Information configuration method and apparatus
Patent Information
- Application Number
- PCT/CN2026/086159
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-03-28
- Filing Date
- 2026-03-26
- Publication Date
- 2026-10-01
Smart Images
Figure CN2026086159_01102026_PF_FP_ABST
Abstract
Description
Information configuration method and device
[0001] This application claims priority to Chinese Patent Application No. 202510392143.X, filed with the State Intellectual Property Office of China on March 28, 2025, entitled "Information Configuration Method and Apparatus", the entire contents of which are incorporated herein by reference. Technical Field
[0002] This application relates to the field of communications, and more particularly to an information configuration method and apparatus. Background Technology
[0003] In existing long term evolution (LTE) and new radio (NR) communication systems, network devices need to obtain downlink channel state information (CSI) to determine the resources, modulation and coding scheme (MCS), and precoding configurations of the downlink data channels of terminal devices.
[0004] In Time Division Duplex (TDD) systems, due to the reciprocity of uplink and downlink channels, network devices can obtain the uplink Channel State Information (CSI) by measuring the uplink reference signal and then infer a relatively accurate downlink CSI, for example, using the uplink CSI as the downlink CSI. However, in Frequency Division Duplex (FDD) systems, uplink and downlink reciprocity cannot be guaranteed. Terminal devices can obtain the downlink CSI by measuring the downlink reference signal, such as the Channel State Information Reference Signal (CSI-RS) or the Synchronizing Signal / Physical Broadcast Channel Block (SSB). Furthermore, terminal devices can also generate CSI reports according to predefined protocols or instructions from the network device, feeding the CSI back to the network device so that the network device can obtain the downlink CSI.
[0005] Artificial intelligence (AI) refers to the ability to endow machines with human-like intelligence, such as enabling machines to use computer hardware and software to simulate certain intelligent human behaviors. Currently, AI has been introduced into wireless communication networks and is widely used in many application scenarios of air interface technology, such as CSI feedback, CSI prediction, beam management, and positioning.
[0006] Currently, when applying AI in CSI prediction scenarios, terminal devices typically utilize measurement resources (such as known CSI-RS) to perform model training and obtain an inference model for subsequent CSI inference. However, the aforementioned inference model may not meet the actual inference requirements of network devices. Summary of the Invention
[0007] This application provides an information configuration method and apparatus that can train a reasoning model that meets the reasoning requirements of a second device as much as possible, thereby improving the matching degree between the reasoning model and the reasoning requirements of the second device and improving the accuracy of subsequent predictions.
[0008] To achieve the above objectives, the embodiments of this application adopt the following technical solutions:
[0009] In a first aspect, an information configuration method is provided, which can be executed by a first device (such as a terminal device) performing model training data collection and / or model training, or by a chip or circuit in the first device. The first device may be a device on the terminal device side, a device on the network device side (e.g., a device on the access network or core network (CN) side). The method includes: receiving first information, the first information being used to indicate a candidate set of inference parameters, the type of the inference parameters included in the candidate set of inference parameters including at least one of the following: temporal behavior of a first resource, the number of the first resources, the time interval of the first resources, the number of at least one prediction instance, the time interval between time units corresponding to two adjacent prediction instances in the at least one prediction instance, the time unit in which the first prediction instance in the at least one prediction instance is located, or the type of output result corresponding to the inference of the first model; the first resource corresponds to a measurement resource used for inference of the first model, the at least one prediction instance is a prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit in the at least one prediction instance; and based on the first information, performing training data collection and / or training of the first model.
[0010] Using the above method, the first device can obtain the inference requirements of the second device. In this case, the first device can perform training data collection and / or training of the first model according to the inference requirements of the second device, thereby training an inference model that meets the inference requirements of the second device as much as possible, improving the matching degree between the inference model and the inference requirements of the second device, and helping to improve the accuracy of subsequent predictions.
[0011] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: receiving second information, the second information being used to indicate a second resource, the second resource including measurement resources for training the first model and / or collecting training data for the first model; wherein the temporal behavior of the first resource is different from the temporal behavior of the second resource, and / or the time interval of the first resource is different from the time interval of the second resource.
[0012] It should be understood that the second resource is a measurement resource actually sent by the second device to the first device for training the first model. It is worth noting that, to meet the inference needs of the second device, the first device may not collect model training data solely based on the second resource. For example, after receiving the second information, the first device may use the second resource and candidate values of various types of inference parameters in the candidate inference parameter set indicated by the first information to collect model training data. During the training data collection phase, the first resource may or may not be a resource actually sent by the second device to the first device. For example, the first resource may not be a resource actually sent by the second device; the parameters of the first resource indicated in the first information may be used to determine the training data needed to meet the inference needs of the second device from the second resource.
[0013] In conjunction with the first aspect, in some implementations of the first aspect, the temporal behavior of the second resource includes periodic or semi-persistent behavior.
[0014] In other words, the temporal behavior of the observed instances in the second resource includes periodicity or semi-persistence. However, the inference requirements of the second device may be that the temporal behavior of the observed instances is aperiodic.
[0015] In conjunction with the first aspect, in some implementations of the first aspect, the temporal behavior of the first resource includes aperiodicity.
[0016] For example, the first device may split the observation instances in the second resource or perform other time-domain related operations to obtain the first resource with a non-periodic time-domain behavior, which is conducive to generating an inference model that meets the inference requirements of the second device.
[0017] In conjunction with the first aspect, in some implementations of the first aspect, the type of the inference parameter includes: the time unit in which the first prediction instance is located, the first information including a first time offset, the first time offset being used to indicate the time offset between the time unit in which the first prediction instance is located and a first reference time unit.
[0018] In other words, when the type of the inference parameter includes the time unit where the first predicted instance is located, the first information includes a first time offset, and the first information indicates the time offset between the time unit where the first predicted instance is located and the first reference time unit through the first time offset.
[0019] The first reference time unit can be any time unit where an observation instance is located. The first reference time unit can be indicated by the second device to the first device, or it can be predefined or preconfigured.
[0020] In conjunction with the first aspect, in some implementations of the first aspect, the type of the inference parameter includes: the time unit in which the first prediction instance is located, the first information including a second time offset, the second time offset being used to indicate the time offset between the time unit in which the first channel state information (CSI) report is located and the first reference time unit, the first channel state information (CSI) report corresponding to the inference of the first model.
[0021] In other words, when the type of the inference parameter includes the time unit where the first prediction instance is located, the first information includes a second time offset. The first information indicates the time offset between the time unit where the first channel state information (CSI) report is located and the first reference time unit through the second time offset, thereby indicating the time offset between the time unit where the first prediction instance is located and the first reference time unit.
[0022] The first Channel State Information (CSI) report corresponds to the inference of the first model, and can be understood as the first CSI report being used to report the inference results of the first model. Optionally, the first model is a model used for CSI prediction. For example, for CSI prediction, the second device can indicate the time offset δ between the time unit where the first prediction instance is located and the time unit where the CSI report is located. In this case, the first device can determine the time offset between the time unit where the first prediction instance is located and the first reference time unit based on the time offset between the time unit where the CSI report is located and the first reference time unit, as well as the time offset δ. The first reference time unit may be the time unit where any one of the multiple observation instances corresponding to one inference is located.
[0023] In conjunction with the first aspect, in some implementations of the first aspect, the first resource includes at least one measurement resource for inference of the first model, and the first reference time unit includes the time unit in which any one of the at least one measurement resource is located.
[0024] Specifically, the first reference time unit may be any one of the observation instances among the multiple observation instances corresponding to one inference. Given a fixed number of observation instances and a fixed time interval between adjacent observation instances, the first device can determine the time offset between the time unit of the first predicted instance and the latest (newest) observation instance among the multiple observation instances corresponding to one inference by using the time offset between the time unit of the first predicted instance and the time unit of any one observation instance. For example, any one observation instance is the latest (newest) observation instance among the multiple observation instances corresponding to one inference. For instance, if the first model requires P observation instances to perform one inference, optionally, the first reference time unit may be the time unit of the latest (newest) observation instance among the P observation instances, and the latest observation instance is the latest observation instance in the corresponding time unit among the P observation instances. In this case, the first time offset can indicate the time offset between the time unit of the first predicted instance and the time unit of the latest (newest) observation instance among the multiple observation instances corresponding to one inference. The first device can train a model that meets the inference requirements of the second device based on this first time offset.
[0025] In conjunction with the first aspect, in some implementations of the first aspect, the number of candidate values of the same type of inference parameter included in the candidate inference parameter set is one or more.
[0026] In this scenario, the first device can utilize the same measurement resources for training the model and / or collecting training data for the model among multiple candidate values, thereby saving training resources.
[0027] In conjunction with the first aspect, in some implementations of the first aspect, the candidate inference parameter set includes multiple sets of candidate inference parameters.
[0028] For example, the second device can indicate multiple sets of candidate inference parameters through the first information, and each set of candidate inference parameters may include different types of inference parameters. In this case, the first device can utilize multiple sets of candidate inference parameters to perform model training data collection and / or model training, which is beneficial for satisfying the various different inference preferences of the second device and saving training resource overhead.
[0029] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: sending third information, the third information being used to indicate whether the ability to train multiple candidate values for the same type of inference parameter is available.
[0030] Through the third information, the second device can determine whether the first device has the ability to process multiple candidate values of the same type of inference parameter. For example, the third information indicates that the first device has the ability to train on multiple candidate values of the same type of inference parameter, which can be understood as the first device having the ability to simultaneously train on multiple candidate values of the same type of inference parameter using a single measurement resource. For example, if the first device has this capability, the second device indicates through the first information that the quantity of the first resource may be a set of values, such as {2, 4, 7}. In one possible implementation, the first device can use the aforementioned multiple values to perform model training data collection and / or model training. The second device does not need to independently configure a measurement resource for each value of the quantity of the first resource, which helps to save measurement resources.
[0031] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: sending fourth information, the fourth information being used to indicate at least one set of candidate inference parameters among the supported plurality of candidate inference parameters.
[0032] In this situation, since the second device can know the processing capability of the first device for inference parameters, the second device can indicate the candidate inference parameters that meet the capability of the first device through the first information. The first device can train a model that meets the inference requirements of the second device without exceeding its own capability.
[0033] In conjunction with the first aspect, in some implementations of the first aspect, the type of the output result includes at least one of the following: Reference Signal Received Power (RSRP), Resource Identifier of Target Reference Signal, Precoding Matrix Indicator (PMI), or Channel Quality Indicator (CQI).
[0034] RSRP is a metric for measuring the signal strength at the receiver. It represents the power of the reference signal transmitted by the second device at the receiver and is used to evaluate the quality of the wireless signal. The resource identifier of the target reference signal typically refers to the identifier of a specific reference signal resource (e.g., the optimal beam or optimal beam candidate). PMI can be used to describe the matrix used by the second device during precoding in a multi-antenna system, helping the first device to report the spatial characteristics of the received signal. CQI is a channel quality metric fed back from the first device to the second device, used to evaluate the channel's signal-to-noise ratio or channel capacity. During the model inference phase, the output results may include various types mentioned above. Selecting the appropriate type of output for different application scenarios is beneficial for optimizing resource allocation and utilization, and improving the performance of the communication system.
[0035] In conjunction with the first aspect, in some implementations of the first aspect, the first information is further used to indicate that the type of the output result includes the resource identifier of the target reference signal, and the type of the inference parameter further includes the number of resource identifiers of the target reference signal.
[0036] The type of output result corresponding to inference is related to the training algorithm. The second device indicates the required type of output result during the training data collection phase, which can help the first device train a model that meets the inference requirements of the second device and avoid wasting training resources.
[0037] In conjunction with the first aspect, in some implementations of the first aspect, the first information further includes the identifier of at least one set of candidate inference parameters among the plurality of candidate inference parameters.
[0038] For example, for both the second and first devices, the multiple sets of candidate inference parameters and their corresponding identifiers can be known information. For instance, the protocol predefines multiple sets of optional inference parameters and their corresponding identifiers, or the second device pre-indicates multiple sets of optional inference parameters and their corresponding identifiers, or the first device pre-reports multiple sets of optional inference parameters and their corresponding identifiers. Thus, the second device can select preferred inference parameters from the multiple sets of optional inference parameters for indication. For example, the second device can use the first information to indicate the identifiers of at least one set of candidate inference parameters. The first device can obtain the corresponding at least one set of candidate inference parameters based on this identifier, thereby performing model training data collection and / or model training.
[0039] In conjunction with the first aspect, in certain implementations of the first aspect, the temporal behavior of the first resource includes at least one of the following: periodic, semi-persistent, or aperiodic.
[0040] For example, the temporal behavior of the first resource is aperiodic. In this case, the temporal behavior of the measurement resource configured by the second device may be any one or more of the above. During the model training data collection phase, the first device can divide the measurement results corresponding to the measurement resource into several groups based on the first information. Each group can be considered to contain several aperiodic measurement results. Each group of measurement results is used as the input data for training, thereby satisfying the reasoning requirements of the second device related to temporal behavior.
[0041] In conjunction with the first aspect, in some implementations of the first aspect, the first information may also include the value or range of the inference parameter.
[0042] In addition, the values or ranges of candidate inference parameters may be predefined by the protocol or reported by the first device. The second device can select the values or ranges of candidate inference parameters that meet the inference requirements and use the first information to give instructions.
[0043] Secondly, another information configuration method is provided, which can be executed by a second device that sends measurement resources or related configurations to the first device, or by a chip or circuit in the second device. The second device may be a device on the terminal device side, a device on the network device side, or a device on the core network (CN) side. The method includes: sending first information, the first information being used to indicate a candidate set of inference parameters, the type of the inference parameters included in the candidate set of inference parameters including at least one of the following: the temporal behavior of a first resource, the number of the first resources, the time interval of the first resources, the number of at least one prediction instance, the time interval between time units corresponding to two adjacent prediction instances in the at least one prediction instance, the time unit in which the first prediction instance in the at least one prediction instance is located, or the type of output result corresponding to the inference of the first model; the first resource corresponds to a measurement resource used for inference of the first model, the at least one prediction instance is a prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit among the at least one prediction instance.
[0044] In conjunction with the second aspect, in some implementations of the second aspect, the method further includes: sending second information, the second information being used to indicate a second resource, the second resource including measurement resources for training the first model and / or collecting training data for the first model; the temporal behavior of the first resource and the temporal behavior of the second resource are different, and / or the time interval of the first resource and the time interval of the second resource are different.
[0045] In conjunction with the second aspect, in some implementations of the second aspect, the method further includes: receiving third information, the third information being used to indicate whether the first device has the capability to support training for multiple candidate values of the same type of inference parameters.
[0046] Thirdly, another information configuration method is provided, which can be executed by a first device performing model inference and / or applicability reporting, or by a chip or circuit in the first device. The method includes: receiving fifth information, the fifth information indicating a range of a third time offset or M values of the third time offset, where M is a positive integer, the third time offset reflecting a time offset between the time unit of a first prediction instance in at least one prediction instance associated with the first task and the time unit of a first observation instance in at least one observation instance associated with the first task; and sending sixth information, the sixth information indicating the applicability of the first task.
[0047] In this scenario, when the second device requests the applicability of a reasoning task from the first device, the first device can determine whether the reasoning task exceeds the model's reasoning capabilities, which helps the first device determine the applicability of the first task.
[0048] In conjunction with the third aspect, in some implementations of the third aspect, the third time offset indicates the time offset between the time unit of at least one CSI report associated with the first task and the second reference time unit, or the time unit of the first prediction instance in at least one prediction instance associated with the first task and the first task can be a CSI reporting task.
[0049] For example, the first task could be a CSI reporting task for inference. A CSI reporting task for inference can be understood as a CSI reporting task that uses an artificial intelligence (AI) model for inference. The time offset between the second reference time units is described.
[0050] In conjunction with the third aspect, in some implementations of the third aspect, the second reference time unit includes any one of the following: the time unit in which the first observation instance in at least one observation instance corresponding to the first task is located; or, the time unit in which the downlink control information (DCI) is located, the DCI being used to trigger the reporting of at least one CSI report associated with the first task.
[0051] For example, the second reference time unit can be any time unit containing any observation instance. For instance, it could be the time unit containing the first observation instance or the time unit containing the latest observation instance. For example, the third time offset indicates the time offset between the time unit containing any predicted instance associated with the first task and the second reference time unit. Based on the third time offset, the first device can determine the time offset between the time unit containing the first predicted instance associated with the first task and the time unit containing the latest observation instance, thereby determining the applicability of the first task. As another example, the third time offset indicates the time offset between the time unit containing at least one CSI report associated with the first task and the second reference time unit. Based on the third time offset and δ, the first device can determine the time offset between the first predicted instance associated with the first task and the time unit containing the latest observation instance, thereby determining the applicability of the first task.
[0052] In conjunction with the third aspect, in some implementations of the third aspect, the second reference time unit is predefined or preconfigured.
[0053] In this situation, it is beneficial to reduce signaling overhead.
[0054] In conjunction with the third aspect, in some implementations of the third aspect, M is greater than or equal to 2.
[0055] In this scenario, the first task may correspond to multiple inference requirements. When the second device uses the fifth information to indicate the multiple inference requirements of the first task to the first device and requests an applicability report, in one possible implementation, the first device reports that the first task is applicable if its model can support all selectable values in the multiple third time offsets; otherwise, the first device reports that it is not applicable. In another possible implementation, the first device may report the applicability of all selectable values in the multiple third time offsets separately.
[0056] In conjunction with the third aspect, in some implementations of the third aspect, the sixth information includes an applicability result, which corresponds to the range of the third time offset or M values of the third time offset, and the applicability result is determined based on the largest time offset value among the third time offsets.
[0057] In this scenario, the first device determines the maximum value of the third time offset based on its range or M possible values, and then determines the applicability based on this maximum value. If the maximum value of the third time offset does not exceed the inference capability of the first device, the applicability result is considered applicable; otherwise, if the maximum value exceeds the inference capability, the applicability result is considered inapplicable. In this case, the first device can use a single applicability result to indicate the applicability of multiple inference tasks, or the applicability of multiple inference requirements within a single inference task, which helps reduce signaling overhead.
[0058] In conjunction with the third aspect, in some implementations of the third aspect, the fifth information includes M values of the third time offset, and the sixth information includes M applicability results, with each of the M applicability results corresponding to one of the M values.
[0059] For example, the requirement for the second device is that the first device performs CSI reporting tasks across multiple time units. The first device can determine and report multiple applicability results corresponding to multiple third time offsets, respectively. In this case, a flexible option is provided for the first device to perform applicability result reporting.
[0060] In conjunction with the third aspect, in some implementations of the third aspect, the sixth information is also used to indicate a fourth time offset, wherein if the third time offset is less than or equal to the fourth time offset, then the first task applies.
[0061] As previously mentioned, the third time offset may include multiple selectable values (indicated by M values). In one possible design, for the multiple selectable values of the third time offset, some values are applicable and some are not. The sixth information can indicate the fourth time offset, so the second device can know that all selectable values less than or equal to the fourth time offset are applicable, and all selectable values greater than the fourth time offset are not applicable. Optionally, the fourth time offset is greater than or equal to the minimum value among the multiple selectable values of the third time offset, and less than or equal to the maximum value among the multiple selectable values of the third time offset. In this case, the first device does not need to report applicability information for each selectable value, which can reduce the reporting overhead of the first device. In another possible design, for the multiple selectable values of the third time offset, all are not applicable. In addition to indicating that the multiple selectable values of the third time offset are not applicable, the sixth information can also indicate the fourth time offset. Then the second device can know that all values less than or equal to the fourth time offset are applicable, and all values greater than the fourth time offset are not applicable. The second device can redetermine a time offset for inference reporting based on the fourth time offset. Optionally, the fourth time offset may be less than the minimum of several selectable values of the third time offset.
[0062] In conjunction with the third aspect, in some implementations of the third aspect, the first task is a dynamically scheduled task.
[0063] In other words, the first task is dynamically triggered based on the instruction information (such as DCI) from the second device. For example, the first task may be an aperiodic CSI reporting task. The first task may also be other inference tasks.
[0064] In conjunction with the third aspect, in some implementations of the third aspect, the first task is a CSI prediction task.
[0065] In other words, the first task is to submit the CSI report corresponding to the CSI prediction results.
[0066] Fourthly, another information configuration method is provided, which can be executed by a second device that sends an applicability request and other related configurations to the first device, or by a chip or circuit in the second device. The method includes: sending fifth information, the fifth information indicating a range of a third time offset or M values of the third time offset, where M is a positive integer, the third time offset reflecting the time offset between the time unit of the first prediction instance in at least one prediction instance associated with the first task and the time unit of the first observation instance in at least one observation instance associated with the first task; and receiving sixth information, the sixth information indicating the applicability of the first task.
[0067] Fifthly, an information configuration apparatus is provided for executing the method in any one of the possible implementations of the first, second, third, or fourth aspects described above. Specifically, the apparatus includes a unit / module for executing the method in any one of the possible implementations of the first, second, third, or fourth aspects described above.
[0068] Sixthly, this application provides yet another information configuration device, including a processor coupled to a memory, which can be used to execute instructions in the memory to implement the methods in any of the possible implementations of the first, second, third, or fourth aspects described above. Optionally, the information configuration device further includes a memory. Optionally, the information configuration device further includes a communication interface, and the processor is coupled to the communication interface.
[0069] In one implementation, the information configuration device is a first device. When the information configuration device is a first device, the communication interface can be a transceiver or an input / output interface.
[0070] In another implementation, the information configuration device is a chip configured in the first device. When the information configuration device is a chip configured in the first device, the aforementioned communication interface can be an input / output interface.
[0071] In one implementation, the information configuration device is a second device. When the information configuration device is a second device, the communication interface can be a transceiver or an input / output interface.
[0072] In another implementation, the information configuration device is a chip configured in the second device. When the information configuration device is a chip configured in the second device, the aforementioned communication interface can be an input / output interface.
[0073] A seventh aspect provides a processor, comprising: an input circuit, an output circuit, and a processing circuit. The processing circuit is configured to receive signals through the input circuit and transmit signals through the output circuit, causing the processor to execute the method in any one of the possible implementations of the first, second, third, or fourth aspects described above.
[0074] In specific implementation, the processor can be a chip, the input circuit can be an input pin, the output circuit can be an output pin, and the processing circuit can be a transistor, gate circuit, flip-flop, and various logic circuits. The input signal received by the input circuit can be received and input by, for example, but not limited to, a receiver, and the signal output by the output circuit can be output to, for example, but not limited to, a transmitter and transmitted by the transmitter. Furthermore, the input circuit and the output circuit can be the same circuit, which is used as the input circuit and the output circuit at different times. This application does not limit the specific implementation of the processor and various circuits.
[0075] Eighthly, a processing apparatus is provided, including a processor and a memory. The processor is configured to read instructions stored in the memory and to receive signals via a receiver and transmit signals via a transmitter to execute the methods in any of the possible implementations of the first, second, third, or fourth aspects described above.
[0076] Optionally, the processor may be one or more, and the memory may be one or more.
[0077] Optionally, the memory may be integrated with the processor, or the memory may be separated from the processor.
[0078] In specific implementation, the memory can be a non-transitory memory, such as read-only memory (ROM), which can be integrated with the processor on the same chip or set on different chips. The embodiments of this application do not limit the type of memory or the way the memory and processor are set.
[0079] It should be understood that the relevant data interaction process, such as sending indication information, can be the process of outputting indication information from the processor, and receiving capability information can be the process of the processor receiving input capability information. Specifically, the processed output data can be output to the transmitter, and the input data received by the processor can come from the receiver. Here, the transmitter and receiver can be collectively referred to as a transceiver.
[0080] The processing device in the eighth aspect can be a chip. The processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in a memory. The memory can be integrated into the processor or located outside the processor and exist independently.
[0081] A ninth aspect provides a communication system, the communication apparatus comprising a first apparatus and a second apparatus. In one possible implementation, the first apparatus is configured to perform the method in any possible implementation of the first aspect, and the second apparatus is configured to perform the method in any possible implementation of the second aspect. In another possible implementation, the first apparatus is configured to perform the method in any possible implementation of the third aspect, and the second apparatus is configured to perform the method in any possible implementation of the fourth aspect.
[0082] In a tenth aspect, a computer program product is provided, the computer program product comprising: a computer program (also referred to as code or instructions), which, when the computer program is run, causes a computer to perform the method in any of the possible implementations of the first, second, third, or fourth aspects described above.
[0083] Eleventhly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program (also referred to as code or instructions) that, when run on a computer, causes the computer to perform the methods in any of the possible implementations of the first, second, third, or fourth aspects described above. Attached Figure Description
[0084] Figure 1 is a schematic diagram of the architecture of a communication system to which the information configuration method provided in the embodiments of this application is applicable;
[0085] Figure 2 is a schematic diagram of the architecture of another communication system to which the information configuration method provided in the embodiments of this application is applicable;
[0086] Figure 3 is a schematic diagram of a possible application architecture in a communication system to which the information configuration method provided in the embodiments of this application is applicable;
[0087] Figure 4 is a schematic diagram of another possible application architecture in the communication system to which the information configuration method provided in the embodiments of this application is applicable;
[0088] Figure 5 is a schematic diagram of a possible neuron structure in a deep neural network;
[0089] Figure 6 is a schematic diagram of a possible structure of a deep neural network;
[0090] Figure 7 shows a possible scenario for spatial beam prediction.
[0091] Figure 8 shows a possible scenario for time-domain beam prediction.
[0092] Figure 9 shows a possible interaction flowchart between terminal devices and network devices;
[0093] Figure 10 is a schematic diagram of an information configuration method provided in an embodiment of this application;
[0094] Figure 11 is a schematic diagram of a possible inference model for CSI prediction;
[0095] Figure 12 is a schematic diagram of another information configuration method provided in an embodiment of this application;
[0096] Figure 13 is a schematic diagram of a possible reasoning scenario provided by an embodiment of this application;
[0097] Figure 14 is a schematic block diagram of an information configuration device provided in an embodiment of this application;
[0098] Figure 15 is a schematic block diagram of another information configuration device provided in an embodiment of this application;
[0099] Figure 16 is a schematic block diagram of another information configuration device provided in an embodiment of this application. Detailed Implementation
[0100] The technical solutions in this application will now be described with reference to the accompanying drawings.
[0101] The technical solutions of this application embodiment can be applied to various communication systems, such as narrowband Internet of Things (NB-IoT) systems, wireless fidelity (WiFi) systems, vehicle-to-everything (V2X) communication systems, device-to-device (D2D) communication systems, vehicle-to-everything (V2X) communication systems, 4th generation (4G) mobile communication systems, 5th generation (5G) or new radio (NR) systems, long term evolution (LTE) systems, LTE frequency division duplex (FDD) systems, LTE time division duplex (TDD) systems, wireless local area network (WLAN) systems, satellite communication systems, future communication systems such as 6th generation (6G) mobile communication systems, or integrated systems of multiple systems. The technical solutions of this application embodiment can also be applied to machine-to-machine (M2M) communication, machine-type communication (MTC), Internet of Things (IoT) communication systems, or other communication systems.
[0102] In a communication system, a network element can send signals to or receive signals from another network element. These signals can include information, signaling, or data. The term "network element" can also be replaced by an entity, network entity, device, communication device, communication module, node, communication node, etc. This application describes the concept of a network element as an example. For instance, a communication system can include at least one terminal device and at least one network device. The network device can send downlink signals to the terminal device, and / or the terminal device can send uplink signals to the network device. It is understood that in this application embodiment, the terminal device can be replaced by a first network element, and the network device can be replaced by a second network element.
[0103] In wireless communication networks, such as mobile communication networks, the services supported by the networks are becoming increasingly diverse, thus requiring increasingly diverse demands. For example, networks need to support ultra-high speeds, ultra-low latency, and / or massive connectivity. This characteristic makes network planning, network configuration, and / or resource scheduling increasingly complex. Furthermore, as network functions become more powerful, such as supporting higher spectrum levels, supporting higher-order multiple-input multiple-output (MIMO) technologies, supporting beamforming, and / or supporting beam management, network energy efficiency has become a hot research topic. These new demands, new scenarios, and new characteristics bring unprecedented challenges to network planning, operation, and efficient operation. To meet these challenges, artificial intelligence (AI) technology can be introduced into wireless communication networks to achieve network intelligence. To support AI technology in wireless networks, AI nodes may also be introduced.
[0104] It should be understood that, in the embodiments of this application, the terminal device may be user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication equipment, user agent, or user device.
[0105] Terminal devices can be devices that provide voice / data, such as handheld devices with wireless connectivity, in-vehicle devices, etc. Currently, examples of terminals include: mobile phones, tablets, laptops, PDAs, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication capabilities, computing devices or other processing devices connected to a wireless modem, wearable devices, terminal devices in 5G networks, or future public land mobile communication networks. Terminal devices in a network (PLMN), etc., are not limited to this in the embodiments of this application.
[0106] In this embodiment, the terminal device can also be a wearable device. Wearable devices, also known as wearable smart devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. For example, a wearable device can be a hardware device that achieves multiple functions through software support, data interaction, and cloud interaction. Broadly defined, wearable smart devices can include devices that are feature-rich, large in size, and can achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses. Broadly defined, wearable smart devices can also include devices that focus on a specific type of application function and require cooperation with other devices (such as smartphones), such as various smart bracelets and smart jewelry for vital sign monitoring.
[0107] In this embodiment, the device for implementing the functions of the terminal device can be the terminal device itself, or it can be any device capable of supporting the terminal device in implementing those functions, such as a chip system. This device can be installed in or used in conjunction with the terminal device. In this embodiment, the chip system can be composed of chips or may include chips and other discrete components. This embodiment only uses the terminal device as an example to illustrate the device for implementing the functions of the terminal device, and does not constitute a limitation on the solution of this embodiment.
[0108] The network device in this application embodiment can be a device for communicating with a terminal device, or it can be a radio access network (RAN) element, such as a base station, that connects the terminal device to the wireless network. A base station can broadly encompass, or be replaced by, various names including: NodeB, evolved NodeB (eNB), next-generation NodeB (gNB), relay station, access point, transmitting and receiving point (TRP), transmitting point (TP), master station, auxiliary station, motor slide retainer (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. A base station can be a macro base station, micro base station, relay node, donor node, or similar entities, or combinations thereof. A base station can also refer to a communication module, modem, or chip installed within the aforementioned equipment or apparatus. A base station can also be a mobile switching center, a device that performs base station functions in D2D, V2X, and M2M communications, or a device that performs base station functions in future communication systems. A base station can support networks using the same or different access technologies. Optionally, access network elements can also be servers, wearable devices, vehicles, or in-vehicle equipment. For example, the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU). The embodiments of this application do not limit the specific technologies or equipment forms used in the network devices.
[0109] The base station in this embodiment can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move according to the location of the mobile base station. In other examples, a helicopter or drone can be configured as a device to communicate with another base station.
[0110] The network device in this application embodiment can be a device including a CU, or a DU, or a device including both CU and DU, or a device with a control plane CU node (central unit-control plane (CU-CP)) and a user plane CU node (central unit-user plane (CU-UP)) and a DU node. For example, the network device may include gNB-CU-CP, gNB-CU-UP, and gNB-DU.
[0111] In some deployments, multiple access network elements collaborate to assist terminals in achieving wireless access, with different access network elements each implementing some of the base station's functions. For example, access network elements can be CU, DU, CU-CP, CU-UP, or RU, etc. CU and DU can be configured separately or included in the same network element, such as in a BBU. RU can be included in radio frequency equipment or radio frequency units, such as in an RRU, AAU, or RRH.
[0112] Access network elements can support one or more types of fronthaul interfaces, with different types of fronthaul interfaces corresponding to DUs and RUs with different functions. If the fronthaul interface between the DU and RU is a common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and RU is another type of interface, relative to the CPRI, some downlink and / or uplink baseband functions, such as, for downlink, precoding, digital beamforming (BF), or one or more of inverse fast Fourier transform (IFFT) / cyclic prefix addition (CP), are moved from the DU to the RU for implementation; and for uplink, digital beamforming (BF), or one or more of fast fourier transform (FFT) / cyclic prefix removal (CP), are moved from the DU to the RU for implementation. In one possible implementation, the interface can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the segmentation between DU and RU differs, corresponding to different categories (Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, and F.
[0113] Taking eCPRI Cat A as an example, for downlink transmission, the DU is configured to implement one or more functions before and after layer mapping (i.e., coding, rate matching, scrambling, modulation, and layer mapping), while other functions after layer mapping (e.g., resource element (RE) mapping, digital beamforming (BF), or one or more functions of inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP)) are moved to the RU. For uplink transmission, the DU is configured to implement one or more functions before and after de-mapping (i.e., decoding, rate matching de-mapping, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and de-RE mapping), while other functions after de-mapping (e.g., digital BF or one or more functions of fast Fourier transform (FFT) / removing CP) are moved to the RU. It is understandable that the functional descriptions of the DU and RU corresponding to various types of eCPRI can be found in the eCPRI protocol, and will not be repeated here.
[0114] In one possible design, the processing unit in the BBU used to implement baseband functions is called the baseband high (BBH) unit, and the processing unit in the RRU / AAU / RRH used to implement baseband functions is called the baseband low (BBL) unit.
[0115] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, the meaning of which will be understood by those skilled in the art. For example, in an open radio access network (ORAN) system, CU may also be called an open CU (O-CU), DU may also be called an open DU (O-DU), CU-CP may also be called an open CU-CP (O-CU-CP), CU-UP may also be called an open CU-CP (O-CU-UP), and RU may also be called an open RU (O-RU). Any of the units among CU (e.g., CU-CP, CU-UP), DU, and RU in the embodiments of this application may be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.
[0116] In this embodiment, the apparatus for implementing the functions of a network device can be a network device itself; it can also be an apparatus capable of supporting the network device in implementing those functions, such as a chip system, hardware circuit, software module, or a hardware circuit plus a software module. This apparatus can be installed in the network device or used in conjunction with the network device. In this embodiment, the example of a network device being used to implement the functions of a network device is provided only and does not constitute a limitation on the solutions described in this embodiment.
[0117] Network devices and / or terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can also be deployed in the air on airplanes, balloons, and satellites. This application does not limit the scenario in which the network devices and terminal devices are located. Furthermore, terminal devices and network devices can be hardware devices, or software functions running on dedicated hardware or general-purpose hardware. For example, virtualization functions instantiated on a platform (e.g., a cloud platform), or entities comprising dedicated or general-purpose hardware devices and software functions. This application does not limit the specific form of the terminal devices and network devices.
[0118] This application will present various aspects, embodiments, or features relating to systems that may include multiple devices, components, modules, etc. It should be understood and appreciated that individual systems may include additional devices, components, modules, etc., and / or may not include all the devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these approaches are also possible.
[0119] Furthermore, in the embodiments of this application, words such as "exemplarily" and "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as an "example" in this application should not be construed as being better or more advantageous than other embodiments or designs. Rather, the use of the word "example" is intended to present the concept in a specific manner.
[0120] In this application, "instruction" can include direct instruction, indirect instruction, explicit instruction, and implicit instruction. When describing a certain instruction information for the purpose of instructing A, it can be understood that the instruction information carries A, directly instructs A, or indirectly instructs A.
[0121] In this embodiment, the information indicated by the instruction information is called the information to be instructed. In specific implementations, there are many ways to indicate the information to be instructed, such as, but not limited to, directly indicating the information to be instructed, such as the information to be instructed itself or its index. It can also indirectly indicate the information to be instructed by indicating other information, where there is a correlation between the other information and the information to be instructed. It can also indicate only a part of the information to be instructed, while the other parts are known or pre-agreed upon. For example, the instruction of specific information can be achieved by using a pre-agreed (e.g., protocol-defined) arrangement of various information, thereby reducing instruction overhead to some extent. Furthermore, the information to be instructed can be sent as a whole or divided into multiple sub-information units, and the sending period and / or timing of these sub-information units can be the same or different.
[0122] In the embodiments of this application, "send" and "receive" indicate the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information being XX, which may include direct transmission via the air interface or indirect transmission by other units or modules via the air interface. "Receive information from YY" can be understood as the source of the information being YY, which may include direct reception from YY via the air interface or indirect reception from YY by other units or modules via the air interface. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface. In other words, sending and receiving can occur between devices, such as between network devices and terminal devices, or within a device, such as between components, modules, chips, software modules, or hardware modules within the device via a bus, wiring, or interface.
[0123] In the embodiments of this application, unless otherwise specified or in case of logical conflict, the terms and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.
[0124] In the embodiments of this application, the terms "first," "second," and "#1," "#2," etc., are merely for descriptive convenience and are used to distinguish objects, and are not intended to limit the scope of the embodiments of this application. They are not used to describe the order or sequence of features. It should be understood that such described objects can be interchanged where appropriate to describe solutions other than those in the embodiments of this application.
[0125] In this application, "predefined" may mean a standard protocol predefined, or it may mean that the devices have agreed or negotiated in advance.
[0126] In the embodiments of this application, the terms "of," "corresponding (relevant)," and "corresponding" can sometimes be used interchangeably. It should be noted that when their distinction is not emphasized, their intended meanings are consistent. Furthermore, "corresponding to" in this application can also be replaced with "as," "determined according to xx," or "used to determine." Similarly, "including" in this application can also be replaced with "as" or "is."
[0127] The network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0128] For ease of understanding, the communication system applicable to the embodiments of this application will be described in detail first with reference to Figures 1 to 4. For example, Figure 1 is a schematic diagram of the architecture of a communication system 100 to which the information configuration method provided in the embodiments of this application applies.
[0129] As shown in Figure 1, the communication system 100 may include at least one network device, such as network device 110. The communication system 100 may also include at least one terminal device, such as terminal device 120 and terminal device 130. Network device 110 and terminal devices (such as terminal device 120 and / or terminal device 130) can communicate via a wireless link. The communication devices in this communication system, for example, network device 110 and terminal device 120, can communicate using multi-antenna technology.
[0130] Figure 2 is a schematic diagram of the architecture of another communication system 200 to which the information configuration method provided in this application embodiment applies. Figure 2 is similar to Figure 1, and the same parts will not be described again. The difference is that the communication system 200 shown in Figure 2 also includes an AI network element 240. The AI network element 240 is used to perform AI-related operations, such as building training datasets or training AI models.
[0131] For example, in communication system 200, network device 210 can send data related to the training of an AI model to AI network element 240, which then constructs a training dataset and trains the AI model. For example, the data related to the training of the AI model may include data reported by terminal devices. AI network element 240 can send the results of operations related to the AI model to network device 210, and then forward them to terminal devices (such as terminal devices 220 and / or 230) via network device 210. The results of the aforementioned operations related to the AI model may include at least one of the following: a trained AI model, model evaluation results, or test results, etc. Exemplarily, a portion of the trained AI model may be deployed on network device 210, and another portion on terminal devices (such as terminal devices 220 and / or 230). Exemplarily, the trained AI model may be deployed on network device 210. Alternatively, the trained AI model may be deployed on terminal devices (such as terminal devices 220 and / or 230).
[0132] It should be understood that Figure 2 is only used as an example of AI network element 240 being directly connected to network device 210. In other scenarios, AI network element 240 can also be connected to terminal devices (such as terminal device 220 and / or terminal device 230). Alternatively, AI network element 240 can be connected to both network device 210 and terminal devices (such as terminal device 220 and / or terminal device 230) simultaneously. Alternatively, AI network element 140 can also be connected to network device 210 through a third-party network element. This application embodiment does not limit the connection relationship between AI network element and other network elements.
[0133] Although not shown in Figure 2, AI network element 240 can also be set as a module in network devices and / or terminal devices. For example, AI network element 240 can also be deployed in network device 210 and / or terminal devices (such as terminal device 220 and / or terminal device 230).
[0134] It should be noted that Figures 1 and 2 are simplified schematic diagrams for ease of understanding. For example, the communication system may also include other devices, such as wireless relay devices and / or wireless backhaul devices, which are not shown in Figures 1 and 2. In practical applications, the communication system may include multiple network devices or multiple terminal devices. The embodiments of this application do not limit the number of network devices and terminal devices included in the communication system.
[0135] In this embodiment, the communication system can further deploy AI nodes to achieve corresponding functions. The AI nodes can be deployed in one or more of the following locations within the communication system: access network elements, terminal devices, or core network elements, etc. Alternatively, the AI nodes can be deployed independently, for example, in a location other than any of the aforementioned devices, such as in the host or cloud server of an over-the-top (OTT) system. The AI nodes can communicate with other devices in the communication system, which can be one or more of the following: access network elements, terminal devices, or core network elements, etc.
[0136] It should be understood that this application does not limit the number of AI nodes. For example, when there are multiple AI nodes, these nodes can be divided based on function, such as different AI nodes being responsible for different functions.
[0137] It should also be understood that an AI node can be a standalone device, or it can be integrated into the same device to implement different functions. It can also be a network element within a hardware device, a software function running on dedicated hardware, or a virtualized function instantiated on a platform (e.g., a cloud platform). This application does not limit the specific form of the AI node described above. For example, an AI node can be an AI network element or an AI module. For ease of understanding, this application uses an AI module as an example to describe the application architecture of the communication system to which the information configuration method provided in this application is applicable.
[0138] Figure 3 is a schematic diagram of a possible application architecture in a communication system to which the information configuration method provided in this application is applicable. Figure 3 includes multiple network elements, such as core network elements, access network elements, terminal devices, or one or more devices in operations, administration, and maintenance (OAM). In Figure 3, each of the above network elements may contain an AI module. In practical application scenarios, the number of AI modules in each network element can be one or more, and this application embodiment does not limit this. As shown in Figure 3, the various network elements in the communication system can be connected through interfaces (e.g., NG or Xn) or air interfaces. For the access network node shown in Figure 3, CU and DU can be deployed together or separately, and the number of access network elements can also be one or more, and this application embodiment does not limit this. When CU and DU are deployed separately, one or more AI modules can also be deployed in CU and / or DU respectively. In addition, CU can be further divided into CU-CP and CU-UP. One or more AI modules can also be deployed in CU-CP and / or CU-UP respectively.
[0139] The AI module is used to implement corresponding AI functions. The AI modules deployed in different network elements can be the same or different. Depending on the configuration parameters, the AI module can implement different functions. The AI model can be configured based on one or more of the following parameters: structural parameters (e.g., at least one of the following: number of neural network layers, neural network width, inter-layer connections, neuron weights, neuron activation function, or bias in the activation function), input parameters (e.g., type and / or dimension of input parameters), or output parameters (e.g., type and / or dimension of output parameters). The bias in the activation function can also be referred to as the neural network bias.
[0140] It should be understood that in the embodiments of this application, model, AI model, inference model and training model may have the same meaning, and the embodiments of this application do not limit this.
[0141] It should also be understood that an AI module can have one or more models. A model can infer an output, which includes one or more parameters. The learning, training, or inference processes of different models can be deployed on different nodes or devices, or they can be deployed on the same node or device. This application embodiment does not impose any restrictions on this.
[0142] It should be noted that the AI module can be a collection of various functions used to implement intelligent network management, and the specific form of the AI module is not limited in the embodiments of this application. For example, the AI module may be a radio access network intelligent controller (RIC) used to control and optimize the functions of the RAN. The application architecture of the communication system in the embodiments of this application is described below using the AI module as an example.
[0143] Figure 4 is a schematic diagram of another possible application architecture in the communication system to which the information configuration method provided in this application embodiment is applicable. As shown in Figure 4, RIC can be further divided into near-real-time RIC (near-RT RIC) and non-real-time RIC (non-RT RIC). Non-real-time RIC mainly processes non-real-time information, such as data that is not sensitive to latency, with latency in the order of seconds. Real-time RIC mainly processes near-real-time information, such as data that is relatively sensitive to latency, with latency in the order of tens of milliseconds.
[0144] The near real-time RIC can be used for model training and inference. For example, it can be used to train an AI model and then use that AI model for inference. The near real-time RIC can obtain network-side and / or terminal-side information from RAN network elements (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminal devices. This information can be used as training data or inference data. Optionally, the near real-time RIC can feed back the inference results to the access network elements and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the near real-time RIC sends the inference results to the DU, and the DU further sends the inference results to the RU.
[0145] The non-real-time RIC is also used for model training and inference. For example, it can be used to train an AI model and then use that model for inference. The non-real-time RIC can obtain network-side and / or terminal-side information from access network elements (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data. Optionally, the non-real-time RIC can feed back the inference results to the access network elements and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU; for example, the non-real-time RIC sends the inference results to the DU, and the DU further sends the inference results to the RU.
[0146] The near real-time RIC and non-real-time RIC can also be configured as separate network elements. Optionally, the near real-time RIC and non-real-time RIC can also be part of other devices. For example, the near real-time RIC can be set in the access network element (e.g., CU, DU), while the non-real-time RIC can be set in the OAM, cloud server, core network element, or other network devices.
[0147] For example, the AI module in Figure 3 may be a RIC, such as a near real-time RIC or a non-real-time RIC. For instance, a near real-time RIC is located in an access network element (e.g., in a CU or DU), while a non-real-time RIC is located in an OAM, a cloud server, a core network element, or other network devices. The RIC can obtain data from multiple terminal devices from access network elements (e.g., CU, CU-CP, CU-UP, DU, and / or RU), reassemble it into a training dataset, and then train the device based on this dataset. Exemplarily, near real-time and non-real-time RICs can also be set up as separate network elements, and network devices can also be either near real-time or non-real-time RICs.
[0148] It should be noted that Figures 3 and 4 are simplified schematic diagrams for ease of understanding. For example, the application architecture in the above communication system may have other deployment methods, which are not shown in Figures 3 and 4. This application embodiment does not limit the deployment method of the AI module in the communication system.
[0149] For ease of understanding, the relevant technologies involved in the embodiments of this application will be further explained below with reference to Figures 5 to 9.
[0150] 1. AI and Machine Learning
[0151] AI refers to the ability to endow machines with human-like intelligence, such as enabling machines to use computer hardware and software to simulate certain intelligent human behaviors. Machine learning (ML) is an important technological approach to achieving AI. In machine learning methods, machines learn (or train) models using training data; this process can be called the model training phase. The model represents the mapping from input to output. The learned model can be used for inference (or prediction), that is, it can be used to predict the output corresponding to a given input; this process can be called the model inference phase. The input can be called an observation instance or measurement resource. The output can be called the inference result, prediction result, or prediction instance. An AI model can be considered a specific method for implementing a certain AI function; the AI model represents the mapping relationship or function between the model's input and output. Machine learning can be divided into supervised learning, unsupervised learning, and reinforcement learning.
[0152] Supervised learning, based on collected sample values and labels, uses machine learning algorithms to learn the mapping relationship between sample values and labels, and expresses this learned mapping relationship using a machine learning model. The process of training the machine learning model is the process of learning this mapping relationship. For example, in signal detection, the noisy received signal is the sample, and the corresponding real constellation point (the original signal) is the label. Machine learning aims to learn the mapping relationship between samples and labels through training, that is, to enable the machine learning model to learn a signal detector. During training, the model parameters are optimized by calculating the error between the model's predicted values and the real labels. Once the mapping relationship is learned, it can be used to predict the sample label of each new sample. The mapping relationship learned in supervised learning can include linear mappings and nonlinear mappings. Based on the type of label, the learning task can be divided into classification tasks and regression tasks.
[0153] Unsupervised learning relies solely on collected sample values, using algorithms to discover inherent patterns within the samples. One type of unsupervised learning algorithm uses the samples themselves as supervisory signals; that is, the model learns the mapping relationship from sample to sample, which is called self-supervised learning. During training, model parameters are optimized by calculating the error between the model's predictions and the samples themselves. Self-supervised learning can be used for signal compression and decompression recovery applications; common algorithms include autoencoders and generative adversarial networks.
[0154] Reinforcement learning is an algorithm that learns problem-solving strategies by interacting with its environment. Unlike supervised and unsupervised learning, reinforcement learning does not have explicit "correct" action labels. The algorithm needs to interact with the environment to obtain reward signals from the environment, and then adjust its decision actions to obtain larger reward signal values. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user based on the total system throughput feedback from the wireless network, aiming to achieve a higher system throughput. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and the optimal decision action. However, because the label of the "correct action" cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action". Reinforcement learning training is achieved through iterative interaction with the environment.
[0155] Deep neural networks (DNNs) are a specific implementation of machine learning. According to the general approximation theorem, neural networks can theoretically approximate any continuous function, thus enabling them to learn arbitrary mappings. Traditional communication systems may require extensive empirical knowledge to design communication modules, while DNN-based deep learning communication systems can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.
[0156] The idea behind DNNs originates from the neuronal structure of the brain. For example, Figure 5 illustrates a possible neuronal structure in a deep neural network. The neuron performs a weighted summation operation on each input, and the multiple weighted sums are output through a non-linear function. As shown in Figure 5, suppose the neuron's input is x = [x0, ..., x...]. n The weights corresponding to this input are w = [w0, ..., w0]. n The bias of the weighted summation is b. Therefore, the result of the weighted summation can be... The neuron may also need to perform further operations on the above calculation results using a nonlinear function f. The nonlinear function can take many forms; for example, it can be a maximum value function max{0,x}. In this case, the effect of a neuron's execution could be...
[0157] Figure 6 illustrates a possible structure in a deep neural network. DNNs typically have a multi-layered structure, as shown in Figure 6. A DNN can include an input layer, hidden layers, and an output layer. Each layer of a DNN can contain multiple neurons. The input layer processes the received values through neurons (e.g., by weighting and summing the data from the output layer) and then passes the results to the hidden layers. Similarly, the hidden layers then pass the calculation results to the final output layer, producing the DNN's output.
[0158] DNNs typically have more than one hidden layer, and these hidden layers usually influence their ability to extract information and fit functions. Increasing the number of hidden layers or widening the width of each layer can improve the function fitting ability of a DNN. The weights in each neuron are the parameters of the DNN network model. The model parameters are optimized through the training process, enabling the DNN network to extract data features and express mapping relationships. DNNs generally use supervised or unsupervised learning strategies to optimize model parameters.
[0159] Based on the way the network is constructed, DNNs can be divided into feedforward neural networks (FNN), convolutional neural networks (CNN), and recurrent neural networks (RNN).
[0160] The characteristic of FNN is that neurons in adjacent layers are completely connected to each other, which makes FNN usually require a lot of storage space and result in high computational complexity.
[0161] CNNs are neural networks specifically designed to process data with a grid-like structure. For example, time-series data (discrete sampling along the time axis) and image data (two-dimensional discrete sampling) can both be considered grid-like data. CNNs do not use all the input information at once for computation; instead, they use a fixed-size window to extract a portion of the information for convolution operations, which significantly reduces the computational cost of model parameters. Furthermore, depending on the type of information extracted by the window (such as people and objects in an image representing different types of information), each window can use different convolution kernels, allowing CNNs to better extract features from the input data.
[0162] Recurrent Neural Networks (RNNs) are a type of distributed neural network (DNN) that utilizes feedback time-series information. Their input includes the current input value and their own output value from the previous time step. RNNs are well-suited for acquiring temporally correlated sequence features, and are particularly applicable to applications such as speech recognition and channel coding / decoding.
[0163] The FNN, CNN, and RNN mentioned above are common neural network structures, all built upon neurons. As introduced above, each neuron performs a weighted summation operation on its input values, and the result is passed through a nonlinear function to produce the output. We call the weights of the weighted summation operation and the nonlinear function in the neural network the parameters of the neural network. Taking a neuron with max{0,x} as the nonlinear function as an example, we perform... The parameters of the operated neuron are: weights w = [w0, ..., wn The parameters of a neural network consist of the parameters of all neurons, including the weighted summation bias b and the nonlinear function max{0,x}.
[0164] 2. CSI Feedback
[0165] In the NR protocol, the downlink CSI configuration and reporting process is as follows: The network device sends the CSI reporting configuration (CSI-ReportConfig) to the terminal device. CSI-ReportConfig may include parameters such as reporting type (reportConfigType) and reporting quantity (reportQuantity). The reporting type indicates the time-domain behavior of the report, and can include periodic, semi-persistent, or aperiodic types. The reporting quantity indicates the measurement quantity to be reported, and can include parameters such as rank indicator (RI), channel quality indicator (CQI), precoding matrix indicator (PMI), and reference signal received power (RSRP).
[0166] For example, a network device sends a CSI-RS to a terminal device. The terminal device performs channel and interference measurements based on the CSI-RS to obtain measurement results. Based on the measurement results, the terminal device determines the reportQuantity in CSI-ReportConfig and reports the downlink CSI to the network device. This downlink CSI may include information such as RI, CQI, PMI, and RSRP measured by the terminal device.
[0167] If the reporting type in CSI-ReportConfig is configured as periodic, the terminal device will report periodically according to the period specified in the radio resource control (RRC) signaling, without needing to send signaling to trigger the reporting each time. If the reporting type in CSI-ReportConfig is configured as semi-persistent, the initial reporting needs to be triggered by signaling, and after the trigger, the terminal device will report periodically according to the specified period. If the reporting type in CSI-ReportConfig is configured as non-periodic, the network device needs to use downlink control information (DCI) to trigger the terminal device to report.
[0168] When the reporting type is configured as semi-persistent, the process of network devices triggering terminal devices to report CSI via signaling is relatively complex. For example, when the network device uses Media Access Control-Control Element (MAC CE) signaling for triggering (or activation), the terminal device uses the Physical Uplink Control Channel (PUCCH) for CSI reporting. As another example, when the network device uses DCI for triggering, the terminal device uses the Physical Uplink Shared Channel (PUSCH) for CSI reporting.
[0169] In the NR protocol, RRC signaling can be used to configure resources for CSI measurements and CSI reports, corresponding to CSI resource configuration (CSI-ResourceConfig) and CSI reporting configuration (CSI-ReportConfig), respectively. CSI-ReportConfig indicates the identifier (CSI-ResourceConfigId) of the resource configuration used for the CSI measurement in that CSI report.
[0170] The resources used for CSI measurements can be Channel State Information Reference Signals (CSI-RS). CSI-RS configuration has three levels: CSI-ResourceConfig, CSI-RS resource set, and CSI-RS resource. CSI-ResourceConfig is the highest-level configuration, defining one or more CSI-RS resource sets. A CSI-RS resource set is an intermediate level, organizing multiple CSI-RS resources together. A CSI-RS resource is the lowest-level configuration, defining the specific CSI-RS resource.
[0171] It should be understood that CSI-RS, the resource of CSI measurement, can be further divided into zero-power (ZP) CSI-RS and non-zero-power (NZP) CSI-RS, which can be referred to as ZP-CSI-RS and NZP-CSI-RS, respectively. ZP-CSI-RS is mainly used for interference estimation, while NZP-CSI-RS can be used for CSI measurement and estimation of transmission channels, interference estimation, and measurement of the RSRP (i.e., L1-RSRP) of layer 1 (L1).
[0172] For ease of description, the configuration of NZP-CSI-RS will be used as an example. Each CSI-ResourceConfig can configure S ≥ 1 NZP-CSI-RS-ResourceSet. For example, the higher-layer signaling parameter nzp-CSI-RS-ResourceSetList contained in the CSI-ResourceConfig specifies S NZP-CSI-RS-ResourceSetIds. Each NZP-CSI-RS-Resource set can configure Ks ≥ 1 NZP-CSI-RS-Resource. For example, the higher-layer signaling parameter nzp-CSI-RS-Resource contained in the NZP-CSI-RS-Resource set specifies Ks NZP-CSI-RS-ResourceIds.
[0173] It should be understood that the time-domain behavior of NZP-CSI-RS-Resource is given by the higher-level parameter resourceType in CSI-ResourceConfig, and can be periodic, semi-persistent, or aperiodic. It should also be understood that all NZP-CSI-RS-Resources associated with the same CSI-ResourceConfig have the same time-domain behavior. Furthermore, the higher-level signaling parameter resourceMapping is used to configure the time-frequency resources for transmitting NZP-CSI-RS-Resources.
[0174] In this embodiment, reference signal resources are used to carry reference signals. The reference signal refers to a signal known to the terminal device. Exemplarily, the reference signal includes one or more of the following: synchronizing signal block (SSB), channel state information reference signal (CSI-RS), tracking reference signal (TRS), phase-tracking reference signal (PTRS), and positioning reference signal (PRS). In other words, the reference signal can be replaced by any one of the known signal, downlink signal, SSB, CSI-RS, TRS, PTRS, and PRS. The reference signal can be used for channel measurement or channel estimation. Reference signal resources can be used to configure the transmission attributes of the reference signal, such as one or more of time-frequency resource locations, port mapping relationships, power factors, and scrambling codes, etc., without specific limitations. The transmitting device can transmit the reference signal based on the reference signal resources, and the receiving device can receive the reference signal based on the reference signal resources. The reference signal resources may include a beam, or have a corresponding relationship with a beam. Furthermore, the reference signal resource may also include time-domain resources and / or frequency-domain resources corresponding to the beam, such as time-frequency resources. The beam can also be referred to as a spatial domain resource. The time-domain behavior of the reference signal resource can be periodic, semi-continuous, or aperiodic. For periodic reference signal resources, one period of transmission can be understood as one transmission occasion. One transmission occasion corresponds to one or more reference signals transmitted within one period.
[0175] 3. Beam management and beam indication
[0176] A beam is a communication resource. In the NR protocol, beams can be represented as spatial filters, or spatial parameters. The beam used to transmit signals can be called the transmission beam (Tx beam), or a spatial domain transmit filter or spatial domain transmit parameter; the beam used to receive signals can be called the reception beam (Rx beam), or a spatial domain receiver filter or spatial domain receive parameter. The transmission beam refers to the distribution of signal strength in different directions in space after the signal is transmitted through the antenna, while the reception beam refers to the distribution of signal strength in different directions in space of the wireless signal received from the antenna.
[0177] Beams can be identified by their identifier (ID). For example, a beam ID can be a Channel State Information Reference Signal Resource Indicator (CSI-RS, CRI), an SSB Resource Indicator (SSBRI), or a bit in a bitmap corresponding to the beam, or an index of the beam within a beam set. For instance, the number of bits in the bitmap is equal to the total number of beams associated with the network device in a single beam management inference task. Beams can be categorized as wide beams and narrow beams. A wide beam refers to a beam with a relatively large radiation range of the transmitting or receiving antenna when transmitting or receiving signals. Wide beams are typically used in applications requiring broadcasting signals to a large area or providing wide coverage. Wide beams can provide a wider coverage area, but the signal strength is relatively weaker. A narrow beam refers to a beam with a relatively small radiation range of the transmitting or receiving antenna. Narrow beams are typically used in applications requiring focusing signals onto a specific target or area. Narrow beams can provide higher signal strength and greater directivity, but the coverage area is relatively small.
[0178] Alternatively, beam can be replaced with signal, downlink beam, transmit beam, transmit beam, thin beam, narrow beam, wide beam, spatial filter, spatial filter, spatial parameters, spatial transmit filter, port, etc.
[0179] NR communication systems typically employ beamforming technology. By weighting the transmitted signal, a narrow beam with more concentrated energy and stronger directionality is formed for each type of channel and signal. At the same transmit power, the propagation distance of a narrow beam may be greater than that of a wide beam. However, the coverage of a narrow beam is limited; a single beam may not cover all users within a cell, nor may it guarantee that every user within the cell will always receive maximum signal energy.
[0180] Therefore, beam scanning technology was introduced. Beam scanning refers to transmitting or receiving beams in a preset manner within a certain time interval to cover a specific spatial area. Currently, the preset method mainly refers to time-division multiplexing, which improves coverage performance by transmitting or receiving narrow beams in different directions at different times to cover a specific spatial area.
[0181] Based on the differences in the weighting strategies used during beamforming, beamforming technology can be further divided into two categories: static beamforming and dynamic beamforming. Static beamforming uses predefined weights, typically determined during the cell planning phase. For example, the number, width, and direction of the beams are fixed. Then, based on information such as cell coverage, user distribution, and system load, the optimal beam is selected for various channels and signals. Dynamic beamforming uses weights calculated based on channel quality, and the beam width and direction are adjusted dynamically according to factors such as UE location and channel status. Beam scanning is primarily for static beamforming using preset weights; dynamic beamforming, using dynamic weights, does not require beam scanning.
[0182] Beam scanning requires combining beam measurement, beam reporting, and beam determination technologies to select an optimal beam pair between the base station and the UE. Specifically, beam scanning finds the most suitable transmit and receive beams, aligning their directions to optimize signal gain and improve communication quality. The beam scanning process can be further divided into three stages: P1, P2, and P3.
[0183] P1: The base station performs a synchronization signal / physical broadcast channel block (SSB) beam scan, while the UE performs a wide beam scan. The base station uses beam scanning to transmit SSB beams from different directions in a time-division manner, broadcasting synchronization messages and system messages. The UE uses beam scanning to receive signals and confirms the received beam. Simultaneously, the UE feeds back the SSB measurement results to the base station, which confirms the transmitted beam. The transmitted and received beams achieve initial alignment. In summary, P1 is used to find an initial beam pair between the base station and the UE.
[0184] P2: The base station performs CSI-RS beam scanning, and the UE's receive beam is fixed. The base station determines the initial connection with the UE through the SSB beam. The base station selects an initial SSB beam based on the random access result, and then performs a more detailed scan around this SSB beam using a narrower CSI-RS beam. The UE feeds back the CSI-RS measurement results to the base station through a measurement report, and the base station confirms the optimal transmit beam. In short, after the initial beam pair is established, in order to obtain higher signal gain, the base station needs to select a CSI-RS beam that is narrower than the SSB beam for beam adjustment. In summary, P2 is used to accurately determine the base station's transmit beam.
[0185] P3: The base station's transmit beam is fixed, and the UE performs narrow beam scanning. For example, if the base station's transmit beam (CSI-RS narrow beam) is fixed, the UE uses beam scanning for signal reception to determine a more accurate receive beam, achieving final alignment between the transmit and receive beams. In short, the UE further adjusts the receive beam through beam scanning, thereby enhancing signal quality. In summary, P3 is used to accurately determine the UE's receive beam.
[0186] In the above process, beam scanning may use SSB or CSI-RS as the reference signal for beam measurement. Therefore, the beam measurement and reporting procedure may be consistent with the CSI configuration and reporting procedure. For example, in P2, the base station can instruct the UE to report the CRI and the corresponding RSRP by configuring the reportQuantity field in CSI-ReportConfig to CRI-RSRP.
[0187] After obtaining the optimal transmission beam through beam scanning, network devices (such as base stations or UEs) need to indicate the beam used for transmission (such as the transmit beam and / or receive beam) through beam indication information. Beam indication information can be one or more of the following: beam number (or number, index, identity, ID, etc.), uplink signal resource number, downlink signal resource number, absolute index of the beam, relative index of the beam, logical index of the beam, index of the antenna port corresponding to the beam, index of the antenna port group corresponding to the beam, index of the downlink signal corresponding to the beam, time index of the downlink synchronization signal block corresponding to the beam, beam pair link (BPL) information, transmit parameters (Tx parameter) corresponding to the beam, receive parameters (Rx parameter) corresponding to the beam, transmit weight corresponding to the beam, weight matrix corresponding to the beam, weight vector corresponding to the beam, receive weight corresponding to the beam, index of transmit weight corresponding to the beam, index of weight matrix corresponding to the beam, index of weight vector corresponding to the beam, index of receive weight corresponding to the beam, receive codebook corresponding to the beam, transmit codebook corresponding to the beam, index of receive codebook corresponding to the beam, and transmit codebook corresponding to the beam. The downlink signal can be one or more of the following: synchronization signal, broadcast channel, broadcast signal demodulation signal, SSB, CSI-RS, cell specific reference signal (CS-RS), user equipment specific reference signal (US-RS), dedicated reference signal (DMRS), downlink data channel demodulation reference signal, and downlink phase noise tracking signal. The uplink signal can be one or more of the following: uplink random access sequence, uplink sounding reference signal (SRS), uplink control channel demodulation reference signal, uplink data channel demodulation reference signal, and uplink phase noise tracking signal.
[0188] Furthermore, beam indication information can also be represented as a transmission configuration indicator (TCI) or a TCI status. A TCI status may include one or more quasi-co-location (QCL) information, which may be used to indicate QCL relationships. For example, each QCL information may include a reference signal (or synchronization signal block) ID and a QCL type. For instance, a terminal device can determine the beam to receive the physical downlink shared channel (PDSCH) based on the TCI status indicated by the network device (typically carried by the physical downlink control channel (PDCCH)).
[0189] The QCL relationship is used to indicate that multiple resources share one or more identical or similar communication characteristics. Multiple resources with a QCL relationship may have identical or similar communication configurations. For example, different signals corresponding to different antenna ports with a QCL relationship may also have the same parameters. In this case, for example, the parameters of one antenna port (also called QCL parameters) can be used to determine the parameters of another antenna port with a QCL relationship to that antenna port. Another example is that two antenna ports have the same parameters, or the parameter difference between two antenna ports is less than a certain threshold. The parameters may include one or more of the following: delay spread, Doppler spread, Doppler shift, average delay, average gain, and spatial Rx parameters. Spatial Rx parameters may include one or more of the following: angle of arrival (AOA), average AOA, AOA spread, angle of departure (AOD), average departure angle AOD, AOD spread, receive antenna spatial correlation parameters, transmit antenna spatial correlation parameters, transmit beam, receive beam, or resource identifier.
[0190] 4. Air Interface AI
[0191] Currently, AI has been introduced into wireless communication networks and is widely used in many application scenarios of air interface technology, such as CSI feedback, CSI prediction, beam management, and positioning. For example, when applying AI in CSI feedback scenarios, an autoencoder architecture can be used for CSI feedback. An autoencoder architecture generally includes an AI encoder and an AI decoder. The AI encoder can be deployed on the terminal device, and the AI decoder can be deployed on the network device. Compared to traditional CSI feedback technology, under the same CSI feedback performance, AI model-based CSI feedback can reduce air interface feedback overhead and the computational complexity of the terminal device, showing greater application potential.
[0192] For example, when applying AI in CSI prediction scenarios, terminal devices or network devices can use prediction models to predict future CSI based on historical CSI and feed this prediction back to the network device. The AI model may reside in the terminal device or the network device. By accurately predicting future CSI, the problem of inaccurate CSI feedback information caused by channel time-varying characteristics can be solved.
[0193] For example, when applying AI in beam management scenarios, terminal devices or network devices can efficiently and accurately identify the best beam using AI models. The AI model may be located in the terminal device or the network device. For example, when applying AI models in positioning scenarios, triangulation can be used for positioning. The terminal device obtains the location information of three surrounding network devices and inputs it into the corresponding AI model. Then, based on the distance, direction, and channel information from the terminal device to the three network devices, the location of the terminal device is obtained.
[0194] Specifically, AI-based beam management includes two typical use cases: spatial domain beam prediction (BM-Case 1) and temporal domain beam prediction (BM-Case 2).
[0195] Figure 7 illustrates a possible scenario for spatial beam prediction. In traditional beam measurement, network or terminal devices need to scan all beams in the selectable beam set using the P1 to P3 process described above to obtain the optimal (maximum RSRP) beam. For systems using massive MIMO, the number of beams in the selectable beam set can be large (e.g., 1024), and the traditional beam scanning process requires significant measurement overhead. With the introduction of AI, network or terminal devices may only measure a portion of the beams in the selectable beam set (all blocks in the beam set in Figure 7) (the bolded blocks in the beam set in Figure 7, representing the measured values). Based on the measured values, the network or terminal device can further use a beam prediction model to predict the Top-k beams in the beam set and output the corresponding Top-k beam IDs, thus significantly reducing beam measurement overhead. The Top-k beams mentioned above may represent the selection of the top k best beam candidates.
[0196] Figure 8 illustrates a possible scenario for time-domain beam prediction. As shown in Figure 8, network devices or terminal devices can use historical beam information (beam measurement information at historical moments) to predict future beam information (beam information at future moments, such as Top-k beam IDs and corresponding RSRPs) based on a beam prediction model. This improves the robustness of beam management in scenarios with rapidly changing channels and avoids frequent beam measurements and handovers. The beam prediction model can be located within the terminal device or network device.
[0197] In both prediction scenarios shown in Figures 7 and 8, the beam prediction model uses the measured values of reference signal resource set B (Set B) to obtain the predicted values of reference signal resource set A (Set A). Set B can be a subset of Set A, or Set B can be different from Set A (e.g., the signal angle corresponding to each reference signal resource in Set B is greater than the signal angle corresponding to each reference signal resource in Set A, i.e., Set B is a wide beam and Set A is a narrow beam), or Set B can be the same as Set A.
[0198] Specifically, AI-based CSI prediction includes a typical use case: time-domain CSI prediction. To address the aging problem of CSI measurements in time-varying channels, AI CSI prediction models can use historical CSI measurement information to predict future CSI information (PMI / CQI, etc.), thereby improving CSI reporting performance in scenarios with rapidly changing channels. AI CSI prediction models are typically located in the terminal device. The principle of time-domain CSI prediction is similar to that of time-domain beamforming, the difference being the different data types of the input and output.
[0199] The lifecycle of an AI model typically involves the following stages: training data collection, model training (or model learning), model information dissemination, model activation / deactivation, model inference (or model reasoning, inference, or prediction), model monitoring or validation, model updates, or inference result dissemination. Air-to-ground (ATO) AI model lifecycle management (LCM) can be based on either model ID or functionality. A model ID is an identifier assigned in some way to identify the model. In model ID-based LCM, the model ID indicates operations performed on the model, such as activation / deactivation / selection / rollback / switching.
[0200] AI features refer to the ability of network devices or terminal devices to use AI, or in other words, the ability of network devices or terminal devices to use AI models. Examples include AI-based CSI feedback and AI-based beam management. Functions may refer to configuration-related AI features, or they may correspond to specific configurations under AI features. For example, AI-based CSI feedback or AI-based beam management under specific configurations. Configuration refers to the configurations supported by the network device or terminal device (this configuration can be based on signaling instructions or predefined). For example, in function-based LCM, the network device can instruct the terminal device to operate AI functions, such as activation / deactivation / selection / fallback / switching, through 3GPP signaling (e.g., RRC, MAC-CE, DCI).
[0201] Specifically, an AI feature may encompass one or more functions. For a single function, there may be one or more models used to implement that function. Furthermore, a single model can also be used to implement multiple functions. The following examples illustrate the relationship between functions and AI features:
[0202] Example 1: One function corresponds to one AI feature. For example, function 1 is AI-based temporal beam prediction, and function 2 is AI-based spatial beam prediction.
[0203] Example 2: One function corresponds to one AI feature and a specific set of RRC configurations. For example, function 1 is AI-based time-domain beam prediction under configuration 1, and function 2 is AI-based time-domain beam prediction under configuration 2. The configuration includes at least one of the following: CSI-RS resource configuration, CSI reporting configuration, beam set configuration, or prediction window configuration, etc.
[0204] Example 3: One function corresponds to one AI feature, a specific set of RRC configurations, and a scene / site identifier. For example, function 1 is AI-based temporal beam prediction under configuration 1 and scene 1, and function 2 is AI-based temporal beam prediction under configuration 1 and scene 2. The configurations will not be elaborated here. The scenes can be: urban areas, suburbs, urban macrocells (UMa), urban microcells (UMi), indoor hotspot cells (InH), and highways, etc.
[0205] 5. Training data collection
[0206] The following explains the collection of training data in AI models.
[0207] During the model training phase, network devices (e.g., base stations) can instruct terminal devices (UEs) to collect training data. For example, the base station can configure CSI reports for training data collection, such as CSI-ReportConfig. For CSI reports used for training data collection, the base station can configure measurement resources, such as CSI-ResourceConfig. The UE performs measurements based on the configured measurement resources, and the measured data can be used for training.
[0208] For example, for training a beam prediction model, a base station can configure a CSI-ResourceConfig for resources in Set A and Set B respectively. A CSI-ResourceConfig can be associated with one or more measurement resources (e.g., NZP-CSI-RS resources or SSB resources), and each measurement resource corresponds to a beam. The base station can configure a CSI-ReportConfig for collecting training data, which is associated with the corresponding CSI-ResourceConfigs for both Set A and Set B. During model training, the measurement results of Set B are used as the model input, and the measurement results of Set A are used as the model's label. The label refers to the target value that the model needs to learn to predict. In this case, the model can adjust its parameters during training to make the output as close to the label as possible, thereby improving the accuracy of the model's predictions.
[0209] For example, when training a CSI prediction model, since the resources to be measured and predicted are the same, the base station can configure a CSI-ResourceConfig. A CSI-ResourceConfig is associated with one or more measurement resources (e.g., NZP-CSI-RS resources). The base station can also configure a CSI-ReportConfig for collecting training data, in which a CSI-ResourceConfig is associated. During model training, the measurement information from the measurement resources in the CSI-ResourceConfig serves as both the model input and the model's label.
[0210] 6. Applicability Request and Reporting
[0211] For beam prediction scenarios based on AI models by terminal devices, the network device needs to know the supported functions / configurations of the terminal device before the terminal device performs inference. Figure 9 illustrates a possible interaction flow 900 between the terminal device and the network device. In interaction flow 900, the terminal device can exchange inference-related capability information with the network device through the following steps:
[0212] Step 901: The network device sends a UECapabilityEnqiry message to the terminal device, and the terminal device receives the message accordingly. This message is used to trigger the terminal device to report supported functions, such as AL / ML functions.
[0213] In step 902, the terminal device sends a UECapabilityInformation message to the network device, and the network device receives the message accordingly. This message contains the supported functionality of the terminal device.
[0214] In step 903, the network device sends an RRCReconfiguration message to the terminal device, and the terminal device receives the message accordingly. This message may contain RRC configuration to request the terminal device to report applicable functions.
[0215] The aforementioned RRC configuration may include an applicability request, which may include one or more CSI-ReportConfigs or one or more sets of inference-related parameters for inference configuration. Inference-related parameters can be selected from information elements in or referenced by CSI-ReportConfig. For example, the inference-related parameters may include: association ID, Set A related information, Set B related information, report content related information, observation instance related information, or prediction instance related information. The association ID is an information element introduced by the standard to ensure the consistency of network-side additional conditions corresponding to model training and inference. If the association IDs corresponding to model training and inference are the same, the network-side additional conditions can be considered to be the same or similar.
[0216] Furthermore, the terminal device determines the applicable functionality based on the RRC configuration described in step 903. For example, the UE may prepare a corresponding model according to the base station's configuration. If the UE does not support a model for a certain configuration, then that configuration is not applicable. It should be noted that the model corresponding to a certain configuration may not be local to the UE, for example, it may be stored on a server in the cloud. In this case, the UE can download the model to its local machine, and then the configuration will be applicable.
[0217] In step 904, the terminal device sends an Applicable functionality report message to the network device, and the network device receives the message accordingly. This message contains the applicable functionality for the terminal device.
[0218] For example, the terminal device can report the applicability of all of the above CSI-ReportConfig or one or more sets of inference-related parameters.
[0219] In step 905, the network device sends an RRCReconfiguration message to the terminal device, which in turn receives the message. This message may contain RRC configuration for the terminal device's model training or inference.
[0220] Step 906: The terminal device performs subsequent operations, such as activation / deactivation / inference / monitoring operations.
[0221] In the above-mentioned interaction process 900, steps 901 and 902 mainly complete the reporting of supported functions, while steps 903 and 904 mainly complete the reporting of applicable functions. The reporting process shown in steps 901 and 902 may be similar to the reporting process of other capabilities of the UE, and will not be described in detail here.
[0222] It should be noted that the meaning of "function" in "supported functions" and "applicable functions" may differ. Supported functions may refer to AI / ML features supported by the terminal device, such as beam prediction and CSI prediction. Applicable functions may refer to the applicable CSI-ReportConfig for inference or an applicable set of inference-related parameters. For example, the terminal device can use the above configuration parameters for inference.
[0223] Furthermore, both the configuration used for inference and the configuration used for determining applicability are RRC configurations, and the parameter types they contain may be partially or entirely the same. The difference between the configuration used for inference and the configuration used for determining applicability is as follows: For the configuration used for inference, the CSI report reported / generated by the terminal device directly corresponds to this configuration, or in other words, the terminal device needs to report a CSI report for this configuration; for the configuration used for determining applicability, the CSI report reported / generated by the terminal device does not directly correspond to this configuration, or in other words, the terminal device does not need to report a CSI report for this configuration. For example, for a periodic CSI report configuration, if this configuration is used for inference, the terminal device needs to report a CSI report for this configuration; if this configuration is used for determining applicability, the terminal device does not need to report a CSI report for this configuration. Alternatively, the difference between the configuration used for inference and the configuration used for determining applicability is as follows: For the configuration used for inference, the information used to trigger / activate the terminal device's reported CSI report corresponds to this configuration; for the configuration used for determining applicability, the information used to trigger / activate the terminal device's reported CSI report does not correspond to this configuration. For example, for an AP / SP CSI reporting configuration, if the configuration is used for inference, the information (DCI / MAC-CE) used to trigger / activate the terminal device's CSI reporting corresponds to that configuration; if the configuration is used for applicability determination, the information (DCI / MAC-CE) used to trigger / activate the terminal device's CSI reporting does not correspond to that configuration. After applicability determination, the network device can further configure CSI reporting for inference, and the information (DCI / MAC-CE) used to trigger / activate the terminal device's CSI reporting corresponds to that inference CSI reporting configuration.
[0224] For CSI prediction scenarios using terminal device models, the network device needs to know the supported functions / configurations of the terminal device before instructing it to perform inference. The process of exchanging inference-related capability information between the terminal device and the network device may be similar to the interaction process 900 described above. The difference lies in the fact that the set of one or more sets of inference-related parameters in the applicability request in step 903 may be different. The inference-related parameters used for CSI prediction may also be selected from information elements in CSI-ReportConfig or information elements referenced by CSI-ReportConfig. The inference-related parameters may include at least one of the following: observation window related information (also known as observation instance or measurement resource related information), prediction window related information (also known as prediction instance related information, such as the number of prediction instances, time interval, and the offset of the first prediction instance relative to the reference time, where the first prediction instance refers to the earliest prediction instance in the time unit).
[0225] For AI models, training data collection is necessary before model training. During the training data collection phase, network devices (e.g., base stations) can configure CSI-ReportConfig for terminal devices (e.g., UEs). CSI-ReportConfig configures measurement resources for training data collection. The terminal device obtains training data based on the measurement results from these resources, which can then be used for model training. Model training can be performed by the terminal device or by other devices serving the terminal device (e.g., servers). However, the model trained by the terminal device may not meet the inference requirements corresponding to the actual business needs of the network device, resulting in the model being unusable. For ease of description, the base station and UE will be used as examples below.
[0226] For beam prediction or CSI prediction, model training requires the collection of a large amount of data. For example, the base station can send measurement resources with periodic or semi-persistent temporal behavior to the UE for training data collection. In this case, the UE can perform measurements based on these measurement resources and use the measurement results as training data for the model. These measurement resources may include at least one observation instance, and the temporal behavior of the measurement resources may refer to the temporal behavior of the observation instance. The temporal behavior may be periodic, semi-persistent, or aperiodic. Taking an observation instance as an example, periodic temporal behavior means that multiple observation instances are sent periodically in time. Semi-persistent temporal behavior means that multiple observation instances maintain periodic transmission after activation. Aperiodic temporal behavior means that multiple observation instances are sent aperiodically in time.
[0227] For example, the model trained by the UE can support inference to obtain Y predicted instances based on a minimum of X observation instances. However, when the base station allocates resources for inference, it can only configure a maximum of K observation instances, such as K observation instances with aperiodic temporal behavior. If K is less than X, it indicates that the UE cannot obtain enough observation instances, and the model trained by the UE cannot be used for base station inference.
[0228] For example, suppose that during the model inference phase, the base station's inference requirement is to make predictions at intervals of d for resources with intervals of m (e.g., the interval between observation instances is m, and the interval between prediction instances is d; or the resource interval corresponding to the model input provided by the base station is m, and the interval corresponding to the model output required by the base station is d). If m is greater than d, for example, in a scenario where the UE uses aperiodic channel state information reference signal (A-CSI-RS) for prediction, the protocol specifies that m can be 1 or 2 time slots, and d can be 1 time slot or m, possibly m is 2 time slots and d is 1 time slot. To meet the base station's inference requirements, the UE may need to use measurement resources with intervals of 2 time slots as model input and resources with intervals of 1 time slot as labels to perform model training. However, since the base station typically configures a CSI-ResourceConfig for training data collection, the UE will obtain both the model input and label values based on this single CSI-ResourceConfig. Since a CSI-ResourceConfig can only be configured for one period, the period of the CSI-ResourceConfig used for training data collection may only be configured for one time slot. When the UE receives a resource with a period configured for one time slot, it may mistakenly believe that the base station's inference requirement is one time slot for m and one time slot for d. In this case, the model trained in this way cannot be used for base station inference.
[0229] For example, in beam prediction or CSI prediction, base stations typically send measurement resources with periodic or half-periodic temporal behavior (referred to as P / SP measurement resources) for training. However, the base station's inference requirements may be based on measurement resources with aperiodic temporal behavior (referred to as AP measurement resources). Generally, for the same prediction capability, a UE requires more AP measurement resources than P / SP measurement resources. Since the base station does not explicitly indicate its inference requirements to the UE, the UE, upon receiving P / SP measurement resources from the base station, may mistakenly assume that the base station's inference requirements are also based on P / SP measurement resources. Suppose the model trained by the UE can infer one prediction instance using four P / SP observation instances. During the model inference phase, the base station might configure four AP measurement resources to instruct the UE to perform inference. However, the UE cannot perform accurate inference using only these four AP measurement resources (the UE may need at least six AP measurement resources), thus the trained model cannot be used for base station inference.
[0230] In summary, if the device (such as the UE) performing model training data collection and / or model training is unaware of the inference needs or inference preferences of the network device (such as the base station), the trained model may become unusable, resulting in a waste of training resources (such as collected training data or measurement resources during the model training phase).
[0231] In view of this, embodiments of this application provide an information configuration method and apparatus. During the model training data collection phase, a first apparatus (e.g., a UE) performing model training data collection and / or model training can receive inference parameters from a set of candidate inference parameters indicated by a second apparatus (e.g., a base station). These inference parameters can indicate the inference requirements of the base station. The first apparatus can use these inference parameters to collect training data during the model training data collection phase, and / or the first apparatus can use these inference parameters to train a model during the model training phase. In this case, the inference model generated by the first apparatus may have a better match with the inference requirements of the second apparatus, thereby improving the accuracy of the inference results.
[0232] In the embodiments of this application, during the model training data collection phase, the first device can be a device for performing model training data collection and / or model training, and the second device can be a device for sending measurement resources or related configurations to the first device. During the model inference phase, the first device can be a device for performing inference using an inference model, and the second device can be a device for sending measurement resources or related configurations to the first device.
[0233] In the embodiments of this application, the first device or the second device may be a device on the terminal device side, a device on the network device side, or a device on the core network (CN) side.
[0234] The devices on the terminal device side may include the terminal device itself, the communication module in the terminal device, or the circuits or chips in the terminal device responsible for communication functions (such as a modem chip, also known as a baseband chip, or a SoC chip or SIP chip containing a modem core, etc.), or may be contained in the AI entity on the terminal device side. The AI entity on the terminal device side can be the terminal device itself, or it can be an AI entity that serves the terminal device, such as a server, such as an OTT server or a cloud server.
[0235] The devices on the network device side can include the network device itself, the communication module within the network device, or the circuits or chips within the network device responsible for communication functions (such as a modem chip, also known as a baseband chip, or a system-on-a-chip (SoC) chip or SIP chip containing a modem core, etc.), or include AI entities on the network device side. AI entities on the network device side can be the network device itself, or AI entities serving the network device, such as RIC, OAM, or servers, such as OTT servers or cloud servers.
[0236] The devices on the core network side may include the core network element itself, the communication module within the core network element, or the circuits or chips (such as modem chips, also known as baseband chips, or system-on-a-chip (SoC) chips or SIP chips containing modem cores) responsible for communication functions within the core network element, or core network-side AI entities. The core network-side AI entities can be the core network element itself, or AI entities serving the core network element, such as servers, like OTT servers or cloud servers.
[0237] In the following text, for ease of description, the information configuration method provided in the embodiments of this application will be described using the example of the first device being the UE and the second device being the base station.
[0238] The information configuration method provided in the embodiments of this application will be described in detail below with reference to Figure 10. It should be understood that the technical solution of this application can be applied to the communication system 100 shown in Figure 1, the communication system 200 shown in Figure 2, and other systems; the embodiments of this application do not limit this application. Without loss of generality, it is assumed that the devices performing model training data collection and model training are both terminal devices, and the information configuration method provided in the embodiments of this application will be described in detail using the interaction process between the terminal device and the base station as an example.
[0239] Figure 10 illustrates an information configuration method 1000 provided in an embodiment of this application, including the following steps:
[0240] Step 1001: The base station sends first information to the UE, and the UE receives the first information accordingly. The first information is used to indicate the candidate set of inference parameters.
[0241] The candidate inference parameter set includes inference parameters of at least one of the following types: temporal behavior of the first resource, number of the first resource, time interval of the first resource, number of at least one prediction instance, time interval between time units corresponding to two adjacent prediction instances in the at least one prediction instance, time unit in which the first prediction instance in the at least one prediction instance is located, or type of output result corresponding to the inference of the first model; the first resource corresponds to the measurement resource used for inference of the first model, the at least one prediction instance is the prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit among the at least one prediction instance.
[0242] In this embodiment, the measurement resources used for inference can also be called observation resources or observation instances. An observation instance is a measurement resource used to obtain input information for inference. In model inference, the UE may infer one or more prediction instances through one or more observation instances. One or more observation instances corresponding to a single prediction can be called an observation window, and one or more prediction instances corresponding to a single prediction can be called a prediction window. A prediction instance may correspond to resources that are actually transmitted or resources that are not actually transmitted. For example, in the model inference stage, measurement resources can be used to obtain the input for model inference, and prediction instances can be inferred based on the model input. Measurement resources can sometimes be simply referred to as resources, and measurement resources can be reference signal resources (such as CSI-RS). A measurement resource can be a CSI-RS or a set of CSI-RS resources, and a CSI-RS resource set contains multiple CSI-RS. For example, for CSI prediction, a measurement resource is a CSI-RS resource; for beam prediction, a measurement resource is a set of CSI-RS resources.
[0243] The first resource corresponds to the measurement resource used for inference in the first model. It can be understood as the first resource being the measurement resource used for inference in the first model. That is, the temporal behavior of the first resource is the same as the temporal behavior of the measurement resource used for inference in the first model; the number of the first resources is the same as the number of measurement resources used for inference in the first model (or the number of measurement resources used for one inference operation of the first model); and the time interval of the first resource is the same as the time interval of the measurement resource used for inference in the first model. The time interval of the measurement resource can be understood as the time interval between two adjacent time units in the temporal domain. Specifically, for periodic measurement resources, the time interval of the measurement resource is the transmission period of the measurement resource.
[0244] The first resource corresponds to the measurement resource used for inference in the first model. It can also be understood that the parameters of the first resource can represent the parameters of the measurement resource used for inference in the first model. For example, the temporal behavior of the first resource can represent the temporal behavior of the measurement resource used for inference in the first model; the quantity of the first resource can represent the quantity of measurement resources used for inference in the first model (or the quantity of the first resource can represent the quantity of measurement resources used for one inference operation of the first model); and the time interval of the first resource can represent the time interval of the measurement resources used for inference in the first model. The measurement resources used for inference in the first model are resources actually sent during the inference phase. The first resource may or may not be an actually sent resource.
[0245] At least one predicted instance is a predicted instance output by the first model's inference. This can be understood as at least one predicted instance being a predicted instance output by the first model in one inference iteration. In other words, the number of at least one predicted instance is the number of predicted instances output by the first model in one inference iteration. Each predicted instance in the at least one predicted instance corresponds to one time unit. Assuming the number of predicted instances output in one inference iteration is N, when N is greater than or equal to 2, there is a one-to-one correspondence between N predicted instances and N time units. Two adjacent predicted instances among the N predicted instances refer to two instances whose corresponding time units are adjacent.
[0246] The type of output result corresponding to the inference of the first model can vary depending on the model type, training algorithm, and application field.
[0247] In this application embodiment, a time unit may refer to a moment, a time period, or a time instance. The time unit can be a second (s), millisecond (ms), microsecond (us), frame, subframe, slot, symbol, or at least one continuous symbol, etc. A time unit can be replaced with a moment or a time period. This application embodiment does not impose any limitations on this.
[0248] In this embodiment, the time interval between two time units can refer to the time interval between two instants or the time interval between two time periods. The time interval between two time periods can be the time interval between the start times of the two time periods or the time interval between the end times of the two time periods. The time unit containing "xx" can refer to the sending time of "xx" or the receiving time of "xx". "xx" can be replaced with information such as "resource", "CSI report", or "DCI". The sending time can be the start time of sending or the end time of sending. The receiving time can be the start time of receiving or the end time of receiving.
[0249] In this embodiment, the inference parameters included in the candidate inference parameter set may be indicated by the first information, or they may be predefined or preconfigured by the protocol. This embodiment does not impose any restrictions on this. For ease of understanding, this embodiment uses the example of inference parameters indicated by the first information for description.
[0250] The aforementioned various types of inference parameters indicate the base station's inference requirements. For example, the number of at least one predicted instance indicates that the base station's inference requirement is the number of predicted instances that can be obtained through a single model inference. In this case, the UE can learn about the base station's inference requirements based on the first information.
[0251] It should also be noted that the aforementioned candidate inference parameters indicate that these inference parameters may not have been applied in the actual model inference stage. In other words, the purpose of these inference parameters is to help the UE train a model that is well-matched to the base station's inference requirements. During the model inference stage, the base station can also send other measurement resources to the UE for model inference. The measurement resources may be the same as or different from the first resource, and this application embodiment does not impose any limitations on this.
[0252] It should also be noted that the first model refers to the inference model generated by the UE through training that meets the inference requirements of the base station. This first model is used to perform the inference tasks indicated by the base station. The first model may correspond to one or more prediction or inference tasks. The number of physical models obtained from training the first model may be one or more, depending on the UE implementation. For ease of description, the first model may also be called a model or an inference model, which is not limited in this document.
[0253] In this application, a model may refer to a task or a function, and is not necessarily an actual physical model at the UE. In this case, "model" can be used interchangeably with "task" or "function".
[0254] Step 1002: Based on the first information, the UE performs training data collection for the first model and / or training of the first model.
[0255] In the embodiments of this application, the candidate inference parameter set indicated by the first information can be used for the collection of training data for the first model and / or the training of the first model. This can be understood as follows: the candidate inference parameter set indicated by the first information can be used for potential model training and / or training data collection; the candidate inference parameter set indicated by the first information can be used as auxiliary information for model training and / or training data collection; the candidate inference parameter set indicated by the first information can be used to determine measurement information (input information) and / or label information in the training data.
[0256] In this embodiment of the application, candidate inference parameters can be understood as preferred inference parameters, potential inference parameters, or target inference parameters.
[0257] For example, the first information is carried in the CSI-ReportConfig used for training data collection; that is, the base station sends the CSI-ReportConfig to the UE, which includes the first information indicating a set of candidate inference parameters. In this embodiment, the set of candidate inference parameters indicated by the first information is not used to report a CSI report. That is, after receiving the set of candidate inference parameters indicated by the first information, the UE does not report CSI information.
[0258] Depending on the type of inference parameters, the UE may perform model training data collection or model training based on the inference requirements of the base station indicated by the first information. For example, the first information may indicate that the temporal behavior of the resource used for inference (the first resource) can include at least one of the following: periodic, semi-persistent, or aperiodic. For instance, the temporal behavior of the first resource is aperiodic. In this case, the temporal behavior of the measurement resources configured by the base station may be any one or more of the above. During the model training data collection phase, the UE can divide the measurement results corresponding to the measurement resources into several groups based on the first information. Each group can be considered to contain several aperiodic measurement results, and each group of measurement results can be used as input data for training, thereby satisfying the base station's inference requirements related to temporal behavior.
[0259] For example, the candidate inference parameter set indicated by the first information includes inference parameters of a type that includes at least one number of predicted instances. For time-domain prediction models, such as time-domain beamforming and CSI prediction, the maximum number of predicted instances that the model can predict after training is fixed. During the training data collection phase, the base station can help the UE train a model that meets the base station's requirements by indicating the required number of predicted instances, thus avoiding situations where the number of predicted instances that the model trained by the UE can predict does not meet the base station's requirements.
[0260] For example, the candidate inference parameter set indicated by the first information includes inference parameters of types including the time interval between time units corresponding to at least two adjacent prediction instances in the prediction instance. For time-domain prediction models, such as time-domain beamforming and CSI prediction, the time interval between prediction instances that the model can support after training is limited. During the training data collection phase, the base station can help the UE train a model that meets the base station's requirements by indicating the required time interval of prediction instances, thus avoiding situations where the time interval of prediction instances that the model trained by the UE can predict does not meet the base station's requirements. In one possible implementation, the time interval between prediction instances in the first information is optional, that is, the base station may not explicitly indicate the required time interval of prediction instances, and the UE can determine the time interval of prediction instances based on the time interval (or period) of the configured training resources. For example, for time-domain beamforming, the UE can determine the time interval of prediction instances based on the time interval (or period) of the configured Set A resources. As another example, for CSI prediction, the UE can determine the time interval of prediction instances based on the time interval (or period) of the configured CSI-RS resources. In this case, it can also be understood that the base station implicitly indicates the time interval of the predicted instance by configuring the time interval (or period) of the training resources. In another possible implementation, the time interval of the first resource is not equal to the time interval of the predicted instance (such as when performing CSI prediction for non-periodic resources), and the time interval of the training resources configured by the base station is the same as the time interval of the first resource. In this case, the time interval of the predicted instance in the first information is mandatory.
[0261] For example, the candidate inference parameter set indicated by the first information includes inference parameters of a type that includes the time unit where the first predicted instance is located. For time-domain prediction models, such as time-domain beamforming and CSI prediction, the time unit where the first predicted instance can be supported after model training is limited. During the training data collection phase, the base station can help the UE train a model that meets the base station's requirements by indicating the required time unit where the first predicted instance is located, thus avoiding situations where the time unit where the first predicted instance can be predicted by the UE's trained model does not meet the base station's requirements. The time unit where the first predicted instance is located indicates the time offset of the first predicted instance relative to the observation instance, used for pairing model input data and label data. In one possible implementation, the time unit where the first predicted instance is located in the first information is optional, that is, the base station may not explicitly indicate the required time unit where the first predicted instance is located, and the UE can determine the time unit where the first predicted instance is located according to preset rules. Preset rules may include: determining the pairing of model input data and label data according to the offset value of the first predicted instance relative to the observation instance predefined in the protocol, or specifying that each model input data is paired with the nearest label data after that input data.
[0262] For example, the candidate inference parameter set indicated by the first information includes inference parameters of types including the temporal behavior of the first resource. Since model training requires collecting a large amount of data, if aperiodic resources are used for training data collection, the base station needs to frequently trigger the transmission of aperiodic resources. Therefore, the base station may prefer to use periodic or semi-persistent resources, while the base station's inference requirements may be for aperiodic resources. During the training data collection phase, by indicating the required temporal behavior of the first resource, the base station can help the UE train a model that meets the base station's requirements, avoiding situations where the temporal behavior of the measurement resources applicable to the UE's trained model does not meet the base station's requirements. In one possible implementation, the temporal behavior of the first resource in the first information is optional; that is, the base station may not explicitly indicate the required temporal behavior of the first resource, and the UE can determine the temporal behavior of the first resource based on the configured temporal behavior of the training resources. For example, for temporal beam prediction, the UE can determine the temporal behavior of the first resource based on the configured temporal behavior of the Set B resource; for CSI prediction, the UE can determine the temporal behavior of the first resource based on the configured temporal behavior of the CSI-RS resource. In this case, it can also be understood that the base station implicitly indicates the temporal behavior of the first resource required by the temporal behavior of the configured training resources.
[0263] For example, the candidate inference parameter set indicated by the first information includes inference parameter types such as the time interval of the first resource. The time interval of the first resource that can be supported after model training is limited. During the training data collection phase, the base station can help the UE train a model that meets the base station's requirements by indicating the required time interval of the first resource, avoiding situations where the time interval of the first resource applicable to the UE's trained model does not meet the base station's requirements. In one possible implementation, the time interval of the first resource in the first information is optional; that is, the base station may not explicitly indicate the required time interval of the first resource, and the UE can determine the time interval of the first resource based on the configured time interval (or period) of the training resources. For example, for time-domain beam prediction, the UE can determine the time interval of the first resource based on the configured time interval (or period) of the Set B resources. As another example, for CSI prediction, the UE can determine the time interval of the first resource based on the configured time interval (or period) of the CSI-RS resources. In this case, it can also be understood that the base station implicitly indicates the required time interval of the first resource through the configured time interval (or period) of the training resources. In another possible implementation, the time interval of the first resource is not equal to the time interval of the predicted instance (such as when performing CSI prediction for aperiodic resources). If the time interval of the training resources configured by the base station is the same as the time interval of the predicted instance, then the time interval of the first resource in the first information is a mandatory option.
[0264] For example, the set of candidate inference parameters indicated by the first information includes the type of inference parameters, including the number of first resources. For time-domain prediction models, such as time-domain beamforming and CSI prediction, the minimum number of first resources required after model training is fixed. During the training data collection phase, the base station can help the UE train a model that meets the base station's requirements by indicating the required number of first resources, avoiding the situation where the number of first resources required by the UE's trained model exceeds the number of first resources that the base station can configure. In one possible implementation, the number of first resources in the first information is optional, meaning the base station may not explicitly indicate the required number of first resources. For example, when the training resources configured by the base station are periodic or semi-persistent, and the required first resources are also periodic or semi-persistent, the UE can determine the number of first resources itself; when the training resources configured by the base station are aperiodic, the UE can determine the number of first resources based on the number of configured aperiodic resources. In another possible implementation, when the training resources configured by the base station are periodic or semi-persistent, and the required first resources are aperiodic, the number of first resources in the first information is mandatory. Generally, for the same prediction capability, the UE requires more aperiodic measurement resources than periodic or semi-persistent measurement resources. If the base station does not additionally indicate the number of primary resources to the UE, the UE, upon receiving periodic or semi-persistent measurement resources, may mistakenly assume that the base station's inference requirement is also for periodic or semi-persistent measurement resources. For example, if the model trained by the UE can infer one prediction instance using four periodic or semi-persistent observation instances, during the model inference phase, the base station might configure four non-periodic measurement resources to instruct the UE to perform inference. However, the UE cannot perform accurate inference using only these four non-periodic measurement resources (the UE may require at least six non-periodic measurement resources), thus rendering the trained model unusable for base station inference.
[0265] As mentioned above, through step 1001, the UE can learn about the base station's inference requirements. In this case, the UE can perform training data collection and / or training of the first model according to the base station's inference requirements, which can train an inference model that meets the base station's inference requirements as much as possible, improve the matching degree between the inference model and the base station's inference requirements, and help improve the performance of subsequent predictions and avoid wasting training resources.
[0266] In some embodiments, the information configuration method 1000 further includes:
[0267] Step 1003: The base station sends second information to the UE, and the UE receives the second information accordingly. The second information is used to indicate second resources, which include measurement resources for training the first model and / or collecting training data for the first model.
[0268] For example, the second information is carried in the CSI-ReportConfig used for training data collection. That is, the base station sends the CSI-ReportConfig to the UE, which includes second information indicating the second resource. The second resource may be a reference signal resource configured by the CSI-ResourceConfig. The second information contained in the CSI-ReportConfig is the identifier of the CSI-ResourceConfig that configures the second resource.
[0269] Specifically, the temporal behavior of the first resource differs from that of the second resource, and / or the time intervals of the first resource and the second resource differ. In other words, the temporal behavior of the measurement resource used for inference of the first model differs from that of the measurement resource used for training the first model, and / or the time intervals of the measurement resource used for inference of the first model differ from those of the measurement resource used for training the first model. For example, the temporal behavior of the measurement resource used for inference of the first model is aperiodic, while the temporal behavior of the measurement resource used for training the first model is periodic or semi-continuous. Another example: the time interval of the measurement resource used for inference of the first model is 10ms, and the time interval of the measurement resource used for training the first model is 5ms.
[0270] Similar to the first resource, the temporal behavior and time interval of the second resource will not be described again. It should be understood that the second resource is a measurement resource actually sent by the base station to the UE for training the first model. It is worth noting that, to meet the base station's inference requirements, the UE may not collect model training data solely based on the second resource. For example, after receiving the second information, the UE may utilize the second resource and candidate values of various types of inference parameters in the candidate inference parameter set indicated by the first information to collect model training data. During the training data collection phase, the first resource may or may not be a resource actually sent by the base station to the UE. For example, the first resource may not be a resource actually sent by the base station; the parameters of the first resource indicated in the first information may be used to determine the training data needed to meet the base station's inference requirements from the second resource.
[0271] For example, the time interval of the second resource might be 5ms. However, the base station's inference requirement might be to infer from observation instances with a time interval of 10ms. In this case, the inference parameters indicated by the first information might include a time interval of 10ms for the first resource. That is, during the inference phase, the time interval of the measurement resources sent by the base station for inference might be 10ms. If the UE is unaware of the base station's inference requirements, it might directly use the measurement results corresponding to the second resource with a time interval of 5ms as the training input data, resulting in a model that cannot support the base station's inference requirements. If the UE is aware of the base station's inference requirements, it might select the measurement results corresponding to the 10ms time interval in the second resource with a time interval of 5ms as the training input data, thus enabling the trained model to support the base station's inference requirements.
[0272] For example, the temporal behavior of the second resource may be periodic or semi-persistent. However, the base station's inference requirements may be for observation instances with aperiodic temporal behavior. In this case, the inference parameters indicated by the first information may include that the temporal behavior of the first resource is aperiodic. That is, during the inference phase, the measurement resources sent by the base station for inference may be aperiodic resources. If the UE is unaware of the base station's inference requirements, it may directly use multiple measurement results corresponding to the periodic or semi-persistent second resource as training input data, and the trained model will not support the base station's inference requirements. If the UE is aware of the base station's inference requirements, it may divide the multiple measurement results of the periodic or semi-persistent second resource into several groups, each group can be considered to contain several aperiodic measurement results, and use each group of measurement results as training input data, so that the trained model can support the base station's inference requirements.
[0273] In both examples above, although the base station indicates a second resource for model training, the UE may further determine the actual data used for model training based on the base station's inference requirements.
[0274] In some embodiments, the temporal behavior of the measurement resource (second resource) configured by the base station is periodic / semi-persistent, and the inference parameters indicated by the base station may indicate that the temporal behavior of the resource (first resource) used for inference is aperiodic. Alternatively, the inference parameters indicated by the base station may indicate that the temporal behavior of the resource (first resource) used for inference may include one or more of the following: periodic, semi-persistent, or aperiodic.
[0275] For example, for beam prediction or CSI prediction, the temporal behavior of the measurement resources configured by the base station for training data collection is periodic or semi-persistent. The base station can indicate, through first information, that the temporal behavior of the resources used for inference (first resource) is aperiodic, or that the temporal behavior of the resources used for inference is both aperiodic and periodic, or that the temporal behavior of the resources used for inference is aperiodic, periodic, and semi-persistent. Then the UE can train the model according to the temporal behavior indicated by the base station. For example, when the base station indicates that the temporal behavior of the resources used for inference is aperiodic, periodic, and semi-persistent, the UE can use the periodic / semi-persistent second resource to train a model that simultaneously supports inference using aperiodic, periodic, and semi-persistent resources.
[0276] It should be understood that the temporal behavior of the resource used for inference, i.e., the first resource, can be either indicated by the first information or predefined or preconfigured. For example, it can be specified that the temporal behavior of the measurement resource used for model training data collection is periodic or semi-persistent. The UE will always collect model training data and / or train the model according to the three cases of the temporal behavior of the measurement resource, including periodic, semi-persistent, and aperiodic. In other words, the UE will by default train a model that supports all three temporal behaviors. In this case, during the model inference stage, the base station can configure the observation instance with any one of the temporal behaviors to the UE, and the UE can use the model to infer prediction instances that support all three temporal behaviors, which is beneficial to improving the adaptability to different application scenarios and meeting the inference requirements of the base station.
[0277] It should also be understood that the various types of inference parameters indicated by the first information may include combinations or superpositions of different types of inference parameters. For example, if the temporal behavior of the second resource configured by the base station is periodic or semi-persistent, the type of inference parameter indicated by the base station through the first information may include the temporal behavior of the first resource being aperiodic. Furthermore, other types of inference parameters may be indicated based on the temporal behavior. For example, the type of inference parameter indicated by the first information may include: the number of first resources with aperiodic temporal behavior or the time interval of the first resources with aperiodic temporal behavior.
[0278] For example, for beam prediction or CSI prediction, the type of inference parameters indicated by the base station through the first information may include the first resource having a non-periodic temporal behavior, and the number of the first resources (the number of observation instances in the first resource) being at most 8. For example, an observation resource may refer to a CSI-RS resource; that is, for CSI prediction, the base station may configure a non-periodic CSI-RS resource set with fewer than 8 CSI-RS resources. As another example, an observation resource may refer to a CSI-RS resource set, and a CSI-RS resource set corresponds to a beam set; that is, for beam prediction, the base station may configure a non-periodic CSI-RS resource set with fewer than 8 CSI-RS resources. The UE performs model training according to the number of non-periodic observation resources indicated by the base station.
[0279] For example, in CSI prediction, the temporal behavior of the second resource configured by the base station is periodic or semi-persistent, with a period of 1ms. The base station may also allow the type of inference parameters indicated by the first information to include a non-periodic temporal behavior of the first resource with an interval of 2ms. In this case, the UE trains the model according to the temporal behavior and period of the measurement resource that meets the inference requirements as indicated by the base station, training a model that supports prediction instances with a non-periodic temporal behavior and a period of 2ms.
[0280] For time-domain prediction models such as BM-Case2 and CSI prediction, the time offset between the time unit of the first predicted instance and the time unit of the observed instance is usually fixed when the UE generates the model and uses it for inference. Furthermore, given a fixed number of predicted instances and a fixed time interval between adjacent predicted instances, the UE can determine the time units of all predicted instances using the time unit of one predicted instance. Therefore, during the model training data collection phase and / or the model training phase, the type of inference parameters indicated by the base station can include the time unit of the first predicted instance, so that the model trained by the UE meets the base station's inference requirements. In this case, there may be multiple ways to indicate the time unit of the first predicted instance.
[0281] In some embodiments, the type of the inference parameters described above includes the time unit in which the first predicted instance is located. In this case, the first information includes a first time offset, which indicates the time offset between the time unit in which the first predicted instance is located and a first reference time unit.
[0282] In one possible implementation, the first resource includes at least one measurement resource for inference of the first model, and the first reference time unit is the time unit of any one of the at least one measurement resource. Specifically, the first reference time unit may be the time unit of any one of the multiple observation instances corresponding to one inference. Given a fixed number of observation instances and a fixed time interval between adjacent observation instances, the UE can determine the time offset between the time unit of the first prediction instance and the latest (newest) observation instance among the multiple observation instances corresponding to one inference by using the time offset between the time unit of the first prediction instance and the time unit of any one observation instance. For example, any one observation instance is the latest (newest) observation instance among the multiple observation instances corresponding to one inference. For instance, if the first model requires P observation instances to perform one inference, optionally, the first reference time unit may be the time unit of the latest (newest) observation instance among the P observation instances, and the latest observation instance is the latest observation instance in the corresponding time unit among the P observation instances. In this case, the first time offset can indicate the time offset between the time unit of the first prediction instance and the time unit of the latest (newest) observation instance among the multiple observation instances corresponding to one inference. The UE can train a model that meets the base station's inference requirements based on this first time offset.
[0283] For example, during the model training data collection phase, measurement resources and prediction resources may each correspond to two independent CSI-ResourceConfigs. For instance, in beam prediction training data collection, measurement resources / observation instances might be used to determine the model training data (model input). Prediction resources / prediction instances might be used to determine the labels for model training. In this case, in addition to the relevant information of the first resource, the base station can indicate the first time offset through the first information. In this case, the first time offset can be the time offset between the time unit where the first prediction instance is located and the time unit where any observation instance is located (e.g., if an observation resource contains multiple CSI-RSs, it can be relative to any CSI-RS within any observation resource). That is, adding this first time offset to the time unit where the observation resource is located gives the time unit where the first prediction instance is located. The first time offset can be represented by the number of time units, or by the number of observation resources and the number of CSI-RSs.
[0284] For example, during the model training data collection phase, measurement resources and prediction resources may correspond to the same CSI-ResourceConfig. For instance, in the collection of training data for CSI prediction, the resources configured in CSI-ResourceConfig are used both to determine the model training data (model input) and to determine the model training labels. In this case, in addition to the relevant information of the first resource, the base station can also indicate the first time offset through the first information. This process can be similar to the one described above and will not be elaborated further here.
[0285] In some embodiments, the type of the inference parameters includes the time unit in which the first prediction instance is located. In this case, the first information includes a second time offset, which indicates the time offset between the time unit in which the first channel state information (CSI) report is located and the first reference time unit, the first CSI report corresponding to the inference of the first model.
[0286] The first Channel State Information (CSI) report corresponds to the inference of the first model. It can be understood that the first Channel State Information (CSI) report is used to report the inference results of the first model.
[0287] For example, for CSI prediction, the base station can indicate the time offset δ between the time unit of the first predicted instance and the time unit of the CSI report. In this case, the UE can determine the time offset between the time unit of the first predicted instance and the first reference time unit based on the time offset between the time unit of the CSI report and the first reference time unit, as well as the time offset δ. The first reference time unit may be any one of the multiple observation instances corresponding to one inference. For example, the first reference time unit is the time unit of the latest (newest) observation instance among the multiple observation instances corresponding to one inference. Then, the UE can determine the time offset between the time unit of the first predicted instance and the time unit of the latest (newest) observation instance among the multiple observation instances based on the second time offset and the time offset δ. It should be noted that once the inference model is determined, the model's inference capability is also determined. However, the base station may not be able to determine the accurate time offset between the time unit of the first channel state information (CSI) report and the first reference time unit before inference. To help the UE train a model that meets the base station's inference requirements, the second time offset can indicate the maximum time offset between the time unit of the CSI report and the first reference time unit during the subsequent model inference phase. In other words, during the subsequent model inference phase, the time offset between the time unit where the CSI report is located and the first reference time unit will not exceed this maximum time offset. In this case, the base station can indicate the maximum time offset of the time unit where the CSI report is located relative to any observation instance through the first information. For example, the base station can indicate the second time offset between the time unit where the CSI report is located and the first reference time unit through the first information. Combined with δ, the UE can determine the maximum time offset of the first predicted instance relative to the observed instance. The UE trains the model based on this maximum time offset, which helps ensure that the base station's inference requirements do not exceed the inference range of the inference model, thereby helping the model trained by the UE to meet the base station's inference requirements.
[0288] It should be noted that in the three examples above, in order to accurately determine the time unit of the first predicted instance during the training data collection phase by using the time offset between the first reference time unit (e.g., the time unit where any observation instance is located) and the time unit where the first predicted instance is located, it is necessary to assume that the time interval between two adjacent observation instances (i.e., the time interval of the measurement resources used to acquire the training input data corresponding to the observation instance) and the time interval between two adjacent predicted instances (i.e., the period of the measurement resources used to acquire the training label data corresponding to the predicted instance) are regular. For example, the time interval between two adjacent observation instances and the time interval between two adjacent predicted instances are the same or are integer multiples of each other. For periodic or semi-continuous measurement resources, the time interval of the measurement resources is also the period of the measurement resources. Therefore, it can be assumed that the period of the observation instance and the period of the predicted instance are the same or are integer multiples of each other.
[0289] In some embodiments, the candidate inference parameter set includes one or more candidate values of the same type of inference parameter. In one possible implementation, the first information also includes the value or range of values of the inference parameter. For example, when there is one candidate value of the same type of inference parameter, the single candidate value can be a specific value or a range of values. For example, when there are multiple candidate values of the same type of inference parameter, the multiple candidate values can be a list of optional specific values. For instance, the type of inference parameter indicated by the first information includes: at least one number of predicted instances. The first information can indicate one candidate value, such as indicating that the number of predicted instances is 4; or the first information can indicate one candidate value, such as indicating that the range of values for the number of predicted instances is 1 to 4 (representing that the values can be 1, 2, 3, 4); or the first information can indicate multiple candidate values, such as indicating that the list of values for the number of predicted instances is {1, 2, 4}.
[0290] To meet the diverse inference needs of base stations, inference parameters may be of various types. Furthermore, to train models that satisfy different inference requirements using the same measurement resource (e.g., a second resource), the number of candidate values for the same type of inference parameter indicated by the base station may be multiple. For example, the first information may indicate multiple specific values for the quantity of the first resource, or it may indicate a set of values or a range of values for the quantity of the first resource. The UE can utilize the same measurement resource to train the model and / or collect training data for the model using multiple candidate values, thereby saving training resources.
[0291] In one possible design, prior to step 1001, method 1000 further includes:
[0292] In step 1004, the UE sends third information to the base station, and the base station receives the third information accordingly. The third information is used to indicate whether the UE has the ability to train on multiple candidate values for the same type of inference parameters.
[0293] Through this third piece of information, the base station can determine whether the UE has the ability to process multiple candidate values for the same type of inference parameter.
[0294] For example, the third information indicates that the UE has the ability to train for multiple candidate values of the same type of inference parameter, which can be understood as the UE having the ability to train for multiple candidate values of the same type of inference parameter simultaneously using a single measurement resource.
[0295] For example, if the UE has this capability, the base station can indicate the quantity of the first resource through the first information, which may be a set of values, such as {2, 4, 7}. In one possible implementation, the UE can use the above multiple values to perform model training data collection and / or model training. The base station does not need to configure a separate measurement resource for each value of the quantity of the first resource, which helps to save measurement resources.
[0296] To save on model training resource overhead, the base station can configure a measurement resource for training, which may correspond to multiple sets of inference parameters. The UE trains a model that supports multiple sets of inference parameters based on this measurement resource.
[0297] In one possible implementation, the set of candidate inference parameters includes multiple sets of candidate inference parameters.
[0298] For example, the base station can indicate multiple sets of candidate inference parameters through first information, each set of candidate inference parameters may include different types of inference parameters. For example, the types of inference parameters included in the candidate inference parameter set may include: the number of at least one predicted instance, and the time interval between the time units corresponding to two adjacent predicted instances in the at least one predicted instance. For CSI prediction, the base station can configure measurement resources with a period of 10ms for training, while indicating two sets of candidate inference parameters. The first set of parameters indicates that the number of at least one predicted instance is 4, and the time interval between the time units corresponding to two adjacent predicted instances in the at least one predicted instance is 20ms. The second set of parameters indicates that the number of at least one predicted instance is 2, and the time interval between the time units corresponding to two adjacent predicted instances in the at least one predicted instance is 80ms. In this case, the UE can utilize multiple sets of candidate inference parameters to perform model training data collection and / or model training, which is beneficial for meeting the base station's various inference preferences and saving training resource overhead.
[0299] In some embodiments, after step 1001, method 1000 further includes:
[0300] In step 1005, the UE sends fourth information to the base station, and the base station receives the fourth information accordingly. The fourth information is used to indicate the inference parameters supported by the UE.
[0301] Through the fourth piece of information, the base station can obtain the UE's support capability for inference parameters. The base station can indicate at least one set of candidate inference parameters based on the UE's support capability, whereby the inference parameters indicated by the base station are the inference parameters that the UE can support. For example, the information on the inference parameters supported by the UE may include: constraints on the relationship between the inference parameters and the configuration parameters of the training resources, or a list of multiple optional inference parameters. The base station can determine the at least one set of candidate inference parameters to be indicated based on the UE's capabilities. For example, regarding the time interval of the resources, the constraints on the relationship between the inference parameters and the configuration parameters of the training resources include: the relationship between the time interval of the measurement resources used for inference and the time interval of the training resources, such as the time interval of the measurement resources used for inference being greater than the time interval of the training resources, or the time interval of the measurement resources used for inference being an integer multiple of the time interval of the training resources. For example, regarding the temporal behavior of the resources, the constraints on the relationship between the inference parameters and the configuration parameters of the training resources include: the relationship between the temporal behavior of the measurement resources used for inference and the temporal behavior of the training resources, such as when the temporal behavior of the measurement resources used for training is periodic or semi-persistent, the temporal behavior of the measurement resources used for inference can be periodic, semi-persistent, or aperiodic.
[0302] For example, the relationship constraints between the inference parameters and the configuration parameters of the training resources for the time interval of the predicted instances include: the relationship between the time interval of the predicted instances of inference and the time interval of the training resources, such as the time interval of the predicted instances of inference being greater than the time interval of the training resources, or the time interval of the predicted instances of inference being an integer multiple of the time interval of the training resources.
[0303] For example, in step 1004, the UE indicates through third information that it has the ability to process multiple candidate values for the same type of inference parameter. In step 1005, the UE indicates through third information that for a specific type of inference parameter, it supports using information on multiple candidate values simultaneously with a single measurement resource. The information on multiple candidate values includes constraints on the relationship between multiple candidate values and the configuration parameters of the training resource, or a list of multiple optional candidate values.
[0304] In this scenario, since the base station can know the UE's processing capability for inference parameters, the base station can indicate candidate inference parameters that meet the UE's capabilities through the first information. The UE can then train a model that meets the base station's inference requirements without exceeding its own capabilities.
[0305] It should be noted that in method 1000, step 1004 can occur before step 1001, step 1002 can occur after steps 1001, 1003, and 1004, step 1005 can occur before step 1001, and steps 1005 and 1004 can be performed simultaneously, or step 1005 can be executed after step 1004. Other steps may have a specific order or may be performed simultaneously; this embodiment does not impose any restrictions on this.
[0306] In another possible implementation, the first information may further include the identifier of at least one set of candidate inference parameters among the plurality of candidate inference parameters.
[0307] For example, for both the base station and the UE, the multiple sets of candidate inference parameters and their corresponding identifiers can be known information. For instance, the protocol may predefine multiple sets of optional inference parameters and their corresponding identifiers, or the base station may pre-indicate multiple sets of optional inference parameters and their corresponding identifiers, or the UE may pre-report multiple sets of optional inference parameters and their corresponding identifiers. Thus, the base station can select preferred inference parameters from the multiple sets of optional inference parameters for indication. For example, the base station can use the first information to indicate the identifiers of at least one set of candidate inference parameters. The UE can obtain the corresponding at least one set of candidate inference parameters based on the identifier, thereby performing model training data collection and / or model training. Furthermore, the values or value ranges of the candidate inference parameters may also be predefined by the protocol or reported by the UE, from which the base station can select the values or value ranges of candidate inference parameters that meet the inference requirements and indicate them using the first information.
[0308] In some embodiments, the type of the output result corresponding to the inference of the first model may include at least one of the following: Reference Signal Received Power (RSRP), Resource Identifier of Target Reference Signal, Precoding Matrix Indicator (PMI), or Channel Quality Indicator (CQI).
[0309] RSRP (Reference Signal Power Ratio) is a metric for measuring the received signal strength. It represents the power of the reference signal transmitted by the base station at the receiver and is used to evaluate the quality of the wireless signal. The resource identifier of the target reference signal typically refers to the identifier of a specific reference signal resource (e.g., the optimal beam or optimal beam candidate). PMI (Position Quality Index) can be used to describe the matrix used by the base station during precoding in a multi-antenna system, helping the UE (User Equipment) report the spatial characteristics of the received signal. CQI (Channel Quality Index) is a channel quality metric reported by the UE to the base station, used to evaluate the channel's signal-to-noise ratio or channel capacity. During the model inference phase, the output results may include the above types. Choosing the appropriate type for output in different application scenarios is beneficial for optimizing resource allocation and utilization, and improving the performance of the communication system. Taking beam prediction as an example, the output results of the first model's inference may include: the resource identifier of the reference signal corresponding to the optimal beam and / or the RSRP of the reference signal corresponding to the optimal beam. When the AI model corresponding to the first model is a regression model, the type of the output result corresponding to the inference of the first model can be the Reference Signal Received Power (RSRP) of the optimal beam. When the AI model corresponding to the first model is a classification model, the type of the output result corresponding to the inference of the first model can be the Resource Identifier (RISA) of the reference signal corresponding to the optimal beam. Taking CSI prediction as an example, the type of the output result corresponding to the inference of the first model can include: Precoding Matrix Indicator (PMI) and / or Channel Quality Indicator (CQI).
[0310] For example, the first information may also indicate that the type of the output result includes the resource identifier of the target reference signal, and the type of the inference parameter may also include the number of resource identifiers of the target reference signal, such as the number of optimal beams reported in beam prediction.
[0311] The type of output result corresponding to inference is related to the training algorithm. The base station indicates the required type of output result during the training data collection phase, which can help the UE train a model that meets the base station's inference requirements and avoid wasting training resources.
[0312] As mentioned earlier, the inference model can be applied to inference tasks, such as CSI inference or reporting tasks. In this embodiment, CSI report and CSI reporting may have the same meaning. For example, in the non-periodic CSI reporting process, the base station can configure a non-periodic CSI-ReportConfig and then trigger a non-periodic CSI report by sending a DCI to the UE. The non-periodic CSI-ReportConfig will configure an offset value list (reportSlotOffsetList) indicating selectable offset values for the time unit where the CSI report is located. This offset value is the time offset between the time unit where the CSI report is located and the time unit where the DCI is located. When the UE receives the DCI, the DCI will indicate the offset value of the time unit where the CSI report triggered by the DCI is located. The offset value indicated by the DCI is selected from the offset value list. Combining the time unit where the DCI is located, the UE can determine the time unit where the CSI report triggered by the DCI is located. A DCI may trigger one CSI report and indicate one offset value, or it may trigger multiple CSI reports and indicate multiple offset values; this embodiment does not limit this.
[0313] For CSI prediction, the current protocol stipulates that in the N4 predicted instances reported by the UE, the time slot of the first predicted instance is l = n + δ, where n is the time unit of the CSI report and δ is the time offset configured by higher-layer parameters. For example, if N4 is 4, it means that the CSI report to be reported by the UE includes 4 predicted instances. In this case, δ ∈ {-n} CSI_ref ,0,1,2} represents the time offset between the time unit where the CSI report is located and the time unit where the first prediction instance among the four prediction instances in the CSI report is located, which is -n respectively. CSI_ref ,0,1,2. When δ=-n CSI_ref When the time unit of the first predicted instance is the same as the time unit of the CSI reference resource corresponding to the CSI report, it indicates that the time unit of the CSI reference resource corresponding to a CSI report is located before the time unit of the CSI report itself, and the time offset between the two time units is -n. CSI_ref .
[0314] As shown in Figure 9, in step 903, for the inference requirement of non-periodic CSI reporting, the base station can send an RRCReconfiguration message to the UE. This message may include an applicability request to request the UE to report applicability. The applicability request contains the CSI-ReportConfig of non-periodic CSI reporting or a set of inference parameters, wherein the inference parameters include information elements in the CSI-ReportConfig.
[0315] As mentioned earlier, once the inference model is determined, its inference capability is also determined. That is, during the model inference phase, the time offset between the time unit of the first predicted instance and the time unit of the latest observed instance is determined. However, in step 903, the base station has not yet sent the DCI used to trigger CSI reporting to the UE. In other words, when the base station requests applicability reporting from the UE, the UE cannot determine the time unit of the CSI report indicated by the subsequent DCI triggering aperiodic reporting, and therefore cannot determine the time unit of the first predicted instance associated with the inference task during subsequent model inference. Furthermore, the UE cannot determine the time offset between the time unit of the first predicted instance associated with the inference task and the time unit of the latest observed instance. In other words, the UE cannot determine whether the time unit of the first predicted instance associated with the inference task will exceed the model's inference capability during subsequent model inference.
[0316] Figure 11 illustrates a possible inference model for CSI prediction. Assume the UE has a model that supports generating two predicted instances from four observation instances. The time interval between all adjacent observation instances or two predicted instances is 5ms, and the time between the first predicted instance and the last observation instance is 5ms. That is, based on this model, the UE can predict the predicted instance up to 10ms after the last observation instance. For example, for a CSI prediction task, the base station's inference requirement is two predicted instances. If the DCI indicates offset 1, the time unit where the CSI report is located corresponds to PUSCH 1 in Figure 11, and the UE can report two predicted instances; the model is applicable. If the DCI indicates offset 2, the time unit where the CSI report is located corresponds to PUSCH 2 in Figure 11, and the UE cannot report two predicted instances, exceeding the model's inference capability; the model is not applicable. However, as mentioned earlier, when the base station requests applicability reporting from the UE, the base station has not yet sent the DCI used to trigger CSI reporting to the UE. The UE cannot determine whether the CSI prediction task exceeds the model's reasoning ability; that is, the UE cannot determine the applicability of aperiodic CSI reporting.
[0317] In view of this, embodiments of this application also provide an information configuration method and apparatus. A first apparatus (such as a UE) performing model inference and / or applicability reporting can receive a time offset between a time unit and a second reference time unit, as indicated by a second apparatus (such as a base station), for at least one CSI report associated with the inference task, or a time offset between the time unit of a first prediction instance among at least one prediction instance associated with the inference task and the second reference time unit. In this case, through the aforementioned time offset and second reference time unit, the first apparatus can determine the time offset between the time unit of the CSI report or the first prediction instance and the time unit of the observation instance, thereby determining whether the inference task exceeds the inference capability of the model, or selecting a model that meets the inference requirements of the second apparatus based on the time offset, which helps the first apparatus determine the applicability of the inference task.
[0318] In the embodiments of this application, the first device may also be a device for performing model inference and / or applicability reporting, and the second device may also be a device for sending an applicability request and other related configurations to the first device. In the following description, for ease of description, the information configuration method provided in the embodiments of this application will be described using the example of a UE as the first device and a base station as the second device.
[0319] The following will describe in detail another information configuration method provided by the embodiments of this application with reference to Figure 12. It should be understood that the technical solution of this application can be applied to the communication system 100 shown in Figure 1, the communication system 200 shown in Figure 2, and other systems; the embodiments of this application do not limit this application. Without loss of generality, it is assumed that the devices performing model inference and / or applicability reporting are all UEs, and the interaction process between the UE and the base station will be used as an example to describe in detail another information configuration method provided by the embodiments of this application.
[0320] Figure 12 illustrates another information configuration method 1200 provided in an embodiment of this application, which includes the following steps:
[0321] Step 1201: The base station sends fifth information to the UE, and the UE receives the fifth information accordingly. The fifth information is used to indicate the range of the third time offset or M values of the third time offset, where M is a positive integer.
[0322] The third time offset reflects the time offset between the time unit where the first prediction instance in at least one prediction instance associated with the first task is located and the time unit where the first observation instance in at least one observation instance associated with the first task is located.
[0323] The third time offset can also be understood as: the third time offset indicates the time offset between the time unit where the first task-associated at least one CSI report is located and the second reference time unit, or the time offset between the time unit where the first prediction instance in the first task-associated at least one prediction instance is located and the second reference time unit.
[0324] In this embodiment, the first task can be a CSI reporting task. For example, the first task can be a CSI reporting task for inference. A CSI reporting task for inference can be understood as a CSI reporting task that uses an artificial intelligence (AI) model for inference. The first task can be a task related to the AI model; "task" can be interchanged with "model" or "function".
[0325] For example, the first task is a dynamically scheduled task; in other words, the first task is dynamically triggered based on indication information from the base station (such as DCI). For instance, the first task is an aperiodic CSI reporting task. The first task may also be other inference tasks, which are not limited in this embodiment.
[0326] For example, the first task is a CSI prediction task; in other words, the first task is to report the CSI prediction results in a CSI report.
[0327] The at least one prediction instance associated with the first task can be understood as a prediction instance output or reported by performing a CSI prediction task once. Each prediction instance in the at least one prediction instance corresponds to a time unit. Assuming the number of prediction instances output by performing one inference is N, when N is greater than or equal to 2, the N prediction instances correspond one-to-one with the N time units. The first prediction instance can be any one of multiple prediction instances. For example, the first prediction instance is the first prediction instance among the at least one prediction instances associated with the first task, and the first prediction instance is the prediction instance with the earliest corresponding time unit among the at least one prediction instances. When the number of prediction instances and the interval between prediction instances are determined, the UE can determine the time unit of the first prediction instance based on the time unit of any other prediction instance.
[0328] The at least one observation instance associated with the first task can be understood as the observation instance required to perform one CSI prediction task, or the observation instance required to infer the at least one prediction instance associated with the first task. Each observation instance in the at least one observation instance corresponds to a time unit. The first observation instance can be any one of multiple observation instances. For example, the first observation instance is the latest observation instance among the at least one observation instances associated with the first task, and the latest observation instance is the observation instance with the latest time unit corresponding to the at least one observation instance. For aperiodic CSI reporting tasks that use aperiodic observation instances for inference, the observation instances associated with the task are all observation instances between the DCI that triggers the aperiodic CSI reporting and the CSI report that carries the reported CSI information. When the number of observation instances and the interval between observation instances are determined, the UE can determine the time unit where the latest observation instance is located based on the time unit where any one observation instance is located.
[0329] At least one CSI report associated with the first task can be understood as a CSI report submitted during the execution of the first task. For example, if the first task is an aperiodic CSI reporting task, the base station configures the first task through CSI-ReportConfig, and then triggers a CSI report based on the configuration of CSI-ReportConfig through DCI, or in other words, triggers the submission of a CSI report. In this case, the CSI report is the CSI report associated with the first task. For example, if the first task is an aperiodic CSI reporting task, the base station configures the first task through the CSI-AperiodicTriggerStateList, and then triggers one of the CSI-AperiodicTriggerStates in the CSI-AperiodicTriggerStateList via DCI. The triggered CSI-AperiodicTriggerState includes at least one CSI-AssociatedReportConfigInfo. Each CSI-AssociatedReportConfigInfo indicates a CSI-ReportConfig and a set of reference signal resources in the CSI-ReportConfig. It can be considered that each CSI-AssociatedReportConfigInfo is used to trigger the reporting of a CSI report, and the CSI report is the CSI report associated with the first task.
[0330] In this scenario, using the aforementioned time offset and second reference time unit, the UE can determine the time unit containing at least one CSI report or the time unit containing the first predicted instance. Using δ, the UE can also calculate the time unit containing the first predicted instance from the time unit containing at least one CSI report. As mentioned earlier, once the inference model is determined, its inference capability is also determined. That is, during the model inference phase, the time offset between the time unit containing the first predicted instance and the time unit containing the latest observed instance is determined. When the base station requests an applicability report from the UE, the UE can determine whether the first task exceeds the inference capability of the UE-side model by using the time offset between the time unit containing the first predicted instance associated with the first task and the time unit containing the latest observed instance. In other words, the UE can determine the applicability of the first task, or the UE can select a model with corresponding inference capabilities to determine the applicability of the first task.
[0331] In step 1202, the UE sends sixth information to the base station, and the base station receives the sixth information accordingly. The sixth information is used to indicate the applicability of the first task.
[0332] In this scenario, when the base station requests the suitability of an inference task from the UE, the UE can determine whether the inference task exceeds the model's inference capabilities, or select a model that meets the base station's requirements based on the time offset, which helps the UE determine suitability.
[0333] The applicability of the first task includes applicable or inapplicable. For example, when the UE determines that the inference task exceeds the inference capability of the model, the UE can send a sixth message indicating that the applicability of the first task is inapplicable; when the UE can select a model with corresponding inference capability based on the time offset, the UE can send a sixth message indicating that the applicability of the first task is applicable.
[0334] In some embodiments, the second reference time unit includes any one of the following: the time unit in which the first observation instance in at least one observation instance corresponding to the first task is located; or, the time unit in which the downlink control information (DCI) is located, the DCI being used to trigger the reporting of at least one CSI report associated with the first task.
[0335] For example, the second reference time unit can be any time unit containing any observation instance. For instance, it could be the time unit containing the first observation instance or the time unit containing the latest observation instance. For example, the third time offset indicates the time offset between the time unit containing any predicted instance associated with the first task and the second reference time unit. Based on the third time offset, the UE can determine the time offset between the time unit containing the first predicted instance associated with the first task and the time unit containing the latest observation instance, thereby determining the applicability of the first task. As another example, the third time offset indicates the time offset between the time unit containing at least one CSI report associated with the first task and the second reference time unit. Based on the third time offset and δ, the UE can determine the time offset between the first predicted instance associated with the first task and the time unit containing the latest observation instance, thereby determining the applicability of the first task.
[0336] For example, the second reference time unit can be the time unit where the DCI is located. For instance, the DCI can be used to trigger an aperiodic CSI report. In one optional design, it can be assumed that the time unit where the DCI is located is close to the time unit where the observation instance is located; in this case, the time unit where the DCI is located can approximate the time unit where the observation instance is located. In another optional design, the UE can determine the time unit where any observation instance is located based on the time unit where the DCI is located and the time offset of the time unit where the first observation instance is located relative to the time unit where the DCI is located. Combining this with the third time offset, the UE can determine the time offset between the first predicted instance associated with the first task and the time unit where any observation instance is located, thereby determining the applicability of the first task.
[0337] The time offset can be represented by the number of time units, or by the number of observation instances or CSI-RS, and this application embodiment does not limit this.
[0338] The time unit in which the DCI is located can be either the time unit in which the base station sends the DCI or the time unit in which the UE receives the DCI. This application embodiment does not limit this.
[0339] In the two examples above, the UE can determine whether there is an applicable model / function based on the time offset between the time unit of the first predicted instance and the time unit of any observed instance. If there is, it reports that the model / function is applicable. If not, it reports that the model / function is not applicable.
[0340] Optionally, in this embodiment, the time interval between two adjacent observation instances and the time interval between two adjacent prediction instances are regular. For example, the time interval between two adjacent observation instances and the time interval between two adjacent prediction instances are the same, or are integer multiples of each other.
[0341] It should be noted that the second reference time unit can be indicated by the base station to the UE. For example, the base station can indicate the second reference time unit through the fifth information. The second reference time unit can also be predefined or preconfigured, which helps to reduce signaling overhead.
[0342] For example, the range of the third time offset indicated by the fifth information, or the M values of the third time offset, can be used to determine the maximum value of the third time offset. For instance, if the range of the third time offset indicated by the fifth information is less than or equal to 50ms, it can be understood that the maximum value of the third time offset is 50ms. As another example, if M=1, and one value of the third time offset indicated by the fifth information is 50ms, this single value can be understood as the maximum value of the third time offset. As yet another example, if M=3, and the three values of the third time offset indicated by the fifth information are {10ms, 30ms, 50ms}, it can be understood that the maximum value of the third time offset is 50ms. Taking the third time offset indicating the time offset between the time unit where the first predicted instance associated with the first task is located and the time unit where the latest observed instance associated with the first task is located as an example, if the maximum value of the third time offset is 50ms, it can be understood that when the base station subsequently triggers CSI inference and reporting, the required time offset between the time unit where the first predicted instance is located and the time unit where the latest observed instance associated with the first task is located will not exceed 50ms. It should be understood that if the maximum value of the third time offset does not exceed the UE's inference capability, then any third time offset less than or equal to that maximum value will not exceed the UE's inference capability. In other words, if the model can support the maximum value of the third time offset, then for any time offset less than or equal to that maximum value indicated by the base station, the UE's inference model can be applied or satisfied. Therefore, by indicating the maximum value of the third time offset, the base station can help the UE determine the applicability of the first task.
[0343] For example, the sixth information includes an applicability result corresponding to a range of the third time offset or M values of the third time offset. In one possible design, the applicability result is determined based on the largest time offset value among the third time offsets. In this case, the UE determines the maximum value of the third time offset based on the range of the third time offset or the M values of the third time offset, and determines applicability based on the maximum value of the third time offset. If the maximum value of the third time offset does not exceed the UE's inference capability, the applicability result is applicable; if the maximum value of the third time offset exceeds the UE's inference capability, the applicability result is not applicable. In this case, the UE can use one applicability result to indicate the applicability of multiple inference tasks, or to indicate the applicability of multiple inference requirements within a single inference task, which helps reduce signaling overhead.
[0344] In one possible design, for the M values of the third time offset, M can be equal to 1. In this case, the UE can determine the applicability of the first task based on the third time offset and report it. In another possible design, M is greater than or equal to 2. In this case, the first task can correspond to multiple inference requirements. When the base station uses the fifth information to indicate multiple inference requirements of the first task to the UE and requests applicability reporting, in one possible implementation, if the UE's model can support all selectable values in the multiple third time offsets, the UE reports that the first task is applicable; otherwise, the UE reports that it is not applicable. In another possible implementation, the UE can report the applicability of all selectable values in the multiple third time offsets separately.
[0345] For example, the fifth information includes M values of the third time offset, and the sixth information includes M applicability results, with each of the M applicability results corresponding to one of the M values. For instance, a base station might require the UE to perform CSI reporting tasks across multiple time units. The UE can determine and report multiple applicability results corresponding to the multiple third time offsets. In this case, a flexible option is provided for the UE to perform applicability result reporting. The base station can select one of the applicable third time offsets reported by the UE for inference reporting.
[0346] To further reduce overhead, the UE can report only one applicability result for multiple third time offsets. For example, if the UE's model can support the maximum value among all selectable values in these third time offsets (or it can be assumed that the UE's model can support all selectable values in these third time offsets), the UE reports that the first task is applicable; otherwise, the UE reports that it is not applicable.
[0347] Figure 13 illustrates a possible inference scenario provided by an embodiment of this application. In Figure 13, it is assumed that the UE has a model that supports generating two prediction instances from four observation instances. For ease of description, the second reference time unit is the time unit where the last observation instance is located. The maximum time offset of the time unit where the first prediction instance associated with the first task is located relative to the time unit where the last observation instance is located can be represented as d1, and the maximum time offset of the time unit where the first task is associated with a CSI report relative to the time unit where the last observation instance is located can be represented as d2. As mentioned above, the time offset between d1 and d2 can be represented by the offset value δ. It is assumed that for the first task, the base station's inference requirement is that the UE's model can obtain at least one prediction instance. In the applicable scenario shown in Figure 13, the time unit of the first prediction instance associated with the first task is between the time unit where the model is capable of generating the first prediction instance and the time unit where the second prediction instance is located, therefore the model's inference capability can support the inference requirement of the first task. In the inapplicable scenario shown in Figure 13, the time unit of the first prediction instance associated with the first task is after the time units where the model is capable of generating the first and second prediction instances, therefore the model's inference capability cannot support the inference requirement of the first task.
[0348] In some embodiments, the sixth information is further used to indicate a fourth time offset, wherein if the third time offset is less than or equal to the fourth time offset, the first task applies.
[0349] As mentioned earlier, the third time offset may include multiple selectable values (indicated by M values). In one possible design, for the multiple selectable values of the third time offset, some values are applicable and some are not. The sixth information can indicate the fourth time offset, so the base station can know that all selectable values less than or equal to the fourth time offset are applicable, and all selectable values greater than the fourth time offset are not applicable. Optionally, the fourth time offset is greater than or equal to the minimum value among the multiple selectable values of the third time offset, and less than or equal to the maximum value among the multiple selectable values of the third time offset. In this case, the UE does not need to report applicability information for each selectable value, which can reduce the reporting overhead of the UE. In another possible design, for the multiple selectable values of the third time offset, all are not applicable. In addition to indicating that the multiple selectable values of the third time offset are not applicable, the sixth information can also indicate the fourth time offset. Then the base station can know that all values less than or equal to the fourth time offset are applicable, and all values greater than the fourth time offset are not applicable. The base station can redetermine a time offset based on the fourth time offset for inference reporting. Optionally, the fourth time offset may be less than the minimum of several selectable values of the third time offset.
[0350] As mentioned earlier, the fifth information may indicate the range of the third time offset. In one possible design, for selectable values within this third time offset range, some values within the range are applicable, some values within the range are not applicable, or all selectable values within the range are not applicable. In this case, the UE should report "not applicable." The sixth information, in addition to indicating that none of the selectable values within the third time offset range are applicable, may also indicate a fourth time offset. Then, the base station can know that all values less than or equal to the fourth time offset are applicable, and all values greater than the fourth time offset are not applicable. The base station can re-determine a time offset based on the fourth time offset for inference reporting, avoiding the base station's assumption that the UE cannot meet any of its inference requirements after the UE reports "not applicable." Optionally, the range where the fourth time offset is less than the third time offset indicates the maximum value of the third time offset.
[0351] In other words, the UE can utilize the inference capabilities of the sixth information reporting model (i.e., the fourth time offset). Once the inference capabilities of the model are known, the base station can allocate different inference tasks for different models, which is beneficial for the full scheduling and utilization of resources.
[0352] It should be noted that the various ways in which the UE reports the applicability results using the sixth information may be indicated by the base station, or may be predefined or preconfigured by the protocol. This application embodiment does not limit this.
[0353] The information configuration method of the present application embodiment has been described in detail above with reference to Figures 1 to 13. The information configuration device of the present application embodiment will be described in detail below with reference to Figures 14 to 16.
[0354] Figure 14 illustrates an information configuration device 1400 provided in an embodiment of this application. The information configuration device 1400 includes a transceiver unit 1410 and a processing unit 1420.
[0355] In one possible implementation, the information configuration device 1400 is used to execute the steps / processes corresponding to the first device (e.g., the UE in the method 1000) in the above method.
[0356] The transceiver unit 1410 is configured to: receive first information, which indicates a candidate set of inference parameters. The type of inference parameters included in the candidate set of inference parameters includes at least one of the following: temporal behavior of a first resource, the number of first resources, the time interval of the first resources, the number of at least one prediction instance, the time interval between time units corresponding to two adjacent prediction instances in the at least one prediction instance, the time unit in which the first prediction instance in the at least one prediction instance is located, or the type of output result corresponding to the inference of the first model; the first resource corresponds to a measurement resource used for inference of the first model, the at least one prediction instance is a prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit among the at least one prediction instance. The processing unit 1420 is configured to: perform training data collection and / or training of the first model based on the first information.
[0357] Optionally, the transceiver unit 1410 is further configured to: receive second information, the second information being used to indicate a second resource, the second resource including measurement resources for training the first model and / or collecting training data for the first model; wherein the temporal behavior of the first resource is different from the temporal behavior of the second resource, and / or the time interval of the first resource is different from the time interval of the second resource.
[0358] Optionally, the temporal behavior of the second resource may include periodic or semi-persistent behavior.
[0359] Optionally, the temporal behavior of the first resource includes aperiodicity.
[0360] Optionally, the type of the inference parameters includes: the time unit where the first predicted instance is located, and the first information includes a first time offset, which is used to indicate the time offset between the time unit where the first predicted instance is located and the first reference time unit.
[0361] Optionally, the type of the inference parameters includes: the time unit in which the first prediction instance is located, the first information including a second time offset, the second time offset being used to indicate the time offset between the time unit in which the first channel state information (CSI) report is located and the first reference time unit, the first channel state information (CSI) report corresponding to the inference of the first model.
[0362] Optionally, the first resource includes at least one measurement resource for the first model inference, and the first reference time unit includes the time unit in which any one of the at least one measurement resource is located.
[0363] Optionally, the set of candidate inference parameters may include one or more candidate values of the same type of inference parameter.
[0364] Optionally, the candidate inference parameter set includes multiple sets of candidate inference parameters.
[0365] Optionally, the transceiver unit 1410 is further configured to: send third information, the third information being used to indicate whether the ability to train on multiple candidate values for the same type of inference parameters is available.
[0366] Optionally, the transceiver unit 1410 is further configured to: send fourth information, the fourth information being used to indicate at least one set of candidate inference parameters among the supported plurality of candidate inference parameters.
[0367] Optionally, the output result may include at least one of the following types: Reference Signal Received Power (RSRP), Resource Identifier of Target Reference Signal, Precoding Matrix Indicator (PMI), or Channel Quality Indicator (CQI).
[0368] Optionally, the first information is further used to indicate that the type of the output result includes the resource identifier of the target reference signal, and the type of the inference parameter further includes the number of resource identifiers of the target reference signal.
[0369] Optionally, the first information may also include the identifier of at least one set of candidate inference parameters among the plurality of candidate inference parameters.
[0370] Optionally, the temporal behavior of the first resource includes at least one of the following: periodic, semi-persistent, or aperiodic.
[0371] In another possible implementation, the information configuration device 1400 is used to execute the steps / processes corresponding to the second device (e.g., the base station in the method 1000) in the above method.
[0372] The transceiver unit 1410 is configured to: send first information, which indicates a candidate set of inference parameters. The type of inference parameters included in the candidate set of inference parameters includes at least one of the following: temporal behavior of a first resource, the number of first resources, the time interval of the first resources, the number of at least one prediction instance, the time interval between time units corresponding to two adjacent prediction instances in the at least one prediction instance, the time unit in which the first prediction instance in the at least one prediction instance is located, or the type of output result corresponding to the inference of the first model; the first resource corresponds to a measurement resource used for inference of the first model, the at least one prediction instance is a prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit among the at least one prediction instance.
[0373] Optionally, the transceiver unit 1410 is further configured to: send second information, the second information being used to indicate a second resource, the second resource including measurement resources for training the first model and / or collecting training data for the first model; the temporal behavior of the first resource and the temporal behavior of the second resource are different, and / or the time interval of the first resource and the time interval of the second resource are different.
[0374] Optionally, the transceiver unit 1410 is further configured to: receive third information, the third information being used to indicate whether the terminal device has the capability to support training for multiple candidate values of the same type of inference parameters.
[0375] In one possible implementation, the information configuration device 1400 is used to execute the steps / processes corresponding to the first device (e.g., the UE in method 1200) in the above method.
[0376] The transceiver unit 1410 is configured to: receive fifth information, which indicates the range of a third time offset or M values of the third time offset, where M is a positive integer, and the third time offset reflects the time offset between the time unit of the first prediction instance in at least one prediction instance associated with the first task and the time unit of the first observation instance in at least one observation instance associated with the first task; and send sixth information, which indicates the applicability of the first task.
[0377] Optionally, the third time offset indicates the time offset between the time unit of at least one CSI report associated with the first task and the second reference time unit, or the time offset between the time unit of the first prediction instance in at least one prediction instance associated with the first task and the second reference time unit.
[0378] Optionally, the second reference time unit includes any one of the following: the time unit in which the first observation instance in at least one observation instance corresponding to the first task is located; or, the time unit in which the downlink control information (DCI) is located, wherein the DCI is used to trigger the reporting of at least one CSI report associated with the first task.
[0379] Optionally, the second reference time unit is predefined or preconfigured.
[0380] Optionally, M is greater than or equal to 2.
[0381] Optionally, the sixth information includes an applicability result, which corresponds to the range of the third time offset or M values of the third time offset, and the applicability result is determined based on the largest time offset value among the third time offsets.
[0382] Optionally, the fifth information includes M values of the third time offset, and the sixth information includes M applicability results, with each of the M applicability results corresponding to one of the M values.
[0383] Optionally, the sixth information is also used to indicate a fourth time offset, and if the third time offset is less than or equal to the fourth time offset, then the first task applies.
[0384] Optionally, the first task is a dynamically scheduled task.
[0385] Optionally, the first task is a CSI prediction task.
[0386] In another possible implementation, the information configuration device 1400 is used to execute the steps / processes corresponding to the second device (e.g., the base station in the method 1200) in the above method.
[0387] The transceiver unit 1410 is configured to: send fifth information, which indicates the range of the third time offset or M values of the third time offset, where M is a positive integer, and the third time offset reflects the time offset between the time unit of the first prediction instance in at least one prediction instance associated with the first task and the time unit of the first observation instance in at least one observation instance associated with the first task; and receive sixth information, which indicates the applicability of the first task.
[0388] It should be understood that the information configuration device 1400 here is embodied in the form of a functional unit. The term "unit" here can refer to an application-specific integrated circuit (ASIC), electronic circuitry, a processor (e.g., a shared processor, a proprietary processor, or a group processor, etc.) and memory for executing one or more software or firmware programs, integrated logic circuitry, and / or other suitable components supporting the described functions. In an alternative example, those skilled in the art will understand that the information configuration device 1400 can be specifically any of the first or second devices in the above embodiments. The information configuration device 1400 can be used to execute the various processes and / or steps corresponding to any of the first or second devices in the above method embodiments; to avoid repetition, these will not be described further here.
[0389] The information configuration device 1400 of each of the above schemes has the function of implementing the corresponding steps of any one of the first or second devices in the above methods; the function can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. For example, the transceiver unit 1410 can be replaced by a receiver and a transmitter, and other units, such as processing units, can be replaced by a processor, respectively executing the transceiver operations and related processing operations in each method embodiment.
[0390] In embodiments of this application, the information configuration device 1400 in FIG14 can also be a chip or a chip system, such as a system on chip (SoC). Correspondingly, the transceiver unit 1410 can be the transceiver circuit of the chip, which is not limited here.
[0391] Figure 15 illustrates another information configuration device 1500 provided in an embodiment of this application. The information configuration device 1500 includes a processor 1510, a transceiver 1520, and a memory 1530. The processor 1510, transceiver 1520, and memory 1530 communicate with each other via an internal connection path. The memory 1530 stores instructions, and the processor 1510 executes the instructions stored in the memory 1530 to control the transceiver 1520 to transmit and / or receive signals.
[0392] It should be understood that the information configuration device 1500 may be specifically any one of the first device or the second device in the above embodiments, and may be used to execute the various steps and / or processes corresponding to any one of the first device or the second device in the above method embodiments. Optionally, the memory 1530 may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information. The processor 1510 may be used to execute instructions stored in the memory, and when the processor 1510 executes instructions stored in the memory, the processor 1510 is used to execute the various steps and / or processes of the above method embodiments corresponding to any one of the first device or the second device. The transceiver 1520 may include a transmitter and a receiver, the transmitter may be used to implement the various steps and / or processes corresponding to the transceiver for performing a transmitting action, and the receiver may be used to implement the various steps and / or processes corresponding to the transceiver for performing a receiving action.
[0393] Figure 16 illustrates another information configuration device 1600 provided in an embodiment of this application. The information configuration device 1600 includes a processor 1610, a memory 1620, a network interface 1630, an input / output interface 1640, and a bus 1650. The processor 1610 is used to execute instructions and process data; the memory 1620 can store content including but not limited to: data files, program code, and operating system; the network interface 1630 (such as a network card) sends and receives messages, enabling the computer to connect to a network (such as a local area network or the Internet) for communication; the input / output interface 1640 is used to connect and manage data exchange between the computer and external devices: input interfaces (such as keyboards and mice) allow users to input data into the computer, while output interfaces (such as monitors and printers) present the processed data to the user. In addition, there are interfaces responsible for connecting storage devices (such as USB ports and hard drive interfaces); the bus 1650 is a set of electrical signal lines used to transmit data, addresses, and control signals within the computer. The bus 1650 connects the processor 1610, memory 1620, network interface 1630, and input / output interface 1640, enabling them to communicate with each other.
[0394] It should be understood that the information configuration device 1600 may be specifically any one of the first device or the second device in the above embodiments, and may be used to execute the various steps and / or processes corresponding to any one of the first device or the second device in the above method embodiments.
[0395] It should be understood that, in the embodiments of this application, the processor of the above-described device can be a central processing unit (CPU), which can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.
[0396] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly manifested as execution by a hardware processor, or as a combination of hardware and software units within the processor. The software units can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor executes the instructions in the memory, combining them with its hardware to complete the steps of the above method. To avoid repetition, detailed descriptions are omitted here.
[0397] This application also provides a computer-readable storage medium for storing a computer program for implementing the method corresponding to either the first or second device in the above embodiments.
[0398] This application also provides a computer program product, which includes a computer program (also referred to as code or instructions), which, when run on a computer, allows the computer to execute the method corresponding to either the first or second device shown in the above embodiments.
[0399] In the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" do not necessarily imply that they are different.
[0400] Furthermore, "at least one" refers to one or more, while "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can mean: a, or b, or c, or a and b, or a and c, or b and c, or a, b, and c, where a, b, and c can be single or multiple.
[0401] Those skilled in the art will recognize that the method steps and units described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the steps and components of each embodiment have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0402] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0403] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, apparatuses, or units, or they may be electrical, mechanical, or other forms of connection.
[0404] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.
[0405] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0406] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0407] It should be understood that the various embodiments of this application can be implemented individually or in combination.
[0408] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. An information configuration method, characterized in that, include: The system receives first information, which indicates a candidate set of inference parameters. The type of inference parameters included in the candidate set of inference parameters includes at least one of the following: the temporal behavior of a first resource, the number of first resources, the time interval of the first resources, the number of at least one prediction instance, the time interval between time units corresponding to two adjacent prediction instances in the at least one prediction instance, the time unit in which the first prediction instance in the at least one prediction instance is located, or the type of output result corresponding to the inference of the first model; the first resource corresponds to a measurement resource used for inference of the first model, the at least one prediction instance is a prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit among the at least one prediction instance. Based on the first information, perform training data collection for the first model and / or training of the first model.
2. The method according to claim 1, characterized in that, The method further includes: Receive second information, the second information being used to indicate a second resource, the second resource including measurement resources for training the first model and / or collecting training data for the first model; Wherein, the temporal behavior of the first resource is different from that of the second resource, and / or the time interval of the first resource is different from that of the second resource.
3. The method according to claim 2, characterized in that, The temporal behavior of the second resource includes periodic or semi-persistent.
4. The method according to any one of claims 1 to 3, characterized in that, The temporal behavior of the first resource includes aperiodicity.
5. The method according to any one of claims 1 to 4, characterized in that, The types of the inference parameters include: the time unit in which the first predicted instance is located. The first information includes a first time offset, which indicates the time offset between the time unit where the first predicted instance is located and the first reference time unit.
6. The method according to any one of claims 1 to 4, characterized in that, The types of the inference parameters include: the time unit in which the first predicted instance is located. The first information includes a second time offset, which indicates the time offset between the time unit where the first channel state information (CSI) report is located and the first reference time unit. The first channel state information (CSI) report corresponds to the inference of the first model.
7. The method according to claim 5 or 6, characterized in that, The first resource includes at least one measurement resource used for the inference of the first model, and the first reference time unit includes the time unit in which any one of the at least one measurement resource is located.
8. The method according to any one of claims 1 to 7, characterized in that, The set of candidate inference parameters includes one or more candidate values of the same type of inference parameter.
9. The method according to claim 8, characterized in that, The candidate inference parameter set includes multiple sets of candidate inference parameters.
10. The method according to any one of claims 1 to 9, characterized in that, The method further includes: Send a third message, which indicates whether the ability to train on multiple candidate values for the same type of inference parameter is available.
11. The method according to any one of claims 1 to 10, characterized in that, The first information also includes the value or range of the inference parameter.
12. An information configuration method, characterized in that, include: Send first information, which indicates a candidate set of inference parameters. The type of inference parameters included in the candidate set of inference parameters includes at least one of the following: the temporal behavior of a first resource, the number of first resources, the time interval of the first resources, the number of at least one prediction instance, the time interval between two adjacent prediction instances in the at least one prediction instance, the time unit in which the first prediction instance in the at least one prediction instance is located, or the type of output result corresponding to the inference of the first model; the first resource corresponds to a measurement resource used for inference of the first model, the at least one prediction instance is a prediction instance output by the inference of the first model, and the first prediction instance is the earliest prediction instance in the corresponding time unit among the at least one prediction instance.
13. The method according to claim 12, characterized in that, The method further includes: Send a second message, the second message being used to indicate a second resource, the second resource including measurement resources for training the first model and / or collecting training data for the first model; The temporal behavior of the first resource is different from that of the second resource, and / or the time interval of the first resource is different from that of the second resource.
14. The method according to claim 12 or 13, characterized in that, The method further includes: Receive third information, which is used to indicate whether the terminal device has the ability to support training for multiple candidate values of the same type of inference parameters.
15. An information configuration method, characterized in that, include: Receive fifth information, the fifth information being used to indicate the range of the third time offset or M values of the third time offset, where M is a positive integer, the third time offset reflecting the time offset between the time unit where the first prediction instance in at least one prediction instance associated with the first task is located and the time unit where the first observation instance in at least one observation instance associated with the first task is located. A sixth message is sent, which indicates the applicability of the first task.
16. An information configuration method, characterized in that, include: Send a fifth message, which is used to indicate the range of the third time offset or M values of the third time offset, where M is a positive integer. The third time offset reflects the time offset between the time unit where the first prediction instance in at least one prediction instance associated with the first task is located and the time unit where the first observation instance in at least one observation instance associated with the first task is located. Receive a sixth message, which indicates the applicability of the first task.
17. The method according to claim 15 or 16, characterized in that, The third time offset indicates the time offset between the time unit of at least one CSI report associated with the first task and the second reference time unit, or the time offset between the time unit of the first prediction instance in at least one prediction instance associated with the first task and the second reference time unit.
18. The method according to claim 17, characterized in that, The second reference time unit includes any one of the following: The time unit in which the first observation instance in at least one observation instance corresponding to the first task is located; or, The time unit in which the downlink control information (DCI) is located, the DCI being used to trigger the reporting of at least one CSI report associated with the first task.
19. The method according to claim 17 or 18, characterized in that, The second reference time unit is predefined or preconfigured.
20. The method according to any one of claims 15 to 19, characterized in that, M is greater than or equal to 2.
21. The method according to any one of claims 15 to 20, characterized in that, The sixth piece of information includes an applicability result, which corresponds to the range of the third time offset or M values of the third time offset, and the applicability result is determined based on the largest time offset value in the third time offset.
22. The method according to any one of claims 15 to 21, characterized in that, The fifth information includes M values of the third time offset, and the sixth information includes M applicability results, with each of the M applicability results corresponding to one of the M values.
23. The method according to any one of claims 15 to 22, characterized in that, The sixth piece of information is also used to indicate a fourth time offset, and if the third time offset is less than or equal to the fourth time offset, then the first task applies.
24. The method according to any one of claims 15 to 23, characterized in that, The first task is a dynamically scheduled task.
25. The method according to any one of claims 15 to 24, characterized in that, The first task is the CSI prediction task.
26. An information configuration device, characterized in that, include: Units for implementing the method according to any one of claims 1 to 25.
27. An information configuration device, characterized in that, include: A processor coupled to a memory for storing programs or instructions that, when executed by the processor, cause the apparatus to perform the information configuration method as described in any one of claims 1 to 25.
28. A computer-readable storage medium, characterized in that, The computer program or instructions are stored thereon, which, when executed, cause the computer to perform the information configuration method as described in any one of claims 1 to 25.
29. A computer program product, characterized in that, The computer program product includes computer program code, which, when run on a computer, causes the computer to implement the information configuration method according to any one of claims 1 to 25.
30. A system, characterized in that, It includes means for implementing the method of any one of claims 1 to 11 and means for implementing the method of any one of claims 12 to 14, or means for implementing the method of any one of claims 15, 17 to 25 and means for implementing the method of any one of claims 16 to 25.