Apparatus, method and computer program for using channel measurement data for primary component carrier

Abstract

An apparatus, method, and computer program are described including obtaining first channel measurement data for a primary component carrier, the primary component carrier being used for communication between a device of a mobile communication system and a network node; providing the first channel measurement data as input to an algorithm to obtain a first embedding output; comparing the first embedding output to embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify a closest reference channel measurement data relative to the first channel measurement data; identifying a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information comprises a carrier aggregation policy, the carrier aggregation policy defining one or more secondary communication channels; and setting a carrier aggregation policy for the primary component carrier for communication between the device of the mobile communication system and the network node in accordance with the carrier aggregation policy of the identified set of reference channel measurement data.

Description

Technical Field

[0001] This manual relates to communication systems. Background Technology

[0002] Carrier aggregation in mobile communication systems can provide multiple carriers with different center frequencies. Resource allocation is still required in such communication systems. Summary of the Invention

[0003] In a first aspect, this specification describes an apparatus comprising components for performing: obtaining first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; providing the first channel measurement data as input to an algorithm to obtain a first embedding output; comparing the first embedding output with embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; identifying a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information includes a carrier aggregation strategy defining one or more secondary communication channels; and setting a carrier aggregation strategy for the primary component carrier, for communication between the device and the network node in the mobile communication system, based on the carrier aggregation strategy of the identified set of reference channel measurement data. Each embedding output obtained from the algorithm may be a lower-dimensional representation of the channel measurement data for the corresponding primary component carrier.

[0004] In some example embodiments, the component also performs the following: identifying the closest reference channel measurement data by identifying the reference channel measurement data among a plurality of reference channel measurement data having the closest embedding output to the first embedding output.

[0005] The algorithm can be implemented using a ternary network with multiple parameters that can be trained using a ternary loss function.

[0006] The secondary communication channels of a carrier aggregation strategy can be arranged in a sorted list.

[0007] Channel measurement data used for the primary component carrier or each primary component carrier may include channel quality information, channel state information, or similar data.

[0008] The component may include: at least one processor; and at least one memory, including computer program code, wherein the at least one memory and the computer program code are configured to cause the execution of the device together with the at least one processor.

[0009] In a second aspect, this specification describes an apparatus comprising components for performing the following: obtaining a plurality of training data sets, wherein each of the plurality of training data sets includes: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for the primary component carrier; and a sorted list of secondary channel identifiers for secondary component carriers used for communication between the device and the network node; identifying one or more reference data sets (e.g., anchor points) from the plurality of training data sets, wherein each reference data set has a combination of primary channel identifiers and secondary channel identifiers that are different from all other reference data sets; and identifying from the plurality of training data sets... For each reference data set, identify one or more matching data sets (e.g., positive), where each matching data set has the same combination of primary channel identifier and secondary channel identifier as the corresponding reference data set; from multiple training data sets, identify one or more non-matching data sets (e.g., negative) for each reference data set, where each non-matching data set has a different combination of primary channel identifier and secondary channel identifier than the corresponding reference data set; and train the trainable parameters of the algorithm by minimizing the loss function until, for all matching data sets, the output of the algorithm responding to the matching data sets is closer to the output of the algorithm responding to the corresponding reference data set than the output of the algorithm responding to any other reference data set.

[0010] In some example embodiments, the loss function is a triplet loss function. The algorithm can be implemented using a triplet network with multiple parameters that can be trained using the triplet loss function. The loss function can be associated with multiple triplets to train the algorithm, where each triplet includes: channel measurement data (e.g., anchor data) for a first reference data set of one or more reference data sets; channel measurement data for a matching data set (e.g., positive) of a corresponding first reference data set; and channel measurement data for a non-matching data set (e.g., negative) of a corresponding first reference data set.

[0011] In some example embodiments, the component is also configured to perform: determining whether the training of trainable parameters of the triplet network has been verified by: providing channel measurement data of each of a plurality of identified matching datasets as input to the algorithm to obtain a plurality of matching embedding outputs; and determining whether each of the plurality of matching embedding outputs is closer to the embedding output of the corresponding reference dataset than the embedding output of any other reference dataset, and if so, determining that the algorithm has been verified.

[0012] The component can also be configured to perform one or more of the following: store the algorithm; store one or more reference data sets; or store a set of embedded outputs, wherein each embedded output is generated by applying first channel measurement data of the main component carrier for the corresponding reference data set to the algorithm.

[0013] Channel measurement data used for the primary component carrier or each primary component carrier may include channel quality information, channel state information, or similar data.

[0014] The component may include: at least one processor; and at least one memory, including computer program code, wherein the at least one memory and the computer program code are configured to cause the execution of the device together with the at least one processor.

[0015] In a third aspect, this specification describes a method comprising: obtaining first channel measurement data for a primary component carrier, the primary component carrier being used for communication between a device and a network node in a mobile communication system; providing the first channel measurement data as input to an algorithm to obtain a first embedding output; comparing the first embedding output with embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; identifying a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information includes a carrier aggregation strategy, the carrier aggregation strategy defining one or more secondary communication channels; and setting a carrier aggregation strategy for the primary component carrier based on the carrier aggregation strategy of the identified set of reference channel measurement data.

[0016] Some example embodiments also include identifying the closest reference channel measurement data by identifying the reference channel measurement data among a plurality of reference channel measurement data having the closest first embedding output.

[0017] The algorithm can be implemented using a ternary network with multiple parameters, which can be trained using a ternary loss function.

[0018] The secondary communication channels of a carrier aggregation strategy can be arranged in a sorted list.

[0019] Channel measurement data used for the primary component carrier or each primary component carrier may include channel quality information, channel state information, or similar data.

[0020] In a fourth aspect, this specification describes a method comprising: obtaining a plurality of training data sets, each of the plurality of training data sets comprising: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for the primary component carrier; and a sorted list of secondary channel identifiers for secondary component carriers used for communication between the device and the network node; identifying one or more reference data sets from the plurality of training data sets, wherein each reference data set has a combination of primary channel identifiers and secondary channel identifiers different from all other reference data sets; identifying one or more matching data sets for each reference data set from the plurality of training data sets, wherein each matching data set has a combination of primary channel identifiers and secondary channel identifiers identical to those of the corresponding reference data set; identifying one or more non-matching data sets for each reference data set from the plurality of training data sets, wherein each non-matching data set has a combination of primary channel identifiers and secondary channel identifiers different from those of the corresponding reference data set; and training trainable parameters of an algorithm by minimizing a loss function until, for all matching data sets, the output of the algorithm responding to the matching data sets is closer to the output of the algorithm responding to the corresponding reference data sets than the output of the algorithm responding to any other reference data set.

[0021] In some example embodiments, the loss function is a triplet loss function. The algorithm can be implemented using a triplet network with multiple parameters that can be trained using the triplet loss function. The loss function can be associated with multiple triplets to train the algorithm, where each triplet includes: channel measurement data (e.g., anchor data) for a first reference data set of one or more reference data sets; channel measurement data for a matching data set (e.g., positive) of a corresponding first reference data set; and channel measurement data for a non-matching data set (e.g., negative) of a corresponding first reference data set.

[0022] In some example embodiments, the method further includes: determining whether the training of trainable parameters of the triplet network has been verified by: providing channel measurement data of each of a plurality of identified matching data sets as input to the algorithm to obtain a plurality of matching embedding outputs; and determining whether each of the plurality of matching embedding outputs is closer to the embedding output of the corresponding reference data set than the embedding output of any other reference data set, and if so, determining that the algorithm has been verified.

[0023] The method may include one or more of the following: storing an algorithm; storing one or more sets of reference data; or storing a set of embedded outputs, wherein each embedded output is generated by applying first channel measurement data of the principal component carrier for the corresponding set of reference data to the algorithm.

[0024] Channel measurement data used for the primary component carrier or each primary component carrier may include channel quality information, channel state information, or similar data.

[0025] In a fifth aspect, this specification describes an apparatus configured to perform any of the methods described with reference to the third or fourth aspect.

[0026] In a sixth aspect, this specification describes computer-readable instructions that, when executed by a computing device, cause the computing device to perform any of the methods described with reference to the third or fourth aspect.

[0027] In a seventh aspect, this specification describes a computer program including instructions for causing a device to perform at least the following: obtaining first channel measurement data for a primary component carrier, the primary component carrier being used for communication between a device and a network node in a mobile communication system; providing the first channel measurement data as input to an algorithm to obtain a first embedding output; comparing the first embedding output with embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; identifying a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information includes a carrier aggregation strategy that defines one or more secondary communication channels; and setting a carrier aggregation strategy for the primary component carrier based on the carrier aggregation strategy of the identified set of reference channel measurement data.

[0028] In an eighth aspect, this specification describes a computer program including instructions for causing a device to perform at least the following: obtaining a plurality of training data sets, wherein each of the plurality of training data sets includes: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for the primary component carrier; and a sorted list of secondary channel identifiers for a secondary component carrier used for communication between the device and the network node; identifying one or more reference data sets from the plurality of training data sets, wherein each reference data set has a combination of primary channel identifiers and secondary channel identifiers that are different from all other reference data sets; from the plurality of For each reference dataset, one or more matching datasets are identified in the training dataset, where each matching dataset has the same combination of primary channel identifier and secondary channel identifier as the corresponding reference dataset; for each reference dataset, one or more non-matching datasets are identified from multiple training datasets, where each non-matching dataset has a different combination of primary channel identifier and secondary channel identifier than the corresponding reference dataset; and the trainable parameters of the algorithm are trained by minimizing a loss function until, for all matching datasets, the output of the algorithm responding to the matching datasets is closer to the output of the algorithm responding to the corresponding reference datasets than the output of the algorithm responding to any other reference dataset.

[0029] In a ninth aspect, this specification describes a computer-readable medium (such as a non-transient computer-readable medium) including program instructions stored thereon for performing at least the following: obtaining first channel measurement data for a primary component carrier, the primary component carrier being used for communication data between a device and a network node in a mobile communication system; providing the first channel measurement data as input to an algorithm to obtain a first embedding output; comparing the first embedding output with embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; identifying a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information includes a carrier aggregation strategy that defines one or more secondary communication channels; and setting a carrier aggregation strategy for the primary component carrier based on the carrier aggregation strategy of the identified set of reference channel measurement data.

[0030] In a tenth aspect, this specification describes a computer-readable medium (such as a non-transient computer-readable medium) including program instructions stored thereon for performing at least the following: obtaining a plurality of training data sets, wherein each of the plurality of training data sets includes: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for said primary component carrier; and a sorted list of secondary channel identifiers for secondary component carriers used for communication between the device and the network node; identifying one or more reference data sets from the plurality of training data sets, wherein each reference data set has a primary channel identifier and a secondary channel identifier different from all other reference data sets. Combination of channel identifiers; identification of one or more matching data sets for each reference data set from multiple training data sets, wherein each matching data set has the same combination of primary and secondary channel identifiers as the corresponding reference data set; identification of one or more non-matching data sets for each reference data set from multiple training data sets, wherein each non-matching data set has a different combination of primary and secondary channel identifiers than the corresponding reference data set; and training the trainable parameters of the algorithm by minimizing a loss function until, for all matching data sets, the output of the algorithm responding to the matching data sets is closer to the output of the algorithm responding to the corresponding reference data set than the output of the algorithm responding to any other reference data set.

[0031] In an eleventh aspect, this specification describes an apparatus comprising: at least one processor; and at least one memory including computer program code that, when executed by the at least one processor, causes the apparatus to: obtain first channel measurement data for a primary component carrier, the primary component carrier being used for communication between devices and network nodes of a mobile communication system; provide the first channel measurement data as input to an algorithm to obtain a first embedding output; compare the first embedding output with embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; identify a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information includes a carrier aggregation strategy that defines one or more secondary communication channels; and set a carrier aggregation strategy for the primary component carrier based on the carrier aggregation strategy of the identified set of reference channel measurement data.

[0032] In a twelfth aspect, this specification describes an apparatus comprising: at least one processor; and at least one memory including computer program code that, when executed by the at least one processor, causes the apparatus to: acquire a plurality of training data sets, wherein each of the plurality of training data sets includes: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for the primary component carrier; and a sorted list of secondary channel identifiers for secondary component carriers used for communication between the device and the network node; and identify one or more reference data sets from the plurality of training data sets, wherein each reference data set has a primary channel different from all other reference data sets. The algorithm includes: a combination of primary channel identifiers and secondary channel identifiers; identifying one or more matching data sets for each reference data set from multiple training data sets, wherein each matching data set has the same combination of primary channel identifiers and secondary channel identifiers as the corresponding reference data set; identifying one or more non-matching data sets for each reference data set from multiple training data sets, wherein each non-matching data set has a different combination of primary channel identifiers and secondary channel identifiers than the corresponding reference data set; and training trainable parameters of the algorithm by minimizing a loss function until, for all matching data sets, the output of the algorithm responding to the matching data sets is closer to the output of the algorithm responding to the corresponding reference data set than the output of the algorithm responding to any other reference data set.

[0033] In a thirteenth aspect, this specification describes an apparatus comprising: a first measurement module (or other components) for acquiring first channel measurement data for a primary component carrier, the primary component carrier being used for communication between a device and a network node in a mobile communication system; a first output (or other components) for providing the first channel measurement data as input to an algorithm to obtain a first embedding output; a comparator (or other components) for comparing the first embedding output with embedding outputs of a plurality of reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; a clustering module (or other components) for identifying a set of reference channel information associated with the identified closest reference channel measurement data, wherein the identified set of reference channel information includes a carrier aggregation strategy defining one or more secondary communication channels; and an output module (or other components) for setting a carrier aggregation strategy for the primary component carrier based on the carrier aggregation strategy of the identified set of reference channel measurement data, for communication between a device and a network node in a mobile communication system.

[0034] In a fourteenth aspect, this specification describes an apparatus comprising: an input module (or other components) for acquiring a plurality of training data sets, wherein each of the plurality of training data sets includes: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for the primary component carrier; and a sorted list of secondary channel identifiers for secondary component carriers used for communication between the device and the network node; a first processor (or other components) for: identifying one or more reference data sets from the plurality of training data sets, wherein each reference data set has a combination of primary channel identifiers and secondary channel identifiers that are different from all other reference data sets; and for acquiring data from the plurality of training data sets. The dataset includes one or more matching datasets for each reference dataset, where each matching dataset has the same combination of primary channel identifier and secondary channel identifier as the corresponding reference dataset; and one or more non-matching datasets for each reference dataset from multiple training datasets, where each non-matching dataset has a different combination of primary channel identifier and secondary channel identifier than the corresponding reference dataset; and a machine learning module (or some other component) for training the trainable parameters of the algorithm by minimizing a loss function until, for all matching datasets, the output of the algorithm responding to the matching datasets is closer to the output of the algorithm responding to the corresponding reference dataset than the output of the algorithm responding to any other reference dataset.

[0035] Embodiments of this disclosure improve the overall data throughput for a device (such as a user equipment UE). Because the probability of selecting the optimal secondary carrier component (CC)(s) is higher, the UE will require fewer CCs to achieve the target throughput, as the bit rate on the optimally selected CC(s)(s) is higher. This also increases the UE's battery life. Furthermore, since the selection of the secondary CC(s)(s) does not require any additional measurements from the secondary CC(s), latency is expected to be reduced. Attached Figure Description

[0036] The exemplary embodiments will now be described by way of example only, referring to the following schematic diagrams, in which:

[0037] Figure 1 This is a block diagram of a system according to an example embodiment;

[0038] Figure 2 This is a flowchart illustrating the algorithm according to an example embodiment;

[0039] Figure 3 This is a block diagram of a system according to an example embodiment;

[0040] Figures 4 to 7This is a flowchart illustrating the algorithm according to an example embodiment;

[0041] Figure 8 and 9 This is a block diagram of a system according to an example embodiment;

[0042] Figure 10 It is a block diagram of a neural network that can be used in some example embodiments;

[0043] Figure 11 This is a block diagram of the components of a system according to an example embodiment; and

[0044] Figure 12A and Figure 12B Tangible media are shown, including a removable non-volatile memory cell storing computer-readable code and an optical disc (CD) that performs operations according to an example embodiment when the computer is running. Detailed Implementation

[0045] The scope of protection sought by the various embodiments of the present invention is defined by the independent claims. Embodiments and features described in the specification that do not fall within the scope of the independent claims (if any) are to be interpreted as examples useful for understanding the various embodiments of the invention.

[0046] In the description and accompanying drawings, similar reference numerals refer to similar elements throughout.

[0047] Figure 1 It is a block diagram of a system according to an example embodiment, generally indicated by reference numeral 10.

[0048] System 10 includes user equipment 12 (such as a mobile communication device or user equipment) and communication node 14 (such as a base station). User equipment 12 has a primary component carrier (PCC) 15. Furthermore, user equipment 12 has one or more secondary component carriers (SCCs) in both the uplink (UL) and downlink (DL) directions. Two secondary component carriers 16a and 16b are... Figure 1 The example is shown below.

[0049] Carrier aggregation (CA) can be useful in mobile communication systems, where high bit-rate communication can be achieved by aggregating multiple component carriers (CCs) at different center frequencies and transmitting them simultaneously in both the downlink (DL) and uplink (UL) on the aggregated CCs. System 10 implements a multi-component carrier-based mobile communication system that can be adapted to technologies such as 4G / LTE and 5G.

[0050] Different user equipment (UEs) may use different carriers as their primary component carriers. The process of adding (multiple) secondary component carriers is called CA configuration, which can be handled at the Radio Resource Control (RRC) layer of the 3GPP protocol stack. The configured (multiple) secondary component carriers can be active or inactive, and CA activation can be handled at the Media Access Control (MAC) layer of the 3GPP protocol stack.

[0051] In some example embodiments, base station 14 may have a set of component carriers (CCs) configured for each of the multiple user equipments (including user equipment 12) communicating with the base station. Therefore, when base station 14 decides to transmit a service on any of the secondary CCs already configured for a given user equipment, that CC should be in a “CA active” state. CA active state means that the relevant user equipment scans the PDCCH (Physical Downlink Control Channel) for any DL / UL grant. It should also be noted that the more bandwidth (more CCs) a user equipment needs to scan, the lower its battery utilization optimization.

[0052] Figure 2 This is a flowchart illustrating an algorithm according to an example embodiment, generally indicated by reference numeral 20. Algorithm 20 can be used to determine, for example, the secondary component carrier configuration of user equipment 12.

[0053] Algorithm 20 begins with operation 22, in which channel information (such as channel state information) of the primary component carrier used for communication between the user equipment and the communication node is obtained (e.g., measured). For example, the channel information may be related to the primary component carrier (PCC) 15 of the system 10 described above.

[0054] In operation 24, one or more similar channels are determined based on the channel information (e.g., channel state information) obtained in operation 22 for the primary component carrier. For example, the channel state information for the primary component carrier may contain channel-related information that can be used to identify other similar channels according to certain metrics (discussed in detail below).

[0055] In operation 26, the secondary component carrier configuration or strategy used for identification of similar channels is applied to the secondary component carrier configuration or strategy of the channel. Using a secondary component carrier configuration or strategy for "similar" channels may result in better performance than randomly or arbitrarily selecting the second component carrier.

[0056] As described in detail below, the operation 24 in which one or more similar channels are identified can be implemented using machine learning algorithms.

[0057] Figure 3This is a block diagram of a system according to an example embodiment, generally indicated by reference numeral 30. System 30 includes a data preparation module 32, an offline training module 34, and an online usage module 36. Once trained, the online usage module 36 can be used to implement the algorithm 20 described above.

[0058] Data preparation module 32 utilizes data for known samples, for which examples (e.g., optimal) secondary component carriers (CCs) for the identified primary component carrier are known and recorded (for training purposes). The data can be organized into a database by measuring the channel quality (or some other channel state metric) on each of the multiple secondary CCs and selecting the optimal CC for activation. Each sample in the database may include identifiers (IDs) for the primary and secondary active component carriers, as well as radio measurements for the primary channel (PCC). For example, such radio measurements may correspond to channel quality measurements in the downlink (e.g., CQI) (which needs to be reported from the user equipment / UE) or channel estimations in the uplink (measured directly by the base station / gNB).

[0059] Offline training module 34 seeks to utilize machine learning (ML) to infer (1) the hidden relationship between channel information (e.g., channel quality) on a specific primary component carrier and (2) the channel quality on a secondary component carrier activated for the corresponding primary component carrier, both of which are determined based on the data processed in operation 32. The ML model is trained to learn an embedding for each sample in the database. Such "embeddings" are lower-dimensional representations of radio measurements, seeking to capture fundamental features of the radio measurements that are crucial in identifying the unknown relationship between channel information (e.g., quality) on the primary and secondary CCs. They are called embeddings because mathematical operations can be performed on them.

[0060] Once the ML model is trained (using offline training module 34), the embeddings (hereinafter referred to as "masters" or "anchors") of certain samples from the training set are stored in a database, which is then used by online usage module 36 for similarity measurement.

[0061] The online usage module 36 enables real-time operations to provide a list of secondary component carriers (CCs) to be activated based on PCC channel quality measurements (available on DL or UL). For a given user equipment (such as user equipment 12 described above), the online usage module 36 can use an ML model to generate an embedding based on the PCC channel measurement data. The anchor / master whose embedding is closest to that of the given user equipment can be identified in the database (thus implementing operation 24 of algorithm 20), and the user equipment is assigned to the cluster belonging to that anchor / master. Subsequently, the secondary CCs of the anchor can be acquired and activated for that user equipment, as discussed in detail below (thus implementing operation 26 of algorithm 20).

[0062] Figure 4 This is a flowchart illustrating an algorithm according to an example embodiment, generally indicated by reference numeral 40. Algorithm 40 begins with operation 41 and then moves to operation 42.

[0063] In operation 42, the marked data is prepared. Operation 42 can be implemented by the data preparation module 32.

[0064] Each element of the tagging data organized in operation 42 includes: the identifier of the primary component carrier (PCC ID), the measured radio conditions of the primary component carrier (PCC measurements, such as channel quality information, channel state information, or similar data), and the identifiers of the secondary component carriers associated with the primary component carrier (SCC IDs, such as SCC1, SCC2, ..., SCCm). Therefore, for each user, there may be m configured secondary CCs, and these listed secondary CCs may be ordered (e.g., the order of the list can determine the order in which the secondary CCs are activated).

[0065] In operation 43, the ML model is trained based on labeled data. Training of the ML model can be implemented by the offline training module 34. The ML model is trained by minimizing a loss function, as discussed in detail below.

[0066] In operation 43, the ML model is used to extract embeddings from PCC measurements against labeled data. For example, operation 43 can be implemented by providing channel measurement data from each of a plurality of identified matching data sets (i.e., obtained from the labeled data processed in operation 42) as input to the ML algorithm to obtain multiple matching embedding outputs.

[0067] In operation 44, the embeddings extracted in operation 43 are clustered. For example, the clustering operation may involve determining, for each embedding output, which of a plurality of reference data sets is closest to the embedding output.

[0068] In operation 45, the cluster determined from the labeled data is compared with the cluster from operation 44. For example, operation 45 can be achieved by determining whether each of the multiple matching embedding outputs is closer to the embedding output of the corresponding reference data set than the embedding output of any other reference data set, and if so, the algorithm is verified.

[0069] If the clusters are equivalent, the ML model is considered to have been trained and the algorithm moves to operation 46. Otherwise, the algorithm returns to operation 43 and the ML model is trained further. Therefore, operation 45 of the algorithm determines whether the training of the trainable parameters of the ML model has been validated, and if the training has not been validated, the training of the trainable parameters is repeated.

[0070] In operation 46 of algorithm 40, model training is considered complete. The algorithm (e.g., in the form of trainable parameters of the ML model) can be stored at this stage along with other data, such as anchor embedding / reference data sets and the set of embedding outputs generated by applying the reference channel measurement data to the algorithm.

[0071] Then Algorithm 40 terminates at operation 47.

[0072] Figure 5 This is a flowchart illustrating an algorithm according to an example embodiment, generally indicated by reference numeral 50. Algorithm 50 can be implemented using the online-use module 36 described above.

[0073] Algorithm 50 starts with operation 51 and then moves to operation 52.

[0074] In operation 52, first channel measurement data for a mobile communication system is acquired (e.g., obtained or received) for use with a primary component carrier (PCC) between a device (such as user equipment 12) and a network node (such as base station 14). The first channel measurement data may include channel quality information, channel state information, or similar data.

[0075] In operation 53, the first channel measurement data obtained in operation 52 is provided as input to the algorithm to obtain a first embedding output. The algorithm is a stored training model 54 (e.g., an ML model stored in operation 46 of the algorithm 40 described above).

[0076] In operation 55, the first embedding output generated in operation 53 is compared with the embedding outputs (e.g., "anchor" embeddings) of multiple reference channel measurement data to identify the closest reference channel measurement data to the first component carrier data. The reference / anchor embeddings are obtained from storage (e.g., a database) 56, and may be those stored in operation 46 of the algorithm 40 described above.

[0077] In operation 57, the carrier aggregation strategy of the reference channel identified in operation 55 is used to assign secondary component carriers for communication between the device and network nodes. Then algorithm 50 terminates at operation 58.

[0078] Figure 6 This is a flowchart illustrating an algorithm according to an example embodiment, generally indicated by reference numeral 60.

[0079] Algorithm 60 begins with operation 62, in which multiple training data are obtained (e.g., received, generated, or otherwise obtained).

[0080] The training data set obtained in operation 62 can be represented as: Data set Each data sample in the dataset can be a CSI vector X. CSI This vector is the radio channel condition measured on the PCC of the user equipment.

[0081] With each X CSI Related to this is the tuple (PCC(X) CSI ), SCC1(S CSI ), ..., SCC m (X CSI )), where PCC(X CSI SCC is the identifier (ID) of the primary component carrier. i (X CSI ) is the ID of the i-th secondary component carrier configured for this user equipment. It is assumed here that there are m configured secondary component carriers for each user, and the list is sorted, i.e., the order of the list determines the activation order of the secondary component carriers (although it should be noted that sorting the list is not necessary for all example embodiments). For convenience, the tuple (PCC(X) is used... CSI ), SCC1(S CSI ), ..., SCC m (X CSI )) is represented as CC(X CSI ).

[0082] In operation 64, the data obtained is prepared (for training the ML model). The data can be prepared by the data preparation module 32 described above.

[0083] In operation 66 of algorithm 60, the ML model is trained using the data prepared in operation 64. The trainable parameters of the ML algorithm are trained by minimizing the loss function until, for all matching datasets, the output of the algorithm responding to the matching dataset is closer to the output of the algorithm responding to the corresponding reference dataset than the output of the algorithm responding to any other reference dataset.

[0084] Figure 7 This is a flowchart illustrating an algorithm according to an example embodiment, generally indicated by reference numeral 70. Algorithm 70 is an example implementation of data preparation operation 64 (and can be implemented using the data preparation module 32 described above).

[0085] Algorithm 70 begins with operation 72, in which one or more reference data sets (or “anchor” data) are identified from multiple training data sets obtained in operation 62. Each reference data set has a combination of a primary channel identifier and a secondary channel identifier that is different from all other reference data sets.

[0086] The reference data set or anchor A may include a data set with unique CC ID tuples. All elements. More precisely, if X 1，CSI ≠X 2，CSI , CC(X 1，CSI )≠CC(X 2，CSI In other words, The elements have a unique combination of associated primary component carriers and secondary component carriers.

[0087] In operation 74, one or more sets (or "positive") of the matching datasets are identified from multiple training datasets for each reference dataset. Each matching dataset has the same combination of primary channel identifier and secondary channel identifier as the corresponding reference / anchor dataset.

[0088] The matching dataset (or positive data) can be defined as follows. For each Positive sets It is being identified. In other words, The elements have the same properties as X. CSI Same CC tuple.

[0089] In operation 76, one or more sets of non-matching data sets (or "no") are identified from the training data for each reference data set. Each non-matching data set has a combination of a primary channel identifier and a secondary channel identifier that is different from the corresponding reference data set.

[0090] A non-matching data set (or negative) can be defined as follows. For each Negative set It is identified. In other words, The elements have the same properties as X. CSI Different CC tuples.

[0091] The model trained in Operation 66 seeks to find hidden relationships between the radio channel conditions of the primary component carrier and those of the associated secondary component carriers. (That is, the radio conditions on the primary and secondary component carriers that have been activated together for a particular user equipment in the past.) For example, suppose PCC1 and SCC1 are activated for a user equipment. There may be some relationship between the radio conditions on PCC1 and SCC1 that makes SCC1 the optimal secondary carrier for activation. This relationship is more obvious if PCC1 and SCC1 are in adjacent frequency bands. However, if they are not in adjacent frequency bands, it is difficult to definitively determine what this relationship might be. Therefore, the neural network attempts to find this relationship and represents it using a lower-dimensional embedding. In other words, the neural network tries to find information about the radio conditions on the PCC that makes a particular SCC the optimal choice.

[0092] Figure 8This is a block diagram of a system according to an example embodiment, generally indicated by reference numeral 80. System 80 includes the data preparation module 32, offline training module 34, and online usage module 36 of system 30 described above. System 80 can be used to implement algorithms 40, 50, 60, and 70 described above.

[0093] The data preparation module 32 includes a CSI measurement module 81, a component carrier identifier (CC ID) module 82, an anchor module 83, a positive module 84, and a negative module 85.

[0094] The offline training module 34 includes a ternary network 91, a ternary loss module 92, and an anchor embedding module 93.

[0095] The online usage module 36 includes a CSI measurement module 101, an ML model 102, a clustering module 103, and a carrier aggregation (CA) strategy module 104.

[0096] As described above, algorithm 60 begins with operation 62, in which multiple sets of training data are obtained. The set of training data obtained in operation 62 can be represented as follows: Data set Each data sample in the dataset can be a CSI vector X. CSI This vector represents the radio channel conditions measured on the PCC of the user equipment. This data can be stored in the CSI measurement module 81. The component carrier identifier associated with the CSI measurement can be stored in the CC ID module 82.

[0097] In operation 64 of algorithm 60, the training data is organized into, for example, anchor, positive, and negative data. For example, the training data can be organized by storing anchor data in anchor module 83, positive data in positive module 84, and negative data in negative module 85.

[0098] For the training set For each sample in the dataset, triples are formed (X... CSI ,p(X CSI ), n(X CSI )),in and The triplet network 91 receives each element of the triplet from the anchor module 83, the positive module 84, and the negative module 85.

[0099] The triplet network 91 is trained using the triplet loss function module 92, which is given below:

[0100]

[0101] Here, α > 0 is a hyperparameter that is fine-tuned during training.

[0102] The output of the neural network is used for XCSI The embedding is denoted as E(X). CSI ).

[0103] The set of all anchor embeddings, by Provided and stored in the database (e.g., anchor embedding module 93).

[0104] As described above, in operation 52 of algorithm 50, first channel measurement data for the primary component carrier (PCC) of mobile communication is obtained. This operation can be implemented using the CSI measurement module 101 of the online usage module, which can output data from Z... CSI The CSI measurement is indicated.

[0105] Next, embed E(Z) CSI (Obtained using learning model 102)

[0106] Cluster module 103 can be used to assign the cluster of its nearest anchor, as follows:

[0107]

[0108] Users can be assigned to anchor C(Z) by the CA policy module 104. CSI The same CA strategy.

[0109] Figure 9 This is a block diagram of a system according to an example embodiment, generally indicated by reference numeral 110. System 110 can be used as part of the triple network 91 discussed above.

[0110] System 110 includes three instances of a neural network that implements the training model 102. Therefore, system 110 includes a first neural network 112, a second neural network 114, and a third neural network 116. These neural networks share weights, as schematically shown in system 110. (Of course, the functionality of system 110 can be implemented without requiring three separate instances of neural networks; for example, a single neural network could be used.)

[0111] As mentioned above, for the training set For each sample in the dataset, the triplet (X) CSI ,p(X CSI ), n(X CSI )) is formed, in which and

[0112] The first neural network 112 receives matched / positive data p(X). CSI The second neural network 114 receives relevant reference / anchor data X. CSI Furthermore, the third neural network 116 receives unmatched / negative data n(X). CSI).

[0113] The output of the second neural network 114 provides for X CSI The embedded output, denoted as E(X) CSI As part of the training algorithm described in detail above, system 110 can be used to compare the outputs of the neural network of the reference / anchor data, positive and negative.

[0114] Figure 10 This is a block diagram of a neural network that can be used in some example embodiments, typically indicated by reference numeral 120. For example, the model described above can be a machine learning model that can be implemented using neural networks (such as neural networks 112, 114, and 116).

[0115] The neural network 120 includes an input layer 121, one or more hidden layers 122, and an output layer 123. During the use of the neural network 120, input such as channel state data can be received at the input layer 121. The hidden layer 122 may include multiple hidden nodes, where processing can be performed based on the received input (e.g., channel state data). At the output layer 123, one or more outputs (e.g., embedded data) associated with the input can be provided.

[0116] The neural network 120 can be trained using inputs (including the labeled data described above) (e.g., see the labeled data for operation 42 of algorithm 40). Training can be performed at base station 14 or elsewhere (such as on a server).

[0117] By training a deep neural network using a triplet neural network architecture, the hidden relationship between the CSI / CQI of the primary control center (CC) and the radio conditions on the secondary CC can be learned. Using a triplet neural network with a triplet loss function allows the network to learn embeddings that are useful for grouping user equipment (UEs) in a certain way (e.g., grouping UEs with similar characteristics together). Therefore, UEs with compatible channel conditions on both the primary and secondary CCs can be clustered together.

[0118] For the sake of completeness, Figure 11 This is a schematic diagram of the components of one or more of the foregoing example embodiments, which are collectively referred to below as processing system 300. Processing system 300 may be, for example, the apparatus mentioned in the following claims.

[0119] Processing system 300 may include one or more of the following: processor 302, memory 304 tightly coupled to the processor and including RAM 314 and ROM 312, user input 310 (such as touchscreen input, hardware keys, and / or voice input mechanisms), and display 318 (at least some of these components may be omitted in some example embodiments). Processing system 300 may include one or more network / device interfaces 308 for connecting to a network / device, such as a wired or wireless modem. Interface 308 may also operate as a connection to other devices, such as devices that are not network-side devices. Therefore, direct connections between devices / devices without network involvement are possible.

[0120] The processor 302 is connected to each of the other components in order to control their operation.

[0121] Memory 304 may include non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD). The ROM 312 of memory 304 stores the operating system 315 and may also store software applications 316. The RAM 314 of memory 304 is used by processor 302 to temporarily store data. The operating system 315 may contain code that, when executed by the processor, implements aspects of the algorithms 20, 40, 50, 60, and 70 described above. Note that in the case of small devices / apparatus, memory may be best suited for small size applications, i.e., hard disk drives (HDDs) or solid-state drives (SSDs) are not always used. Memory 304 may include computer program code such that at least one memory 304 and the computer program can be configured to cause the device to execute together with at least one processor 302.

[0122] The processor 302 can take any suitable form. For example, it can be a microcontroller, multiple microcontrollers, a processor, or multiple processors.

[0123] The processing system 300 can be a standalone computer, server, console, or its network. The processing system 300 and the necessary structural components can all be internal to a device (such as an IoT device), i.e., embedded in a very small size.

[0124] In some example embodiments, the processing system 300 may also be associated with external software applications. These may be applications stored on a remote server device / device and may run partially or exclusively on the remote server device / device. These applications may be referred to as cloud-hosted applications. The processing system 300 may communicate with the remote server device / device to utilize the software applications stored there.

[0125] Figure 12A and Figure 12BTangible media are shown, namely a removable storage unit 365 and an optical disc (CD) 368, storing computer-readable code that, when run by a computer, can execute the methods according to the example embodiments described above. The removable storage unit 365 may be a memory stick, such as a USB memory stick, having an internal memory 366 for storing computer-readable code. The internal memory 366 may be accessed by a computer system via a connector 367. The CD 368 may be a CD-ROM or DVD or the like. Other forms of tangible storage media may be used. Tangible media can be any device / apparatus capable of storing data / information that can be exchanged between devices / apparatus / networks.

[0126] Embodiments of this disclosure allow for the discovery of hidden relationships between radio conditions of different CCs. By finding suitable representations for the radio conditions of CCs, similar users (similarity based on proximity of representations in Euclidean space) are clustered together. Each cluster has a primary UE with a known optimal carrier aggregation (CA) strategy and a certain number of active CCs. All UEs in a given cluster will follow the carrier aggregation strategy of the primary UE of that cluster. Since the objective of this invention is to optimally select the secondary CCs to be activated on a given UE without any additional radio measurements from the secondary CCs, both UE and cell throughput can be maximized while reducing latency (because no additional measurements are required). UE battery life will also be improved because the target throughput is expected to be achieved using the minimum number of CCs.

[0127] Embodiments of the present invention can be implemented in software, hardware, application logic, or a combination of software, hardware, and application logic. The software, application logic, and / or hardware can reside on memory or any computer medium. In example embodiments, the application logic, software, or instruction set is maintained on any of a variety of conventional computer-readable media. In the context of this document, "memory" or "computer-readable medium" can be any non-chronological medium or component that can contain, store, communicate, propagate, or transmit instructions for use by or in connection with an instruction execution system, apparatus, or device, such as a computer.

[0128] In relevant contexts, references to “computer-readable medium,” “computer program product,” “tangible computer program,” or “processor” or “processing circuit system” should be understood to include not only computers with different architectures (such as single-processor / multi-processor architectures and sequencer / parallel architectures), but also special-purpose circuits such as field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), signal processing devices / apparatus, and other devices / apparatus. References to computer programs, instructions, code, etc., should be understood to refer to software used in programmable processor firmware, such as programmable content of hardware devices / apparatus as instructions for a processor or for the configuration or configuration settings of fixed-function devices / apparatus, gate arrays, programmable logic devices / apparatus, etc.

[0129] If necessary, the different functions discussed herein can be executed in different orders and / or concurrently with each other. Furthermore, one or more of the functions described above can be optional or can be combined, if needed. Similarly, it will be understood that... Figure 2 and Figures 4 to 7 The flowchart is merely an example, and the various operations described therein can be omitted, reordered, and / or combined.

[0130] It should be understood that the above exemplary embodiments are purely illustrative and do not limit the scope of the invention. Other variations and modifications will be apparent to those skilled in the art after reading this specification.

[0131] Furthermore, the disclosure of this application should be understood to include any novel feature or any novel combination of features or any generalization thereof that is expressly or implicitly disclosed herein, and during the implementation of this application or any application derived therefrom, new claims may be formulated to cover any such feature and / or combination of such features.

[0132] Although various aspects of the invention are set forth in the independent claims, other aspects of the invention include other combinations of features from the described exemplary embodiments and / or dependent claims with features of the independent claims, and not only the combinations expressly listed in the claims.

[0133] It should also be noted in this document that while various examples have been described above, these descriptions should not be considered limiting. Rather, several changes and modifications may be made without departing from the scope of the invention as defined by the appended claims.

Claims

1. An apparatus for communication, comprising components for performing the following: First channel measurement data is obtained for the main component carrier, which is used for communication between devices and network nodes in a mobile communication system. The first channel measurement data is provided as input to the algorithm to obtain the first embedding output; The first embedding output is compared with the embedding output of multiple reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; A set of reference channel information associated with the identified closest reference channel measurement data is identified, wherein the identified set of reference channel information includes a carrier aggregation strategy that defines one or more secondary communication channels; as well as Based on the carrier aggregation strategy of the identified reference channel measurement data set, a carrier aggregation strategy is set for the primary component carrier, for communication between the device and the network node of the mobile communication system.

2. The apparatus of claim 1, wherein the component further performs the function of identifying the closest reference channel measurement data by identifying a reference channel measurement data among the plurality of reference channel measurement data having the closest reference channel measurement data to the first embedding output.

3. The apparatus of claim 1, wherein the algorithm is implemented using a ternary network having multiple parameters that can be trained using a ternary loss function.

4. The apparatus of claim 1, wherein the secondary communication channels of the carrier aggregation strategy are arranged in a sorted list.

5. The apparatus according to any one of the preceding claims, wherein each embedding output obtained from the algorithm is a lower-dimensional representation of the channel measurement data of the corresponding principal component carrier.

6. An apparatus for communication, comprising components for performing the following: Obtain multiple sets of training data, each set of which includes: First channel measurement data for the main component carrier, which is used for communication between devices and network nodes in a mobile communication system; The main channel identifier used for the main component carrier; And a sorted list of secondary channel identifiers for secondary component carriers used for communication between the device and the network node; One or more reference data sets are identified from the plurality of training data sets, wherein each reference data set has a combination of a primary channel identifier and a secondary channel identifier that is different from all other reference data sets; One or more matching data sets are identified for each reference data set from the plurality of training data sets, wherein each matching data set has the same combination of primary channel identifier and secondary channel identifier as the corresponding reference data set; For each reference data set, one or more non-matching data sets are identified from the plurality of training data sets, wherein each non-matching data set has a combination of primary channel identifier and secondary channel identifier that is different from the corresponding reference data set; as well as The trainable parameters of the algorithm are trained by minimizing the loss function until, for all matching datasets, the output of the algorithm responding to the matching dataset is closer to the output of the algorithm responding to the corresponding reference dataset than the output of the algorithm responding to any other reference dataset.

7. The apparatus of claim 6, wherein the loss function is a triplet loss function, and wherein the algorithm is implemented using a triplet network having a plurality of parameters that can be trained using the triplet loss function.

8. The apparatus of claim 7, wherein the loss function is associated with a plurality of triples to train the algorithm, wherein each triple comprises: Channel measurement data of the first reference data set of the one or more reference data sets; Channel measurement data for one of the matching data sets in the corresponding first reference data set; as well as Channel measurement data for one of the non-matching data sets in the corresponding first reference data set.

9. The apparatus of claim 8, wherein the component is further configured to perform: determining whether the training of the trainable parameters of the triplet network is verified by: Channel measurement data for each of a plurality of identified matching datasets is provided as input to the algorithm to obtain a plurality of matching embedding outputs; and Determine whether each of the plurality of matching embedding outputs is closer to the corresponding embedding output of the reference data set than the embedding output of any other reference data set, and if so, determine that the algorithm is verified.

10. The apparatus of claim 6, wherein the component is further configured to perform one or more of the following: Store the algorithm; Store the one or more reference data sets; or A set of embedded outputs is stored, wherein each embedded output is generated by applying first channel measurement data of the main component carrier for the corresponding reference data set to the algorithm.

11. The apparatus of claim 6, wherein the channel measurement data for the primary component carrier or each primary component carrier includes channel quality information, channel state information, or similar data.

12. The apparatus according to any one of claims 6 to 11, wherein the component comprises: At least one processor; as well as At least one memory, including computer program code, the at least one memory and the computer program code being configured to cause the execution of the device together with the at least one processor.

13. A method of communication, comprising: First channel measurement data is obtained for the main component carrier, which is used for communication between devices and network nodes in a mobile communication system. The first channel measurement data is provided as input to the algorithm to obtain the first embedding output; The first embedding output is compared with the embedding output of multiple reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; A set of reference channel information associated with the identified closest reference channel measurement data is identified, wherein the identified set of reference channel information includes a carrier aggregation strategy that defines one or more secondary communication channels; as well as Based on the carrier aggregation strategy of the identified reference channel measurement data set, a carrier aggregation strategy for the primary component carrier is set.

14. A method of communication, comprising: A plurality of training data sets are obtained, each of the plurality of training data sets comprising: first channel measurement data for a primary component carrier used for communication between a device and a network node in a mobile communication system; a primary channel identifier for the primary component carrier; and a sorted list of secondary channel identifiers for a secondary component carrier used for communication between the device and the network node. One or more reference data sets are identified from the plurality of training data sets, wherein each reference data set has a combination of a primary channel identifier and a secondary channel identifier that is different from all other reference data sets; One or more matching data sets are identified for each reference data set from the plurality of training data sets, wherein each matching data set has the same combination of primary channel identifier and secondary channel identifier as the corresponding reference data set; Identify one or more non-matching data sets for each reference data set from the plurality of training data sets, wherein each non-matching data set has a combination of primary channel identifier and secondary channel identifier that is different from the corresponding reference data set; and The trainable parameters of the algorithm are trained by minimizing the loss function until, for all matching datasets, the output of the algorithm responding to the matching dataset is closer to the output of the algorithm responding to the corresponding reference dataset than the output of the algorithm responding to any other reference dataset.

15. A computer program product comprising instructions for causing a device to perform at least the following: First channel measurement data is obtained for the main component carrier, which is used for communication between devices and network nodes in a mobile communication system. The first channel measurement data is provided as input to the algorithm to obtain the first embedding output; The first embedding output is compared with the embedding output of multiple reference channel measurement data input to the algorithm to identify the closest reference channel measurement data relative to the first channel measurement data; A set of reference channel information associated with the identified closest reference channel measurement data is identified, wherein the identified set of reference channel information includes a carrier aggregation strategy that defines one or more secondary communication channels; as well as Based on the carrier aggregation strategy of the identified reference channel measurement data set, a carrier aggregation strategy for the primary component carrier is set.

Images