WO2020087260A1

WO2020087260A1 - Method and apparatus for sinr prediction for link adaption

Info

Publication number: WO2020087260A1
Application number: PCT/CN2018/112608
Authority: WO
Inventors: Dezheng YAN; Gang Xu; Biao Yang; Kefeng Liu
Original assignee: Nokia Shanghai Bell Co., Ltd.
Priority date: 2018-10-30
Filing date: 2018-10-30
Publication date: 2020-05-07
Also published as: CN112913275A

Abstract

Methods and apparatus for SINR prediction for link adaption are provided. The method comprises a step of receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements, a step of training one or more prediction models with the historic measurement data based on one or more neutral networks, and a step of predicting SINR with the prediction models and the historic measurement data. With a deep learning network, a Deep-Learning-Assistant (DLA) SINR prediction is provided. It will simplify the online calculation in field, by shifting many challenges in traditional approach to neutral training stage. Besides, it will improve the accuracy of SINR prediction, and then improve the accuracy of MCS selection.

Description

[Title established by the ISA under Rule 37.2] METHOD AND APPARATUS FOR SINR PREDICTION FOR LINK ADAPTION

TECHNICAL FIELD

Various example embodiments relate to cellular radio implementation, especially to a technique for SINR prediction for link adaption.

BACKGROUND

In operator’s cellular network, link adaptation is an important component which is desired to improve system capacity, peak data rate and coverage reliability by the adaptation of transmission settings to the radio channel conditions, and proper SINR (Signal to Interference plus Noise Ratio) estimation/prediction for transmitting timing is the key for link adaption to achieve its purposes as much as possible.

For the scheme overview, high level modulation and coding scheme and less HARQ (Hybrid Automatic Repeat Request) retransmissions will get high data throughput if radio channel is good, and respectively, low level modulation and coding scheme and more HARQ retransmissions will drop data throughput if radio channel is poor.

Besides, for a traditional LTE (Long Term Evolution) link adaption PUSCH (Physical Uplink Shared Channel) /SRS (Sounding Reference Signal) and AN (Access Network) feedback, initial MCS (Modulation Coding Scheme) selection is based on SINR to MCS lookup table, where SINR is calculated with averaging historic measurements from PUSCH/SRS/PUCCH (Physical Uplink Control Channel) . MCS selection is then updated with historic ACK/NACK information on transmitted PDUs (Protocol Data Units) .

Historic SINR measurement may not reflect latest channel radio, especially in high speed train scenario where fast fading and path loss are changed quickly. During evolution to LTE-Advanced Pro and 5G, interference may be varying rapidly further with the deployment of more spectral efficiency technologies like higher-order MU-MIMO (multi-user MIMO) . Interference among paired UEs may vary in TTI level.

Consequently, it’s very challenge to select right MCS for high throughput in those scenarios. Inaccurate MCS leads to high BLER (Block Error Ratio) , low transmission efficiency and throughput, bad experience to users. For example, smaller MCS results in lower throughput, and higher MCS may result in re-transmission and bandwidth waste if it cannot be received correctly due to wrong prediction on varying interference from many factors.

SUMMARY

Various example embodiments relate to a method for SINR prediction for link adaption, comprising the steps of:

receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

training one or more prediction models with the historic measurement data based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data.

Other various example embodiments relate to a first apparatus for SINR prediction for link adaption, comprises:

at least one processor; and

at least one memory including computer program code,

the at least one memory and the computer program code configured to, with the at least one processor, cause the first apparatus to perform operation of at least the following:

training one or more prediction models with the historic measurement data based on one or more neutral networks.

Other various example embodiments relate to a second apparatus for SINR prediction for link adaption, comprises:

at least one processor; and

at least one memory including computer program code,

the at least one memory and the computer program code configured to, with the at least one processor, cause the second apparatus to perform operation of at least the following:

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

sending the predicting SINR to one or more link adaptation modules in cellular network.

Other various example embodiments relate to a central unit of radio access nodes, for SINR prediction for link adaption, wherein, the central unit comprises the above first apparatus.

Other various example embodiments relate to a distributed unit of radio access nodes, for SINR prediction for link adaption, wherein, the central unit comprises the above second apparatus.

Other various example embodiments relate to a system for SINR prediction for link adaption, wherein, the system comprises the above first apparatus and the above second apparatus.

Other various example embodiments relate to a non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least:

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

As such, the implementation of various example embodiments of the subject matter proposes methods and apparatuses for SINR prediction for link adaption. With a deep learning network, a Deep-Learning-Assistant (DLA) SINR prediction is provided. It will simplify the online calculation in field, by shifting many challenges in traditional approach to neutral training stage. Besides, it will improve the accuracy of SINR prediction, and then improve the accuracy of MCS selection.

BRIEF DESCRIPTION OF THE DRAWINGS

Other features, purposes and advantages of the subject matter will become more explicit by means of reading the detailed statement of the non-restrictive embodiments made with reference to the accompanying drawings.

FIG. 1 is a block diagram illustrating an exemplary embodiment of a first apparatus for SINR prediction for link adaption.

FIG. 2 is a block diagram illustrating an exemplary embodiment of a second apparatus for SINR prediction for link adaption.

FIG. 3 is a block diagram illustrating an exemplary embodiment of a deployment for SINR prediction for link adaption.

FIG. 4 shows an exemplary embodiment of SINR prediction implantation using DNN.

FIG. 5 is a flow chart illustrating an exemplary embodiment of a method for SINR prediction for link adaption.

DETAILED DESCRIPTION AND PREFERRED EMBODIMENT

The word "exemplary" is used herein to mean "serving as an example, instance, or illustration. "

Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. All of the embodiments described in this Detailed Description are exemplary embodiments provided to enable persons skilled in the art to make or use the disclosure and not to limit the scope of the disclosure.

Referring to FIG. 1, an exemplary embodiment of a first apparatus 10 for SINR prediction modeling for link adaption, comprising means for performing:

means 110 for receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

means 120 for training one or more prediction models with the historic measurement data based on one or more neutral networks.

Specifically, the related measurements includes but not limited to RSSI (Received Signal Strength Indication) , PHR (Power Headroom Report) , CQI (Channel Quality Indicator) . For example, the radio access nodes send SINR, RSSI, PHR, CQI which collected from a plurality of TTIs before to the first apparatus 10.

The prediction models could predict SINR and related measurements on the next sequence. For example, the prediction models are trained based on historic SINR and related measurements from TTI (n) and before, and the prediction models could predict the SINR and related measurements for TTI (n+1) .

In another embodiment, the neutral networks comprise any one or more DNNs (Deep Neural Networks) . Specifically, the DNNs includes but not limited to CNN (Convolutional Neural Network) , RNN (Recurrent Neural Network) , LSTM (Long Short-Term Memory) or other neutral networks.

In another embodiment, the first apparatus 10 further comprises: means for converting the historic measurement data into input data for one or more neutral networks. For example, according to the format requirement of neural networks, the first apparatus 10 converts the historic measurement data in format into adaptive input data for neural networks.

Then, the means for training one or more prediction models trains one or more prediction models with the input data based on the one or more neutral networks.

In another embodiment, the first apparatus 10 further comprises: means for obtaining one or more prediction measurement data predicted by the prediction models and real measured SINR corresponding to the prediction measurement data; and means for updating the prediction models based on the prediction measurement data and real measured SINR.

Specifically, the means for updating the prediction models compares the predict SINR and real measured SINR first, and determines whether updating the prediction models or not; if the predict SINR and real measured SINR are not same, or the difference between the predict SINR and real measured SINR exceeds the predetermined threshold, the means will update the prediction models by the historic measurement data which is used as input data and the real measured SINR which is used as a model answer. Herein, the historic measurement data includes but not limited to SINR, RSSI, PHR, CQI, etc.

Therefore, the first apparatus 10 could realize a self-corrective feedback with online learning, which updates learned models based on predicted SINR data and real measured SINR data later.

Those skilled in the art could understand that it will simplify the online calculation in field, by shifting many challenges in traditional approach to offline training stage. Additionally, none latency sensitivity online training and learning is also further options for better fine tuning on results later.

Referring to FIG. 2, an exemplary embodiment of a second apparatus 20 for SINR prediction for link adaption, comprising means for performing:

means 210 for receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

means 220 for obtaining one or more prediction models based on one or more neutral networks;

means 230 for predicting SINR with the prediction models and the historic measurement data;

means 240 for sending the predicting SINR to one or more link adaptation modules in cellular network.

Specifically, the means 220 could obtain the one or more prediction models from the apparatus 10, or other apparatus which could provide the prediction models. The means 230 takes the historic measurement data as input information. The prediction models could calculate a result as the predicting SINR.

Herein, the means preceding comprises:

at least one processor; and

at least one memory including computer program code, that at least one memory and computer program code configured to , with the at least one processor, cause the performance of the apparatus.

Referring to FIG. 3, FIG. 3 is a block diagram illustrating an exemplary embodiment of a deployment for SINR prediction for link adaption.

The first apparatus is deployed on Central Unit (CU) to train the model with batch measurement data collected from Radio Access Node (RAN) , and the second apparatus is deployed on DU to close RAN (UP) and ensure quick delivery of model output. This deployment could support Deep-Learning-Assistant (DLA) SINR prediction or intelligent Link Adaption.

RAN transmits batch sampled measurement data to the first apparatus on CU. The first apparatus trains models, and the second apparatus on DU predicts SINR.

The second apparatus transmits predict results to RAN, and RAN sends current measurement data to the second apparatus.

The second apparatus transmits the current measurement data to the first apparatus as model feedback, and the first apparatus updates the models with the model feedback. Then the update model is send to the second apparatus.

Referring to FIG. 4, FIG. 4 shows an exemplary embodiment of SINR prediction implantation using DNN.

The embodiment takes use of RNN /LSTM as the deep learning model. It includes two parts:

One is a feature extractor with prediction, which learns features from raw data automatically.

The other is a self-corrective feedback with online learning, which updates learned model based on predicted SINR data and real measured SINR data later.

The parameters n and k could be determined by product requirements. For example, in an embodiment, the n could be defined 5 or 10.

Referring to FIG. 5, FIG. 5 is a flow chart illustrating an exemplary embodiment of a method for SINR prediction for link adaption.

at step510, receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

at step520, training one or more prediction models with the historic measurement data based on one or more neutral networks;

at step530, predicting SINR with the prediction models and the historic measurement data.

Specifically, the related measurements includes but not limited to RSSI (Received Signal Strength Indication) , PHR (Power Headroom Report) , CQI (Channel Quality Indicator) . For example, the radio access nodes send SINR, RSSI, PHR, CQI which collected from a plurality of TTIs before.

In another embodiment, the method further comprises the following steps:

converting the historic measurement data into input data for one or more neutral networks. For example, according to the format requirement of neural networks, it converts the historic measurement data in format into adaptive input data for neural networks.

Then, at step520, it trains one or more prediction models with the input data based on the one or more neutral networks.

In another embodiment, the method further comprises the following steps:

obtaining one or more prediction measurement data predicted by the prediction models and real measured SINR corresponding to the prediction measurement data; and updating the prediction models based on the prediction measurement data and real measured SINR.

Specifically, it compares the predict SINR and real measured SINR first, and determines whether updating the prediction models or not; if the predict SINR and real measured SINR are not same, or the difference between the predict SINR and real measured SINR exceeds the predetermined threshold, the prediction models will be updated by the historic measurement data which is used as input data and the real measured SINR which is used as a model answer. Herein, the historic measurement data includes but not limited to SINR, RSSI, PHR, CQI, etc.

Therefore, the method could realize a self-corrective feedback with online learning, which updates learned models based on predicted SINR data and real measured SINR data later.

At step530, it takes the historic measurement data as input information. The prediction models could calculate a result as the predicting SINR.

Also, a non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least:

And, a non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least:

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

For a person skilled in the art, it is apparent that the present disclosure is not limited to the details of the above exemplary embodiments, and the present disclosure can be implemented in other specific forms without departing from the spirit or essential characteristics of the present disclosure. Therefore, the embodiments should be regarded as exemplarily and not restrictive, and the scope of the present disclosure is defined by the appended claims rather than the above description, and therefore it is intended that the claims All changes that come within the meaning and range of equivalency of the disclosure are encompassed by the disclosure. Any reference signs in the claims should not be regarded as limiting the involved claims. In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or devices recited in the system claims may also be implemented by one unit or device through software or hardware. Words such as first, second, etc., are used to indicate a name and do not indicate any specific order.

Aspects of various embodiments are specified in the claims. Those and other aspects of various embodiments are specified in the following numbered clauses:

1. A method for SINR prediction for link adaption, comprising the steps of:

predicting SINR with the prediction models and the historic measurement data.

2. The method of clause 1, wherein the method further comprises:

a step of obtaining one or more prediction measurement data predicted by the prediction models and real measured SINR corresponding to the prediction measurement data;

a step of updating the prediction models based on the prediction measurement data and real measured SINR.

3. The method of

clause

1 or 2, wherein the method further comprises:

a step of converting the historic measurement data into input data for one or more neutral networks;

wherein, the step of training one or more prediction models with the historic measurement data based on one or more neutral networks comprises:

training one or more prediction models with the input data based on the one or more neutral networks.

4. The method of any one of clauses 1 to 3, wherein the neutral networks comprises any one or more DNNs (Deep Neural Networks) .

5. A method for SINR prediction modeling for link adaption in first apparatus, comprising the following steps:

6. The method of clause 5, wherein the method further comprises:

7. The method of clause 5 or 6, wherein the method further comprises:

8. The method of any one of clause s 5 to 7, wherein the neutral networks comprises any one or more DNNs (Deep Neural Networks) .

9. A method for SINR prediction for link adaption in second apparatus, comprising the following steps:

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

10. A first apparatus for SINR prediction modeling for link adaption, comprising means for performing:

11. The apparatus of claim 10, wherein the means are further configured to perform:

obtaining one or more prediction measurement data predicted by the prediction models and real measured SINR corresponding to the prediction measurement data;

updating the prediction models based on the prediction measurement data and real measured SINR.

12. The apparatus of claim 10 or 11, wherein the means are further configured to perform:

converting the historic measurement data into input data for one or more neutral networks;

wherein, the means for training one or more prediction models with the historic measurement data based on one or more neutral networks performing:

13. The apparatus of any one of claims 10 to 12, wherein the neutral networks comprises any one or more DNNs (Deep Neural Networks) .

14. The apparatus of any one of claims 10 to 13, wherein the means comprises:

at least one processor; and

15. A second apparatus for SINR prediction for link adaption, comprising means for performing:

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

16. The apparatus of claim 15, wherein the means comprises:

at least one processor; and

17. A central unit of radio access nodes, for SINR prediction modeling for link adaption, wherein, the central unit comprises the first apparatus of anyone of clauses 10 to 14.

18. A distributed unit of radio access nodes, for SINR prediction for link adaption, wherein, the central unit comprises the second apparatus of anyone of clauses 15 or 16.

19. A system for SINR prediction for link adaption, wherein, the system comprises a first apparatus of anyone of clauses 10 to 14 and a second apparatus of anyone of clauses 15 or 16.

20. A non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least: :

21. A non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least:

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

Claims

A method for SINR prediction for link adaption, comprising the steps of:

receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

training one or more prediction models with the historic measurement data based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data.
The method of claim 1, wherein the method further comprises:

a step of obtaining one or more prediction measurement data predicted by the prediction models and real measured SINR corresponding to the prediction measurement data;

a step of updating the prediction models based on the prediction measurement data and real measured SINR.
The method of claim 1 or 2, wherein the method further comprises:

a step of converting the historic measurement data into input data for one or more neutral networks;

wherein, the step of training one or more prediction models with the historic measurement data based on one or more neutral networks comprises:

training one or more prediction models with the input data based on the one or more neutral networks.
The method of any one of claims 1 to 3, wherein the neutral networks comprises any one or more DNNs (Deep Neural Networks) .
A first apparatus for SINR prediction modeling for link adaption, comprises:

at least one processor; and

at least one memory including computer program code,

the at least one memory and the computer program code configured to, with the at least one processor, cause the first apparatus to perform operation of at least the following:

receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

training one or more prediction models with the historic measurement data based on one or more neutral networks.
The apparatus of claim 5, wherein the operation further comprises:

obtaining one or more prediction measurement data predicted by the prediction models and real measured SINR corresponding to the prediction measurement data;

updating the prediction models based on the prediction measurement data and real measured SINR.
The apparatus of claim 5 or 6, wherein the operation further comprises::

converting the historic measurement data into input data for one or more neutral networks;

wherein, the means for training one or more prediction models with the historic measurement data based on one or more neutral networks performing:

training one or more prediction models with the input data based on the one or more neutral networks.
The apparatus of any one of claims 5 to 7, wherein the neutral networks comprises any one or more DNNs (Deep Neural Networks) .
A second apparatus for SINR prediction for link adaption, comprises:

at least one processor; and

at least one memory including computer program code,

the at least one memory and the computer program code configured to, with the at least one processor, cause the second apparatus to perform operation of at least the following:

receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

sending the predicting SINR to one or more link adaptation modules in cellular network.
A central unit of radio access nodes, for SINR prediction modeling for link adaption, wherein, the central unit comprises the first apparatus of anyone of claims 5 to 8.
A distributed unit of radio access nodes, for SINR prediction for link adaption, wherein, the central unit comprises the second apparatus of claim 9.
A system for SINR prediction for link adaption, wherein, the system comprises a first apparatus of anyone of claims 5 to 8 and a second apparatus of claim 9.
A non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least:

receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

training one or more prediction models with the historic measurement data based on one or more neutral networks.
A non-transitory computer-readable medium comprising computer instructions for causing an apparatus to perform at least:

receiving historic measurement data collected from radio access nodes, wherein, the measurement data comprises SINR and related measurements;

obtaining one or more prediction models based on one or more neutral networks;

predicting SINR with the prediction models and the historic measurement data;

sending the predicting SINR to one or more link adaptation modules in cellular network.