Methods, data processing devices, and media for configuring access points using computers

Abstract

This invention provides a method for configuring a first access point in a wireless telecommunications network, wherein the first access point is one of a plurality of access points. The method includes the following steps: obtaining time-series performance data of each of the plurality of access points; using dynamic time warping technology to determine a similarity value between the time-series performance data of each of the plurality of access points and the time-series performance data of each other of the plurality of access points; identifying a plurality of clusters of the plurality of access points based on the similarity value, wherein the first access point is a member of a first cluster of the plurality of clusters; identifying the configuration of the first access point based on the association between the first access point and the first cluster; and prompting the identified configuration of the first access point.

Description

Technical Field

[0001] This invention relates to a method for reconfiguring a wireless telecommunications network. Background Technology

[0002] Wireless telecommunications networks may experience business growth due to an increase in users and / or an increase in the amount of data consumed by one or more users. To prevent any problems that may occur if this business exceeds the network's capacity, network operators can upgrade the network. Network upgrades can include adding new transceivers, using new radio resources, using new backhaul resources, using new technologies, using new protocols, and / or adding additional processing capabilities. Network operators can forecast usage growth in the network and upgrade the network to handle the forecasted growth. Growth can be forecasted at the access point level, allowing each access point to be upgraded independently of other access points in the network. If the usage growth at an access point is underestimated, users at the upgraded access point may experience poor service as business demand exceeds the capacity of the upgraded access point. If the usage growth is overestimated, the network operator wastes resources on upgrades that might otherwise have been used to upgrade other access points in the network.

[0003] Network operators can determine the growth rate of each access point by reviewing its historical performance data and identifying its growth rate. However, this requires significant storage and processing resources when there are many access points in the network. A simplified approach to forecasting access point growth in a network is to estimate a general growth rate (e.g., 10%) and apply this general growth rate to every access point in the network. This general approach reduces the resources required to determine the growth rate of each access point, but it is more likely to underestimate or overestimate the growth rate.

[0004] Other network configurations suffer from the same problem: network operators must use significant resources to identify the configuration for each access point based on data from each access point, or apply a generic configuration across multiple access points, which risks being unsuitable for one or more of these access points. This can be applied, for example, to network operators determining when to switch access points to power-saving mode. Summary of the Invention

[0005] According to a first aspect of the present invention, a computer-implemented method for configuring a first access point in a wireless telecommunications network is provided, the first access point being one of a first plurality of access points. The method includes the steps of: obtaining time-series performance data of each of the first plurality of access points; determining a similarity value between the time-series performance data of each of the first plurality of access points and the time-series performance data of each other of the first plurality of access points using dynamic time warping techniques; identifying a first plurality of clusters among the first plurality of access points based on the similarity value; identifying the configuration of the first access point based on its association with the first cluster among the plurality of clusters; and enabling the first access point to implement the identified configuration. This method can be defined as a method for configuring a first access point in a wireless telecommunications network, wherein the first access point is one of a plurality of access points, and the method includes the following steps: obtaining time-series performance data of each of the plurality of access points; using dynamic time warping techniques to determine a similarity value between the time-series performance data of each of the plurality of access points and the time-series performance data of each other of the plurality of access points; identifying a plurality of clusters of the plurality of access points based on the similarity value, wherein the first access point is a member of a first cluster of the plurality of clusters; identifying the configuration of the first access point based on the association between the first access point and the first cluster; and prompting the identified configuration of the first access point.

[0006] The first wireless telecommunications network may include a second plurality of access points. The step of determining a similarity value is performed between each access point in the second plurality of access points and each other access point in the second plurality of access points. The step of identifying a first plurality of clusters may be performed for the second plurality of access points. The method may further include the following steps: generating time-series performance data for each cluster in the first plurality of clusters; determining a similarity value between the time-series performance data of each access point in the first plurality of access points and the time-series performance data generated for each cluster in the first plurality of clusters; and associating each access point in the first plurality of access points with a cluster in the first plurality of clusters based on the determined similarity value. In other words, the first plurality of access points may include a first subset of access points, and the step of determining the similarity value may be performed between each access point in the first subset of access points and each other access point in the first subset of access points, and the step of identifying the first plurality of clusters may be performed for the first subset of access points. The method may also include the following steps: generating time series performance data for each cluster in the first plurality of clusters; determining a similarity value between the time series performance data of each access point in the first plurality of access points and the time series performance data generated for each cluster in the first plurality of clusters; and associating each access point in the first plurality of access points with a cluster in the first plurality of clusters based on the determined similarity value.

[0007] The step of generating time-series performance data for each of the first plurality of clusters can be based on the average of the time-series performance data for each access point in the cluster.

[0008] The configuration can be determined using a machine learning method that takes the association between the first access point and the first cluster in the first plurality of clusters as input, wherein the machine learning method can be trained on a dataset that identifies the historical performance data of each of the first plurality of clusters.

[0009] The machine learning method may also use one or more of the following as input: the density of sites associated with the first access point, the average value of performance data over a specific time period, the height of the antenna of the first access point, the set of carriers used by the first access point, and the supplier of the antenna of the first access point, wherein the dataset used to train the machine learning method also identifies the corresponding historical metric for each of the first plurality of clusters.

[0010] Time-series performance data may include one or more of the following: measures of connected users and measures of resource usage.

[0011] Time series performance data can include multiple performance metrics, and dynamic time warping techniques can be multivariate dynamic time warping techniques.

[0012] The identified configuration can be one or more of the following: the capacity of the first access point, the switching parameters of the first access point, and the energy-saving mode of the first access point.

[0013] The wireless telecommunications network may be a cellular telecommunications network, and the first access point may be the first sector of the first base station.

[0014] According to a second aspect of the invention, a computer program including instructions is provided, which, when executed by a computer, causes the computer to perform the steps of the first aspect of the invention. The computer program may be stored on a computer-readable media.

[0015] According to a third aspect of the present invention, a data processing apparatus is provided, the data processing apparatus comprising a processor adapted to perform the first aspect of the present invention. Attached Figure Description

[0016] To better understand the present invention, embodiments thereof will now be described by way of example only with reference to the accompanying drawings, wherein:

[0017] Figure 1 This is a schematic diagram of a wireless telecommunications network according to the first embodiment of the present invention;

[0018] Figure 2This is a flowchart of the first embodiment of the method of the present invention;

[0019] Figure 3 This is a flowchart of a second embodiment of the method of the present invention; and

[0020] Figure 4 This is a flowchart illustrating the first and second embodiments of the present invention. Detailed Implementation

[0021] Now refer to Figure 1 A first embodiment of a wireless telecommunications network 100 is described. In this first embodiment, the wireless telecommunications network 100 is a cellular telecommunications network having a first base station 110 located at a first cell site and a second base station 120 located at a second cell site. Both the first base station 110 and the second base station 120 are three-sector base stations (i.e., they each have a first sector, a second sector, and a third sector). Each sector of each base station has a different coverage area. Furthermore, each sector of each base station can communicate using a single carrier or multiple carriers.

[0022] Figure 1 A controller 150 is also shown. The controller 150 is configured to receive data from each base station; in this embodiment, the controller includes:

[0023] ● The average number of users connected to each carrier in each sector of the base station. Each value of the average number of users is associated with a specific point in time (i.e., it is timestamped) to indicate the average number of users connected to a specific carrier in a specific sector of a specific base station at that point in time. In this implementation, each data point is recorded and timestamped at hourly intervals to represent the average number of users connected to each carrier in each sector of the base station within that hour. Such a dataset of values ​​can cover a specific time period (e.g., a specific week), such that each value of the average number of users is for a specific point in time within that time period;

[0024] ● The density of cell sites where the base station is located is based on the proximity of the cell site to its four nearest neighbors (e.g., “dense city”, “city”, “suburb”, and “rural”).

[0025] ● Physical resource block (PRB) usage data includes a count of PRBs used by each carrier in each sector of the base station within a time period. In this implementation, the PRB usage data is also recorded and timestamped at hourly intervals to represent the count of PRBs used by each carrier in each sector of the base station within that hour. Such a dataset of values ​​can cover a specific time period (e.g., one year), which may differ from the time period covered by the dataset of average user numbers.

[0026] ● The height of each antenna in each sector of the base station;

[0027] ● The set of carriers transmitted in each sector of the base station; and

[0028] ●The manufacturer of each antenna in each sector of the base station.

[0029] The controller 150 includes a communication interface for receiving the data, a memory for storing the data, and a processor for implementing embodiments of the method of the present invention to determine the configuration of each sector of each base station in network 100 based on the data. The data may be stored in the memory along with a timestamp indicating when the base station recorded the data. The data may also be processed by the processor to determine new metrics. Reference will now be made to... Figure 2 A first implementation of the method is described.

[0030] In the first step (S101) of this first embodiment, the controller 150 obtains data indicating the average number of users connected to each carrier of each sector of each base station at multiple time points (e.g., every hour) within a specific time period (e.g., a day).

[0031] In step S103, controller 150 processes the data to determine the average number of users connected to each sector of each base station at each time point within the time period. This is achieved by retrieving the value of the average number of connected users per carrier of a specific sector at a specific time point (e.g., the value of the average number of users connected to each carrier of a specific sector of a specific base station for 0100 on the first day of the week), and summing these values ​​to determine the average number of users connected to that sector of the base station at that time point. This is repeated at each time point where a value of the average number of users connected to that sector of the base station exists. The processed data can then be represented as a sector-specific vector v, where each element of the vector is the average number of users connected to that sector at a specific time point, and the vector covers each time point between the start and end of the time period. This vector can be described as a time series of the average number of users in a specific sector of a specific base station within a specific time period. This process is repeated for all sectors of all base stations. Therefore, after this step, controller 150 obtains a time series of the average number of users in each sector of each base station within the time period.

[0032] In step S105, controller 150 processes each time series vector to determine a similarity value with each other time series vector. This is achieved using dynamic time warping techniques to identify similar time series pairs that occur simultaneously or are temporally offset (in other words, any similar patterns in two time series do not necessarily occur simultaneously to identify the two time series as similar). More specifically, a first time series vector V1 with n time points and a second time series vector V2 with m time points are processed to compute a similarity matrix S. The similarity matrix is ​​a two-dimensional matrix of size n x m. Each entry S[i, j] of the similarity matrix is ​​computed by measuring the distance between the i-th point V1 in the first time series (where i is a set from 1 to n) and the j-th point V2 in the second time series (where j is a set from 1 to m), and determining the value of S[i, j] as the sum of this distance and the minimum of S[i-1, j], S[i, j-1], and S[i-1, j-1], such that each entry monotonically increases from the previous entries of the similarity matrix. The similarity value between two vectors is determined as the value of S[n, m]. The following pseudocode includes more details about the dynamic time warping method used in this first embodiment:

[0033] DTW(v1, v2){

[0034] / / Where, vector v1 = (a1, ..., a2) n v2 = (b1, ..., b) m () are time series with n and m time points respectively.

[0035] Suppose that a two-dimensional data matrix S is a storage of similarity metrics, such that S[0, ..., n, 0, ..., m] and i, j are circular indices, and the cost is an integer.

[0036]

[0037]

[0038] Note that this specific time warping method requires that both time series have the same initial time point (i.e., a1 and b1 occur simultaneously) and the same final time point (i.e., a...). n and b m (occurring simultaneously).

[0039] Following step S105, controller 150 has calculated the similarity value between each time series vector and each other time series vector. In step S107, controller 150 performs a clustering process to identify multiple time series clusters, where each time series vector cluster comprises one or more time series vectors. In this embodiment, the clustering process is based on Ward's method, such that if the similarity values ​​of the time series vectors are sufficiently close (i.e., relative to a threshold), the time series vector is clustered with one or more other time series vectors. The threshold can be changed to vary the number of clusters generated by the clustering process. In one implementation, an operator can supervise the clustering process and set the threshold such that the resulting clusters represent a specific use case. In another implementation, the threshold can be calculated to identify a specific number of clusters.

[0040] In this implementation, the clustering process is implemented using the `scipy.cluster.hierarchy.ward` function of the Python programming language, as described in detail at https: / / docs.scipy.org / doc / scipy / reference / generated / scipy.cluster.hierarchy.ward.html. Each cluster is identified by a cluster identifier.

[0041] Following step S107, controller 150 has identified clusters for each time-series vector (associated with a specific sector of a base station). In step S109, controller 150 estimates the service growth rate for each sector of each base station in the network based on the identified clusters. In this embodiment, the sector growth rate is based on a function using the following inputs (collected in step S101):

[0042] ● Cluster identifier for the sector;

[0043] ● The maximum PRB usage of the sector in the previous calendar year (which is the maximum sum of PRB usage for each carrier in that sector);

[0044] ● The average PRB usage of the sector on January 1st of the current calendar year (which is the average of the sum of the PRB usage of each carrier in the sector on January 1st);

[0045] ● The density of cell sites where the base station of this sector is located (e.g., "dense city", "city", "suburb" and "rural");

[0046] ● Height of the sector antenna;

[0047] ● The set of carriers transmitted by this sector; and

[0048] ● Manufacturer of sector antennas.

[0049] The function is developed using a supervised machine learning model, such as a decision tree, based on a labeled training dataset (a mapping between these inputs to a specific sector and the known growth rate of that sector). The trained decision tree can also be applied to another test dataset and checked to determine if the function is outputting the accurate growth rate for each sector in the test dataset. When applied to the test dataset, the decision tree can be retrained one or more times after checking the output growth rate. The trained and tested function can also be applied as a final test to a validation dataset to ensure that the function performs as expected. Then, in step S109, the function can be used to estimate the service growth rate for each sector of each base station in the network based on the identified clusters.

[0050] In step S111, controller 150 prompts network reconfiguration based on the estimated traffic growth rate of each sector of each base station in the network. This reconfiguration can be an upgrade of the capacity of each sector, enabling each sector to handle the anticipated additional traffic volume before the next upgrade (the additional traffic volume is the growth rate multiplied by the time until the next upgrade). The reconfiguration can also be timed so that the upgrade is performed before traffic increases beyond the current capacity of the sector.

[0051] The above-described implementation provides the benefit of calculating growth forecasts based on behavioral trends. That is, by identifying which sectors of a base station are members of a specific cluster, growth forecasts can be based on the business growth trends of that cluster. For example, if a base station's sectors are clustered with other sectors that primarily serve office worker customers, growth forecasts can be based on the expected business growth of such customers. Therefore, behavioral changes experienced by some sectors in a cluster can alter the growth rate forecasts for other sectors in the cluster, even if those other sectors have not yet experienced those behavioral trends. Thus, these forecasts based on cluster identifiers are more accurate and responsive than existing methods that apply a general growth rate to each sector.

[0052] Another benefit of upgrading a sector based on its cluster identifier is that the upgrade can be applied to customers associated with that cluster. For example, if a sector clusters with other sectors that primarily serve commuters, it can be upgraded to increase its reach (reducing the likelihood of users served by that sector being switched) by using lower-frequency transceivers. Conversely, if a sector clusters with other sectors that primarily serve fixed customers (e.g., suburban or corporate park customers), it can be upgraded to increase its throughput by using higher-frequency transceivers.

[0053] Furthermore, this clustering is performed based on the similarity between time series vectors that do not necessarily occur simultaneously to identify two time series as similar. This allows time series with similar trends that are offset over time to be identified as similar, thereby increasing the likelihood that those time series vectors are associated with appropriate clusters.

[0054] In the above implementation, a similarity value is calculated between each time series vector and every other time series vector. However, as the number of time series vectors increases, the computational resource requirements may exceed the computational resources of the controller 150. This is addressed in the second implementation below (in... Figure 3 (As shown in the image)

[0055] In the first step (S201) of this second embodiment, the controller 150 obtains data indicating the average number of users connected to each carrier in each sector of each base station at multiple time points (e.g., per hour) within a specific time period (e.g., one day). In step S203, the controller 150 processes the data to obtain a set of time-series vectors, each time-series vector representing the average number of users in a specific sector of a specific base station within that time period.

[0056] In step S204, controller 150 identifies a subset of time-series vectors from the set of time-series vectors. The subset of time-series vectors is selected such that the count of the time-series vectors in the subset can be processed in step S205 within a time frame less than a threshold (set by the operator based on their computing resources). In step S205, controller 150 processes each time-series vector in the subset to determine a similarity value with each other time-series vector in the subset. Step S205 of this second embodiment uses the same dynamic time warping technique detailed in step S105 of the first embodiment described above. Once these similarity values ​​have been calculated, controller 150 performs a clustering process (in step S207) to identify multiple time-series clusters, where each time-series vector cluster comprises one or more time-series vectors from the subset of time-series vectors. Step S207 of this second embodiment also uses Ward's method to perform the clustering, as discussed in step S107 of the first embodiment described above. After step S207, controller 150 has identified clusters of time-series vectors in the subset of time-series vectors (associated with a specific sector of the base station). Each cluster is associated with a cluster identifier.

[0057] In step S209, controller 150 creates a common time series vector for each of the multiple time series clusters. The common time series vector of a time series cluster is based on all time series vector members of that time series cluster. Each element of the common time series vector is the average of the corresponding elements of each time series vector, where each element of the common time series vector and its corresponding element in each time series vector within the time series cluster are related to the same time point. Therefore, the common time series vector of a time series cluster is the average of the time series vectors within that time series cluster.

[0058] In step S211, controller 150 processes each time series vector in the time series vector set (i.e., all time series vectors calculated in step S203) to determine a similarity value with the ordinary time series vectors of each time series cluster. This is also performed using the same dynamic time warping technique detailed in step S105 of the first embodiment described above. Controller 150 associates each time series vector in the time series vector set with a cluster identifier based on these similarity values. In this embodiment, the association causes each time series vector in the time series vector set to be associated with the cluster identifier of the time series cluster with which it has the largest similarity value.

[0059] In step S213, controller 150 estimates the traffic growth for each sector of each base station in the network based on the identified clusters. This is performed using the same technique detailed in step S109 above. In step S215, controller 150 initiates a reconfiguration in the network based on the estimated traffic growth.

[0060] In the above implementation, the sector's cluster identifier is used in conjunction with PRB usage data to determine the appropriate upgrade for that sector (based on its estimated growth) and when that upgrade should be performed. Using performance metrics with the cluster identifier to determine appropriate network reconfiguration in machine learning algorithms is unnecessary. If a performance metric is used, it indicates PRB usage, which is also unnecessary (and could alternatively or additionally represent data throughput or the number of connected users). Furthermore, the cluster identifier can be used for other network configurations. For example, if a sector is a member of a cluster characterized by very low utilization during a specific time period, the network operator can apply reconfiguration (such as switching to power-saving mode) or physical operations (such as upgrading the sector) to that sector during that time period. In another example, if a sector is a member of a cluster characterized by commuting traffic, the network operator can apply reconfiguration to change handover parameters, resulting in lower user handover frequencies.

[0061] Furthermore, in the above embodiments, similarity calculations are performed on the time-series vector of the average number of connected users. Those skilled in the art will understand that the average number of connected users is merely an example metric, and other metrics can be used (as long as they are associated with points in time, allowing the creation of time-series metrics), such as data throughput or physical resource blocks. Therefore, the present invention can be derived from... Figure 4 The flowchart shows that the flowchart includes the following steps: obtaining time series performance data of each of the first plurality of access points (step S301); using dynamic time warping technology to determine the similarity value between the time series performance data of each of the first plurality of access points and the time series performance data of each other of the first plurality of access points (step S303); based on the similarity value, identifying the first plurality of clusters of the first plurality of access points, wherein the first access point is a member of the first cluster of the first plurality of clusters (step S305); identifying the reconfiguration of the first access point based on the association between the first access point and the first cluster (step S307); and prompting the identified reconfiguration of the first access point (step S309).

[0062] In a further enhancement of the first and second embodiments described above, multiple measures can be used in similarity calculations using multivariate dynamic time warping techniques. This can be achieved by: 1) obtaining the time series for each measure and performing dynamic time warping (i.e., independent multivariate dynamic time warping) between each pair of time series for each measure, and summing the similarity values; or 2) obtaining time series representing multiple measures and performing dynamic time warping between each pair of multivariate time series (where points are aligned in a multidimensional matrix). Once the similarity values ​​between each pair of sectors with multivariate time series have been calculated, the process can continue to cluster the sectors and determine the network configuration of the sectors based on the clusters, as described above in the first and second embodiments. Principal component analysis (PCA) can be applied to multivariate time series to reduce the number of input variables while retaining a large amount of information. The output of PCA can be used as input in dynamic time warping techniques.

[0063] In the above embodiments, each sector of the base station is considered a separate access point, and a growth forecast is predicted for that access point. However, those skilled in the art will recognize that growth forecasts can be considered for a single base station. Furthermore, those skilled in the art will understand that the present invention is not limited to cellular telecommunications networks. That is, it can be applied to any wireless telecommunications network with multiple access points.

[0064] Those skilled in the art will understand that any combination of features is possible within the scope of the claimed invention.

Claims

1. A computer-implemented method of configuring a first access point in a wireless telecommunication network, the wireless telecommunication network comprising a first plurality of access points and a second plurality of access points, the first access point being one of the first plurality of access points in the wireless telecommunication network, the second plurality of access points being a subset of the first plurality of access points, the method comprising the steps of: obtaining time series performance data for each access point in the second plurality of access points; determining, using a dynamic time warping technique, a similarity value between the time series performance data for each access point in the second plurality of access points and the time series performance data for each other access point in the second plurality of access points; based on the similarity values, identifying a first plurality of clusters of access points in the second plurality of access points: generating time series performance data for each cluster in the first plurality of clusters; determining a similarity value between the time series performance data for each access point in the first plurality of access points and the time series performance data generated for each cluster in the first plurality of clusters; based on the time series performance data for each access point in the first plurality of access points having a determined maximum similarity value with the time series performance data generated for a cluster in the first plurality of clusters, respectively associating each access point in the first plurality of access points with the cluster in the first plurality of clusters, wherein the first access point is associated with a first cluster in the first plurality of clusters based on the time series performance data for the first access point having a maximum similarity value with the time series performance data generated for the first cluster in the first plurality of clusters; based on the association of the first access point with the first cluster in the first plurality of clusters, identifying a configuration of the first access point associated with characteristics of the first cluster; and causing the first access point to implement the identified configuration.

2. The computer-implemented method of claim 1, wherein, The step of generating time series performance data for each cluster in the first plurality of clusters is based on an average of the time series performance data for each access point in the cluster.

3. The computer-implemented method of claim 1, wherein, The configuration is determined using a machine learning method that uses the association of the first access point with the first cluster in the first plurality of clusters as input, wherein the machine learning method is trained on a dataset that identifies historical performance data for each cluster in the first plurality of clusters.

4. The computer-implemented method of claim 3, wherein, The machine learning method further uses one or more of the following as input: a density of sites associated with the first access point, an average of performance data over a particular time period, a height of an antenna of the first access point, a set of carriers used by the first access point, and a vendor of the antenna of the first access point, wherein the dataset used to train the machine learning method further identifies corresponding historical metrics for each cluster in the first plurality of clusters.

5. The computer-implemented method of claim 1, wherein, The time series performance data comprises one or more of a measure of connected users and a measure of resource usage.

6. The computer-implemented method of claim 1, wherein, The time series performance data comprises a plurality of performance metrics and the dynamic time warping technique is a multivariate dynamic time warping technique.

7. The computer-implemented method of claim 1, wherein, The identified configuration is one or more of: a capacity of the first access point, a handover parameter of the first access point, and a power saving mode of the first access point.

8. The computer-implemented method of claim 1, wherein, The wireless telecommunications network is a cellular telecommunications network.

9. The computer-implemented method of claim 8, wherein, The first access point is a first sector of a first base station.

10. A computer readable carrier medium comprising a computer program comprising instructions which, when executed by a computer, cause the computer to carry out the computer-implemented method according to any one of claims 1 to 9.

11. A data processing apparatus comprising a processor adapted to carry out the computer-implemented method according to any one of claims 1 to 9.

Images