WO2024032872A1

WO2024032872A1 - Methods and nodes for predicting azimuth values of cells in communications networks

Info

Publication number: WO2024032872A1
Application number: PCT/EP2022/072279
Authority: WO
Inventors: Tahar ZANOUDA
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 2022-08-08
Filing date: 2022-08-08
Publication date: 2024-02-15

Abstract

A computer implemented method performed in a communications network comprises i) obtaining (202) first geo-located signal strength measurements from first devices connected to a first cell in the communications network; ii) obtaining (204) a first location, l, of a first antenna associated with the first cell; iii) providing (206) the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process, wherein the model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell; and iv) receiving (208) as output from the model, a predicted azimuth of the first cell.

Description

METHODS AND NODES FOR PREDICTING AZIMUTH VALUES OF CELLS IN COMMUNICATIONS NETWORKS

Technical Field

This disclosure relates to methods, nodes and systems in a communications network. More particularly but non-exclusively, the disclosure relates to predicting azimuth values of cells in communications networks.

Background

The increasing market for emerging connected services and the rise of mobile traffic has exerted pressure on mobile operators to monitor their networks and infrastructure continuously in order to address new users’ requirements. Mobile operators divide the network into coverage areas, often referred as sectors, to analyze and optimize the deployment setup. The geographic positioning of sectors has an impact on network performance optimization processes.

A communications network is comprised of a set of interconnected (hardware) nodes located on physical sites, each node having an antenna that provides coverage in a geographical area referred to as a cell. A sector typically constitutes the “area” covered by a cell. However, one sector can also be an overlap of two or more different cell coverage areas (each cell has a different frequency and cell range, but cells with different frequencies can be cover the same geographic area). Thus, a sector can refer to a cell (area) or the union of more than one cell area. A sector can be defined with respect to the constituent cell(s), directional antenna of the cell(s), and the associated geographical coverage area(s). A sector can thus be characterized by geographic position of the respective antenna (latitude, longitude), antenna azimuth and cell range. Traditionally, antenna elements are generally co-located in the same site, but have different characteristics such as cell range and frequency. Typically, 360-degree site coverage layout is divided horizontally into 60, 90, or 120-degree sectors. Other sector configurations can be introduced, but this is a common setup in industry.

Some operators store antenna parameters in a centralized database, the antenna operators often schedule field visits on a regular basis to monitor configuration correctness and to adjust antenna parameters. Such manual processes are time consuming and require skilled engineers. Moreover, antenna parameters are prone to human error and environment influence. Extreme weather conditions, network upgrade and maintenance visits can, unintentionally, change the antenna parameters. For example, an antenna can be impacted by heavy wind, or knocked during maintenance, causing a change in antenna azimuth. This can make it hard to maintain and track correct values. Summary

As noted above, accurate cell information (antenna location, azimuth, range etc.) is important for coverage planning purposes and network reliability.

Existing solutions to azimuth inference have various limitations:

- Manual data entry requires a field visit from an Engineer. It is an error prone and costly process. It is often hard to establish a process to monitor and validate antenna azimuth correctness.

Sector inference using on-site naming conventions rely on the assumption that cells are configured with hierarchical naming convention and this assumption is not always valid. Several mobile network providers use a hash names of cells or use a different naming scheme that is hard to reverse engineer. Without knowing antenna azimuth, it is hard to understand the characteristics of the area that is covered by cells (e.g., a cell directed toward a train station might have different traffic behaviour compared to the one directed toward a mall or a park). Some operators use a comprehensive naming convention (e.g. north, south, etc), but it is not enough to infer the sector accurately.

- Azimuth estimation methods that solely use handover relations and location of neighboring cells can have limitations as a handover relation doesn’t necessarily mean that two cells are spatially directed towards each other.

Azimuth estimation methods using network mobility statistics can allow cell azimuth estimation. However, the predictions are based on imprecise information. For example, cell handover relation doesn’t necessary provide information on cell direction. Moreover, the geolocation accuracy is low since user position is estimated when a handover is performed.

There are many efforts that use drive tests (whereby an Engineer drives around collecting information in order to directly measure cell azimuth) to collect granular data. However, the complexity of the setup (i.e., special equipment, staff, etc.), and constrained executed time span make the method inefficient and costly for large-scale processes.

- Sector inference methods that use crowdsourced data to calculate the shape of geo-crowdsourced data using a convex hull algorithm tend to produce irregular sector shapes due to data sparsity (e.g., due to lifetime of deployed cells).

It is an object of the methods and systems herein to address some of the aforementioned issues.

Thus, according to a first aspect herein, there is a computer implemented method performed in a communications network. The method comprises: i) obtaining first geo-located signal strength measurements from first devices connected to a first cell in the communications network. The method further comprises: ii) obtaining a first location, /, of a first antenna associated with the first cell, and iii) providing the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process. The model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell. The method further comprises: iv) receiving as output from the model, a predicted azimuth of the first cell.

According to a second aspect there is a computer implemented method. The method comprises: obtaining a training dataset comprising training examples, each training example comprising: a) example geo-located signal strength measurements of example devices connected to an example cell in a communications network, b) a corresponding example location of an antenna associated with the example cell and c) a ground truth azimuth of the example cell. The method then comprises training a model using a machine learning process to predict azimuth value of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell, using the training dataset.

According to a third aspect there is a node in a communications network. The node comprises: a memory comprising instruction data representing a set of instructions, and a processor configured to communicate with the memory and to execute the set of instructions. The set of instructions, when executed by the processor, cause the processor to: i) obtain first geo-located signal strength measurements for first devices connected to a first cell in the communications network, ii) obtain a first location, /, of a first antenna associated with the first cell, and iii) provide the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process. The model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell. The processor is further caused to: iv) receive as output from the model, a predicted azimuth of the first cell.

According to a fourth aspect there is a node in a communications network. The node comprises: a memory comprising instruction data representing a set of instructions (306), and a processor configured to communicate with the memory and to execute the set of instructions. The set of instructions, when executed by the processor, cause the processor to: obtain a training dataset comprising training examples, each training example comprising: a) example geo-located signal strength measurements of example devices connected to an example cell in a communications network, b) a corresponding example location of an antenna associated with the example cell and c) a ground truth azimuth of the example cell. The processor is further cause to train a model using a machine learning process to predict azimuth value of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell, using the training dataset. According to a fifth aspect there is a computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out a method according to the first or second aspect.

According to a sixth aspect there is a carrier containing a computer program according to the fifth aspect, wherein the carrier comprises one of an electronic signal, optical signal, radio signal or computer readable storage medium.

According to a seventh aspect, there is a computer program product comprising non transitory computer readable media having stored thereon a computer program according to the fifth aspect. There is thus provided methods and nodes for Cell Azimuth Estimation and Sector Shape Inference. From the predicted azimuth, sector geographic shape can be inferred allowing the geographic characteristics of the site and deployment to be understood. The methods and nodes herein make use of the rising ubiquity of passive network data, which provides new opportunities to monitor network usage at a detailed level of granularity. In light of the challenges presented in the Background Section above, the methods herein leverage anonymized crowdsourced geo-signal strength measurements (passive network data) to estimate antenna azimuth. From the azimuth, the sector geographical region can further be calculated. This allows verification of the azimuth and sector geographic information without the need for costly site visits. Furthermore, the azimuth values can be monitored over time, enabling more accurate signal coverage maps to be determined which can help ensure service levels remain high. Furthermore, using the methods herein, the azimuth and coverage area can be monitored over time, without the need for repeated visits. Such monitoring may be used, e.g. to trigger maintenance and/or configuration processes at an appropriate time. The processes described herein may be performed independently of the underlying hardware/software configuration (e.g. independent of network infrastructure vendor), which provides scalable and adaptable methods of azimuth monitoring that can be used in any network around the globe.

Brief Description of the Drawings

For a better understanding and to show more clearly how embodiments herein may be carried into effect, reference will now be made, by way of example only, to the accompanying drawings, in which:

Fig. 1a shows a node in a communications network according to some embodiments herein;

Fig. 1 b shows three example node configurations in a communications network according to some embodiments herein; Fig. 2 shows a method in a node in a communications network according to some embodiments herein;

Fig. 3 shows a node in a communications network according to some embodiments herein;

Fig. 4 shows a method in a node in a communications network according to some embodiments herein;

Fig. 5 illustrates a sector according to some embodiments herein;

Fig. 6 shows a method of training a model according to some embodiments herein;

Fig. 7 shows a system according to some embodiments herein; and

Fig. 8 shows another system according to some embodiments herein.

Detailed Description

The disclosure herein relates to a communications network (or telecommunications network). A communications network may comprise any one, or any combination of: a wired link (e.g. ASDL) or a wireless link such as Global System for Mobile Communications (GSM), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), New Radio (NR), WiFi, Bluetooth or future wireless technologies. The skilled person will appreciate that these are merely examples and that the communications network may comprise other types of links. A wireless network may be configured to operate according to specific standards or other types of predefined rules or procedures. Thus, particular embodiments of the wireless network may implement communication standards, such as Global System for Mobile Communications (GSM), Universal Mobile Telecommunications System (UMTS), Long Term Evolution (LTE), and/or other suitable 2G, 3G, 4G, or 5G standards; wireless local area network (WLAN) standards, such as the IEEE 802.11 standards; and/or any other appropriate wireless communication standard, such as the Worldwide Interoperability for Microwave Access (WiMax), Bluetooth, Z-Wave and/or ZigBee standards.

Fig. 1a illustrates a network node 100 in a communications network according to some embodiments herein. Generally, the node 100 may comprise any component or network function (e.g. any hardware or software module) in the communications network suitable for performing the functions described herein. For example, a node may comprise equipment capable, configured, arranged and/or operable to communicate directly or indirectly with a UE (such as a wireless device) and/or with other network nodes or equipment in the communications network to enable and/or provide wireless or wired access to the UE and/or to perform other functions (e.g., administration) in the communications network. Examples of nodes include, but are not limited to, access points (APs) (e.g., radio access points), base stations (BSs) (e.g., radio base stations, Node Bs, evolved Node Bs (eNBs) and NR NodeBs (gNBs)). Further examples of nodes include but are not limited to core network functions such as, for example, core network functions in a Fifth Generation Core network (5GC).

The node 100 is configured (e.g. adapted, operative, or programmed) to perform any of the embodiments of the method 200 as described below. It will be appreciated that the node 100 may comprise one or more virtual machines running different software and/or processes. The node 100 may therefore comprise one or more servers, switches and/or storage devices and/or may comprise cloud computing infrastructure or infrastructure configured to perform in a distributed manner, that runs the software and/or processes.

The node 100 may comprise a processor (e.g. processing circuitry or logic) 102. The processor 102 may control the operation of the node 100 in the manner described herein. The processor 102 can comprise one or more processors, processing units, multi-core processors or modules that are configured or programmed to control the node 100 in the manner described herein. In particular implementations, the processor 102 can comprise a plurality of software and/or hardware modules that are each configured to perform, or are for performing, individual or multiple steps of the functionality of the node 100 as described herein.

The node 100 may comprise a memory 104. In some embodiments, the memory 104 of the node 100 can be configured to store program code or instructions 106 that can be executed by the processor 102 of the node 100 to perform the functionality described herein. Alternatively or in addition, the memory 104 of the node 100, can be configured to store any requests, resources, information, data, signals, or similar that are described herein. The processor 102 of the node 100 may be configured to control the memory 104 of the node 100 to store any requests, resources, information, data, signals, or similar that are described herein.

It will be appreciated that the node 100 may comprise other components in addition or alternatively to those indicated in Fig. 1a. For example, in some embodiments, the node 100 may comprise a communications interface. The communications interface may be for use in communicating with other nodes in the communications network, (e.g. such as other physical or virtual nodes). For example, the communications interface may be configured to transmit to and/or receive from other nodes or network functions requests, resources, information, data, signals, or similar. The processor 102 of node 100 may be configured to control such a communications interface to transmit to and/or receive from other nodes or network functions requests, resources, information, data, signals, or similar.

Turning now to Fig. 1 b, which shows some examples of how node 100 may be implemented in certain embodiments of the described solution including: 1) a special-purpose network device XX502 that uses custom processing circuits such as application-specific integrated-circuits (ASICs) and a proprietary operating system (OS); and 2) a general purpose network device XX504 that uses common off-the-shelf (COTS) processors and a standard OS which has been configured to provide one or more of the features or functions disclosed herein.

Special-purpose network device XX502 includes hardware XX510 comprising processor(s) XX512, and interface XX516, as well as memory XX518 having stored therein software XX520. In one embodiment, the software XX520 implements modules to perform the method 200 described below. During operation, the software XX520 may be executed by the hardware XX510 to instantiate a set of one or more software instance(s) XX522. Each of the software instance(s) XX522, and that part of the hardware XX510 that executes that software instance (be it hardware dedicated to that software instance, hardware in which a portion of available physical resources (e.g., a processor core) is used, and/or time slices of hardware temporally shared by that software instance with others of the software instance(s) XX522), form a separate virtual network element XX530A-R. Thus, in the case where there are multiple virtual network elements XX530A-R, each operates as one of the network devices from the preceding figures.

Returning to Fig. 1 b, the example general purpose network device XX504 includes hardware XX540 comprising a set of one or more processor(s) XX542 (which are often COTS processors) and interface XX546 , as well as memory XX548 having stored therein software XX550. During operation, the processor(s) XX542 execute the software XX550 to instantiate one or more sets of one or more applications XX564A-R. While certain embodiments do not implement virtualization, alternative embodiments may use different forms of virtualization. For example, in certain alternative embodiments virtualization layer XX554 represents the kernel of an operating system (or a shim executing on a base operating system) that allows for the creation of multiple instances XX562A-R called software containers that may each be used to execute one (or more) of the sets of applications XX564A-R. In this embodiment, software containers XX562A-R (also called virtualization engines, virtual private servers, or jails) are user spaces (typically a virtual memory space) that may be separate from each other and separate from the kernel space in which the operating system is run. In certain embodiments, the set of applications running in a given user space, unless explicitly allowed, may be prevented from accessing the memory of the other processes. In other such alternative embodiments virtualization layer XX554 may represent a hypervisor (sometimes referred to as a virtual machine monitor (VMM)) or a hypervisor executing on top of a host operating system; and each of the sets of applications XX564A-R may run on top of a guest operating system within an instance XX562A-R called a virtual machine (which in some cases may be considered a tightly isolated form of software container that is run by the hypervisor). In certain embodiments, one, some or all of the applications are implemented as unikernel(s), which can be generated by compiling directly with an application only a limited set of libraries (e.g., from a library operating system (LibOS) including drivers/libraries of OS services) that provide the particular OS services needed by the application. As a unikernel can be implemented to run directly on hardware XX540, directly on a hypervisor (in which case the unikernel is sometimes described as running within a LibOS virtual machine), or in a software container, embodiments can be implemented fully with unikernels running directly on a hypervisor represented by virtualization layer XX554, unikernels running within software containers represented by instances XX562A-R, or as a combination of unikernels and the above-described techniques (e.g., unikernels and virtual machines both run directly on a hypervisor, unikernels and sets of applications that are run in different software containers).

The instantiation of the one or more sets of one or more applications XX564A-R, as well as virtualization if implemented are collectively referred to as software instance(s) XX552. Each set of applications XX564A-R, corresponding virtualization construct (e.g., instance XX562A-R) if implemented, and that part of the hardware XX540 that executes them (be it hardware dedicated to that execution and/or time slices of hardware temporally shared by software containers XX562A-R), forms a separate virtual network element(s) XX560A-R. [0026] The virtual network element(s) XX560A-R perform similar functionality to the virtual network element(s) XX530A-R. This virtualization of the hardware XX540 is sometimes referred to as network function virtualization (NFV)). Thus, NFV may be used to consolidate many network equipment types onto industry standard high volume server hardware, physical switches, and physical storage, which could be located in for example data centers and customer premise equipment (CPE). However, different embodiments of the invention may implement one or more of the software container(s) XX562A-R differently. While embodiments of the invention are illustrated with each instance XX562A-R corresponding to one VNE XX560A-R, alternative embodiments may implement this correspondence at a finer level granularity; it should be understood that the techniques described herein with reference to a correspondence of instances XX562A-R to VNEs also apply to embodiments where such a finer level of granularity and/or unikernels are used.

The third exemplary ND implementation in Fig. 1 b is a hybrid network device XX506, which includes both custom ASICs/proprietary OS and COTS processors/standard OS in a single node or a single card within a node. In certain embodiments of such a hybrid network device, a platform virtual machine (VM), such as a VM that that implements the functionality of the special-purpose network device XX502, could provide for para-virtualization to the hardware present in the hybrid network device XX506.

Turning back to Fig. 1a, briefly, in one embodiment, the node 100 is configured to: i) obtain first geo-located signal strength measurements from first devices connected to a first cell in the communications network; ii) obtain a first location, /, of a first antenna associated with the first cell; iii) provide the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process, wherein the model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell; and iv) receive as output from the model, a predicted azimuth of the first cell.

Coverage, or access to the communications network is provided by antennae, e.g. radio antennae. A cellular base-station (otherwise referred to herein as a base station) may refer to any equipment or hardware configured to, or capable of providing radio access to the communications network. Examples of base stations (BSs) include but are not limited to e.g., radio base stations, Node Bs, evolved Node Bs (eNBs) and NR NodeBs (gNBs).

Each base-station has one or more antennae. Each antenna (or antenna array) is associated with (e.g. serves) one or more cells. Each cell corresponds to an individual radio coverage area provided by an antenna (or antenna array) in a particular frequency band. In general, cells may be overlapping. For example, cells may be overlapped so as to form contiguous coverage.

The direction in which an antenna points is known as the cell azimuth. Cell/Antenna Azimuth, as used herein, refers to the clockwise antenna direction on the horizontal axis compared to north. The range of cell/antenna azimuth values is between 0-360. The cell azimuth bisects the cell (e.g. lies in a direction through the centre of the cell from the antenna).

Thus, the first cell corresponds to a first coverage area served by a first antenna. The first antenna may be comprised, for example in an outdoor sectorized base-station for a 4G or 5G network.

As used herein, a "sector" may be a base service area, for example, a sector may represent the smallest service area where users are served in a geographical region by a cell. A sector is generally a coverage area associated with one cell, or the coverage area representing the union of the coverage areas of two or more cells. A sector may be limited by geographical boundaries, licensed amount of frequency spectrum, radio propagation conditions, and can be associated with a geographic area characterized by a set of sociodemographic (population density, age groups etc.) and economic (level of demand for certain SLA, QoS etc.) characteristics that can evolve over time. A sector can be served by one cell, that can be provided by more than one Communications Service Provider (CSP). However, different cells with different frequencies can provide coverage to the same area.

For a cell in a sectorized base-station, a sector can be characterized by geographic position of the antenna (latitude, longitude), cell range, and centered on its antenna azimuth. Antenna downtilt can have an impact on sector shape calculations. However, cell range, in this context, refers to the outreach of signal range that is often affected by maximum cell range (depends on HW product and technology supported), and antenna downtilt. Traditionally, antenna elements are generally co-located in the same site, but have different characteristics such as cell range and frequency. Hence, a set of cells can be allocated in same area which lead to overlapped shapes of different sectors. In 4G and 5G mobile networks, the most common setup for outdoor networks is sectorized base-station with three sectors for the same frequency, where 360-degree site coverage layout is divided horizontally into 120-degree coverage areas for the same frequency.

Fig. 5 illustrates an example of a sector shape. Fig 5 shows a cellular basestation 502 at a location, I. The sector area associated with one cell can be calculated using antenna location (latitude, longitude), cell azimuth, and cell range.

The term device (otherwise referred to herein as a wireless device or User Equipment, (UE) is a device capable, configured, arranged and/or operable to communicate wirelessly with network nodes and/or other wireless devices. Examples of devices include, but are not limited to, a smart phone, a mobile phone, a cell phone, a voice over IP (VoIP) phone, a wireless local loop phone, a desktop computer, a personal digital assistant (PDA), a wireless cameras, a gaming console or device, a music storage device, a playback appliance, a wearable terminal device, a wireless endpoint, a mobile station, a tablet, a laptop, a laptop- embedded equipment (LEE), a laptop-mounted equipment (LME), a smart device, a wireless customer-premise equipment (CPE), a vehicle-mounted wireless terminal device, etc.. A device may support device-to-device (D2D) communication, for example by implementing a 3GPP standard for sidelink communication, vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), vehicle-to-everything (V2X) and may in this case be referred to as a D2D communication device. As yet another specific example, in an Internet of Things (loT) scenario, a device may represent a machine or other device that performs monitoring and/or measurements, and transmits the results of such monitoring and/or measurements to another UE and/or a network node. The device may in this case be a machine-to-machine (M2M) device, which may in a 3GPP context be referred to as an MTC device. As one particular example, the a device may implement the 3GPP narrow band internet of things (NB-loT) standard. Particular examples of such machines or devices are sensors, metering devices such as power meters, industrial machinery, or home or personal appliances (e.g. refrigerators, televisions, etc.) personal wearables (e.g., watches, fitness trackers, etc.). In other scenarios, a UE may represent a vehicle or other equipment that is capable of monitoring and/or reporting on its operational status or other functions associated with its operation. A device 700 as described above may represent the endpoint of a wireless connection, in which case the device may be referred to as a wireless terminal. Furthermore, a device as described above may be mobile, in which case it may also be referred to as a mobile device or a mobile terminal.

Turning now to Fig. 2 which shows a method 200 in a node in a communications network. The method 200 may be performed by a node such as the node 100 described above. Briefly, in a first step 202 the method 200 comprises i) obtaining first geo-located signal strength measurements from first devices connected to a first cell in the communications network. In a second step 204 the method comprises: ii) obtaining a first location, /, of a first antenna associated with the first cell. In a third step 206 the method comprises: iii) providing the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process, wherein the model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell. In a fourth step 208 the method comprises and iv) receive as output from the model, a predicted azimuth of the first cell.

In step 202 of the method 200, first geo-located signal strength measurements are obtained for (or from) first devices connected to a first cell in the communications network. In other words, a first plurality of geo-located signal strength measurements are obtained, as measured by a first plurality of devices connected to the first cell. The first geo-located signal strength measurements may comprise signal strength measurements and corresponding locations at which the signal strength was measured, in other words, [location, signal strength] tuples. In some embodiments, the first geo-located signal strength measurements comprise a list of [location, signal strength] tuples obtained for the first devices.

Geo-located signal strength measurements may be Reference Signal Received Power, RSRP, Reference Signal Received Quality, RSRQ, Signal to Interference & Noise Ratio, SINR, measurements and/or any other type of measurement that may be used to indicate signal strength of the cell.

In some embodiments, step 202 may comprise obtaining geo-located signal strength measurements in the format shown in Table 1.2 of Appendix I.

In some embodiments, the first geo-located signal strength measurements are obtained from the first devices as part of the normal signal strength reporting procedures. Such data may be considered passive network data. In other embodiments, dedicated measurements may be made, for example, step 202 of the method 200 may comprise initiating requests to the first devices connected to the first cell to make signal strength measurements and the resulting measurements may be reported back to the node 100.

In some embodiments, the locations of the devices (at which the signal strength measurements were made) are converted into unique geographic identifiers corresponding to a region on the Earth in which a respective device was located when a respective signal strength measurement was made. For example, the locations may be geo-hashed. Put another way, an Index may be obtained for each device. An example method is described in the paper by Sahr, K., White, D., & Kimerling, A. J. (2003) entitled: “Geodesic discrete global grid systems"’ Cartography and Geographic Information Science, 30(2), 121-134. In some embodiments, the H3 method from Sahr et al. (2003) is used to index (latitude, longitude) to spatial hexagon and generate the geo index (geo hash). The framework comprises a global grid system that is suitable for analyzing large spatial data sets, by partitioning areas of the Earth into identifiable grid tiles. The resolution/hezagon size reflects the size of homogenous hexagons used to divide the earth. The choice of hexagon size can be fine-tuned during the training process. In the experiments described below with respect to Appendix III, resolution 8 was used, however, it will be appreciated that different resolutions may be more appropriate, depending on the particular conditions.

In step 204, a first location, /, of a first antenna associated with the first cell is obtained. The first antenna may be part of a first cellular base station. In other words, the first antenna may be a first cellular base-station antenna associated with the cell. The first antenna is associated with the first cell. In other words, the first antenna generates or provides the first cell.

The first location of the first antenna may be given as a pair of co-ordinates, e.g. as a latitude and longitude. This may be obtained, for example from site configuration data (e.g. network site inventory data) for the antenna associated with the first cell. Example site configuration data is given in Table 1.1 of Appendix I.

In some embodiments, step 204 may comprise converting the first location (e.g expressed as co-ordinates) into a unique geographic identifier. In other words, the latitude and longitude of the antenna maybe geo-hashed. The first location may be geo-hashed in the same manner (e.g using the same process) as was used to geo-hash the locations of the first devices in step 202.

In step 206 the method comprises providing the first geo-located signal strength measurements and I (the first location of the first antenna associated with the first cell) as input to a model trained using a machine learning process. The model has been trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell.

In other words, the model has been trained to take as input geo-located signal strength measurements of devices connected to a respective cell and a location of an antenna associated with the respective cell. Based on the inputs, the model is trained to output azimuth of the respective cell.

In step 208, the method 200 comprises receiving as output from the model, a predicted azimuth of the first cell. The predicted azimuth of the first cell is predicted based on the input data (e.g. the first geo-located signal strength measurements and I (the first location of the first antenna associated with the first cell). In other words, the model processes the input data to predict the azimuth of the first cell and provides the predicted azimuth of the first cell as output. As used herein, a model trained using a machine learning process may alternatively be referred to as a machine learning model. The skilled person will be familiar with machine learning (ML) and machine learning processes for use in training machine learning models. ML is an approach that allows a programmer to implement a program by finding patterns in data samples. A program or model that is obtained through Machine Learning is called a Machine Learning model. ML models can be trained to perform tasks such as classification (e.g. label prediction) or regression (e.g. prediction of a value) tasks. A dataset of samples used to train the model is also known as a training set. Training data comprises training examples (each training example comprising an example input and a corresponding “correct” ground truth output). The model is trained on the training data, using the machine learning process.

A machine learning process comprises a procedure that is run on the training data to create the machine learning model. The machine learning process comprises procedures and/or instructions through which training data, may be processed or used in a training process to generate the machine learning model. The machine learning process learns from the training data. For example, the process may be used to determine how one set of parameters in the training data (input parameters of the model) are correlated with another set of parameters in the training data (output parameters of the model). The machine learning process may be used to fit the model to the training data.

Examples of machine learning processes include but are not limited to, e.g. algorithms for classification, such as k-nearest neighbors, algorithms for regression, such as linear regression or logistic regression, and algorithms for clustering, such as k-means.

The model, or machine learning model, may comprise both data and procedures for how to use the data to e.g. make the predictions described herein. The model is what is output from the machine learning (e.g. training) process, e.g. a collection of rules or data processing steps that can be performed on the input data in order to produce the output. As such, the model may comprise e.g. rules, numbers, and any other algorithm-specific data structures or architecture required to e.g. make predictions.

Different types of models take different forms. Some examples of machine learning processes and models that may be used herein include, but are not limited to: linear regression processes that produce models comprising a vector of coefficients (data) the values of which are learnt through training; decision tree processes that produce models comprising trees of if/then statements (e.g. rules) comprising learnt values; and neural network models comprising a graph structure with vectors or matrices of weights and biases with specific values, the values of which are learnt using machine learning processes such as backpropagation and gradient descent. The skilled person will appreciate that many different model types may be trained to predict azimuth from geo-located signal strength measurements and cell antenna location. For example, the model may be a decision tree model, such as a gradient boosted decision tree. In some embodiments the model is a Random Forest model or an XGBoost model.

XGBoost is an implementation of a gradient boosted decision tree, and is described in the paper by T. Chen and C. Guestrin, entitled: ‘‘XGBoost: A scalable tree boosting system," in Proc. 22^nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, 2016, pp. 785-794. XGBoost is optimized to make predictions from structured or tabular data and as such is well suited to predicting azimuth from a list of geo-located signal strength measurements.

Although particular model types have been indicated herein, it will be appreciated that other types of model may equally be used, for example, the model may be a neural network model or any other model that can be trained to predict cell azimuth in the manner described herein.

The inputs to the model are geo-located signal strength measurements of devices connected to the respective cell and the location of an antenna associated with (e.g. serving) the respective cell. The geo-located signal strength measurements may comprise: at least one of: Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), and/or Signal to Interference & Noise Ratio (SINR) measurements. The locations of the devices may be expressed in terms of a unique geographic identifier corresponding to a region on the Earth in which a respective device was located when a respective signal strength measurement was made. In other words, the cell/antenna location and/or the locations of the devices may be geo-hashed.

The model may be trained to take further inputs, such as, for example, distances between the devices and the location of the first antenna; and/or bearings of the devices with respect to the first location of the first antenna. A such, the method 200 may further comprise, determining first distances between the location I and the first devices and/or determining bearings of the first devices with respect to the location of the first cell and providing these as input to the model.

In some embodiments, the model takes as input: Location of first cell, and a list of [Location, RSRQ, RSRP, SINR/SNR, Cell- measurement Bearing, Cell- measurement Distance] for each device of the plurality of devices.

A full description of the manner in which it is trained is given below with respect to Figs 3 and 4 and the detail therein will be understood to apply equally to trained model in the method 200.

Turning back to the method 200, the predicted azimuth of the first cell obtained in step 208 can be used in a wide range of tasks. As an example, the predicted azimuth may be used to verify site inventory data for the first cell. For example, the method 200 may further comprise comparing the predicted azimuth of the first cell to a reported azimuth for the first cell obtained from network site inventory data to determine whether the network site inventory data is correct.

In some embodiments, steps i), ii), iii) and iv) may be repeated (e.g. at regular time intervals or following significant events, such as maintenance or weather incidents) in order to monitor the predicted azimuth of the first cell over time. As such, the predictions, or drift in the predictions may be used in maintenance planning and/or to initiate Engineering works.

In other embodiments, the predicted azimuth of the first cell may be used to infer other shape features of the cell or a sector associated with the cell.

For example, cell range may be inferred from the maximum distance at which a device is connected to the cell. As such, in some embodiments, the method 200 may further comprise determining distances between positions of the first devices and the first location, I, and estimating a range, r, of the first cell from a maximum of the determined distances.

The angular extent of a cell (e.g. the angle covered or spanned by the cell) depends on the site coverage layout, which is typically divided horizontally into 60, 90, or 120- degree sectors. Hence, the number of sectors in the site can be inferred from the number of cells with the same frequency.

Thus, the method 200 may further comprise obtaining a number of cells, n, being served by the first antenna with the same frequency as the first cell. This may be obtained, for example, from site inventory data. An angular extent of the first cell may then be determined by dividing 360 degrees by n.

The outer edges of a geographic coverage area of the first cell can then be determined from the predicted azimuth of the first cell and the angular extent. This can be performed, for example, by determining a first point, xi (as illustrated in Fig. 5), wherein the first point is determined at a distance r from the first location of the first antenna, and at an angle of +360 degrees/ 2n from the azimuth. A second point, x₂ (as illustrated in Fig. 5), can be determined wherein the second point is determined at a distance /-from the first location of the first antenna, and at an angle of -360 degrees/ 2n from the azimuth. The first point and the second point represent a first outer extent of the first cell and a second outer extent of the first cell respectively.

The geographic coverage area of the first cell can then be determined from the points, I, xi, and x₂. For example, the geographic coverage area may be determined as an area bounded by: a first straight line between xi and /; a second straight line between x₂ and /; and a third line between x1 and x2. As noted above, typically, for outdoor base-stations, a 360-degree site coverage layout is commonly divided horizontally into 120-degree sectors with the same frequency. However, the boundaries of adjacent sectors need to have a slight overlap to enable a smooth handover for floating UEs. In addition, the inconsistent nature of the radio propagation and the attenuation of electromagnetic waves with respect to distance, diverse geographic distribution and weather conditions. In order to analyze and optimize coverage with respect to network configuration, sectors are often represented in different shapes. In some embodiments herein, the site coverage area is represented as a circle with the base-station at the centre, and each sector coverage area as pie-shape area. However, coverage regions can also be represented using other shapes such squares, hexagons, rectangles, or irregular shapes. Using the predicted azimuth for the first cell and the estimate range, the sector coverage shape can be represented using e.g. triangles or pie-shape sectors. The number of sectors (i.e., pie slices in coverage layout) depends on the number of cells with the same frequency.

In order to represent coverage area as pie-shape areas, the following steps may be performed to make the third line a curved line between x1 and x2:

- Set number of points (n_points) that we need to draw in pie-shaped boundaries. E.g. 100

- Generate points (n_points) between x1 and x2 using the following formula. For each point degree pd. between x1 & x2,

Draw pie-shaped area connecting base-station location, x1 , and boundary points (boundary_pointi for / between 0 and n_points ).

In order to represent coverage area as a rectangular areas (e.g. where the third line is a straight line), the following step may be performed:

Draw a rectangular shape between base-station location, x1 , x2.

Turning now to an example, in one embodiment, azimuth of the first cell is estimated using antenna physical information and geo-located signal measurements. The maximum possible distance between cell/antenna location and the location of the signal strength measurements is calculated. Signal strength measurement distance refers to the geographical distance between signal measurement’s locations and cell/antenna location.

In order to calculate sector shape region (as illustrated in Fig. 5), the following values are estimated or calculated:

Estimate cell/antenna azimuth based on the method 200 described above.

Estimate cell range based on the maximum possible distance.

- Calculate number of sectors. A sector typically constitutes the “area” covered by many cells with different frequency bands. Typically, 360-degree site coverage layout is divided horizontally into 60, 90, or 120-degree sectors with the same frequency. Hence, number of sectors in the site can be inferred from the number of cells with the same frequency.

- Antenna physical information (latitude, longitude).

Based on this information, a geographic shape for each sector can be generated, where coverage area span between three points

Retrieve cell location from the associated with antenna location (Cell latitude, Cell longitude)

- second point coordinates, which can be calculated using (Cell latitude, Cell longitude) given a bearing = (estimated azimuth - 360/ 2*number of sectors) and a distance = cell range.

- third point coordinates, which can be calculated using (Cell latitude, Cell longitude) given a bearing = (estimated azimuth + 360/ 2*number of sectors) and a distance = cell range.

In order to represent coverage area as pie-shape or rectangular areas, the steps outlined above may be performed to make the third line a curved line between x1 and x2.

In this way, a sector shape can be inferred, which can be used in various tasks such as telecom network maintenance, site planning, base-station configuration optimization, network coverage analysis and network coverage planning.

Turning now to other embodiments, the following embodiments relate to example methods of training the model used in the method 200 described above. Fig. 3, in some embodiments, a node 300 may be used to train the model. The node 300 may be configured to perform the method 400 described below. The node 300 may comprise a processor 302, and a memory 304. The memory 304 may comprise a set of instructions 306 that when executed by the processor 302 cause the processor to perform the functionality described herein. Processors, memory and instructions were all described above with respect to the node 100 and the detail therein will be understood to apply equally to the node 300.

Examples of nodes were given above with respect to the node 100 and the examples therein will be appreciated to apply equally to the node 300. For example, the node 300 may be an access point (APs) (e.g., radio access point), base-station (BSs) (e.g., radio base station, Node B, evolved Node B (eNB) and NR NodeB (gNB)). Further examples of nodes include but are not limited to core network functions such as, for example, core network functions in a Fifth Generation Core network (5GC). In some embodiments, the node 300 may be a node in the cloud.

The node 300 is configured to obtain a training dataset comprising training examples, each training example comprising: a) example geo-located signal strength measurements of example devices connected to an example cell in a communications network, b) a corresponding example location of an antenna associated with the example cell and c) a ground truth azimuth of the example cell; and train a model using a machine learning process to predict azimuth value of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell, using the training dataset.

Fig. 4 shows a method 400 in a node in a communications network according to some embodiments, herein. The method 400 may be performed by the node 300 described above, or the node 100 described above. In other words, the training of the model (using the method 400) and the inference of the model (using the method 200) may be performed by the same node, or by different nodes in the communications network.

Briefly, the method 400 comprises, in a first step 402, obtaining a training dataset comprising training examples, each training example comprising: a) example geo-located signal strength measurements of example devices connected to an example cell in a communications network, b) a corresponding example location of an antenna associated with the example cell and c) a ground truth azimuth of the example cell. In a step 404, the method comprises training a model using a machine learning process to predict azimuth value of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell, using the training dataset.

Considering a cell as a “pie shape” or wedge, as noted above, the azimuth is on the axis of symmetry of sector area shape. The propagation of electromagnetic waves attenuates with respect to distance, so signal strength is correlated with both distance and azimuth line. The signal strength tails off either side of the azimuth line. Thus, signal strength is correlated to cell azimuth and as such, a plurality of signal strength values, obtained at different locations in the cell can be used to predict the location of the cell azimuth.

The training dataset may be compiled of data relating to example cells with known cell direction e.g. known azimuth values. Cell range and/or cell shape may also be known for the example cells in the training dataset. A threshold density of training examples may be introduced, in order to eliminate cases where there are too few samples to accurately estimate cell azimuth. In such cases, the results can lead to biased results. The choice of a threshold can be fine-tuned. However, by considering a long historic interval, these cases can be avoided. It is noted that it can be more difficult to collect sufficient data for newly deployed sites (e.g. when mobile operator starts rolling out a new 5G site, in ML literature, this is often referred to as a “Cold Start problem” as it relates to an estimation a new event or phenomena).

The skilled person will appreciate that the accuracy of the model will be dependent on the range and variety of training data provided to the model in the training phase. Generally, cells from different regions may be used as training data in order to assess the scalability of the method. For the most complete training dataset, data from all cells for each mobile operator in the country may be accrued. However it will be appreciated that a model may be trained on small datasets. As described below with respect to Appendix III, a model may be trained with good accuracy on 6 weeks’ work of geo-located signal strength measurements from three sites.

The example geo-located signal strength measurements in the training data set can be acquired from Drive Tests (CTR) or can be acquired from third-parties that benchmark signal strength and network coverage globally. The raw data can be anonymized to preserve privacy. Each data point encompasses a set of attributes such as global cell identifier, signal strength measurement values, timestamp of observation, and the longitude and latitude information of samples. Table 1.2 of Appendix I shows an example of geo-located signal strength measurements.

The example locations of the example devices may be converted into a unique geographic identifier corresponding to a region on the Earth in which a respective example device was located when a respective signal strength measurement was made. E.g. the example device locations may be geo-hashed, as described above with respect to the method 200.

The antenna location and ground truth azimuth values for each example cell may be obtained from network site inventory data. Network site inventory data is a dataset comprising information about site physical location and (a selected set of) cell configuration attributes, e.g. obtained when the cell was configured. Example network site inventory data is given in Table 1.1 of Appendix I.

The location (latitude, longitude) of the example cells in the training data may be converted into a unique geographic identifier corresponding to a region on the Earth in which the example cell is located. In other words, the model may be trained to take a geohash of cell/ antenna location as input (e.g. rather than the location in coordinate pairs form). Geo-hashing was described above and the detail therein will be understood to apply equally to the method 400.

It is noted that the efficiency training of the model is improved when geo-hashed locations are used, compared to latitude and longitude value pairs, as each geo-hash coordinate directly corresponds to a single categorical geographical location, whereas latitude and longitude are co-dependent continuous variables, from which location can only be obtained in combination.

In some embodiments, the model is further trained to take bearing of each device with respect to the cell antenna and/or the distance between the cell antenna and each device as input. A such, the method 400 may further comprise, for each training example, determining distances between the example devices and the example location of an antenna associated with the respective example cell; and/or determining bearings of the example devices with respect to the location of the antenna associated with the respective example cell. The model may further take as input the distances and/or bearings and may take the distances and bearings into account when predicting cell azimuth.

The experiments described below with respect to Appendix III showed that using the distances and/or bearings helps enhance the azimuth estimation. This may be because points can be clustered or grouped in a place with respect to the location of base-station. The group of measurements share approximately similar bearing. At the other hand, the propagation of the signal and signal strength is correlated with distance and location with respect of azimuth line.

In one example embodiment, training proceeds according to Fig. 6. In this embodiment, an XGBoost model is used, however it will be appreciated that the steps in Fig. 6 could equally be applied to the training of other types of model.

In step 602: the method 600 comprises obtaining a plurality of example geo-located signal strength measurements for different example cells (according to step 402 of the method 400 described above).

In step 604: Live network inventory data is obtained to obtain the corresponding example locations of the antennae associated with each example cell and the ground truth ground truth azimuth of the example cell

In step 606: Each geo-located signal strength measurement is mapped (e.g. attributed to, or associated with) an example cell using e.g. the field “Global Cell ID”.

In step 608: cell/antenna location and antenna direction (e.g. cell azimuth) are extracted from network inventory data. The cell/antenna location is used to calculate distance between geolocated signal strength measurement and the cell, cell/antenna direction is used as ground truth (e.g. example correct output) and is only needed during training of the model and validation of the results. This is the feature that the model is trained to predict from the input parameters.

Once the measurements are grouped by each cell in the network (in step 606), the following data processing tasks are performed:

612: Index cell/antenna location: the H3 method (as described in the paper by Sahr & Kimerling (2003) cited above) is used to index cell/antenna location (latitude, longitude) to a spatial hexagon and generate geo index (geo hash). The hexagon resolution can be fine-tuned, depending on country size, and density of deployed base-stations. In this example, resolution 8 is used. It will be appreciated that other geo-hashing methods (with other resolutions) may equally be used. For example, the spatial resolution chosen may depend on the size of the cells under consideration.

614: Calculate geographic distance between sample location and cell/antenna location. This can be, for example, a Euclidean distance 616: Calculate “bearing azimuth” between measurement location and cell/antenna location for each example device. This is also referred to herein as determining bearings of the example devices with respect to the example cell/antenna location. Bearing azimuth, in this context, refers to the angular direction of signal measurement’s location corresponding to the cell/antenna location. The values can vary between 0 and 360.

Signal strength values such RSRQ (Reference Signal Received Quality), RSRP (Reference Signal Received Power), RSSI (Received Signal Strength Indicator), and SINR/SNR (Signal- to-noise ratio of the given signal) are also extracted and in step 618, the features above are stored for model training. Appendix II shows an example of raw data and features for model training.

In the training phase, the constructed dataset which includes model features and known (e.g. ground-truth) example cell/antenna azimuth can be divided into different folds (in step 624) where one fold is used as a test/validation dataset, and the remaining (k-1) folds are used to train the model to estimate cell/antenna azimuth (in steps 626 and 628).

The output of the model (e.g. the output data type) is an estimated azimuth value, that can vary between 0-360 degrees.

As an example, in one experiment, Deep Neural Network (DNN), XGBoost and Random Forest models were trained, using the process described in Fig. 6. The models were trained on the same training data set, comprising 6 weeks’ worth of geo-located signal measurements data taken from two different geographical regions. The inputs to the models and an example line of training data is shown in Appendix II. The results of the experiments showed that XGBoost outperforms other methods, with an average accuracy greater than 0.8 R². XGBoost outperforms DNN-based methods for tabular datasets. It was noticed that (as with most machine learning models) the accuracy improved with increasing number of geolocated signal measurements (e.g. more training data results in increased accuracy). This may have implications for newly deployed sites where less data is available, however for older sites that may be more prone to azimuth drift over time, good accuracy may be achieved. It is also noted that the training data is collected passively in the network, and thus the accuracy of the model can be improved over time, e.g. through the use of continuous training on new data as it is received. The results of the aforementioned experiments are shown in Appendix III.

Turning now to Fig. 7 which shows a system 700 according to some embodiments herein. In this embodiment, network data 702, (comprising cell site and antenna location) is collected from network configuration data. Geo-located signal strength measurements 704 are obtained for devices. The collected data is passive or crowd sourced data. A data Pre-processing and fusion module extracts Global ID from cell configuration data and the geo-located signal measurements (according to step 608 described above) and groups the geo-located signal measurements according to cell in step 608a. A feature engineering Block calculates the distance 614 between each device and the location of the cell associated with it. The bearings of the devices with respect to the associated cell/antenna location are determined 702. Any other statistical features for geolocated signal measurement are determined in 702. The data from steps 614, 616 and 702 can then be used as training data with which to train the model according to the method 400 described above.

Alternatively, the output from steps 608a (geo-located signal strength measurements for an individual cell and cell antenna location) can be fed into the trained model and used to estimate azimuth direction according to the method 200 described above. This, in this manner, the system 700 can be used both to train and use the model.

Fig. 8 puts the system 700 into the context of the wider communications network. In Fig. 8, live device data 802 is obtained from the lie customer network and fed into a site data collection module. The site data collection module ingests the data 810, decrypts it 812 and it is then sent through a data broker 814 and Parser 816 before being sent to a database 820.

External data comprising geo-located signal strength measurements 804 and geospatial datasets 806 are also collected, passed through a data ingestion module 818 and stored in database 820. The data in database 820 is used to infer sector shape according to the principles described herein. For example, it may be pre-processed 822 to obtain the features such as distances between a respective device and the antenna associated with the device that is serving the respective device, and bearings between the cell antenna and the respective devices. The processed data can be used as training data to train a model using the method 400 described above and/or the trained model can be used to estimate azimuth of a new cell using the method 200 described above. Cell range 822 can be estimated as described above, and the predicted azimuth and cell range can be used to calculate sector shape 825. The determined sector shapes can then be exported in 826 for use in e.g. planning and/or monitoring processes. Estimating cell range and sector shape was described above with respect to the method 200 and the detail therein will be understood to apply equally to the system shown in Fig. 8.

There is thus provided methods and nodes that can be used to predict azimuth of a cell, using geo-located signal measurements. These can be collected in a passive manner. As has been shown herein, the volume of passive data collected in this manner can be leveraged to train machine learning models to predict cell azimuth with high levels of accuracy.

Turning to other embodiments, there is also provided a computer program product comprising a computer readable medium, the computer readable medium having computer readable code embodied therein, the computer readable code being configured such that, on execution by a suitable computer or processor, the computer or processor is caused to perform the method or methods described herein. Thus, it will be appreciated that the disclosure also applies to computer programs, particularly computer programs on or in a carrier, adapted to put embodiments into practice. The program may be in the form of a source code, an object code, a code intermediate source and an object code such as in a partially compiled form, or in any other form suitable for use in the implementation of the method according to the embodiments described herein.

It will also be appreciated that such a program may have many different architectural designs. For example, a program code implementing the functionality of the method or system may be sub-divided into one or more sub-routines. Many different ways of distributing the functionality among these sub-routines will be apparent to the skilled person. The sub-routines may be stored together in one executable file to form a self-contained program. Such an executable file may comprise computer-executable instructions, for example, processor instructions and/or interpreter instructions (e.g. Java interpreter instructions). Alternatively, one or more or all of the sub-routines may be stored in at least one external library file and linked with a main program either statically or dynamically, e.g. at runtime. The main program contains at least one call to at least one of the sub-routines. The subroutines may also comprise function calls to each other.

The carrier of a computer program may be any entity or device capable of carrying the program. For example, the carrier may include a data storage, such as a ROM, for example, a CD ROM or a semiconductor ROM, or a magnetic recording medium, for example, a hard disk. Furthermore, the carrier may be a transmissible carrier such as an electric or optical signal, which may be conveyed via electric or optical cable or by radio or other means. When the program is embodied in such a signal, the carrier may be constituted by such a cable or other device or means. Alternatively, the carrier may be an integrated circuit in which the program is embedded, the integrated circuit being adapted to perform, or used in the performance of, the relevant method.

Variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. A computer program may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.

Table 1.1: Example of cell configuration attributes

Table 1.2: Example of geo-located signal strength measurements

Appendix III

Claims

1 . A computer implemented method performed in a communications network, the method comprising: i) obtaining (202) first geo-located signal strength measurements for first devices connected to a first cell in the communications network; ii) obtaining (204) a first location, /, of a first antenna associated with the first cell; iii) providing (206) the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process, wherein the model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell; and iv) receiving (208) as output from the model, a predicted azimuth of the first cell.

2. A method as in claim 1 further comprising: determining distances between positions of the first devices and the first location, I; and estimating a range, r, of the first cell from a maximum of the determined distances.

3. A method as in claim 2 further comprising: obtaining a number of cells, n, being served by the first antenna with the same frequency as the first cell; determining an angular extent of the first cell as 360 degrees divided by n; and determining outer edges of a geographic coverage area of the first cell from the predicted azimuth of the first cell and the angular extent.

4. A method as in claim 3 further comprising: determining a first point, xi, wherein the first point is determined at a distance r from the first location of the first antenna, and at an angle of:

+360 degrees/ 2n from the azimuth.

5. A method as in claim 4 further comprising: determining a second point, x₂, wherein the second point is determined at a distance r from the first location of the first antenna, and at an angle of:

-360 degrees/ 2n from the azimuth.

6. A method as in claim 5 wherein the first point and the second point represent a first outer extent of the first cell and a second outer extent of the first cell respectively.

7. A method as in claim 5 or 6 further comprising: determining the geographic coverage area of the first cell from the points, I, xi, and x₂.

8. A method as in claim 7 wherein the geographic coverage area is determined as an area bounded by: a first straight line between xi and /; a second straight line between x₂ and I; and a third line between xi and x₂.

9. A method as in any one of the preceding claims further comprising: comparing the predicted azimuth of the first cell to a reported azimuth for the first cell obtained from network site inventory data to determine whether the network site inventory data is correct.

10. A method as in any one of the preceding claims further comprising: repeating steps i), ii), iii) and iv) to monitor the predicted azimuth of the first cell over time.

11. A computer implemented method comprising: obtaining (402) a training dataset comprising training examples, each training example comprising: a) example geo-located signal strength measurements of example devices connected to an example cell in a communications network, b) a corresponding example location of an antenna associated with the example cell and c) a ground truth azimuth of the example cell; and training (404) a model using a machine learning process to predict azimuth value of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell, using the training dataset.

12. A method as in claim 11 further comprising: determining distances between the example devices and the location of the antenna associated with the respective example cell; and/or determining bearings of the devices with respect to the location of the antenna associated with the respective example cell; and wherein the model further takes as input the distances and/or bearings.

13. A method as in any one of the preceding claims wherein the model is a decision tree model.

14. A method as in claim 13 wherein the model is an XGBoost model.

15. A method as in any one of the preceding claims wherein the geo-located signal strength measurements comprise: at least one of: Reference Signal Received Power, RSRP, Reference Signal Received Quality, RSRQ, and/or Signal to Interference & Noise Ratio, SINR, measurements; and a unique geographic identifier corresponding to a region on the Earth in which a respective device was located when a respective signal strength measurement was made.

16. A node (100) in a communications network, the node comprising: a memory (104) comprising instruction data representing a set of instructions (106); and a processor (102) configured to communicate with the memory and to execute the set of instructions, wherein the set of instructions, when executed by the processor, cause the processor to: i) obtain first geo-located signal strength measurements for first devices connected to a first cell in the communications network; ii) obtain a first location, /, of a first antenna associated with the first cell; iii) provide the first geo-located signal strength measurements and the first location as input to a model trained using a machine learning process, wherein the model is trained to predict azimuth of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell; and iv) receive as output from the model, a predicted azimuth of the first cell.

17. A node as in claim 16 wherein the set of instructions, when executed by the processor, cause the processor to perform the method of any one of claims 2 to 10, 12 to 15.

18. A node (300) in a communications network, the node comprising: a memory (304) comprising instruction data representing a set of instructions (306); and a processor (302) configured to communicate with the memory and to execute the set of instructions, wherein the set of instructions, when executed by the processor, cause the processor to: obtain a training dataset comprising training examples, each training example comprising: a) example geo-located signal strength measurements of example devices connected to an example cell in a communications network, b) a corresponding example location of an antenna associated with the example cell and c) a ground truth azimuth of the example cell; and train a model using a machine learning process to predict azimuth value of a cell from: geo-located signal strength measurements of devices connected to the respective cell and a location of an antenna associated with the respective cell, using the training dataset.

19. A node as in claim 18, wherein the set of instructions, when executed by the processor, cause the processor to perform the method of any one of claims 12 to 15.

20. A node (100; 300) in a communications network, the node being configured to perform the method of any one of claims 1 to 15.

21 . A computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out a method according to any of claims 1 to 15.

22. A carrier containing a computer program according to claim 21 , wherein the carrier comprises one of an electronic signal, optical signal, radio signal or computer readable storage medium.

23. A computer program product comprising non transitory computer readable media having stored thereon a computer program according to claim 21 .