[go: up one dir, main page]

CN112235821B - Pilot frequency signal intensity prediction method based on AI model - Google Patents

Pilot frequency signal intensity prediction method based on AI model Download PDF

Info

Publication number
CN112235821B
CN112235821B CN202010998126.8A CN202010998126A CN112235821B CN 112235821 B CN112235821 B CN 112235821B CN 202010998126 A CN202010998126 A CN 202010998126A CN 112235821 B CN112235821 B CN 112235821B
Authority
CN
China
Prior art keywords
pilot frequency
rsrp
model
pci
cell
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010998126.8A
Other languages
Chinese (zh)
Other versions
CN112235821A (en
Inventor
余秋星
黄晶
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Honglingtong Information Technology Co ltd
Original Assignee
Hangzhou Honglingtong Information Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Honglingtong Information Technology Co ltd filed Critical Hangzhou Honglingtong Information Technology Co ltd
Priority to CN202010998126.8A priority Critical patent/CN112235821B/en
Publication of CN112235821A publication Critical patent/CN112235821A/en
Application granted granted Critical
Publication of CN112235821B publication Critical patent/CN112235821B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W24/00Supervisory, monitoring or testing arrangements
    • H04W24/06Testing, supervising or monitoring using simulated traffic
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B17/00Monitoring; Testing
    • H04B17/30Monitoring; Testing of propagation channels
    • H04B17/391Modelling the propagation channel
    • H04B17/3912Simulation models, e.g. distribution of spectral power density or received signal strength indicator [RSSI] for a given geographic region

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Electromagnetism (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a pilot frequency signal strength prediction method based on an AI model. The method comprises the following steps: step 1, data acquisition, step 2, sample pretreatment, step 3, obtaining a prediction model, step 4, model detection, and step 5, model-based prediction and use: and 6: the invention has the advantages that: the pilot frequency signal strength is accurately predicted, and the pilot frequency measurement process is avoided under the scene that pilot frequency switching is carried out on UE to select a pilot frequency target cell or a plurality of auxiliary carriers are configured during carrier aggregation. The method and the device have the advantages that the signal strength information of the pilot frequency adjacent cell is obtained without depending on the measurement time slot, and meanwhile, the problems that in the prior art, the pilot frequency signal strength is inaccurate to predict due to too large grid granularity and the prediction model is invalid due to the fact that grids are not updated regularly are solved.

Description

Pilot frequency signal intensity prediction method based on AI model
Technical Field
The invention belongs to the technical field of mobile communication, and particularly relates to an AI (Artificial Intelligence) model-based pilot signal intensity prediction method.
Background
At present, with the gradual popularization of smart phones, the rapid development of mobile communication requirements is driven, the traffic of a Long Term Evolution (LTE) wireless network as a main communication technology is increasing, and a single wireless frequency band, that is, a single carrier cannot meet the social requirements, so an operator usually deploys more wireless frequency bands to provide a larger communication capacity, for example, the same area is covered by multiple carriers to improve the communication capacity of a system. Due to the frequency difference and deployment difference of different frequency bands, the signal strength of different frequency bands at the same position may have a large difference, and therefore, a User Equipment (UE) needs to measure the signal strength of each frequency band to ensure the success rate of inter-frequency handover. Since the user equipment is constrained by cost, power consumption and other aspects, it is difficult to measure the signal strength of multiple frequency bands simultaneously, and the signal strength of each frequency band is often measured by turns for multiple times. For inter-frequency measurement, in the LTE standard, a base station configures a measurement time slot (measurement gap) lasting 6 milliseconds to a user equipment in a serving cell in a certain frequency band, so that the user equipment switches to a specified inter-frequency to measure signal strength, and at this time, the user equipment interrupts signal reception and transmission with the current serving cell. In order to measure the neighboring cell signals of multiple different frequency bands, the ue needs to use multiple measurement timeslots. The interaction with the current cell is interrupted due to the pilot frequency measurement, so that the uplink and downlink transmission rate of the user equipment is greatly reduced, and the service experience of the user is poor. Especially for the LTE TDD system, the uplink and downlink rate losses are about 20% and 40%, respectively, which seriously reduces the user experience.
In order to solve the above problems, a prior art solution provided by patent publication No. CN111356142 is: dividing the coverage areas of a target cell and two adjacent cells into RSRP (Reference Signal Receiving Power), identifying each RSRP value interval by a fixed step length (such as 5dB), carrying out grid division on the coverage area of the target cell and forming grid information, namely, the grid is the RSRP value interval of three cells, and establishing a grid fingerprint library for selecting and establishing a double-link auxiliary cell. The scheme has the defects that when the area in which users are concentrated is limited by a fixed step length (such as 5dB) and the number of adjacent cells is limited, the grid division granularity is large, the characteristic difference of different grids is large, and the problem of low prediction accuracy is brought; if the fixed step length is too small, excessive grids can be caused, so that the storage space is too large and the searching time is too long; in the area with few users, the grid does not have reasonable statistical characteristics due to the fact that the number of samples is too small, and therefore the grid is not available or prediction accuracy is not high after the grid is used.
Disclosure of Invention
It is an object of the present invention to provide an AI model-based pilot signal strength prediction method capable of overcoming the above-mentioned technical problems,
The method comprises the following steps:
step 1, data acquisition:
AI model information among different frequency points of a Cell is constructed through signal strength measurement information of the same frequency and different frequencies reported by UE, and PCI (Physical Cell Identifier) and RSRP information of a serving Cell and adjacent cells of the same frequency, which are measured by the UE at the same time, are associated with the PCI and RSRP of the different frequency points measured by the UE;
step 2, sample pretreatment:
sample pre-processing includes de-duplicating samples and filling the unmeasured RSRP values of co-or inter-frequency neighbor cells with-140 dBm (decibel-milliwatts).
The data format of the preprocessed data is as follows:
{[f1,PCI1_1,RSRP1_1],…,[f1,PCI1_N1,RSRP1_N1],
[f2,PCI2_1,RSRP2_1],…,[f2,PCI2_N2,RSRP2_N2],
[fK,PCIK_1,RSRPK_1],…,[fK,PCIK_NK,RSRPK_NK]}
wherein, f1, f2,. f K represents K frequency points, and corresponding PCI and RSRP are respectively N1, N2, … and NK groups;
step 3, obtaining a prediction model:
the PCI and RSRP of the co-frequency cells with the frequency points of fj are used as characteristic input, the co-frequency cells comprise PCI and RSRP information of 1 serving cell and 1-N cells, the PCI and RSRP of the pilot frequency cells are output of model prediction, and the model can predict 1-M pilot frequency cells with strongest RSRP;
specifically, the method comprises the following steps: the input data is:
{[fi,PCI1_1,RSRP1_1],…,[fi,PCI1_N1,RSRP1_Ni]}
the output data is:
{[fj,PCI1_1,RSRP1_1],…,[fj,PCI1_N1,RSRP1_Nj]
[fm,PCI1_1,RSRP1_1],…,[fm,PCI1_N1,RSRP1_Nm]}
wherein, fj, …, fk represent different k-j +1 frequency points;
Model training randomizes the samples used, using a set proportion of sample data (e.g., 75%) therein for training, with the remainder of the samples, e.g., 25%, being used as a measure of the predicted performance of the model.
The training of the AI model adopts common AI algorithms such as GBDT (Gradient Boosting Decision Tree), linear models and the like, the AI algorithm can perform multiple rounds of iteration according to sample data, and multiple Decision Tree models are automatically generated according to characteristics;
step 4, model detection:
the model detection detects the accuracy rate and the recall ratio, wherein the definition of the accuracy rate and the recall ratio is respectively the following formula (1) and (2):
the accuracy rate is that the predicted value is A and the number of the sample attributes is A/the number of the samples with the predicted value is A is 100%; … … (1) of the raw materials,
the recall ratio is the number of the predicted value A and the sample attribute A/the number of the sample attributes A in all samples is 100%; … … (2) of the first and second substrates,
in the above formulas (1) and (2), a represents PCI and RSRP information of the inter-frequency cell;
when the accuracy rate and the recall ratio exceed a preset threshold (such as 70%), the model can pass the detection and be used online, and the next step is carried out; otherwise, turning to step 1, the model needs to acquire data again for retraining and detecting.
And 5, model-based prediction and use:
after the model is used online, predicting PCI and RSRP information of a pilot frequency cell according to a plurality of groups of PCI and RSRP information of cells based on a certain frequency point reported by UE, and selecting a pilot frequency adjacent cell according to a use scene;
and step 5.1, after predicting the PCI and RSRP information of 1 or more (1-M) pilot frequency cells based on the AI model, the method can be used for selecting a target cell when the UE performs a big data service so as to ensure the user experience of the UE.
According to PCI and RSRP information of the same-frequency cells reported by UE, predicting PCI and RSRP information of a plurality of pilot frequency cells under the same pilot frequency point, and after rejecting pilot frequency cell information with RSRP of-140 dBm, calculating SINR (signal to interference plus noise ratio) under one or a plurality of pilot frequency points, as shown in the following formula (3):
Figure BDA0002693319200000031
wherein, the SINRjSignal to interference plus noise ratio, RSRP, of the jth pilot frequency pointj,0Is the strongest RSRP under the pilot frequency point jj,kThe reference signal is the kth RSRP except the strongest RSRP under the pilot frequency point j;
when the UE is detected to be doing big data service, comparing the signal to interference plus noise ratio of the current service cell with the signal to interference plus noise ratio information of each pilot frequency point, namely in the process of duration T, when the signal to interference plus noise ratio of the current service cell is lower than a first threshold Th1 and the signal to interference plus noise ratio of the pilot frequency point exceeds a second threshold Th2, selecting the pilot frequency cell { fx, PCIx } which meets the conditions and has the largest signal to interference plus noise ratio, and directly switching the UE to the pilot frequency cell { fx, PCIx } so as to ensure the user experience of the UE;
Step 5.2, the method for selecting a SCC (secondary Carrier component) by a conventional CA (Carrier Aggregation), wherein the UE performs pilot frequency measurement by using a measurement time slot, and for pilot frequency measurement under multiple frequency points, after predicting PCI and RSRP information of 1 or more (1 to M) pilot frequency cells based on an AI model, the UE is used for selecting the SCC under the CA, and meanwhile, the pilot frequency measurement process can be avoided to ensure that multiple SCCs are quickly selected while ensuring user experience of the UE;
according to PCI and RSRP information of the same-frequency cells reported by the UE, predicting PCI and RSRP information of a plurality of pilot frequency cells under the same pilot frequency point, and then selecting the cell with the strongest RSRP and exceeding a third threshold Th3 as the SCC of the UE for each pilot frequency point, so that a plurality of SCCs of the UE are quickly determined;
calculating the signal to interference plus noise ratio of the pilot frequency points according to the predicted information of the PCI and RSRP of the pilot frequency points; in the process of the duration T, the signal to interference plus noise ratio of the current service cell is lower than a first threshold Th1, and the signal to interference plus noise ratio of the pilot frequency point exceeds a second threshold Th2, and the pilot frequency cell with the highest signal to interference plus noise ratio is selected as a target cell for pilot frequency switching;
according to the predicted pilot frequency point PCI and RSRP information, when the carriers are aggregated, for each pilot frequency point, selecting 1 or more pilot frequency cells with the strongest RSRP and exceeding a third threshold Th3 as auxiliary carriers of the UE;
And 6: updating and maintaining the model:
after the model is used on line, data are periodically collected to carry out on-line detection, and when the accuracy and recall ratio of the model are found to be reduced to a threshold value through on-line detection, incremental data are collected to carry out updating training so as to prevent the model from being invalid.
The invention has the following advantages:
1. the method of the invention accurately predicts the pilot frequency signal strength, and avoids the pilot frequency measurement process under the scene of selecting the pilot frequency target cell by carrying out pilot frequency switching on the UE or configuring a plurality of auxiliary carriers during carrier aggregation.
2. The method of the invention can rapidly select the adjacent cell or a plurality of auxiliary carriers for pilot frequency switching without influencing user experience, thereby solving the problems of inaccurate pilot frequency signal strength prediction caused by too large grid granularity and inaccurate pilot frequency signal strength prediction caused by prediction model aging brought by not updating grids in the prior art.
3. The method of the invention obtains the signal strength information of the pilot frequency adjacent cell without depending on the measurement time slot, and simultaneously solves the problems of inaccurate pilot frequency signal strength prediction caused by too large grid granularity and inaccurate pilot frequency signal strength prediction caused by invalid prediction model due to no regular grid updating in the prior art.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the method of the present invention comprises the following steps:
step 1, data acquisition:
establishing AI model information among different frequency points of a Cell through signal strength measurement information of same frequency and different frequency reported by UE, and associating PCI (Physical Cell Identifier) and RSRP (reference signal received power) information of a serving Cell and a neighboring Cell of the same frequency, which are measured by the UE at the same time, with PCI and RSRP of different frequency points measured by the UE;
step 2, sample pretreatment:
sample preprocessing comprises the steps of carrying out de-repeatability on a sample, and filling the RSRP value of an identical-frequency adjacent cell or an different-frequency adjacent cell which is not measured to be-140 dBm (decibel-milliwatt);
the data format of the preprocessed data is as follows:
{[f1,PCI1_1,RSRP1_1],…,[f1,PCI1_N1,RSRP1_N1],
[f2,PCI2_1,RSRP2_1],…,[f2,PCI2_N2,RSRP2_N2],
[fK,PCIK_1,RSRPK_1],…,[fK,PCIK_NK,RSRPK_NK]}
wherein, f1, f2, fK represents K frequency points, and the corresponding PCI and RSRP are respectively N1, N2, …, and NK group;
step 3, obtaining a prediction model:
the PCI and RSRP of the co-frequency cells with the frequency points of fj are used as characteristic input, the co-frequency cells comprise PCI and RSRP information of 1 serving cell and 1-N cells, the PCI and RSRP of the pilot frequency cells are output of model prediction, and the model can predict 1-M pilot frequency cells with strongest RSRP;
Specifically, the method comprises the following steps: the input data is:
{[fi,PCI1_1,RSRP1_1],…,[fi,PCI1_N1,RSRP1_Ni]}
the output data is:
{[fj,PCI1_1,RSRP1_1],…,[fj,PCI1_N1,RSRP1_Nj]
[fm,PCI1_1,RSRP1_1],…,[fm,PCI1_N1,RSRP1_Nm]}
wherein, fj, …, fk represent different k-j +1 frequency points;
the model training randomizes the sample, wherein a set proportion of sample data (such as 75%) is used for training, and the rest of sample (such as 25%) is used for detecting the model predictive performance;
the AI model is trained by common AI algorithms such as GBDT (Gradient Boosting Decision Tree), linear models and the like, the AI algorithm performs multiple iterations according to sample data, and multiple Decision Tree models are automatically generated according to characteristics;
step 4, model detection:
the model detection detects the accuracy and the recall ratio, wherein the definition of the accuracy and the recall ratio is respectively as the following formula (1) and (2):
the accuracy rate is that the predicted value is A and the number of the sample attributes is A/the number of the samples with the predicted value is A is 100%; … … (1) in the form of a powder,
the recall ratio is the number of the predicted value A and the sample attribute A/the number of the sample attributes A in all samples is 100%; … … (2) of the first and second substrates,
in the above formulas (1) and (2), a represents PCI and RSRP information of the inter-frequency cell;
when the accuracy rate and the recall ratio exceed a preset threshold (such as 70%), the model can pass the detection and be used online, and the next step is carried out; otherwise, turning to the step 1, the model needs to acquire data again for retraining and detecting;
And 5, model-based prediction and use:
after the model is used online, predicting PCI and RSRP information of a pilot frequency cell according to a plurality of groups of PCI and RSRP information of cells based on a certain frequency point reported by UE, and selecting a pilot frequency adjacent cell according to a use scene;
step 5.1, after predicting the PCI and RSRP information of 1 or more (1-M) pilot frequency cells based on the AI model, the method can be used for selecting a target cell when the UE performs a big data service so as to ensure the user experience of the UE;
according to PCI and RSRP information of the same-frequency cells reported by UE, predicting PCI and RSRP information of a plurality of pilot frequency cells under the same pilot frequency point, and after rejecting pilot frequency cell information with RSRP of-140 dBm, calculating SINR (signal to interference plus noise ratio) under one or a plurality of pilot frequency points:
Figure BDA0002693319200000061
wherein, the SINRjSignal to interference plus noise ratio, RSRP, of the jth pilot frequency pointj,0Is the strongest RSRP under the pilot frequency point jj,kThe reference signal is the kth RSRP except the strongest RSRP under the pilot frequency point j;
when the UE is detected to be doing big data service, comparing the signal to interference plus noise ratio of the current service cell with the signal to interference plus noise ratio information of each pilot frequency point, namely in the process of duration T, when the signal to interference plus noise ratio of the current service cell is lower than a first threshold Th1 and the signal to interference plus noise ratio of the pilot frequency point exceeds a second threshold Th2, selecting the pilot frequency cell { fx, PCIx } which meets the conditions and has the largest signal to interference plus noise ratio, and directly switching the UE to the pilot frequency cell { fx, PCIx } so as to ensure the user experience of the UE;
Step 5.2, the method for selecting SCC (secondary Carrier component) by conventional CA (Carrier Aggregation), wherein the UE performs pilot frequency measurement by using a measurement time slot, and for pilot frequency measurement under multiple frequency points, after predicting PCI and RSRP information of 1 or more (1 to M) pilot frequency cells based on an AI model, the UE is used for selecting SCC under CA, and meanwhile, the pilot frequency measurement process can be avoided to ensure that multiple SCCs are quickly selected while user experience of the UE is ensured;
according to PCI and RSRP information of the same-frequency cells reported by the UE, predicting PCI and RSRP information of a plurality of pilot frequency cells under the same pilot frequency point, and then selecting the cell with the strongest RSRP and exceeding a third threshold Th3 as the SCC of the UE for each pilot frequency point, so that a plurality of SCCs of the UE are quickly determined;
calculating the signal to interference plus noise ratio of the pilot frequency points according to the predicted information of the PCI and RSRP of the pilot frequency points; in the process of the duration T, the signal to interference plus noise ratio of the current service cell is lower than a first threshold Th1, and the signal to interference plus noise ratio of the pilot frequency point exceeds a second threshold Th2, and the pilot frequency cell with the highest signal to interference plus noise ratio is selected as a target cell of pilot frequency switching;
according to the predicted pilot frequency point PCI and RSRP information, when the carriers are aggregated, for each pilot frequency point, selecting 1 or more pilot frequency cells with the strongest RSRP and exceeding a third threshold Th3 as auxiliary carriers of the UE;
And 6: updating and maintaining the model:
after the model is used on line, data are periodically collected to carry out on-line detection, and when the accuracy and recall ratio of the model are found to be reduced to a threshold value through on-line detection, incremental data are collected to carry out updating training so as to prevent the model from being invalid.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the present disclosure should be covered within the scope of the present invention claimed.

Claims (2)

1. An AI model-based pilot frequency signal strength prediction method is characterized by comprising the following steps:
step 1, data acquisition:
AI model information among different frequency points of a cell is constructed through signal strength measurement information of the same frequency and different frequency reported by UE, and PCI and RSRP information of a serving cell and adjacent cells of the same frequency measured by the UE at the same time are associated with PCI and RSRP of different frequency points measured by the UE;
step 2, sample pretreatment:
the sample pretreatment comprises the steps of carrying out de-repeatability on the sample, and filling the RSRP value of the same-frequency adjacent cell or different-frequency adjacent cell which is not measured to-140 dBm;
The data format of the preprocessed data is as follows:
{[f1,PCI1_1,RSRP1_1],…,[f1,PCI1_N1,RSRP1_N1],
[f2,PCI2_1,RSRP2_1],…,[f2,PCI2_N2,RSRP2_N2],
[fK,PCIK_1,RSRPK_1],…,[fK,PCIK_NK,RSRPK_NK]}
wherein, f1, f2, fK represents K frequency points, and the corresponding PCI and RSRP are respectively N1, N2, …, and NK group;
step 3, obtaining a prediction model:
the PCI and RSRP of the co-frequency cells with the frequency points of fj are used as characteristic input, the co-frequency cells comprise PCI and RSRP information of 1 serving cell and 1-N cells, the PCI and RSRP of the pilot frequency cells are output of model prediction, and the model can predict 1-M pilot frequency cells with strongest RSRP;
specifically, the method comprises the following steps: the input data is:
{[fi,PCI1_1,RSRP1_1],…,[fi,PCI1_N1,RSRP1_Ni]}
the output data is:
{[fj,PCI1_1,RSRP1_1],…,[fj,PCI1_N1,RSRP1_Nj]
[fm,PCI1_1,RSRP1_1],…,[fm,PCI1_N1,RSRP1_Nm]}
wherein, fj, …, fk represent different k-j +1 frequency points;
randomizing the adopted sample by using the model training, wherein sample data with a set proportion is used for training, and the rest sample is used for detecting the prediction performance of the model;
step 4, model detection:
the model detection detects the accuracy and the recall ratio, wherein the definition of the accuracy and the recall ratio is respectively as the following formula (1) and (2):
the accuracy rate is that the predicted value is A and the number of the sample attributes is A/the number of the samples with the predicted value is A is 100%; … … (1) in the form of a powder,
the recall ratio is the number of the predicted value A and the sample attribute A/the number of the sample attributes A in all samples is 100%; … … (2) of the first and second substrates,
In the above formulas (1) and (2), a represents PCI and RSRP information of the inter-frequency cell;
when the accuracy rate and the recall ratio exceed the preset threshold, the model can pass the detection and be used online, and the next step is carried out; otherwise, turning to the step 1, the model needs to acquire data again for retraining and detecting;
and 5, model-based prediction and use:
after the model is used on line, predicting PCI and RSRP information of a pilot frequency cell according to a plurality of groups of PCI and RSRP information which is reported by UE and is based on a certain frequency point, and selecting a pilot frequency adjacent cell according to a use scene;
and 6: updating and maintaining the model:
after the model is used online, data are periodically collected for online detection, and when the accuracy and recall ratio of the model are found to be reduced below threshold values through online detection, incremental data are collected for updating training to prevent the model from failing.
2. The AI-model-based pilot signal strength prediction method according to claim 1, wherein said step 5 comprises the steps of:
step 5.1, after predicting PCI and RSRP information of 1 or more pilot frequency cells based on the AI model, the method can be used for selecting a target cell when UE (user equipment) performs a big data service so as to ensure the user experience of the UE;
According to PCI and RSRP information of the same-frequency cells reported by UE, predicting PCI and RSRP information of a plurality of pilot frequency cells under the same pilot frequency point, eliminating pilot frequency cell information with RSRP being-140 dBm, and then calculating SINR (signal to interference plus noise ratio) under one or a plurality of pilot frequency points:
Figure FDA0002693319190000021
wherein, the SINRjSignal to interference plus noise ratio, RSRP, of the jth pilot frequency pointj,0Is the strongest RSRP under the pilot frequency point jj,kThe reference signal is the kth RSRP except the strongest RSRP under the pilot frequency point j;
when the UE is detected to be doing big data service, comparing the signal to interference plus noise ratio of the current service cell with the signal to interference plus noise ratio information of each pilot frequency point, namely in the process of duration T, when the signal to interference plus noise ratio of the current service cell is lower than a first threshold Th1 and the signal to interference plus noise ratio of the pilot frequency point exceeds a second threshold Th2, selecting the pilot frequency cell { fx, PCIx } which meets the conditions and has the largest signal to interference plus noise ratio, and directly switching the UE to the pilot frequency cell { fx, PCIx } so as to ensure the user experience of the UE;
step 5.2, the method for selecting SCC, i.e. secondary carrier, by conventional CA, i.e. carrier aggregation, includes that UE performs pilot frequency measurement by using a measurement time slot, and for pilot frequency measurement under multiple frequency points, after predicting PCI and RSRP information of 1 or more pilot frequency cells based on an AI model, the UE is used for selecting SCC under CA, and simultaneously, the pilot frequency measurement process can be avoided to ensure user experience of UE while quickly selecting multiple SCCs;
According to PCI and RSRP information of the same-frequency cells reported by the UE, predicting PCI and RSRP information of a plurality of pilot frequency cells under the same pilot frequency point, and then selecting the cell with the strongest RSRP and exceeding a third threshold Th3 as the SCC of the UE for each pilot frequency point, so that a plurality of SCCs of the UE are quickly determined;
calculating the signal to interference plus noise ratio of the pilot frequency points according to the predicted information of the PCI and RSRP of the pilot frequency points; in the process of the duration T, the signal to interference plus noise ratio of the current service cell is lower than a first threshold Th1, and the signal to interference plus noise ratio of the pilot frequency point exceeds a second threshold Th2, and the pilot frequency cell with the highest signal to interference plus noise ratio is selected as a target cell for pilot frequency switching;
and according to the predicted information of the pilot frequency points PCI and RSRP, selecting 1 or more pilot frequency cells with the strongest RSRP and exceeding a third threshold Th3 as auxiliary carriers of the UE for each pilot frequency point during carrier aggregation.
CN202010998126.8A 2020-09-21 2020-09-21 Pilot frequency signal intensity prediction method based on AI model Active CN112235821B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010998126.8A CN112235821B (en) 2020-09-21 2020-09-21 Pilot frequency signal intensity prediction method based on AI model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010998126.8A CN112235821B (en) 2020-09-21 2020-09-21 Pilot frequency signal intensity prediction method based on AI model

Publications (2)

Publication Number Publication Date
CN112235821A CN112235821A (en) 2021-01-15
CN112235821B true CN112235821B (en) 2022-06-28

Family

ID=74108424

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010998126.8A Active CN112235821B (en) 2020-09-21 2020-09-21 Pilot frequency signal intensity prediction method based on AI model

Country Status (1)

Country Link
CN (1) CN112235821B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116017616A (en) * 2021-10-21 2023-04-25 维沃移动通信有限公司 Cell reselection method, device and related equipment
CN116419332A (en) * 2021-12-29 2023-07-11 中国移动通信有限公司研究院 Channel measurement method, device, terminal and readable storage medium
WO2024161191A1 (en) * 2023-02-03 2024-08-08 Telefonaktiebolaget Lm Ericsson (Publ) Serving carrier prediction for mobility management
CN116506050A (en) * 2023-05-08 2023-07-28 中国电信股份有限公司北京研究院 Different frequency information prediction method, device, equipment and medium
WO2025184806A1 (en) * 2024-03-05 2025-09-12 Oppo广东移动通信有限公司 Wireless communication method, terminal device, and network device

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6252861B1 (en) * 1998-03-26 2001-06-26 Lucent Technologies, Inc. Methods and apparatus for interfrequency handoff in a wireless communication system
CN105592495B (en) * 2014-10-24 2020-08-04 南京中兴软件有限责任公司 Pilot frequency measurement threshold configuration method and device
EP3433953B1 (en) * 2016-03-21 2020-12-02 Telefonaktiebolaget LM Ericsson (PUBL) Target carrier radio predictions using source carrier measurements
CN108271165B (en) * 2016-12-30 2021-04-27 中国移动通信集团浙江有限公司 Method and system for predicting signal coverage state of base station network
CN107249200B (en) * 2017-05-10 2019-10-11 西安交通大学 A Switching Method Using Fuzzy Prediction Model
CN111385037A (en) * 2018-12-28 2020-07-07 上海交通大学 Real-time prediction method of indoor available frequency spectrum

Also Published As

Publication number Publication date
CN112235821A (en) 2021-01-15

Similar Documents

Publication Publication Date Title
CN112235821B (en) Pilot frequency signal intensity prediction method based on AI model
US20190068443A1 (en) Automatically optimize parameters via machine learning
CN106792752B (en) Base station signal coverage self-optimization method and system
US20100246529A1 (en) Wireless communication terminal and wireless communication system
CN109151881B (en) Network load balancing optimization method based on user data
JP7486659B2 (en) Energy saving processing method and processing device for base station
CN103929781A (en) Cross-layer interference coordination optimization method in ultra-dense heterogeneous networks
Farooq et al. Machine learning aided holistic handover optimization for emerging networks
CN100399858C (en) Selecting method of switching target community in TDD/CDMA system
US10986553B2 (en) Terminal device, communication system, and communication control method
WO2017166968A1 (en) Cell handover method and apparatus, and base station device
WO2018184667A1 (en) Apparatus and method for performing network optimization
CN103002495A (en) Evaluation method and device for wireless network structure
CN113133069A (en) Method and device for determining target cell, electronic equipment and storage medium
CN103139825A (en) Method, device and base station for dividing limbic users in long term evolution (LTE) system
CN105933923B (en) A kind of super-intensive networking cell cluster-dividing method
CN114363924B (en) 5G non-resident problem automatic root cause analysis method
EP2485516B1 (en) Radio coverage in mobile telecommunications systems
KR20130057987A (en) Method and apparatus for neighbor cell optimization in wireless communication network
CN114650598B (en) Method, network device and storage medium for configuring communication cell coordination relationship
CN109391310B (en) Beam configuration method and beam configuration device
CN115499884A (en) Cell service migration method and device and computer readable storage medium
CN102781015A (en) Method and system for quantitatively evaluating complexity of network structures
CN103458511B (en) A kind of method and device of division of cells edge band
CN114258071B (en) Measurement reporting method and device, storage medium and terminal

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CB03 Change of inventor or designer information

Inventor after: Yu Qiuxing

Inventor after: Zhao Xianming

Inventor after: Dong Zhenjiang

Inventor after: Fang Bingyi

Inventor after: Huang Jing

Inventor before: Yu Qiuxing

Inventor before: Huang Jing

CB03 Change of inventor or designer information