Base station paging method based on artificial intelligence

Abstract

The invention discloses a base station paging method based on artificial intelligence. The method comprises the following steps: S1, starting reporting of UE position information and measurement information; s2, preprocessing the user position information, and clustering the user position information by using a clustering analysis method; s3, based on the clustering result of the step S2, constructing a Markov model to predict the activity track of the user; s4, determining the initial state of the user by using the constructed Markov model, predicting the position where the user appears in the future based on the state transition matrix, and gradually predicting the activity track of the user in a period of time in the future; and S5, issuing paging information according to the predicted user activity track, preparing paging air interface resources in advance for each cell, and initiating random access to complete the paging process when the terminal receives the paging information and confirms matching. The paging resource is prepared in advance after the UE activity track is obtained based on the artificial intelligence technology, the paging time delay is reduced, paging is assisted by the neighbor cell in an extreme scene, and the paging success rate is improved.

Description

Technical Field

[0001] The invention belongs to the field of communication technology, and in particular relates to a base station paging method based on artificial intelligence. Background Art

[0002] The current LTE / NR 5G system paging uses the 5G core network to take the intersection of the tracking area supported by the UE when registering and the PLMN and tracking area list supported by the gNodeB, and sends the NG port PAGING to the gNodeB that supports the tracking area where the UE is located. This message carries the pagingDRX parameter, which is used to calculate the UE identity (5G-S-TMSI) at the paging time. After receiving the paging information, the gNodeB matches the TAI carried in the NG paging message with the cell TA configured by the base station, and selects the cell with the matching TA for air interface paging. Using the paging UE_ID and paging parameters, the PO is calculated and the UE identity (5G-S-TMSI) is sent to the UE through air interface Paging. The UE uses DRX to reduce power consumption. In each DRX cycle, the UE will only open the receiver in the slot of its own PO to listen to the PDCCH and read the paging message packet. If the UE does not detect the PDCCH encrypted with P-RNTI, or the decoded UE identity is inconsistent with its own UE representation (5G-S-TMSI), the UE immediately turns off the receiver and enters the DRX sleep period to save power; if the UE decodes its own paging message, that is, its own UE identity (5G-S-TMSI) is consistent with that carried in the paging message, the random access process is initiated.

[0003] Current technical problems:

[0004] (1) The paging process is complicated, and there is a large amount of redundant paging, which increases unnecessary overhead of the base station;

[0005] (2) Paging delay is large;

[0006] (3) In extreme scenarios, the UE may not be paged, which may lead to a paging storm. Summary of the invention

[0007] The main purpose of the present invention is to solve the problems of complex paging process and large paging delay.

[0008] To achieve the above object, the present invention provides a base station paging method based on artificial intelligence, comprising the following steps:

[0009] S1. Enable UE location information and measurement information reporting, periodically report the latitude and longitude of the area, local and neighboring area measurements, and cell information, and the base station receives and transmits the information to the designated server;

[0010] S2. Preprocessing the collected user location information, and clustering the user location information using a clustering analysis method to obtain a final user location clustering result;

[0011] S3, based on the result of user location clustering in step S2, construct a Markov model to predict the user's activity trajectory by determining the state space, calculating the state transition probability and constructing the state transition matrix;

[0012] S4. Use the constructed Markov model to determine the user's initial state, and predict the user's future location based on the state transition matrix, and gradually predict the user's activity trajectory in the future; evaluate the prediction results, and adjust and optimize the model based on the evaluation results;

[0013] S5. According to the predicted user activity trajectory, a cell containing the user is selected to send paging information. Each cell prepares paging air interface resources in advance. When the terminal receives the paging message and confirms a match, a random access process is initiated to complete the paging process.

[0014] Further, the specific implementation of step S2 is:

[0015] S201, collecting user location information, including latitude and longitude location information;

[0016] S202, pre-processing the user's location information, and deleting abnormal or erroneous data;

[0017] S203: Perform cluster analysis on the location information to obtain clustering results.

[0018] Furthermore, the specific method of step S203 is as follows:

[0019] (1) Calculate the distance based on the longitude and latitude reported in step S1

[0020] a. Convert longitude and latitude from degrees to radians: radians = degrees × (π÷180);

[0021] b. Calculate the latitude difference and longitude difference according to and The latitude and longitude values ​​of two points are used to calculate the latitude difference and longitude difference

[0022]

[0023] c. Substitute the earth's radius, latitude difference, and longitude difference into the Haversine formula to calculate the distance between the two points; the Haversine formula is as follows:

[0024]

[0025] Where d is the distance between two points, r is the radius of the earth, and are the latitudes of the two points, is the latitude difference, is the difference in longitude;

[0026] (2) Determine the number of initial clusters

[0027] Determine the final number of clusters N required based on the city or regional site planning plan;

[0028] (3) Obtaining the center of mass

[0029] Randomly select the centroids of N cluster samples as the initial centroids O1, O2, ... O N ; According to the reported longitude and latitude of the user information, the distance between all collected sample points and N centroids is calculated by the Haversine formula, and all users are divided into N categories; sorting is performed according to the calculated distance, and the sample points are placed in the category with the smallest distance. After classification, the centroid is recalculated, and the distance between samples in each category is averaged to obtain the new centroid O1′, O2′…O′ N After multiple iterations, the final clustering result O is obtained. final1 ,O final2 ,…O finalN .

[0030] Further, the specific implementation method of step S3 is:

[0031] I. Determine the state space: Based on the clustering result of step S2 final1 ,O final2 ,…O finalN , determine the state space of the model;

[0032] II. Calculate state transition probability: Based on the user's location information, calculate the probability that the user will transfer from one location to another;

[0033] III. Construct a state transfer matrix: The state transfer probability calculated in II is constructed into a state transfer matrix, where the rows and columns of the matrix represent different states, and the elements in the matrix represent the probability of transitioning from one state to another.

[0034] Further, the specific implementation method of step S4 is:

[0035] a. Initial state determination: Determine the initial state of the model based on the position reported by the user at the last moment;

[0036] b. Predicting future states: Using the state transition matrix of step S3, according to the transition size of each state, predict the user's future location, and gradually predict the user's activity trajectory in the future;

[0037] c. Result evaluation and adjustment: Evaluate the prediction results and calculate the prediction accuracy index by comparing them with the actual activity trajectory. The specific formula is:

[0038] Accuracy = Number of correctly predicted samples / Total number of samples

[0039] According to the evaluation results, the model is adjusted and optimized. When the accuracy is lower than the threshold T, the model is adjusted, that is, the number of clusters N is changed to M and steps S1-S4 are repeated.

[0040] Beneficial effects:

[0041] After obtaining the UE activity trajectory based on artificial intelligence technology, paging resources are prepared in advance to reduce paging delays. In extreme scenarios, neighboring cells assist in paging, improve the paging success rate, and do not waste a lot of paging resources. The present invention also reduces unnecessary resource consumption by optimizing the paging process, providing strong support for the further development of 5G communication systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of the base station paging method provided by the present invention;

[0043] Figure 2 It is a base station paging structure diagram provided by the present invention;

[0044] Figure 3 It is a user movement trajectory diagram provided by the present invention;

[0045] Figure 4 This is a user movement trajectory prediction graph provided by the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] The application principle of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0048] Embodiment 1:

[0049] A base station paging method based on artificial intelligence is as follows:

[0050] S1. Get the user's location information

[0051] Enable UE location information and measurement information reporting, and users periodically report the latitude and longitude of the area, local and neighboring area measurements, cell information, etc.

[0052] The base station transmits the data to the designated server in the format of Table 1 according to the periodically reported information;

[0053] Table 1 User location information

[0054]

[0055] Where Gnb id indicates the ID of the base station, Cell id indicates the number of the cell under the base station, and RSRP indicates the signal strength of the cell under the base station. For example: cell 11 of base station 1111.

[0056] S2. Processing of user location information data

[0057] S201. Collect the location information of the user in step S1, including the latitude and longitude location and other information.

[0058] S202: pre-process the data, that is, delete abnormal values ​​and data with incorrect format.

[0059] S203: Perform cluster analysis on the location information to obtain cluster results. The specific method is as follows:

[0060] (1) Calculate the distance based on the longitude and latitude reported in step S1

[0061] a. Convert longitude and latitude from degrees to radians; latitude and longitude are usually measured in degrees, but radians are required in the formula. Specifically: radians = degrees × (π÷180).

[0062] b. Calculate the latitude difference and longitude difference according to and The latitude and longitude values ​​of two points are used to calculate the latitude difference and longitude difference

[0063]

[0064] c. Substitute the radius of the earth, the latitude difference, and the longitude difference into the Haversine formula to calculate the distance between the two points. The Haversine formula is:

[0065]

[0066] Where d is the distance between two points, r is the radius of the earth (usually an average value of about 6371 kilometers), and are the latitudes of the two points, is the latitude difference, It is the longitude difference.

[0067] (2) Determine the number of initial clusters

[0068] The final number of clusters N required is determined based on the site planning scheme for a certain city / region. That is, when operators plan and build networks in a certain city, they will have a clear plan, that is, the number of base stations to be built in the region is determined. Assuming that a certain region plans to have M base stations, N physical sites are randomly selected as the number of initial clusters, and the locations of these N base stations are recorded as the initial centroids in the next step (3). Of course, N here is <M。

[0069] (3) Obtaining the center of mass

[0070] Randomly select the centroids of N samples as the initial centroids O1, O2, ... O N . Based on the reported longitude and latitude, the distance between all collected sample points and N centroids is calculated using the Haversine formula, and all users are divided into N categories. The samples are sorted according to the calculated distances, and the sample points are placed in the category with the smallest distance. After classification, the centroid is recalculated, and the average distance between samples in each category is calculated to obtain the new centroid O1′, O2′…O′ N After multiple iterations, the final clustering result O is obtained. final1 ,O final2 ,…O finalN .

[0071] S3. Constructing a Markov model

[0072] The Markov model is a probability-based random process model that assumes that the state of the system at a certain moment is only related to the state at the previous moment, and has nothing to do with the earlier state. In the prediction of user activity trajectory, the user's location is regarded as a state, and the user's transfer between different locations is regarded as a state transfer. The details are as follows:

[0073] I. Determine the state space: Based on the clustering result of step S2 final1 ,O final2 ,…O finalN , determine the state space of the model, that is, all possible user locations. When the core network pages a terminal, the core network can obtain the base station ID and location information of the terminal's last access, so that the category can be obtained according to the classification result in step S2. Assume that the core network wants to page UE1, obtain the latitude and longitude of the base station where UE1 was last located and the base station ID. According to the clustering method in step S2, the category is 1, that is, state 1. According to the historical probability of users in category 1 (state 1) moving to the other 5 categories, they are shown in Table 2.

[0074] II. Calculate the state transition probability: Based on the user's location information, calculate the probability of the user transferring from one location to another. This can be done by counting the number of historical transfers of the user and then calculating the transition probability.

[0075]

[0076] Where N1 represents the total number of class 1 (state 1) samples, n 11 It indicates the number of times state 1 is converted to state 1, that is, the number of times the state is still in state 1 the next time it was in class 1 (state 1) the last time. The calculation of other state transition probabilities is similar. For example Figure 4 As shown, P 11 It means the probability of going from the last state 1 to the next state 1 is 0.25.

[0077] III. Constructing the state transfer matrix: The state transfer probability calculated in II is constructed into a state transfer matrix, where the rows and columns of the matrix represent different states, and the elements in the matrix represent the probability of transitioning from one state to another, as shown in the matrix in Table 2.

[0078] Table 2 is the state transfer matrix

[0079]

[0080] S4. Predicting the user's activity trajectory

[0081] a. Initial state determination: Determine the initial state of the model based on the position reported by the user at the last moment.

[0082] b. Predicting future states: Using the state transition matrix of step S3, according to the transition size of each state, predict the possible future location of the user, and gradually predict the user's activity trajectory in the future.

[0083] c. Result evaluation and adjustment: The prediction results can be evaluated by comparing them with the actual activity trajectory and calculating indicators such as prediction accuracy. The specific formula is as follows.

[0084] Accuracy = Number of correctly predicted samples / Total number of samples

[0085] According to the evaluation results, the model can be adjusted and optimized. When the accuracy is lower than the threshold T, T can be set according to the industry's common threshold, such as 0.9. To adjust the model, the specific method is to change the number of clusters N to M and continue to repeat the above steps S1-S4. If the accuracy is low, the number of initial clusters can be reduced. For example, if the initial setting is 10 categories and the accuracy is low, it can be set to 9 categories for retraining, and then observe whether the accuracy is improved, and keep repeating the experiment.

[0086] S5. Sending paging information

[0087] According to the predicted user activity trajectory in step S4, 5 cells are selected to send paging information. Each cell prepares paging air interface resources in advance. Specific resources include boost power, spectrum resource allocation, etc. The paging message carries the terminal IMSI / UE ID information. When the terminal receives the paging message and finds that the UE ID / IMSI is its own, it initiates a random access process.

[0088] The explanations of the technical terms in the above content are shown in Table 3.

[0089] Table 3 List of proper nouns

[0090]

[0091] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A base station paging method based on artificial intelligence, characterized in that: The following steps are involved: S1. Enable UE location information and measurement information reporting, periodically report the latitude and longitude of the area, local and neighboring area measurements, and cell information, and the base station receives and transmits the information to the designated server; S2. Preprocessing the collected user location information, and clustering the user location information using a clustering analysis method to obtain a final user location clustering result; S3, based on the result of user location clustering in step S2, construct a Markov model to predict the user's activity trajectory by determining the state space, calculating the state transition probability and constructing the state transition matrix; S4. Determine the initial state of the user using the constructed Markov model, and predict the user's future location based on the state transition matrix, and gradually predict the user's activity trajectory in the future; evaluate the prediction results, and adjust and optimize the model based on the evaluation results; S5. According to the predicted user activity trajectory, a cell containing the user is selected to send paging information. Each cell prepares paging air interface resources in advance. When the terminal receives the paging message and confirms a match, a random access process is initiated to complete the paging process.

2. The base station paging method based on artificial intelligence according to claim 1, characterized in that: Specific implementation of step S2: S201, collecting user location information, including latitude and longitude location information; S202, pre-processing the user's location information, and deleting abnormal or erroneous data; S203: Perform cluster analysis on the location information to obtain clustering results.

3. The base station paging method based on artificial intelligence according to claim 2, characterized in that: The specific method of step S203 is as follows: (1) Calculate the distance based on the longitude and latitude reported in step S1 a. Convert longitude and latitude from degrees to radians: radians = degrees × (π÷180); b. Calculate the latitude difference and the longitude difference ▽φ, according to and The latitude and longitude values ​​of two points are used to calculate the latitude difference and longitude difference ▽φ: c. Substitute the earth's radius, latitude difference, and longitude difference into the Haversine formula to calculate the distance between the two points; the Haversine formula is as follows: Where d is the distance between two points, r is the radius of the earth, and are the latitudes of the two points, is the latitude difference, ▽φ is the longitude difference; (2) Determine the number of initial clusters Determine the final number of clusters N required based on the city or regional site planning plan; (3) Obtaining the center of mass Randomly select the centroids of N cluster samples as the initial centroids O1, O2, ... O N ; According to the reported longitude and latitude of the user information, the distance between all collected sample points and N centroids is calculated by the Haversine formula, and all users are divided into N categories; sorting is performed according to the calculated distance, and the sample points are placed in the category with the smallest distance. After classification, the centroid is recalculated, and the distance between samples in each category is averaged to obtain the new centroid O1′, O2′…O′ N After multiple iterations, the final clustering result O is obtained. final1 ,O final2 ,…O finalN .

4. The base station paging method based on artificial intelligence according to claim 1, characterized in that: Specific implementation method of step S3: I. Determine the state space: Based on the clustering result of step S2 final1 ,O final2 ,…O finalN , determine the state space of the model; II. Calculate state transition probability: Based on the user's location information, calculate the probability that the user will transfer from one location to another; III. Construct a state transfer matrix: The state transfer probability calculated in II is constructed into a state transfer matrix, where the rows and columns of the matrix represent different states, and the elements in the matrix represent the probability of transitioning from one state to another.

5. The base station paging method based on artificial intelligence according to claim 1, characterized in that: Specific implementation method of step S4: a. Initial state determination: Determine the initial state of the model based on the position reported by the user at the last moment; b. Predicting future states: Using the state transition matrix of step S3, according to the transition size of each state, predict the user's future location, and gradually predict the user's activity trajectory in the future; c. Result evaluation and adjustment: Evaluate the prediction results and calculate the prediction accuracy index by comparing them with the actual activity trajectory. The specific formula is: Accuracy = Number of correctly predicted samples / Total number of samples According to the evaluation results, the model is adjusted and optimized. When the accuracy is lower than the threshold T, the model is adjusted, that is, the number of clusters N is changed to M and steps S1-S4 are repeated.

Images