures, the network device may include a centralized unit (CU) node and a distributed unit (DU) node, and the centralized unit and the distributed unit may also be arranged geographically separately.

Network devices and terminal devices can each use one or more antennas for multiple-input multiple-output (MIMO) transmission. MIMO transmission can be single-user MIMO (SU-MIMO) or multi-user MIMO (MU-MIMO). Depending on the form and number of antenna combinations, MIMO transmission can be 2D-MIMO, 3D-MIMO, FD-MIMO or massive-MIMO, or it can be diversity transmission, precoded transmission or beamforming transmission, etc.

Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

The present application is described with reference to the flowchart and/or block diagram of the method, device (system) and computer program product according to the embodiment of the present application. It should be understood that each process and/or box in the flowchart and/or block diagram, and the combination of the process and/or box in the flowchart and/or block diagram can be implemented by computer executable instructions. These computer executable instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the function specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.

These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the processor-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These processor-executable instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims (35)

1. A model monitoring and processing method, comprising:

the first node receives the model monitoring related information sent by the second node;

the first node monitors relevant information according to the model and at least one of the following is executed:

Model monitoring;

a distribution mode of the decision model;

the training mode of the decision model.

2. The method of claim 1, wherein the model monitoring related information comprises at least one of:

the first information is used for assisting model monitoring;

And the second information is used for assisting the distribution mode and/or training mode of the decision-making model.

3. The method as recited in claim 2, further comprising:

The first node sends first configuration information to the second node before receiving the first information, wherein the first configuration information comprises at least one of the following:

a first indication for indicating the second node to send the first information;

A second instruction, configured to instruct the second node to acquire the first information;

and third information, wherein the third information is used for assisting the second node to acquire the first information.

4. The method as recited in claim 2, further comprising:

And before receiving the first information, the first node receives a third instruction sent by the second node, wherein the third instruction is used for informing the first node that the second node has acquired the first information.

5. The method as recited in claim 2, further comprising:

the first node receives second configuration information sent by the second node before receiving the second information, wherein the second configuration information comprises at least one of the following items:

a fourth indication, configured to instruct the first node to send the first information;

A fifth indication for instructing the first node to acquire the first information;

fourth information, the fourth information is used for assisting the first node to acquire the first information.

6. The method as recited in claim 2, further comprising:

the first node sends a sixth indication to the second node before receiving the second information, wherein the sixth indication is used for notifying the second node that the first node has acquired the first information.

7. The method as recited in claim 2, further comprising:

the first node sends third configuration information to the second node before receiving the second information, wherein the third configuration information comprises at least one of the following:

A seventh instruction, configured to instruct the second node to send the second information;

and an eighth instruction, wherein the eighth instruction is used for instructing the second node to acquire the second information.

8. The method as recited in claim 2, further comprising:

The first node receives a ninth instruction sent by the second node before receiving the second information, wherein the ninth instruction is used for notifying the first node that the second node has acquired the second information.

9. The method as recited in claim 2, further comprising:

The first node sends the first information to the second node before receiving the second information.

10. The method of claim 2 or 3 or 4 or 9, wherein the first information comprises at least one of:

an input of a Channel State Information (CSI) generation section;

An output of the CSI reconstruction section;

Compressed CSI;

Beam assistance information.

11. The method of claim 10, wherein the input of the CSI generating part or the output of the CSI reconstructing part comprises at least one of:

An original channel;

Precoding matrix indication PMI;

a feature vector;

A precoder matrix.

12. The method of claim 10, wherein the beam assistance information comprises at least one of:

Beam shape related information;

a radiation pattern;

A beam width;

a base station encodes an index;

Receiving a beam angle;

A terminal position;

The base station transmits the antenna angle.

13. The method of claim 2 or 5 or 6 or 7 or 8 or 9, wherein the second information comprises at least one of:

Monitoring result information by a model;

Monitoring an algorithm;

model identification information;

Function identification information;

Time information.

14. The method of claim 1, wherein the distribution of the decision model comprises at least one of:

The decision model is distributed in a centralized type, a full-distributed type or a half-distributed type;

Model assignment within the decision semi-distributed.

15. The method of claim 1, wherein the training of the decision model comprises training the decision model in a centralized or distributed manner.

16. A model monitoring and processing method, comprising:

The second node transmits the model monitoring related information to the first node;

Wherein the model monitoring related information is used for at least one of:

Model monitoring;

a distribution mode of the decision model;

the training mode of the decision model.

17. The method of claim 16, wherein the model monitoring related information comprises at least one of:

the first information is used for assisting model monitoring;

And the second information is used for assisting the distribution mode and/or training mode of the decision-making model.

18. The method as recited in claim 17, further comprising:

The second node receives first configuration information from the first node before sending the first information, the first configuration information including at least one of:

a first indication for indicating the second node to send the first information;

A second instruction, configured to instruct the second node to acquire the first information;

and third information, wherein the third information is used for assisting the second node to acquire the first information.

19. The method as recited in claim 17, further comprising:

The second node sends a third indication to the first node before sending the first information, the third indication being used to inform the first node that the second node has acquired the first information.

20. The method as recited in claim 17, further comprising:

The second node sends second configuration information to the first node before sending the second information, wherein the second configuration information comprises at least one of the following:

a fourth indication, configured to instruct the first node to send the first information;

A fifth indication for instructing the first node to acquire the first information;

fourth information, the fourth information is used for assisting the first node to acquire the first information.

21. The method as recited in claim 17, further comprising:

The second node receives a sixth instruction sent by the first node before sending the second information, wherein the sixth instruction is used for notifying the second node that the first node has acquired the first information.

22. The method as recited in claim 17, further comprising:

The second node receives third configuration information sent by the first node before sending the second information, wherein the third configuration information comprises at least one of the following items:

A seventh instruction, configured to instruct the second node to send the second information;

and an eighth instruction, wherein the eighth instruction is used for instructing the second node to acquire the second information.

23. The method as recited in claim 17, further comprising:

the second node sends a ninth indication to the first node before sending the second information, the ninth indication being used to inform the first node that the second node has acquired the second information.

24. The method as recited in claim 17, further comprising:

and the second node receives the first information sent by the first node before sending the second information.

25. A model monitoring and processing device, comprising: a memory, transceiver, processor;

a memory for storing program instructions; a transceiver for transceiving data under control of the processor; a processor for reading the program instructions in the memory and performing the following operations:

receiving model monitoring related information sent by a second node;

and monitoring relevant information according to the model, and executing at least one of the following steps:

Model monitoring;

a distribution mode of the decision model;

the training mode of the decision model.

26. The apparatus of claim 25, wherein the model monitoring related information comprises at least one of:

the first information is used for assisting model monitoring;

And the second information is used for assisting the distribution mode and/or training mode of the decision-making model.

27. The apparatus of claim 26, wherein the processor is further configured to:

transmitting first configuration information to the second node before receiving the first information, the first configuration information including at least one of:

a first indication for indicating the second node to send the first information;

A second instruction, configured to instruct the second node to acquire the first information;

and third information, wherein the third information is used for assisting the second node to acquire the first information.

28. The apparatus of claim 26, wherein the processor is further configured to:

Before receiving the second information, receiving second configuration information sent by the second node, where the second configuration information includes at least one of the following:

a fourth instruction, configured to instruct a first node to send the first information;

a fifth indication for instructing a first node to acquire the first information;

fourth information, the fourth information is used for assisting the first node to acquire the first information.

29. A model monitoring and processing device, comprising:

the first receiving module is used for receiving the model monitoring related information sent by the second node;

the processing module is used for monitoring related information according to the model and executing at least one of the following:

Model monitoring;

a distribution mode of the decision model;

the training mode of the decision model.

30. A model monitoring and processing device, comprising: a memory, transceiver, processor; a memory for storing program instructions; a transceiver for transceiving data under control of the processor; a processor for reading the program instructions in the memory and performing the following operations:

Sending model monitoring related information to the first node;

Wherein the model monitoring related information is used for at least one of:

Model monitoring;

a distribution mode of the decision model;

the training mode of the decision model.

31. The apparatus of claim 30, wherein the model monitoring related information comprises at least one of:

the first information is used for assisting model monitoring;

And the second information is used for assisting the distribution mode and/or training mode of the decision-making model.

32. The apparatus of claim 31, wherein the processor is further configured to:

before transmitting the first information, receiving first configuration information from the first node, the first configuration information including at least one of:

A first indication, wherein the first indication is used for indicating a second node to send the first information;

A second instruction, wherein the second instruction is used for instructing a second node to acquire the first information;

And third information, wherein the third information is used for assisting the second node to acquire the first information.

33. The apparatus of claim 31, wherein the processor is further configured to:

Transmitting second configuration information to the first node before transmitting the second information, the second configuration information including at least one of:

a fourth indication, configured to instruct the first node to send the first information;

A fifth indication for instructing the first node to acquire the first information;

fourth information, the fourth information is used for assisting the first node to acquire the first information.

34. A model monitoring and processing device, comprising:

the first sending module is used for sending the model monitoring related information to the first node;

Wherein the model monitoring related information is used for at least one of:

Model monitoring;

a distribution mode of the decision model;

the training mode of the decision model.

35. A processor-readable storage medium, characterized in that the processor-readable storage medium stores a computer program for causing the processor to execute the information indicating method according to any one of claims 1 to 15 or the information indicating method according to any one of claims 16 to 24.