Systems and methods for wireless signal configuration by a neural network

Patent metadata

Field	Value
Publication number	US-12562824-B2
Application number	US-202418662342-A
Country	US
Kind code	B2
Filing date	May 13, 2024
Priority date	May 15, 2019
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

A wireless network can generate candidate signal configurations for physical transmissions to or from a user equipment (UE) in a radio environment. The generation of candidate signal configurations can be performed using a first neural network that is associated with the UE. These signal configurations can then be evaluated using a second neural network that is associated with the radio environment. The second neural network can be trained using measurements from previous physical transmissions in the radio environment. The trained second neural network generates a reward value that is associated with the candidate signal configurations. The first neural network is then trained using the reward values from the second neural network to produce improved candidate signal configurations. When a signal configuration that produces a suitable reward value is generated, this signal configuration can be used for the physical transmission in the radio environment.

First claim

The invention claimed is:

1 . A method for obtaining a wireless signal configuration, the method comprising:
collecting a plurality of data samples, each data sample, among the plurality of data samples, providing information associated with a respective physical transmission in a radio environment; training, using a subset of the plurality of data samples, a first neural network associated with the radio environment, the training resulting in a trained first neural network; obtaining, from a second neural network specific to a single user equipment (UE):
a first candidate wireless signal configuration for a scheduled transmission to or from the single UE in the radio environment;
information associated with a state of the single UE; and
information associated with the radio environment;
inputting, into the trained first neural network:
the first candidate wireless signal configuration;
the information associated with the state of the single UE; and
the information associated with the radio environment;
obtaining, from the trained first neural network, a value for an evaluation metric, the evaluation metric representing a predicted effectiveness of the first candidate wireless signal configuration; and adjusting, based on the value for the evaluation metric, the first candidate wireless signal configuration to obtain a second candidate wireless signal configuration.

2 . The method of claim 1 , wherein the first neural network includes first weights, the method further comprising updating the first weights.

3 . The method of claim 2 , wherein the updating acts to adapt the first neural network to changing conditions in the radio environment.

4 . The method of claim 2 , wherein the training the first neural network comprises acting to optimize a value of a reward.

5 . The method of claim 4 , further comprising determining the value of the reward.

6 . The method of claim 5 , wherein determining the value of the reward comprises evaluating a reward function.

7 . The method of claim 6 , wherein the reward function associates the value of the reward with the information associated with the state of the single UE, including one or more of: throughput; latency; or error rate.

8 . The method of claim 6 , wherein determining the value of the reward comprises re-evaluating the reward function responsive to changing conditions in the radio environment.

9 . The method of claim 6 , wherein the updating the first weights comprises using a policy gradient model to update the first weights based on the reward function.

10 . The method of claim 1 , wherein the data samples include information associated with a geographic location of the single UE.

11 . The method of claim 1 , wherein the data samples include environmental factors.

12 . The method of claim 11 , wherein the environmental factors comprise weather conditions.

13 . The method of claim 11 , wherein the environmental factors comprise time of day.

14 . The method of claim 11 , further comprising adjusting, based on the environmental factors, the reward function.

15 . A method for obtaining a wireless signal configuration, the method comprising:
collecting a plurality of data samples, each data sample, among the plurality of data samples, providing information associated with a respective physical transmission in a radio environment; training, using a subset of the plurality of data samples, a first neural network associated with the radio environment, the training resulting in a trained first neural network; obtaining, from a second neural network specific to a user equipment (UE):
a first candidate wireless signal configuration for a scheduled transmission to or from the UE in the radio environment;
information associated with a state of the UE; and
information associated with the radio environment;
inputting, into the trained first neural network:
the information associated with the state of the UE; and
the information associated with the radio environment;
obtaining, from the trained first neural network, a value for an evaluation metric, the evaluation metric representing a predicted effectiveness of the first candidate wireless signal configuration; and adjusting, based on the value for the evaluation metric, the first candidate wireless signal configuration to obtain a second candidate wireless signal configuration, wherein the evaluation metric comprises a weighted sum of values for the information associated with the state of the UE, the weighted sum defined by a plurality of weights, wherein each weight, among the plurality of weights, is associated with a corresponding one of the values for the information associated with the state of the UE.

16 . The method of claim 15 , wherein the information associated with the state of the UE comprises one or more of: predicted throughput; latency; or error rate.

17 . The method of claim 15 , wherein the method is carried out by a reinforcement learning agent, the method further comprising:
obtaining a measurement of performance associated with the reinforcement learning agent; and adjusting, based on the measurement, a given weight, among the plurality of weights.

18 . The method of claim 15 , further comprising adjusting a given weight, among the plurality of weights, responsive to changing conditions in the radio environment.

19 . The method of claim 1 , further comprising storing the second candidate wireless signal configuration and an associated evaluation metric in a database for future reference and retraining of the first neural network and the second neural network.

20 . A system for wireless signal configuration optimization, comprising:
at least one processor coupled with a memory; storing instructions that, when executed by the at least one processor, cause the system to perform the method of claim 1 ; and a database storing the plurality of data samples and the second candidate wireless signal configuration.

21 . The system of claim 20 , wherein the instructions, when executed by the at least one processor, cause the system to:
collect new data samples collected from further physical transmissions in the radio environment; and retrain, using the new data samples, the first neural network to obtain a retrained first neural network defined by retrained neural network parameters.

22 . The system of claim 20 , wherein the instructions, when executed by the at least one processor, cause the system to update the database with the new data samples and the retrained neural network parameters.

23 . The system of claim 20 , wherein the memory further stores a reinforcement learning model for the second neural network; and
the instructions, when executed by the at least one processor, cause the system further to:
provide, to the second neural network, feedback from the first neural network;
provide, to the second neural network, an adjusted reward function; and
obtain, from the second neural network, an updated first candidate wireless signal configuration, the updated first candidate wireless signal configuration based on applying the reinforcement learning model to the feedback and the adjusted reward function.

Assignees

• Ge Yiqun
• Shi Wuxian
• Tong Wen
• Huawei Tech Co Ltd

Inventors

• Ge Yiqun
• Shi Wuxian
• Tong Wen

Classifications

• H04L25/03165

  using neural networks · CPC title

• H04B17/373

  Predicting channel quality {or other radio frequency [RF]} parameters · CPC title

• H04W24/10

  Scheduling measurement reports {; Arrangements for measurement reports} · CPC title

• H04L25/0254

  using neural network algorithms · CPC title

• G06N3/09

  Supervised learning · CPC title