Adjacency conflict prediction

Abstract

The neighbor relationship conflict prediction is performed by operations including: receiving, from the terminal's serving MCG, a measurement report for the terminal including a plurality of signal measurements associated with an SCG represented by a PCI and an ARFCN; identifying an unlisted SCG among the plurality of signal measurements; identifying one or more neighboring MCGs within a threshold distance of the serving MCG; counting the number of SCGs in the NRT of each of the neighboring MCGs that have the PCI and ARFCN of the unlisted SCG; and applying a classification model to the counted number of SCGs and the MCG-PCI-ARFCN identifiers representing the serving MCG, PCI, and ARFCN to obtain a binary value indicating whether providing the serving MCG and the plurality of neighboring MCGs to the unlisted SCG will result in a PCI conflict.

Description

[Technical field]

[0001] Priority claims and cross references This application claims priority to Provisional Application No. 63 / 229,039, filed August 3, 2021, and Nonprovisional Application No. 17 / 542,242, filed December 3, 2021, both of which are incorporated by reference in their entireties herein. [Background technology]

[0002] In the early stages of introducing 5G technology into existing cellular networks, such as by using the E-UTRAN New Radio-Dual Connectivity (EN-DC) architecture, the network environment is expected to change frequently. The number of available cells is expected to increase due to the frequency characteristics and data offloading of 5G cells. These changes are expected to cause neighbor relation tables, which contain information to perform handoffs to each of some neighboring cells, to change more frequently than before the adoption of 5G technology. In addition, the coverage area of ​​5G cells is smaller than that of 4G cells, and the number of devices connected to the network is still steadily increasing, so the number of handovers is expected to increase exponentially.

[0003] In the early stages of 5G deployment, E-UTRAN will support Multiple Radio - Dual Connectivity (MR-DC) via EN-DC, where a terminal is connected to one 4G cell (eNB) acting as a master node and one 5G cell (en-gNB) acting as a secondary node. The eNB is connected to an Evolved Packet Core (EPC) via an S1 interface and to an en-gNB via an X2 interface. The en-gNB may also be connected to the EPC via an S1-U interface and to other en-gNBs via an X2-U interface.

[0004] Since the coverage of 5G cells is limited compared to 4G cells, more 5G cells than 4G cells are used to cover an area. When a terminal moves from the coverage area of ​​one cell to the coverage area of ​​another cell, it connects to the other cell and disconnects from the first cell. This procedure is called handover. For this handover to occur, neighbor relations should be provisioned in the respective neighbor relation tables of the eNB. Summary of the Invention

[0005] Aspects of the present disclosure are best understood from the following detailed description when read in conjunction with the accompanying drawings. It should be noted that, according to standard practice in the industry, various features have not been drawn to scale. In fact, dimensions of various features may be arbitrarily increased or decreased for clarity of discussion. [Brief description of the drawings]

[0006] [Figure 1] FIG. 2 is a schematic diagram of a data flow for adjacency conflict prediction in accordance with at least one embodiment of the present invention. [Diagram 2] 1 is an operational flow for neighbor relationship conflict prediction in accordance with at least one embodiment of the present invention. [Diagram 3]FIG. 2 is a schematic diagram of data flow for continuous classification model training in accordance with at least one embodiment of the present invention. [Figure 4] 1 is an operational flow for training a continuous classification model in accordance with at least one embodiment of the present invention. [Diagram 5] FIG. 2 is a schematic diagram of data flow for training an initial classification model in accordance with at least one embodiment of the present invention. [Figure 6] 1 is an operational flow for training an initial classification model in accordance with at least one embodiment of the present invention. [Figure 7] FIG. 2 is a schematic diagram of data flow for neighborhood table population in accordance with at least one embodiment of the invention. [Figure 8] 1 is an operational flow for neighborhood table population in accordance with at least one embodiment of the present invention. [Figure 9] FIG. 2 is a schematic diagram of data flow for training a neighborhood decision regression model in accordance with at least one embodiment of the present invention. [Figure 10] 1 is an operational flow for training a neighborhood decision regression model in accordance with at least one embodiment of the present invention. [Figure 11] FIG. 2 is a block diagram of an example hardware configuration for adjacency conflict prediction, in accordance with at least one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] The following disclosure provides many different embodiments or examples for implementing different features of the provided subject matter. To simplify the disclosure, certain examples of components, values, operations, materials, arrangements, and the like are described below. Of course, these are merely examples and are not intended to be limiting. Other components, values, operations, materials, arrangements, and the like are contemplated. In addition, the disclosure may repeat reference numbers and / or letters in various examples. This repetition is for the purposes of brevity and clarity and does not in itself dictate a relationship between the various embodiments and / or configurations discussed.

[0008] At least some embodiments use ensemble machine learning to suggest the addition of 5G cells to neighboring cell neighbor relations tables or to foresee that such additions may result in Physical Cell Identifier (PCI) conflicts.

[0009] In at least some embodiments, when new neighboring 5G cells are proactively discovered, a proposal to add any newly discovered neighboring 5G cells to the neighbor relationship table is issued if such addition is predicted not to cause a PCI conflict.

[0010] According to certain standards, the target cell or secondary cell group (Secondary Cell Group (MCG)) is selected as the anchor cell or master cell group (MCG). The existing neighboring cell relationship to the Target Cell Group (SCG) uses the eNB (controlling the anchor cell) to know the Cell Global Identity (CGI) and PCI of the target cell, and the anchor cell has an entry in the neighboring relationship table to identify the target cell. For each cell with which the eNB has a neighboring relationship, the neighboring relationship table contains a Target Cell Identifier (TCI) that identifies the target cell.

[0011] In at least some embodiments, an eNB serving cell configured for neighbor relationship conflict prediction instructs each connected terminal to perform measurements on neighboring cells as part of normal call procedures. In at least some embodiments, the network sends a measurement command (measurement control) to the terminal to perform cell detection and signal measurements of the terminal's surrounding cells. The terminal detects and measures the neighboring cell signals and reports them to the eNB serving cell that sent the measurement command. In at least some embodiments, neighbor relationship conflict prediction is performed upon receipt of each measurement report. In at least some embodiments, the neighbor relationship conflict prediction process starts in response to the terminal reporting a 5G PCI in a measurement report. In at least some embodiments, the frequency of the PCI is extended from the measurement ID reported by the terminal in the measurement report. In at least some embodiments, a check is made in the network configuration to understand whether the en-gNB corresponding to the PCI and frequency is already present in the neighbor relationship table of the reported eNB. If the relationship already exists, the entry is discarded and the neighbor relationship conflict prediction process stops. Otherwise, the neighbor relationship conflict prediction process proceeds. In response to detecting a 5G PCI from a measurement report, at least some embodiments verify whether a relationship is already provisioned and then perform a conflict prediction for the neighbor relationship before adding it to the neighbor relationship table of the eNB as well as any neighboring eNBs or proposing a conflict.

[0012] In at least some embodiments, a first level of neighbor relations is provisioned at the initial configuration of the system, such that the neighbor relations table of each MCG is initially provisioned with some known nearby 5G cells. In at least some embodiments, once the system is operational, further neighbor relations can be added through terminal-assisted neighbor relations conflict prediction and eNB configuration update procedures via the X2AP interface. At least some embodiments estimate the distance between various LTE cells.

[0013] In at least some embodiments, neighbor relationship conflict prediction is performed in real-time to capture and provide any neighbors reported by the terminal, hi at least some embodiments, the network configuration is updated frequently to help prevent any conflicts and to help ensure that the correct relationships are established for coverage.

[0014] 1 is a schematic diagram of the data flow for adjacency conflict prediction in accordance with at least one embodiment of the present invention, including an apparatus 100, an MCG 130, a database of adjacency tables 132, and a database of MCG geographic locations 133.

[0015] The MCG 130 is a node of a cellular network. In at least some embodiments, the MCG 130 operates in an EN-DC network. In the EN-DC network, the MCG 130 is an eNB node configured for 4G LTE technology, and the MCG 130 facilitates connection to a gNB node configured for 5G technology. In at least some embodiments, the MCG 130 consults a Neighbor Relation Table (NRT) that lists neighboring 5G cells supported by the MCG 130, neighboring 5G cells in conflict with the MCG 130, and neighboring 4G nodes provided by the MCG 130 for handoff. In at least some embodiments, the MCG 130 receives measurement reports from connected terminals for which the MCG 130 is the serving MCG. In at least some embodiments, the measurement reports include signal strength measurements, PCI, and ARFCN for each signal detected by the terminal. In at least some embodiments, the MCG 130 forwards the measurement reports, such as the measurement report 131, to the device 100.

[0016] The device 100 includes an identification section, a counting section 115, and a classification model 102. In at least some embodiments, the device receives measurement reports, such as measurement report 131, and outputs a list of recommendations 136, each of which includes information representing the PCI and ARFCN of the 5G cell, the serving MCG and neighboring MCGs, and an indication of whether the 5G cell should be provided in the neighbor relationship tables of the serving MCG and neighboring MCGs, or whether the 5G cell should be noted as having "contention" in the neighbor relationship tables. In at least some embodiments, the device 100 communicates with a database of neighbor relationship tables 132 and a database of MCG geographic locations 133 through a cellular network or another network, such as the Internet. In at least some embodiments, the database of neighbor relationship tables 132 is a constantly updated central source of configuration information for each cell in the cellular network, including the neighbor relationship tables. In at least some embodiments, the database of MCG geographic locations 133 is a daily updated source of site-level information for each cell in the cellular network, including geographic locations. In at least some embodiments, information may be retrieved from databases 132 and 133 by running a query for specific information or by obtaining a complete “dump” of the information and then reviewing the information on device 100.

[0017] The identification section 112 is configured to identify new 5G cells by PCI and ARFCN and to identify neighboring MCGs of the serving MCG. In at least some embodiments, the identification section 112 is configured to compare the measurement report with an adjacency table to identify any combination of PCI and ARFCN that was not found in the adjacency table of the serving MCG associated with the measurement report. In at least some embodiments, the identification section 112 is configured to retrieve the adjacency table of the MCG from the database of adjacency tables 132. In at least some embodiments, the identification section 112 is configured to output the PCI and ARFCN of any such unfounded combination with the serving MCG identifier to the counting section 115, such as PCI-ARFCN-NEARBY MCG 113. In at least some embodiments, the identification section 112 is configured to identify any MCGs that are in the neighborhood of the serving MCG. In at least some embodiments, the identification section 112 refers to a table of neighboring MCGs. In at least some embodiments, the identification section 112 is configured to populate the table of neighboring MCGs by retrieving MCG geographic locations from the database of MCG geographic locations 133. In at least some embodiments, the identification section 112 is configured to train a machine learning model to identify neighboring MCGs based on geographic coordinates and populate the table of neighboring MCGs according to the results. In at least some embodiments, the identification section 112 is configured to output MCG identifiers for any neighboring MCGs to the classification model 102 along with any combination of PCI and ARFCN that was not found in the adjacency table of the serving MCG, such as PCI-ARFCN-MCG 114.

[0018] The counting section 115 is configured to count 5G cells that are in the vicinity of the serving MCG and have the same PCI and ARFCN. In at least some embodiments, the counting section 115 is configured to count 5G cells that appear in the neighboring MCG's neighbor relation table. In at least some embodiments, the counting section 115 is configured to retrieve the neighboring MCG's neighbor relation table from the database of neighbor relation tables 132. In at least some embodiments, the counting section 115 is configured to output the number of 5G cells with the same PCI and ARFCN, such as PCI count 116, that are configured for input to the classification model 102.

[0019] The classification model 102 is a machine learning model trained to predict whether a PCI conflict will occur in response to provisioning a new 5G cell based on the PCI count. In at least some embodiments, the classification model 102 is trained to predict further based on geographic location. In at least some embodiments, the classification model 102 receives a geographic location in the form of a latitude and longitude from a database 133 of MCG geographic locations. In at least some embodiments, the classification model 102 receives as input a concatenation of the PCI count 116, the latitude, and the longitude. In at least some embodiments, the classification model 102 receives as input a concatenation of the PCI count 116, the latitude, the longitude, and the frequency type. In at least some embodiments, the frequency type is either Sub-6 or mmWAVE. In at least some embodiments, the classification model 102 receives as an input an index value of the PCI-ARFCN-MCG 113. In at least some embodiments, the classification model 102 outputs a PCI conflict prediction 103, which is a binary value representing a prediction of whether a PCI conflict will occur or not in response to providing a 5G cell associated with the PCI and ARFCN to the serving MCG and any neighboring MCGs. In at least some embodiments, the classification model 102 is a Support Vector Machine (SVM) classification model. In at least some embodiments, the classification model 102 is trained by solving a primal optimization problem using a Limited-memory Broyden-Fletcher-Goldfarb-Shanno (LBFGS) algorithm.

[0020] 2 is an operational flow for neighbor relationship conflict prediction in accordance with at least one embodiment of the present invention. The operational flow provides a method for neighbor relationship conflict prediction. In at least some embodiments, the method is performed by a controller of an apparatus that includes sections for performing certain operations, such as the controller and apparatus shown in FIG. 10, described below.

[0021] At S240, the controller or a section thereof receives a measurement report. In at least some embodiments, the controller receives a measurement report from an MCG, such as measurement report 131 of Figure 1. In at least some embodiments, the controller receives a measurement report for a terminal from a serving MCG for the terminal, the measurement report including a plurality of signal measurements, each signal measurement being associated with an SCG represented by a PCI and an ARFCN.

[0022] In S241, an identification section, such as the identification section 112 shown in FIG. 1, or a subsection thereof, identifies any new SCGs by PCI and ARFCN. In at least some embodiments, the identification section identifies any PCI and ARFCN combinations that were not found in the neighbor relations table of the serving MCG associated with the measurement report. In at least some embodiments, the identification section identifies any unlisted SCGs among the multiple signal measurements, where an unlisted SCG is represented by a PCI and associated ARFCN that does not correspond to any single SCG in the Neighbor Relations Table (NRT) of the serving MCG. If the identification section is able to identify any such unlisted SCGs, the operational flow proceeds to identifying neighboring MCGs in S242. If the identification section is unable to identify such unlisted SCGs, such as when the neighbor relations table already has an entry for each SCG in the measurement report, the operational flow proceeds to S249, where the measurement report is discarded just before the operational flow ends.

[0023] In S242, the identification section or a subsection thereof identifies MCGs in the measurement report that are in the vicinity of the serving MCG. In at least some embodiments, the identification section identifies one or more neighboring MCGs within a threshold distance of the serving MCG. In at least some embodiments, the threshold distance is one kilometer. In at least some embodiments, the identification section references a table of neighboring MCGs that pairs an MCG with an indication of the distance and / or an indication of whether the distance is within the threshold distance.

[0024] In S243, a counting section, such as counting section 115 shown in FIG. 1, or a subsection thereof, counts SCGs with the same PCI and ARFCN in the vicinity of the serving MCG. In at least some embodiments, the counting section counts the number of SCGs in the NRT of each neighboring MCG that have the PCI and ARFCN of an unlisted SCG. In at least some embodiments, the counting section examines each neighbor relationship table and increments the PCI count each time a matching SCG is found.

[0025] At S244, the application section of the controller or a subsection thereof applies a classification model, such as classification model 102 shown in FIG. 1, to the final PCI count. In at least some embodiments, the application section applies the classification model to the number of counted SCGs and the MCG-PCI-ARFCN identifiers representing the serving MCG, PCI, and ARFCN to obtain a predicted value, the predicted value being a binary value indicating whether or not providing the serving MCG and multiple neighboring MCGs to an unlisted SCG will result in a PCI conflict. In at least some embodiments, the MCG-PCI-ARFCN identifiers are index values ​​that do not form any basis for conflict prediction. In at least some embodiments, the application section further applies the classification model to the geographic location of the serving MCG to obtain a prediction, such that the conflict prediction is based on the PCI count and geographic location. In at least some embodiments, the application section further applies the classification model to the frequency type of the ARFCN to obtain a prediction, such that the conflict prediction is based on the PCI count, geographic location, and frequency type. In at least some embodiments, the application section causes the classification model to output a binary prediction value representing a prediction of whether or not a PCI conflict will occur in response to providing a 5G cell associated with the PCI and ARFCN to the serving MCG and any neighboring MCGs.

[0026] In S246, the controller or a portion of the controller determines whether the output of the classification model indicates a prediction that no conflict will occur. If the controller determines that the output of the classification model indicates a prediction that no conflict will occur, the operational flow proceeds to providing an adjacency relation in S246. If the controller determines that the output of the classification model indicates a prediction that a conflict will occur, the operational flow proceeds to notating "conflict" in S248.

[0027] At S247, the controller or a portion thereof causes the SCG to be provisioned to the serving MCG and neighboring MCGs. In at least some embodiments, the controller causes the SCG to appear as a handoff option in the neighbor relationship table of the serving MCG and each neighboring MCG. In at least some embodiments, the controller enables the serving MCG and neighboring MCGs to perform handoff to the SCG for any connected terminal capable of an EN-DC connection.

[0028] At S248, the controller, or a portion thereof, causes the SCG to appear as having a "conflict" in the neighbor relationship tables of the serving MCG and each neighboring MCG. In at least some embodiments, the controller enables the serving MCG and neighboring MCGs to recognize the SCG as having a PCI conflict with the cellular network.

[0029] Figure 3 is a schematic diagram of a data flow for continuous classification model training in accordance with at least one embodiment of the present invention. The diagram includes a training section 319, a classification model 302, and a database of adjacency tables 332. In at least some embodiments, the training section 319 is a component of an apparatus for adjacency conflict prediction, such as the apparatus 100 shown in Figure 1. The classification model 302 and the database of adjacency tables 332 are substantially similar in structure and function to the classification model 102 and the database of adjacency tables 132, respectively, described with respect to Figure 1, except where described differently.

[0030] The training section 319 is configured to train the classification model 302. In at least some embodiments, the training section 319 is configured to continually train the classification model 302 using new training samples created by validating predictions with the updated adjacency table. In at least some embodiments, the training section 319 maintains a copy of input data to the classification model 302, such as input data 305, such as data entered at S244 shown in FIG. 2. In at least some embodiments, the training section 319 is configured to check the database of adjacency tables 332 and apply appropriate labels to the input data to form the training samples. In at least some embodiments, the training section 319 is configured to train the classification model 302 using a batch of new training samples, such as training sample 306, to update the classification model 302.

[0031] 4 is an operational flow for continuous classification model training in accordance with at least one embodiment of the present invention. The operational flow provides a method for continuous classification model training. In at least some embodiments, the method is performed by a training section, such as training section 319 shown in FIG. 3, to train a machine learning classification model, such as classification model 302 shown in FIG. 3.

[0032] At S450, the training section or a subsection thereof stores copies of the input data as training samples to the classification model. In at least some embodiments, the stored input data includes MCG-PCI-ARFCN identifiers and PCI counts corresponding to the MCG-PCI-ARFCN identifiers. In at least some embodiments, the training section stores the counted numbers and the MCG-PCI-ARFCN identifiers as training samples. In at least some embodiments, the stored input data further includes geographic location and / or frequency type.

[0033] At S452, the training section or a subsection thereof compares the neighbor relationship table of the MCG identified in the training sample with the PCI and ARFCN identified in the training sample. In at least some embodiments, the training section checks the neighbor relationship table to determine whether the 5G cell associated with the PCI and ARFCN has been provisioned in the neighbor relationship table or has been marked as "PCI conflict". In at least some embodiments, the comparison at S452 or the determination at S453 is performed after the storage at S450 by a period ranging from one day to one week, or by the time it takes for the provisioning of the neighbor table to be approved by an authority and implemented in a configuration database such as the neighbor relationship table 332 shown in FIG. 3.

[0034] At S453, the training section or a subsection thereof determines whether a match is found in the adjacency table. In at least some embodiments, the training section determines whether the PCI and ARFCN are included in the NRT of the serving MCG. If a match is found in the adjacency table, the operational flow proceeds to S454, where a label indicating "add" is applied to the training sample. If no match is found in the adjacency table, the operational flow proceeds to S455, where a label indicating "conflict" is applied to the training sample. In at least some embodiments, the training section may proceed to S455 in response to determining that the adjacency table includes a notation that the matching 5G cell has a "PCI conflict". In at least some embodiments, the training section labels the training sample with a ground truth prediction value indicative of the result of the determination.

[0035] In at least some embodiments, the classification model is trained once per time period ranging from 15 minutes to one day. In at least some embodiments, the determination in S453, the labeling in S454 or S455, and the training in S459 are performed in batches once per time period ranging from 15 minutes to one day, and on multiple stored training samples per batch. In at least some embodiments, each batch of training data includes tens or hundreds of stored unlabeled training samples.

[0036] At S456, the training section or a subsection thereof determines whether all of the stored unlabeled training samples have been processed. If the training section determines that all of the stored unlabeled training samples have been processed, operational flow proceeds to training at S459. If the training section determines that unprocessed training samples remain, operational flow proceeds to selecting the next sample at S457 before returning to the comparison at S452.

[0037] In S459, the training section or a subsection thereof trains the classification model using one or more training samples. In at least some embodiments, the training section begins training the classification model after all training samples of a batch have been labeled. In at least some embodiments, the training section trains the classification model using the labeled training samples resulting from the labeling in S454 and S455 before proceeding to the next sample (S457).

[0038] 5 is a schematic diagram of data flow for training an initial classification model in accordance with at least one embodiment of the present invention. The diagram includes an identification section 512, a counting section 515, a training section 519, an initialized classification model 502, a database of adjacency tables 532, and a database of MCG geographic locations 533. In at least some embodiments, the training section 519 is a component of an apparatus for adjacency conflict prediction, such as the apparatus 100 shown in FIG. 1. In at least some embodiments, the functions of the training section 519 described below and the functions of the training section 319 described with respect to FIG. 3 are performed by a single training section of the apparatus. The identification section 512, the counting section 515, the initialized classification model 502, the database of adjacency tables 532, and the database of MCG geographic locations 533 are substantially similar in structure and function to the identification section 112, the counting section 115, the classification model 102, the database of adjacency tables 132, and the database of MCG geographic locations 133 described with respect to FIG. 1, respectively, except as described differently.

[0039] The training section 519 is configured to train the initialized classification model 502 to make an adjacency conflict prediction using the current configuration information. In at least some embodiments, the training section 519 is configured to arbitrarily select a PCI-ARFCN-MCG combination and have the identification section 512 and the counting section 515 provide an associated PCI count, such as PCI count 516. In at least some embodiments, the training section 519 is configured to retrieve an MCG geographic location from the database 533 and combine the PCI-ARFCN-MCG, PCI count, and MCG geographic location into a training sample. In at least some embodiments, the training section 519 is configured to label the training sample based on the PCI count. In at least some embodiments, the training section 519 is configured to generate one or more batches of training samples and train the initialized classification model 502 using the one or more batches until the classification model 502 accurately outputs a PCI conflict prediction, such as PCI conflict prediction 503.

[0040] 6 is an operational flow for training an initial classification model in accordance with at least one embodiment of the present invention. The operational flow provides a method of initial classification model training. In at least some embodiments, the method is performed by a training section, such as training section 519 shown in FIG. 5, to train an initialized machine learning classification model, such as initialized classification model 502 shown in FIG. 5.

[0041] At S660, the training section, or a subsection thereof, initializes the classification model. In at least some embodiments, the training section randomly assigns weight values ​​to each node of the classification model to serve as a starting point for training to begin.

[0042] At S661, the training section or a subsection thereof generates initial training samples of MCG-PCI-ARFCN combinations. In at least some embodiments, the training section generates a plurality of initial training samples, each of which includes an arbitrarily selected MCG-PCI-ARFCN identifier.

[0043] At S662, the training section or a subsection thereof counts the matching SCGs provided in the MCG adjacency table that are in the neighborhood of the MCG identified in the training sample. In at least some embodiments, the training section causes the identification section to identify, for each initial training sample, one or more neighboring MCGs within a threshold distance of the MCG represented by the MCG-PCI-ARFCN identifier. In at least some embodiments, the training section causes the counting section to count, for each initial training sample, the number of SCGs in the NRT of each neighboring MCG having a PCI and ARFCN represented by the MCG-PCI-ARFCN identifier. In at least some embodiments, the training section receives the PCI counts from the counting section.

[0044] In S663, the training section or a subsection thereof determines whether the PCI count resulting from the counting in S662 is less than a threshold PCI count. In at least some embodiments, the threshold PCI count is two. If the PCI count is less than the threshold PCI count, operational flow proceeds to S664, where a label indicating "add" is applied to the training sample. If the PCI count is not less than two, operational flow proceeds to S665, where a label indicating "conflict" is applied to the training sample. In at least some embodiments, the training section labels the training sample with one of a ground truth prediction indicating a PCI conflict in response to the counted number being equal to or greater than the threshold count, and a ground truth prediction indicating no PCI conflict in response to the counted number being less than the threshold count.

[0045] At S666, the training section or a subsection thereof determines whether all of the training samples generated at S661 have been processed. If the training section determines that all of the training samples generated at S661 have been processed, operational flow proceeds to training at S669. If the training section determines that unprocessed training samples remain, operational flow proceeds to selecting the next sample at S667 before returning to counting at S662.

[0046] In S669, the training section, or a subsection thereof, trains the initialized classification model using the multiple training samples generated in S661.

[0047] 7 is a schematic diagram of data flow for neighbourhood table population in accordance with at least one embodiment of the present invention. The diagram includes a pairing section 770, a distance calculation section 774, a neighbourhood determination section 777, a neighbourhood table 779, and a database of MCG geographic locations 733. In at least some embodiments, the pairing section 770, the distance calculation section 774, the neighbourhood determination section 777, and the neighbourhood table 779 are included in an identification section of a device such as the identification section 112 and device 100 shown in FIG. 1. The database of MCG geographic locations 733 is substantially similar in structure and function to the database of MCG geographic locations 133 as described with respect to FIG. 1, except where described differently.

[0048] The pairing section 770 pairs one MCG with another MCG among multiple MCGs in the cellular network. In at least some embodiments, the pairing section 770 pairs each MCG with each of the other MCGs until all possible combinations of MCGs are present in the MCG pairing. In at least some embodiments, the pairing section retrieves a geographic location of each MCG in the MCG pairing and outputs the geographic location, such as geographic location 771, to a distance calculation section 774. In at least some embodiments, each geographic location is represented as a latitude and longitude. In at least some embodiments, the pairing section 770 is configured to record the MCG pairing, such as MCG pairing 772, in a neighborhood table 779.

[0049] The distance calculation section 774 is configured to calculate a distance between two geographic locations. In at least some embodiments, the distance calculation section 774 is configured to receive two geographic locations, such as geographic location 771, from the pairing section 770. In at least some embodiments, the distance calculation section 774 is configured to apply the Pythagorean theorem to the difference between the latitudes and the difference between the longitudes of the two geographic locations. In at least some embodiments, the distance calculation section 774 is configured to output a resulting distance, such as distance 775, to the proximity determination section 777.

[0050] The neighborhood determination section 777 is configured to determine whether the MCGs paired are located near each other for the neighbor relationship conflict prediction. In at least some embodiments, the neighborhood determination section 777 is configured to receive a distance, such as distance 775, from the distance calculation section 774. In at least some embodiments, the neighborhood determination section 777 is configured to determine whether the distance is within a threshold distance. In at least some embodiments, the neighborhood determination section 777 is configured to apply the distance to a machine learning regression model trained to associate the distance with the result of the determination of whether the MCGs are located near each other. In at least some embodiments, the neighborhood determination section 777 is configured to record the result of the determination of whether the pair of MCGs are located near each other, such as the neighborhood determination result 778, in association with the MCG pairing, such as the MCG pairing 772, in the neighborhood table 779.

[0051] 8 is an operational flow for neighborhood table clustering in accordance with at least one embodiment of the present invention. The operational flow provides a method for neighborhood table clustering. In at least some embodiments, the method is performed by an identification section, such as identification section 112 shown in FIG.

[0052] At S880, the identification section or a subsection thereof obtains a geographic location for each of a plurality of MCGs in the cellular network. In at least some embodiments, the identification section obtains a latitude and longitude for each MCG in the cellular network.

[0053] In S881, the identification section or a subsection thereof forms a plurality of MCG pairings, each pairing including two MCGs in the cellular network. In at least some embodiments, the identification section forms a plurality of MCG pairs, each MCG pair including a first MCG of the plurality of MCGs and a second MCG of the plurality of MCGs. In at least some embodiments, the identification section pairs each MCG with each of the other MCGs until all possible combinations of MCGs in the cellular network are present in the MCG pairings.

[0054] In S883, the identification section or a subsection thereof calculates the distance between the geographic locations of the MCG pairings. In at least some embodiments, the identification section calculates, for each MCG pairing, the distance between the first MCG and the second MCG based on the geographic location of the first MCG and the geographic location of the second MCG. In at least some embodiments, the identification section applies Pythagoras' theorem to the difference between the latitudes and the difference between the longitudes of the geographic locations of the MCG pairings.

[0055] In S884, the identification section or a subsection thereof determines whether the distance calculated in S883 is within a threshold distance. In at least some embodiments, the identification section applies a machine learning regression model to the distance calculated in S883, the model being trained to relate the distance to the outcome of a determination of whether MCGs are located near each other. In at least some embodiments, the identification section applies the regression model to the calculated distance to obtain a proximity value, the proximity value being a binary value indicating whether the calculated distance is within a threshold distance.

[0056] In S886, the identification section or a subsection thereof adds the MCG pairing and the result of the proximity determination in S884 to a table, such as the proximity table 779 shown in FIG. 7. In at least some embodiments, the identification section stores data representing the first MCG corresponding to the second MCG, the geographic location of the first MCG, and a proximity value. In at least some embodiments, the identification section records the result of the determination whether the pairing of MCGs is located near each other in the proximity table in association with the MCG pairing. In at least some embodiments, the identification section stores the proximity table such that the identification of one or more neighboring MCGs is performed by accessing the stored data to retrieve one or more second MCGs corresponding to the serving MCG as the stored first MCG and a proximity value representing that the calculated distance is within a threshold distance.

[0057] In S888, the identification section or a subsection thereof determines whether all pairings have been processed. If the identification section determines that unprocessed MCG pairings remain, operational flow proceeds to selecting the next pairing in S889 before returning to distance calculation in S883. If the identification section determines that all MCG pairings have been processed, operational flow ends.

[0058] In at least some embodiments, the identification section performs the acquisition in S880 and the formation in S881, and the repetition of the calculation in S883 and storage in S886 is performed once every period ranging from 6 hours to 1 week to keep the neighborhood table updated.

[0059] FIG. 9 is a schematic diagram of data flow for training a neighborhood decision regression model, according to at least one embodiment of the present invention. The diagram includes a pairing section 970, a distance calculation section 974, a neighborhood decision regression model 907, a training section 919, and a database of MCG geographic locations 933. In at least some embodiments, the training section 919 is a component of an apparatus for neighbor relationship conflict prediction, such as the apparatus 100 shown in FIG. 1. In at least some embodiments, the functions of the training section 919 described below and the functions of the training section 319 described with respect to FIG. 3 and the training section 519 described with respect to FIG. 5 are performed by a single training section of the apparatus. The pairing section 970 and the distance calculation section 974 are substantially similar in structure and function to the pairing section 770 and the distance calculation section 774, respectively, described with respect to FIG. 7, except where differently described. The database of MCG geographic locations 933 is substantially similar in structure and function to the database of MCG geographic locations 133 as described with respect to FIG. 1, except where differently described.

[0060] The neighborhood decision regression model 907 is a machine learning model trained to associate distance with the outcome of a decision of whether MCGs are located near each other. In at least some embodiments, the neighborhood decision regression model 907 receives the distance as an input. In at least some embodiments, the neighborhood decision regression model 907 receives the MCG pairing as an input index value. In at least some embodiments, the neighborhood decision regression model 907 outputs a binary value indicating whether the calculated distance is within a threshold distance or not. In at least some embodiments, the neighborhood decision regression model 907 is a logistic regression machine learning model trained using a one-vs-rest approach.

[0061] The training section 919 is configured to train the neighborhood determination regression model 907 to associate distances with results of determining whether the MCGs are located near each other. In at least some embodiments, the training section 919 is configured to receive MCG pairings, such as MCG pairing 972, from the pairing section 970. In at least some embodiments, the training section 919 is configured to receive distances, such as distance 975, associated with the MCG pairings from the distance calculation section 974, and combine the distances and the MCG pairings into training samples. In at least some embodiments, the training section 919 is configured to label the training samples based on the distances. In at least some embodiments, the training section 919 is configured to combine the training samples of all the MCG pairings to train the neighborhood determination regression model 907 on one or more batches of all the MCG pairings until the neighborhood determination regression model 907 accurately outputs a result of the neighborhood determination, such as the neighborhood determination result 908.

[0062] 10 is an operational flow for training a neighborhood decision regression model according to at least one embodiment of the present invention. The operational flow provides a method of initial regression model training. In at least some embodiments, the method is performed by a training section, such as training section 919 shown in FIG. 9, to train an initialized machine learning regression model, such as neighborhood decision regression model 907 shown in FIG. 9.

[0063] At S990, the training section, or a subsection thereof, initializes the regression model. In at least some embodiments, the training section randomly assigns weight values ​​to each node of the regression model to serve as a starting point for beginning training.

[0064] In S991, the training section or a subsection thereof generates initial training samples of MCG pairings. In at least some embodiments, as the iterations of S991 proceed, the training section generates multiple initial training samples, each training sample including one of the multiple pairings.

[0065] At S992, the training section or a subsection thereof calculates the distance between the geographic locations of the MCG pairing. In at least some embodiments, the training section causes the identification section to apply the Pythagorean theorem to the difference between the latitudes and the difference between the longitudes of the geographic locations of the MCG pairing.

[0066] At S993, the training section or a subsection thereof determines whether the distance calculated at S992 is within a threshold distance. In at least some embodiments, the threshold distance is one kilometer. If the calculated distance is less than the threshold distance, the operational flow proceeds to S994, where a label indicating "inside" is applied to the training sample. If the calculated distance is equal to or greater than one kilometer, the operational flow proceeds to S995, where a label indicating "outside" is applied to the training sample. In at least some embodiments, the training section labels each training sample with a ground truth neighborhood value indicating whether the calculated distance is within the threshold distance.

[0067] At S996, the training section or a subsection thereof determines whether all of the training samples generated in the iterations of S991 for each MCG pairing have been processed. If the training section determines that all of the training samples for all MCG pairings have been processed, operational flow proceeds to training at S999. If the training section determines that unprocessed training samples remain, operational flow proceeds to selecting the next training sample at S997 before returning to generation at S991.

[0068] In S999, the training section or a subsection thereof trains the initialized regression model using the multiple training samples generated in the iterations of S991.

[0069] FIG. 11 is a block diagram of an exemplary hardware configuration for adjacency conflict prediction in accordance with at least one embodiment of the present invention.

[0070] An exemplary hardware configuration includes a device 1100 that communicates with a network 1136 and interacts with an input device 1139. The device 1100 may be a computer or other computing device that receives input or commands from the input device 1139. The device 1100 may be a host server that connects directly to the input device 1139 or indirectly through the network 1136. In some embodiments, the device 1100 is a computer system that includes two or more computers. In some embodiments, the device 1100 is a personal computer that runs applications for a user of the device 1100.

[0071] The device 1100 includes a controller 1110, a storage unit 1120, a communication interface 1135, and an input / output interface 1138. In some embodiments, the controller 1110 includes a processor or programmable circuitry that executes instructions to cause the processor or programmable circuitry to perform operations according to the instructions. In some embodiments, the controller 1110 includes analog or digital programmable circuits, or any combination thereof. In some embodiments, the controller 1110 includes physically separate storage devices or circuits that interact through communication. In some embodiments, the storage unit 1120 includes a non-volatile computer readable medium that can store executable and non-executable data for the controller 1110 to access during execution of instructions. The communication interface 1135 transmits and receives data from a network 1136, thereby enabling communication with the MCG 1130. The input / output interface 1138 connects various input / output units via parallel ports, serial ports, keyboard ports, mouse ports, monitor ports, etc., to accept commands and present information.

[0072] The controller 1110 includes an identification section 1112, a counting section 1115, an application section 1117, and a training section 1119. The storage unit 1120 includes a neighborhood table 1122, model parameters 1124, training samples 1126, and training parameters 1128.

[0073] The identification section 1112 is circuitry or instructions of the controller 1110 configured to identify unlisted SCGs from the measurement report and neighboring MCGs. In at least some embodiments, the identification section 1112 is configured to compare the measurement report to an adjacency table to identify any PCI and ARFCN combinations that were not found in the adjacency table of the serving MCG associated with the measurement report. In at least some embodiments, the identification section 1112 utilizes information in the storage unit 1120, such as neighborhood 1122. In at least some embodiments, the identification section 1112 includes subsections for performing additional functions, as described in the flow diagrams above. In at least some embodiments, such subsections may be referenced by names associated with their functions.

[0074] The counting section 1115 is a circuit or instruction of the controller 1110 configured to count matching SCGs. In at least some embodiments, the counting section 1115 is configured to count 5G cells that are in the vicinity of a serving MCG, such as MCG 1130, and have the same PCI and ARFCN. In at least some embodiments, the counting section 1115 utilizes information in a database in communication with the device 1100. In at least some embodiments, the counting section 1115 includes subsections for performing additional functions, as described in the flow charts above. In at least some embodiments, such subsections may be referenced by names associated with their functions.

[0075] The apply section 1117 is circuitry or instructions of the controller 1110 configured to apply the model to input data to obtain output data. In at least some embodiments, the apply section 1117 is configured to feed the input data to an input layer of the model, calculate intermediate values ​​of an intermediate layer, and read output data from an output layer of the model. In at least some embodiments, the apply section 1117 utilizes information in the storage unit 1120, such as model parameters 1124. In at least some embodiments, the apply section 1117 includes subsections for performing additional functions, as described in the preceding flow charts. In at least some embodiments, such subsections may be referenced by names associated with their functions.

[0076] The training section 1119 is circuitry or instructions of the controller 1110 configured to train the model. In at least some embodiments, the training section 1119 is configured to adjust the weights of the model based on the loss function until the loss function converges to a solution or training is otherwise considered complete. In at least some embodiments, the training section 1119 stores information in training samples 11126 and utilizes information in the storage unit 1120, such as model parameters 1124 and training parameters 1128. In at least some embodiments, the training section 1119 includes subsections for performing additional functions, as described in the preceding flow charts. In at least some embodiments, such subsections are referenced by names associated with the corresponding functions.

[0077] In at least some embodiments, the apparatus is another device capable of processing logical functions to perform the operations herein. In at least some embodiments, the controller and the storage unit need not be completely separate devices, and in some embodiments share circuitry or one or more computer-readable media. In at least some embodiments, the storage unit includes a hard drive that stores both computer-executable instructions and data accessed by the controller, and the controller includes a combination of a central processing unit (CPU) and RAM, and the computer-executable instructions can be copied in whole or in part for execution by the CPU during performance of the operations herein.

[0078] In at least some embodiments where the apparatus is a computer, a program installed on the computer can cause the computer to function as an apparatus of the embodiments described herein or to perform operations associated with the apparatus of the embodiments described herein, and in at least some embodiments, such a program is executable by a processor to cause the computer to perform certain operations associated with some or all of the blocks of the flowcharts and block diagrams described herein.

[0079] Various embodiments of the invention are described with reference to flow charts and block diagrams, where the blocks may represent (1) steps of a process in which an action is performed, or (2) sections of a controller responsible for performing an action. Particular steps and sections are implemented by dedicated circuitry, programmable circuitry provided with computer readable instructions stored on a computer readable medium, and / or a processor provided with computer readable instructions stored on a computer readable medium. In some embodiments, the dedicated circuitry includes digital and / or analog hardware circuitry, and may include integrated circuits (ICs) and / or discrete circuits. In some embodiments, the programmable circuitry includes reconfigurable hardware circuitry, such as field-programmable gate arrays (FPGAs), programmable logic arrays (PLAs), comprising logical AND, OR, XOR, NAND, NOR, and other logic operations, flip-flops, registers, memory elements, and the like.

[0080] Various embodiments of the invention include systems, methods, and / or computer program products. In some embodiments, the computer program product includes a computer readable storage medium (or media) having computer readable program instructions for causing a processor to perform aspects of the invention.

[0081] In some embodiments, a computer readable storage medium includes a tangible device capable of holding and storing instructions for use by an instruction execution device. In some embodiments, a computer readable storage medium includes, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of computer readable storage media includes portable computer diskettes, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, mechanically encoded devices such as punch cards or raised structures with instructions recorded in grooves, and any suitable combination of the above. As used herein, computer-readable storage media should not be interpreted as being transitory signals per se, such as electric waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (e.g., light pulses passing through a fiber optic cable), or electrical signals transmitted through wires.

[0082] In some embodiments, the computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to the respective computing / processing device or to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. In some embodiments, the network can include copper transmission cables, optical transmission fiber, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in the respective computing / processing device.

[0083] In some embodiments, the computer readable program instructions for performing the operations described above are either assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, and traditional procedural programming languages ​​such as the "C" programming language or similar programming languages. In some embodiments, the computer readable program instructions execute entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In some embodiments, in the latter scenario, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or a connection to an external computer may be made (e.g., via the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), performs aspects of the invention by executing computer readable program instructions and utilizing state information of the computer readable program instructions to individualize the electronic circuitry.

[0084] Although the embodiments of the present invention have been described above, the technical scope of the invention described in the claims is not limited to the scope described in the above embodiments. It is clear to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that embodiments with such modifications or improvements can also be included in the technical scope of the present invention.

[0085] It should be noted that the order of operations, procedures, steps, stages, and other processes in the devices, systems, programs, and methods shown in the claims, embodiments, and drawings is not explicitly stated as "before" or "prior to," and that the processes may be performed in any order as long as the output of a previous process is not used in a subsequent process. Even if the operational flow in the claims, embodiments, and drawings is described using "first," "next," and the like for convenience, this does not mean that the processes must be performed in this order.

[0086] According to at least one embodiment of the invention, the neighbor relation conflict prediction includes receiving a measurement report for the terminal from a serving Master Cell Group (MCG) of the terminal, the measurement report including a plurality of signal measurements, each of the signal measurements being associated with a Secondary Cell Group (SCG) represented by a Physical Cell Identifier (PCI) and an Absolute Radio-Frequency Channel Number (ARFCN), and identifying an SCG not listed in the plurality of signal measurements, the unlisted SCG being identified in a Neighbor Relations Table (NRR) of the serving Master Cell Group (MCG). identifying one or more nearby MCGs within a threshold distance of the serving MCG, where the nearby MCGs are represented by a PCI and associated ARFCN that do not correspond to any single SCG in the Listed SCGs Table (NRT); counting the number of SCGs in the NRT of each of the nearby MCGs that have the PCI and ARFCN of the unlisted SCG; and applying a classification model to the counted number of SCGs and the MCG-PCI-ARFCN identifiers representing the serving MCG, PCI, and ARFCN to obtain a predicted value, the predicted value being a binary value indicating whether providing the serving MCG and multiple nearby MCGs to the unlisted SCG will result in a PCI conflict.

[0087] Some embodiments include instructions in a computer program, a method implemented by a processor executing the instructions of the computer program, and an apparatus for implementing the method. In some embodiments, the apparatus includes a controller including circuitry configured to perform operations in the instructions.

[0088] The above outlines features of some embodiments so that those skilled in the art may better understand aspects of the present disclosure. It should be appreciated that those skilled in the art may readily use this disclosure as a basis for designing or modifying other processes and structures for carrying out the same purposes and / or achieving the same advantages of the embodiments introduced herein. Those skilled in the art should also appreciate that such equivalent constructions do not depart from the spirit and scope of the present disclosure, and that various changes, substitutions, and alterations can be made herein without departing from the spirit and scope of the present disclosure.

Claims

1. A computer-readable medium containing instructions executable by a computer to cause the computer to perform operations, the operations including: receiving a measurement report for a terminal from a serving master cell group (MCG) of the terminal, the measurement report including a plurality of signal measurements, each of the signal measurements being associated with a secondary cell group (SCG) represented by a physical cell identifier (PCI) and an absolute radio frequency channel number (ARFCN); identifying an SCG that is not listed in the plurality of signal measurements, the unlisted SCG being represented by a PCI and associated ARFCN that does not correspond to any single SCG in a Neighbor Relation Table (NRT) of the serving MCG; identifying one or more neighboring MCGs within a threshold distance of the serving MCG; Counting the number of SCGs in the NRT of each of the neighboring MCGs that have the PCI and the ARFCN of the unlisted SCG; applying a classification model to the counted number of SCGs and an MCG-PCI-ARFCN identifier representing the serving MCG, the PCI, and the ARFCN to obtain a predicted value, the predicted value being a binary value indicating whether or not providing the serving MCG and the neighboring MCGs to the unlisted SCG will result in a PCI conflict; 13. A computer readable medium comprising:

2. 2. The computer-readable medium of claim 1, wherein the instructions are further configured to cause the computer to perform applying the classification model to at least one of a geographic location of the serving MCG or a frequency type of the ARFCN to obtain the predicted value.

3. The instructions cause the computer to: storing the counted number and the MCG-PCI-ARFCN identifier as training samples; determining whether the PCI and the ARFCN are included in the NRT of the serving MCG; labeling the training samples with ground truth predictions indicative of the outcome of the determination; training the classification model using the labeled training samples; The computer-readable medium of claim 1 , further configured to cause the computer to perform the following steps:

4. The computer-readable medium of claim 3 , wherein the determining is performed a period of time after the storing, the period of time ranging from 1 day to 10 days.

5. The determining, labeling, and training are performed in batches, one per time period ranging from 15 minutes to one day, and each batch is performed on a plurality of stored training samples. The computer-readable medium of claim 3.

6. The instructions cause the computer to: initializing the classification model; generating a plurality of initial training samples, each of the plurality of initial training samples including an arbitrarily selected MCG-PCI-ARFCN identifier; for each of the plurality of initial training samples, identifying one or more neighboring MCGs within the threshold distance of the MCG represented by the MCG-PCI-ARFCN identifier; For each of the plurality of initial training samples, counting the number of SCGs in the NRT of each of the neighboring MCGs having the PCI and the ARFCN represented by the MCG-PCI-ARFCN identifier; labeling the training samples with one of a ground truth prediction indicating a PCI conflict in response to the counted number being equal to or greater than a threshold count, or a ground truth prediction indicating no PCI conflict in response to the counted number being less than the threshold count; training the initialized classification model using the plurality of initial training samples; The computer-readable medium of claim 3 , further configured to cause the computer to perform the following steps:

7. The computer-readable medium of claim 1 , wherein the classification model is a support vector machine classification model.

8. The instructions cause the computer to: obtaining a geographic location for each of a plurality of MCGs in a cellular network; forming a plurality of MCG pairings, each of the plurality of MCG pairings including a first MCG of the plurality of MCGs and a second MCG of the plurality of MCGs; For each of the plurality of MCG pairings, calculating a distance between the first MCG and the second MCG based on the geographic location of the first MCG and the geographic location of the second MCG; storing data representative of the first MCG corresponding to the second MCG, the geographic location of the first MCG, and a proximity value, the proximity value being a binary value indicating whether the calculated distance is within the threshold distance; and further configured to cause the the identifying one or more neighboring MCGs includes accessing the stored data to retrieve one or more second MCGs corresponding to the serving MCG as the stored first MCG and the proximity value representing that the calculated distance is within the threshold distance; The computer-readable medium of claim 1 .

9. 9. The computer readable medium of claim 8, wherein the obtaining, forming, calculating, and storing are performed once every period ranging from six hours to one week.

10. The computer-readable medium of claim 8 , wherein the instructions are further configured to cause the computer to apply a regression model to the calculated distance to obtain the neighborhood value.

11. The instructions cause the computer to: initializing the regression model; generating a plurality of initial training samples, each of the plurality of initial training samples comprising one of the plurality of MCG pairings; labeling each of the plurality of initial training samples with a ground truth proximity value indicating whether the calculated distance is within the threshold distance; training the initialized regression model using the plurality of initial training samples; The computer readable medium of claim 10 further configured to:

12. 1. A method comprising: receiving a measurement report for a terminal from a serving master cell group (MCG) of the terminal, the measurement report including a plurality of signal measurements, each of the signal measurements being associated with a secondary cell group (SCG) represented by a physical cell identifier (PCI) and an absolute radio frequency channel number (ARFCN); identifying an SCG that is not listed in the plurality of signal measurements, the unlisted SCG being represented by a PCI and associated ARFCN that does not correspond to any single SCG in a Neighbor Relation Table (NRT) of the serving MCG; identifying one or more neighboring MCGs within a threshold distance of the serving MCG; counting the number of SCGs in the NRT of each of the neighboring MCGs that have the PCI and the ARFCN of the unlisted SCG; applying a classification model to the counted number of SCGs and an MCG-PCI-ARFCN identifier representing the serving MCG, the PCI, and the ARFCN to obtain a predicted value, the predicted value being a binary value indicating whether or not providing the serving MCG and the neighboring MCGs to the unlisted SCG will result in a PCI conflict; A method comprising:

13. The method of claim 12 , wherein the applying further comprises applying the classification model to at least one of a geographic location of the serving MCG or a frequency type of the ARFCN to obtain the predicted value.

14. storing the counted number and the MCG-PCI-ARFCN identifier as training samples; determining whether the PCI and the ARFCN are included in the NRT of the serving MCG; labeling the training samples with ground truth predictions indicative of the outcome of the determination; training the classification model using the labeled training samples; The method of claim 12 further comprising:

15. The method of claim 14, wherein the determining is performed a period of time after the storing, the period of time ranging from 1 day to 10 days.

16. The determining, labeling, and training are performed in batches, one per time period ranging from 15 minutes to one day, and each batch is performed on a plurality of stored training samples. The method of claim 14.

17. initializing the classification model; generating a plurality of initial training samples, each of the plurality of initial training samples including an arbitrarily selected MCG-PCI-ARFCN identifier; for each of the plurality of initial training samples, identifying one or more neighboring MCGs within the threshold distance of the MCG represented by the MCG-PCI-ARFCN identifier; For each of the plurality of initial training samples, counting the number of SCGs in the NRT of each of the neighboring MCGs having the PCI and the ARFCN represented by the MCG-PCI-ARFCN identifier; labeling the training samples with one of a ground truth prediction indicating a PCI conflict in response to the counted number being equal to or greater than a threshold count, and a ground truth prediction indicating no PCI conflict in response to the counted number being less than the threshold count; training the initialized classification model using the plurality of initial training samples; The method of claim 14 further comprising:

18. The method of claim 12 , wherein the classification model is a support vector machine classification model.

19. obtaining a geographic location for each of a plurality of MCGs in a cellular network; forming a plurality of MCG pairings, each of the plurality of MCG pairings including a first MCG of the plurality of MCGs and a second MCG of the plurality of MCGs; For each of the plurality of MCG pairings, calculating a distance between the first MCG and the second MCG based on the geographic location of the first MCG and the geographic location of the second MCG; storing data representative of the first MCG corresponding to the second MCG, the geographic location of the first MCG, and a proximity value, the proximity value being a binary value indicating whether the calculated distance is within the threshold distance; Further comprising: the identifying one or more neighboring MCGs includes accessing the stored data to retrieve one or more second MCGs corresponding to the serving MCG as the stored first MCG and the proximity value representing that the calculated distance is within the threshold distance. The method of claim 12.

20. An apparatus comprising: a controller including a circuit, the circuit comprising: receiving a measurement report for a terminal from a serving master cell group (MCG) of the terminal, the measurement report including a plurality of signal measurements, each of the signal measurements being associated with a secondary cell group (SCG) represented by a physical cell identifier (PCI) and an absolute radio frequency channel number (ARFCN); identifying an SCG that is not listed in the plurality of signal measurements, the unlisted SCG being represented by a PCI and associated ARFCN that does not correspond to any single SCG in a Neighbor Relation Table (NRT) of the serving MCG; identifying one or more neighboring MCGs within a threshold distance of the serving MCG; Counting the number of SCGs in the NRT of each neighboring MCG that has the PCI and the ARFCN of the unlisted SCG; applying a classification model to the counted number of SCGs and an MCG-PCI-ARFCN identifier representing the serving MCG, the PCI, and the ARFCN to obtain a predicted value, the predicted value being a binary value indicating whether or not providing the serving MCG and the neighboring MCGs to the unlisted SCG will result in a PCI conflict; An apparatus configured to:

Images