Beam Management Method

Abstract

The present disclosure provides a beam management method, including: a communication device, such as a terminal device or an access network device, maps reference signal measurements to input adaptation information by using an input adaptation layer, where a beam pattern corresponding to the reference signal measurements is a first beam pattern; and obtains a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. This method can reduce system overhead.

Description

[Technical Field]

[0001] The present application relates to the field of communications technology, and in particular to beam management methods and apparatus. [Background technology]

[0002] In wireless communication networks, e.g., mobile communication networks, the services supported by the network are becoming increasingly diverse, and therefore the requirements that need to be met are becoming increasingly diverse. For example, networks need to be capable of supporting very high rates, very low latency, and / or very large connections. This characteristic makes network planning, network configuration, and / or resource scheduling increasingly complex. Furthermore, as networks have increasingly powerful capabilities, e.g., supporting increasingly higher spectrum and new technologies such as high-order multiple-input multiple-output (MIMO) technology, beamforming, and / or beam management, network energy conservation has become a hot research topic. These new requirements, scenarios, and characteristics pose unprecedented challenges to network planning, operation and maintenance, and efficient operation. To address these challenges, artificial intelligence techniques can be introduced into wireless communication networks to implement network intelligence. Based on this, how to effectively implement artificial intelligence in networks is a worthy research issue. Summary of the Invention

[0003] The present disclosure provides a beam management method and apparatus for reducing overhead in beam management processing.

[0004] According to a first aspect, a beam management method is provided. The method can be implemented in a reference signal receiver. For example, the method can be implemented by a terminal device, an access network device, a module of the access network device (e.g., a DU or a quasi-real-time RIC), or a non-real-time RIC.

[0005] In a possible implementation, the method includes a step of mapping reference signal measurements to input adaptation information by using an input adaptation layer, wherein the beam pattern corresponding to the reference signal measurements is a first beam pattern; and a step of obtaining a first beam prediction result by using a beam prediction model, wherein the input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern.

[0006] In a possible implementation, the method includes a step of obtaining a first beam prediction result by using a beam prediction model, wherein an input of the beam prediction model includes a reference signal measurement amount and a beam pattern corresponding to the reference signal measurement amount is a second beam pattern, and a step of mapping the first beam prediction result to a second beam prediction result by using an output adaptation layer.

[0007] In a possible implementation, the method includes the steps of mapping reference signal measurements to input adaptation information by using an input adaptation layer, where the beam pattern corresponding to the reference signal measurements is a first beam pattern; obtaining a first beam prediction result by using a beam prediction model, where the input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern; and mapping the first beam prediction result to a second beam prediction result by using an output adaptation layer.

[0008] According to the above method, when a system supports multiple sparse beam patterns, a small number of beam prediction models and input adaptation layers and / or output adaptation layers whose scales are smaller than those of the beam prediction models can be configured for beam management without configuring one beam prediction model for each sparse beam pattern. According to this method, the storage overhead of the beam prediction models can be reduced. When information about the beam prediction models needs to be exchanged between different network elements, the signaling overhead can be further reduced.

[0009] In a possible design, the first beam prediction result includes the top K1 beams in the full beam corresponding to the first beam pattern or the second beam pattern, where K1 is a positive integer.

[0010] In a possible design, the second beam prediction result includes the top K2 beams in the full beam corresponding to the first beam pattern or the second beam pattern, where K2 is a positive integer.

[0011] According to this method, the top beam in the full beam can be predicted by sweeping a sparse beam pattern and using a beam prediction model without obtaining the top beam in the full beam by sweeping the full beam, thereby reducing latency overhead and reference signal overhead in the beam management process.

[0012] In a possible design, information regarding a beam prediction model is received. According to the method, the beam prediction model can be flexibly configured.

[0013] In a possible design, the beam prediction models are included in a candidate beam prediction model set, where each beam prediction model in the candidate beam prediction model set corresponds to one beam pattern, and the beam prediction model corresponds to a second beam pattern. Optionally, information about each beam prediction model in the candidate beam prediction model set is agreed upon in a protocol or received from a transmitter. Optionally, correspondence between the beam prediction models in the candidate beam prediction model set and the (sparse) beam patterns is agreed upon in a protocol or received from the transmitter. Optionally, information indicating the second beam pattern is received. According to this method, signaling overhead for configuring beam prediction models can be reduced.

[0014] In a possible design, information about the input adaptation layer or information about the output adaptation layer is received. According to the method, the input adaptation layer and / or the output adaptation layer may be flexibly configured.

[0015] In a possible design, the input adaptation layer or the output adaptation layer is obtained through training.

[0016] Optionally, an ideal beam prediction result is obtained based on a measurement corresponding to a full beam. Measurements corresponding to sparse beams of the first beam pattern are mapped to input adaptation information by using an input adaptation layer. An actual beam prediction result is obtained based on the input adaptation information and a beam prediction model. Parameters of the input adaptation layer are adjusted based on the ideal beam prediction result and the actual beam prediction result to enable a difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold.

[0017] Optionally, an ideal beam prediction result is obtained based on measurements corresponding to the full beam. An actual beam prediction result is obtained based on measurements of the sparse beam of the second beam pattern, a beam prediction model, and an output adaptation layer. Parameters of the output adaptation layer are adjusted based on the ideal beam prediction result and the actual beam prediction result to enable a difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold.

[0018] Optionally, an ideal beam prediction result is obtained based on a measurement corresponding to a full beam. Measurements corresponding to sparse beams of the first beam pattern are mapped to input adaptation information by using an input adaptation layer. An actual beam prediction result is obtained based on the input adaptation information, a beam prediction model, and an output adaptation layer. At least one of parameters of the input adaptation layer and parameters of the output adaptation layer is adjusted based on the ideal beam prediction result and the actual beam prediction result to enable a difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold.

[0019] According to this method, the input adaptation layer and / or the output adaptation layer can be flexibly adjusted based on the channel conditions, so that the input adaptation layer and / or the output adaptation layer can better match the current channel conditions and achieve better matching performance.

[0020] According to a second aspect, a beam management method is provided. The method can be implemented in a reference signal transmitter. For example, the method can be implemented by a terminal device, an access network device, a module of the access network device (e.g., a DU or a quasi-real-time RIC), or a non-real-time RIC.

[0021] In a possible implementation, the method includes sending information about an input adaptation layer configured to perform adaptation on reference signal measurements to obtain inputs for a beam prediction model, where a beam pattern corresponding to the reference signal measurements is a first beam pattern, and the inputs for the beam prediction model match a second beam pattern, and the first beam pattern is different from the second beam pattern.

[0022] In a possible implementation, the method includes sending information about an output adaptation layer configured to map a first beam prediction result output by the beam prediction model to a second beam prediction result, the first beam prediction result being different from the second beam prediction result.

[0023] In a possible implementation, the method includes sending information related to an input adaptation layer and information related to an output adaptation layer. The input adaptation layer is configured to perform adaptation on reference signal measurements to obtain an input for a beam prediction model, where a beam pattern corresponding to the reference signal measurements is a first beam pattern, and the input for the beam prediction model matches a second beam pattern, where the first beam pattern is different from the second beam pattern. The output adaptation layer is configured to map the first beam prediction result output by the beam prediction model to a second beam prediction result, where the first beam prediction result is different from the second beam prediction result.

[0024] In a possible design, the method further includes indicating a beam prediction model from the candidate beam prediction model set. Each beam prediction model in the candidate beam prediction model set corresponds to one beam pattern. Optionally, information about each beam prediction model in the candidate beam prediction model set is agreed upon in a protocol, or the method includes sending information about each beam prediction model in the candidate beam prediction model set. Optionally, correspondence between beam prediction models in the candidate beam prediction model set and the (sparse) beam patterns is agreed upon in a protocol, or the method includes sending information about the correspondence between beam prediction models in the candidate beam prediction model set and the (sparse) beam patterns. Optionally, the method includes sending information indicative of the second beam pattern.

[0025] According to a third aspect, there is provided a communications device, the communications device being adapted to implement the method according to the first aspect.

[0026] In an optional implementation, the apparatus may include modules in one-to-one correspondence with the methods / operations / steps / actions described in the first aspect. The modules may be implemented through hardware circuits, software, or a combination of hardware circuits and software. In an optional implementation, the communication apparatus includes a baseband device and a radio frequency device. In another optional implementation, the communication apparatus includes a processing unit (which may also be referred to as a processing module) and a transceiver unit (which may also be referred to as a transceiver module). The transceiver unit may implement a sending function and a receiving function. When the transceiver unit implements the sending function, the transceiver unit may be referred to as a sending unit (which may also be referred to as a sending module). When the transceiver unit implements the receiving function, the transceiver unit may be referred to as a receiving unit (which may also be referred to as a receiving module). The sending unit and the receiving unit may be the same functional module, and the functional module may be referred to as a transceiver unit, and the functional module may implement the sending function and the receiving function. Alternatively, the sending unit and the receiving unit may be different functional modules, and the transceiver unit is a collective term for these functional modules.

[0027] For example, the processing unit is configured to map the reference signal measurement to input adaptation information by using an input adaptation layer, where the beam pattern corresponding to the reference signal measurement is a first beam pattern, and obtain a first beam prediction result by using a beam prediction model, where the input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. The reference signal is received by the receiving unit.

[0028] For example, the processing unit is configured to obtain a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes a reference signal measurement amount and a beam pattern corresponding to the reference signal measurement amount is a second beam pattern, and to map the first beam prediction result to the second beam prediction result by using an output adaptation layer.

[0029] For example, the processing unit is configured to map a reference signal measurement to input adaptation information by using an input adaptation layer, where the beam pattern corresponding to the reference signal measurement is a first beam pattern, obtain a first beam prediction result by using a beam prediction model, where the input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern, and map the first beam prediction result to a second beam prediction result by using an output adaptation layer.

[0030] For a description of the first beam prediction result and the second beam prediction result, please refer to the first embodiment, and the details will not be described again here.

[0031] In a possible design, the receiving unit is configured to receive information related to a beam prediction model.

[0032] In a possible design, the beam prediction models are included in a candidate beam prediction model set, where each beam prediction model in the candidate beam prediction model set corresponds to one beam pattern, and the beam prediction model corresponds to a second beam pattern. Optionally, information about each beam prediction model in the candidate beam prediction model set is agreed upon in a protocol or received from the transmitter by the receiving unit. Optionally, correspondence between the beam prediction models in the candidate beam prediction model set and the (sparse) beam patterns is agreed upon in a protocol or received from the transmitter by the receiving unit. Optionally, the receiving unit is configured to receive information indicating the second beam pattern.

[0033] In a possible design, the receiving unit is configured to receive information about the input adaptation layer or information about the output adaptation layer.

[0034] In a possible design, the processing unit is configured to obtain the input adaptation layer or the output adaptation layer through training. For a specific training method, see the first aspect. Details will not be described again here.

[0035] In another possible implementation, a communication device includes a processor configured to implement the method described in the first aspect. The device may further include a memory configured to store instructions and / or data. The memory is coupled to the processor. When executing program instructions stored in the memory, the processor can implement the method described in the first aspect. The device may further include a communication interface, which is used by the device to communicate with another device. For example, the communication interface may be a transceiver, a circuit, a bus, a module, a pin, or another type of communication interface, and the other device may be a model inference node, etc. The functionality of the processor is similar to that of the processing unit described above, and the functionality of the communication interface is similar to that of the transceiver unit described above. Details will not be described again here.

[0036] According to a fourth aspect, there is provided a communications device, the communications device being adapted to implement the method according to the second aspect.

[0037] In an optional implementation, the apparatus may include modules in one-to-one correspondence with the methods / operations / steps / actions described in the second aspect. The modules may be implemented through hardware circuits, software, or a combination of hardware circuits and software. In an optional implementation, the communication apparatus includes a baseband device, or includes a baseband device and a radio frequency device. In another optional implementation, the communication apparatus includes a processing unit (which may also be referred to as a processing module) and a transceiver unit (which may also be referred to as a transceiver module). The transceiver unit may implement a sending function and a receiving function. When the transceiver unit implements the sending function, the transceiver unit may be referred to as a sending unit (which may also be referred to as a sending module). When the transceiver unit implements the receiving function, the transceiver unit may be referred to as a receiving unit (which may also be referred to as a receiving module). The sending unit and the receiving unit may be the same functional module, and the functional module may be referred to as a transceiver unit, and the functional module may implement the sending function and the receiving function. Alternatively, the sending unit and the receiving unit may be different functional modules, and the transceiver unit is a collective term for these functional modules.

[0038] For example, the sending unit is configured to send information about an input adaptation layer. The input adaptation layer is configured to perform adaptation on reference signal measurements to obtain an input of a beam prediction model, where a beam pattern corresponding to the reference signal measurements is a first beam pattern, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. The information about the input adaptation layer is determined by the processing unit.

[0039] For example, the sending unit is configured to send information about an output adaptation layer. The output adaptation layer is configured to map a first beam prediction result output by the beam prediction model to a second beam prediction result, where the first beam prediction result is different from the second beam prediction result. The information about the output adaptation layer is determined by the processing unit.

[0040] For example, the sending unit is configured to send information about the input adaptation layer and information about the output adaptation layer. The input adaptation layer is configured to perform adaptation on reference signal measurements to obtain inputs for the beam prediction model, where a beam pattern corresponding to the reference signal measurements is a first beam pattern, and the inputs for the beam prediction model match a second beam pattern, where the first beam pattern is different from the second beam pattern. The output adaptation layer is configured to map a first beam prediction result output by the beam prediction model to a second beam prediction result, where the first beam prediction result is different from the second beam prediction result. The information about the input adaptation layer and the information about the output adaptation layer are determined by the processing unit.

[0041] In a possible design, the sending unit is further configured to indicate a beam prediction model from the candidate beam prediction model set. Each beam prediction model in the candidate beam prediction model set corresponds to one beam pattern. Optionally, information about each beam prediction model in the candidate beam prediction model set is agreed upon in a protocol, or the sending unit is further configured to send information about each beam prediction model in the candidate beam prediction model set. Optionally, correspondence between the beam prediction models in the candidate beam prediction model set and the (sparse) beam patterns is agreed upon in a protocol, or the sending unit is further configured to send the correspondence between the beam prediction models in the candidate beam prediction model set and the (sparse) beam patterns. Optionally, the sending unit is further configured to send information indicating the second beam pattern.

[0042] In another possible implementation, a communication device includes a processor configured to implement the method described in the second aspect. The device may further include a memory configured to store instructions and / or data. The memory is coupled to the processor. When executing program instructions stored in the memory, the processor can implement the method described in the second aspect. The device may further include a communication interface, which is used by the device to communicate with another device. For example, the communication interface may be a transceiver, a circuit, a bus, a module, a pin, or another type of communication interface, and the other device may be a model inference node, etc. The functionality of the processor is similar to that of the processing unit described above, and the functionality of the communication interface is similar to that of the transceiver unit described above. Details will not be described again here.

[0043] According to a fifth aspect, a computer-readable storage medium is provided, comprising instructions which, when run on a computer, enable the computer to perform the method of either the first or second aspect.

[0044] According to a sixth aspect, there is provided a chip system. The chip system includes a processor and may further include a memory, and is configured to implement the method of either the first or second aspect. The chip system may include a chip, or may include a chip and another individual component.

[0045] According to a seventh aspect, a computer program product is provided, comprising instructions which, when run on a computer, enable the computer to carry out the method of either the first or second aspect.

[0046] According to an eighth aspect, there is provided a communication system, the system comprising the apparatus of the third aspect and the apparatus of the fourth aspect. [Brief explanation of the drawings]

[0047] [Figure 1] 1 is a diagram of the architecture of a communication system 1000 according to the present disclosure. [Figure 2A] FIG. 1 is a diagram of a neuron structure according to the present disclosure. [Figure 2B] FIG. 1 is a diagram of an FNN network according to the present disclosure. [Figure 3A] FIG. 1 is an exemplary diagram of an application framework for AI in a communication system according to the present disclosure. [Figure 3B] FIG. 1 is an exemplary diagram of an application framework for AI in a communication system according to the present disclosure. [Figure 3C] FIG. 1 is an exemplary diagram of an application framework for AI in a communication system according to the present disclosure. [Figure 3D] FIG. 1 is an exemplary diagram of an application framework for AI in a communication system according to the present disclosure. [Figure 4] FIG. 1 illustrates an AI-based beam management method according to the present disclosure. [Figure 5] FIG. 1 is an exemplary diagram of a top beam according to the present disclosure. [Figure 6] A diagram showing four different sparse beam patterns for a full beam containing 64 beams in accordance with the present disclosure. [Figure 7] FIG. 1 illustrates a beam management method according to the present disclosure. [Figure 8] FIG. 2 is an exemplary diagram of the structure of an input adaptation layer according to the present disclosure. [Figure 9] FIG. 10 is an exemplary diagram of the structure of an output adaptation layer according to the present disclosure. [Figure 10] FIG. 1 illustrates a beam prediction procedure according to the present disclosure. [Figure 11] FIG. 1 illustrates a beam prediction procedure according to the present disclosure. [Figure 12A] FIG. 1 illustrates a training method for an input adaptation layer according to the present disclosure. [Figure 12B] FIG. 1 illustrates a training method for an output adaptation layer according to the present disclosure. [Figure 12C] FIG. 1 illustrates a training method for input and / or output adaptation layers according to the present disclosure. [Figure 13] 1 is an exemplary diagram of an apparatus according to the present disclosure. [Figure 14] 1 is an exemplary diagram of an apparatus according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0048] FIG. 1 is a diagram of the architecture of a communication system 1000 to which the present disclosure is applicable. As shown in FIG. 1, the communication system includes a radio access network (RAN) 100 and a core network (CN) 200. Optionally, the communication system 1000 may further include the Internet 300. The radio access network 100 may include at least one access network device (e.g., 110a and 110b in FIG. 1 ) and may further include at least one terminal device (e.g., 120a to 120j in FIG. 1 ). The terminal device is connected to the access network device in a wireless manner. The access network device is connected to the core network in a wireless or wired manner. The core network device and the access network device may be different physical devices independent of each other, or the functions of the core network device and the logical functions of the access network device may be integrated into the same physical device, or part of the functions of the core network device and part of the functions of the access network device may be integrated into one physical device. The physical presence of the core network device and the access network device is not limited in this disclosure. The terminal devices may be connected to each other in a wireless manner. The access network devices may be connected to each other in a wired or wireless manner. Figure 1 is merely a diagram and does not limit the present disclosure. For example, the communication system may further include another network device, such as a wireless relay device and a wireless backhaul device.

[0049] An access network device may be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next generation NodeB (gNB) in a fifth generation (5G) mobile communication system, an access network device in an open radio access network (O-RAN), a next generation NodeB in a sixth generation (6G) mobile communication system, a base station in a future mobile communication system, an access node in a wireless fidelity (Wi-Fi) system, etc. 5G is sometimes referred to as new radio (NR). Alternatively, an access network device may be a module or unit that completes part of the functions of a base station. For example, the access network device may be a central unit (CU), a distributed unit (DU), a central unit control plane (CU-CP) module, or a central unit user plane (CU-UP) module. The access network device may be a macro base station (e.g., 110a in FIG. 1), a micro base station or an indoor base station (e.g., 110b in FIG. 1), or may be a relay node, a donor node, etc. The particular technology and the particular device form used by the access network device are not limited by this disclosure.

[0050] In the present disclosure, an apparatus configured to implement the functions of an access network device may be an access network device, or may be an apparatus capable of supporting an access network device in implementing the functions, such as a chip system, a hardware circuit, a software module, or a combination of a hardware circuit and a software module. The apparatus may be attached to or used together with an access network device. In the present disclosure, a chip system may include a chip, or may include a chip and another individual component. For ease of explanation, the following will describe the technical solutions provided in the present disclosure by using an example in which the apparatus configured to implement the functions of an access network device is an access network device.

[0051] (1) Protocol layer structure

[0052] Communications between an access network device and a terminal device may conform to a specific protocol layer structure. For example, the protocol layer structure may include a control plane protocol layer structure and a user plane protocol layer structure. For example, the control plane protocol layer structure may include at least one of a radio resource control (RRC) layer, a packet data convergence protocol (PDCP) layer, a radio link control (RLC) layer, a media access control (MAC) layer, or a physical (PHY) layer. For example, the user plane protocol layer structure may include at least one of a service data adaptation protocol (SDAP) layer, a PDCP layer, an RLC layer, a MAC layer, and a physical layer.

[0053] The protocol layer structure between the access network device and the terminal device may be considered an access stratum (AS) structure. Optionally, a non-access stratum (NAS) may further exist above the AS and be used by the access network device to forward information from the core network device to the terminal device, or used by the access network device to forward information from the terminal device to the core network device. In this case, it may be considered that there is a logical interface between the terminal device and the core network device. Optionally, the access network device may forward information between the terminal device and the core network device through transparent transmission. For example, NAS messages may be mapped to RRC signaling or included in the RRC signaling as an element of the RRC signaling.

[0054] Optionally, the protocol layer structure between the access network device and the terminal device may further include an artificial intelligence (AI) layer used for transmitting data related to AI functions.

[0055] An access network device may include a CU and a DU. This design is sometimes referred to as a CU and DU split. Multiple DUs may be controlled in a centralized manner by one CU. For example, the interface between the CU and the DU is called an F1 interface. The control plane (CP) interface may be F1-C, and the user plane (UP) interface may be F1-U. The specific names of the interfaces are not limited in this disclosure. The CU and the DU may be divided according to the protocol layers of the wireless network. For example, the functions of the PDCP layer and the protocol layers above the PDCP layer (e.g., the RRC layer and the SDAP layer) are configured in the CU, and the functions of the protocol layers below the PDCP layer (e.g., the RLC layer, the MAC layer, and the PHY layer) are configured in the DU. In another example, the functions of the protocol layers above the PDCP layer are configured in the CU, and the functions of the PDCP layer and the protocol layers below the PDCP layer are configured in the DU. This is not limited.

[0056] The division of the processing functions of the CU and the DU based on the protocol layer is merely one example, and the processing functions of the CU and the DU may alternatively be divided in a different manner. For example, the CU or DU may be divided to have more protocol layer functions. In another example, the CU or DU may be divided to have some of the processing functions of the protocol layers. For example, some of the RLC layer functions and functions of protocol layers above the RLC layer are configured in the CU, and the remaining RLC layer functions and functions of protocol layers below the RLC layer are configured in the DU. In another example, the division of the functions of the CU or DU may alternatively be performed based on service type or other system requirements. For example, the division may be performed based on latency. Functions whose processing time must meet latency requirements are configured in the DU, and functions whose processing time does not need to meet latency requirements are configured in the CU.

[0057] Optionally, the CU may have one or more functions of a core network.

[0058] Optionally, a radio unit (RU) of the DU may be remotely located. The RU has radio frequency functions. For example, the DU and the RU may be separated at the PHY layer. For example, the DU may implement upper layer functions of the PHY layer, and the RU may implement lower layer functions of the PHY layer. When transmitting, the PHY layer functions may include at least one of adding cyclic redundancy check (CRC) bits, channel coding, rate matching, scrambling, modulation, layer mapping, precoding, resource mapping, physical antenna mapping, or radio frequency transmission. When receiving, the PHY layer functions may include at least one of CRC checking, channel decoding, rate dematching, descrambling, demodulation, layer demapping, channel detection, resource demapping, physical antenna demapping, or radio frequency reception. The upper layer functions of the PHY layer may include some of the functions of the PHY layer. For example, some of the functions are closer to the MAC layer. The lower layer functions of the PHY layer may include another part of the functions of the PHY layer. For example, some of the functions are closer to radio frequency functions. For example, upper layer functions of the PHY layer may include adding CRC bits, channel coding, rate matching, scrambling, modulation, and layer mapping, while lower layer functions of the PHY layer may include precoding, resource mapping, physical antenna mapping, and radio frequency transmission. Alternatively, upper layer functions of the PHY layer may include adding CRC bits, channel coding, rate matching, scrambling, modulation, layer mapping, and precoding, while lower layer functions of the PHY layer may include resource mapping, physical antenna mapping, and radio frequency transmission. For example, upper layer functions of the PHY layer may include CRC checking, channel decoding, rate dematching, decoding, demodulation, and layer demapping, while lower layer functions of the PHY layer may include channel detection, resource demapping, physical antenna demapping, and radio frequency reception.Alternatively, the upper layer functions of the PHY layer may include CRC checking, channel decoding, rate de-matching, decoding, demodulation, layer demapping, and channel detection, and the lower layer functions of the PHY layer may include resource demapping, physical antenna demapping, and radio frequency reception.

[0059] Optionally, the functions of the CU may be further divided, and the control plane and the user plane may be split and implemented by using different entities. The split entities are a control plane CU entity (i.e., a CU-CP entity) and a user plane CU entity (i.e., a CU-UP entity), respectively. The CU-CP entity and the CU-UP entity may be separately connected to the DU. In the present disclosure, an entity may be understood as a module or unit, and may exist in the form of a hardware structure, a software module, or a combination of a hardware structure and a software module. This is not limited thereto.

[0060] Optionally, any one of the CU, CU-CP, CU-UP, DU, and RU may be a software module, a hardware structure, or a combination of a software module and a hardware structure. This is not limited. Different entities may exist in the same form or different forms. For example, the CU, CU-CP, CU-UP, and DU are software modules, and the RU is a hardware structure. For simplicity, not all possible combinations are listed herein. Modules and methods implemented by modules also fall within the scope of protection of the present disclosure. For example, when the method in the present disclosure is implemented by an access network device, the method may be specifically implemented by at least one of the CU, CU-CP, CU-UP, DU, RU, or near-real-time RIC described below.

[0061] In the present disclosure, when the function of the access network device is completed by a module of the access network device, for example, when a DU sends a signal such as a reference signal to a terminal device, it can be understood that the destination of the signal is the terminal device. The sending is performed logically, and the reference signal is not limited to being physically sent by the DU directly to the terminal device.

[0062] In the present disclosure, module A sending information to the terminal includes the following: module A sends information to the terminal through the air interface. Optionally, module A may perform baseband and / or intermediate radio frequency operations on the information. Alternatively, module A delivers the information to module B, which sends the information to the terminal. When sending information to the terminal, module B may transparently transmit the information, segment the information and then send it, or multiplex the information with other information and then send it. Optionally, module B may perform baseband and / or intermediate radio frequency operations on the information and then send it. Optionally, module B may encapsulate the information into a data packet. Optionally, module B may further add a packet header and / or padding bits to the data packet.

[0063] In the present disclosure, when the function of an access network device is completed by a module of the access network device, for example, when a DU receives a signal such as a reference signal from a terminal device, the source of the signal can be understood to be the terminal device. The transmission is performed logically, and the reference signal is not limited to being physically sent directly by the terminal device to the DU.

[0064] In the present disclosure, receiving information from a terminal device may be understood as the source of the information being the terminal device. For example, module A receiving information from a terminal device includes: module A receiving information from a terminal through an air interface; Optionally, module A may perform baseband and / or intermediate radio frequency operations on the information; alternatively, module B receiving information from a terminal through an air interface and delivering the information to module A; module B delivering information to module A includes: module B transparently delivering the received information to module A, combining multiple received segments into information and then delivering the information to module A, or extracting information from multiplexed information and then delivering the information to module A; Optionally, module B may perform baseband and / or intermediate radio frequency operations on the received information and then send out the information; Optionally, the information received by module B is encapsulated in a data packet; Optionally, the data packet includes a packet header and / or padding bits, etc.

[0065] The above-mentioned module A or B may be one module or multiple modules sequentially coupled. This is not limited. For example, module A is a DU module and module B is an RU module. In another example, module A is a CU-CP module and module B is a DU module and an RU module.

[0066] The terminal device may also be referred to as a terminal, user equipment (UE), mobile station, mobile terminal, etc. The terminal device may be widely used in various scenarios for communication. For example, the scenarios include, but are not limited to, at least one of enhanced mobile broadband (eMBB), ultra-reliable low-latency communication (URLLC), massive machine-type communication (mMTC), device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, machine-type communication (MTC), internet of things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, etc. The terminal device may be a mobile phone, a tablet computer, a computer with wireless transceiver capabilities, a wearable device, a vehicle, an unmanned aerial vehicle, a helicopter, an airplane, a ship, a robot, a robotic arm, a smart home device, etc. The particular technology used by the terminal device and the particular device form are not limited by this disclosure.

[0067] In the present disclosure, an apparatus configured to implement the functions of a terminal device may be a terminal device, or may be an apparatus capable of supporting a terminal device in implementing functions, such as a chip system, a hardware circuit, a software module, or a hardware circuit combined with a software module. The apparatus may be attached to or used together with a terminal device. For ease of explanation, the following describes the provided technical solution by using an example in which the apparatus configured to implement the functions of a terminal device is a terminal device.

[0068] In the present disclosure, the access network device and / or the terminal device may be at a fixed location or may be mobile. The access network device and / or the terminal device may be handheld or vehicle-mounted, deployed on land, including indoors or outdoors, on water, or on airborne aircraft, balloons, and satellites. The application scenarios of the access network device and the terminal device are not limited in the present disclosure. The access network device and the terminal device may be deployed in the same scenario or in different scenarios. For example, the access network device and the terminal device are both deployed on land. Alternatively, the access network device is deployed on land and the terminal device is deployed on water. Examples are not provided one by one.

[0069] The roles of an access network device and a terminal device may be relative. For example, helicopter or unmanned aerial vehicle 120i in FIG. 1 may be configured as a mobile access network device. In the case of terminal device 120j accessing wireless access network 100 via 120i, terminal device 120i is an access network device. However, with respect to base station 110a, 120i is a terminal device. In other words, 110a and 120i communicate with each other according to a wireless air interface protocol. 110a and 120i may alternatively communicate with each other according to an interface protocol between base stations. In this case, 120i is also an access network device with respect to 110a. Therefore, both access network devices and terminal devices may be collectively referred to as communication devices. 110a and 110b in FIG. 1 may be referred to as communication devices having the functionality of access network devices, and 120a to 120j in FIG. 1 may be referred to as communication devices having the functionality of terminal devices.

[0070] In conventional communication systems, communication is mainly implemented by using frequency spectrum in low-frequency bands and mid-frequency bands below 6 gigahertz (GHz). However, the spectrum resources in the low-frequency bands and mid-frequency bands below 6 GHz are relatively scarce. Therefore, in 5G communication systems, high-frequency bands (e.g., millimeter wave (mmW) frequency bands) are introduced for wireless communication to introduce more resources and improve communication rates. To address the weak penetration ability and strong path fading effect of high-frequency signals, when transmitting high-frequency signals, a transmitter may improve signal transmission quality by using beamforming technology. To improve the performance of beamforming technology, the present disclosure introduces artificial intelligence (AI) into beamforming technology.

[0071] Artificial intelligence enables machines to have human intelligence. For example, machines can simulate some intelligent human behavior using computer software and hardware. Machine learning methods can be used to implement artificial intelligence. In machine learning methods, a machine obtains a model through learning (or training) by using training data. The model represents a mapping from input to output. The model obtained through learning can be used for inference (or prediction). Specifically, the model can be used to predict an output corresponding to a given input. The output may be called an inference result.

[0072] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning.

[0073] With regard to supervised learning, based on collected sample values ​​and sample labels, a mapping relationship between the sample values ​​and the sample labels is learned by using a machine learning algorithm, and the learned mapping relationship is expressed by using an AI model. The process of training a machine learning model is a process of learning the mapping relationship. In the training process, sample values ​​are input into the model to obtain predicted values ​​of the model, and model parameters are optimized by calculating the error between the predicted values ​​of the model and the sample labels (ideal values). After the mapping relationship is learned, new sample labels can be predicted by using the learned mapping relationship. The mapping relationship learned through supervised learning may include linear mapping or nonlinear mapping. Learning tasks can be classified into classification tasks and regression tasks based on the type of label.

[0074] In unsupervised learning, the internal patterns of samples are autonomously explored by using an algorithm based on collected sample values. In a specific type of unsupervised learning algorithm, the samples are used as supervised signals, in other words, the model learns the mapping relationship between the samples, which is called self-supervised learning. During training, the model parameters are optimized by calculating the error between the model's predicted values ​​and the samples. Self-supervised learning can be used for signal compression and decompression restoration. Common algorithms include autoencoders, generative adversarial networks, etc.

[0075] Unlike supervised learning, reinforcement learning is an algorithm that learns a policy to solve a problem by interacting with the environment. Unlike supervised learning and unsupervised learning, reinforcement learning does not have a clear "correct" action label. The algorithm needs to interact with the environment to obtain a reward signal fed back by the environment and adjust its decision action to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts each user's downlink transmit power based on the total system throughput fed back by the wireless network, hoping to obtain a higher system throughput. The goal of reinforcement learning is also to learn a mapping relationship between the environmental status and the optimal decision action. However, the label for the "correct action" cannot be obtained in advance. Therefore, the network cannot be optimized by calculating the error between the action and the "correct action." Reinforcement learning training is implemented through repeated interactions with the environment.

[0076] A neural network (NN) is a specific model in machine learning technology. According to the universal approximation theorem, a neural network can theoretically approximate any continuous function, and as a result, the neural network has the ability to learn any mapping. In traditional communication systems, extensive expertise is required to design communication modules. However, in a neural network-based deep learning communication system, implicit pattern structures can be automatically discovered from large datasets and mapping relationships between data can be established to obtain better performance than traditional modeling methods.

[0077] The idea of ​​neural networks comes from the neuron structure of the brain. For example, each neuron performs a weighted sum operation on its inputs and outputs a result based on an activation function. Figure 2A shows a diagram of a neuron structure. If the inputs of a neuron are x=[x0,x1,...,x n], and the weights corresponding to each input are w=[w,w1,...,w n ], where n is a positive integer and w i and x i w can be of a variety of possible types, such as a fractional number, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. i is x i is used as the weight of x i The offset for performing a weighted sum on the input values ​​based on the weights is, for example, b. There can be multiple forms of activation functions. For example, if the activation function of a neuron is y = f(z) = max(0,z) and the output of the neuron is

[0078]

number

[0079] In another example, the activation function of a neuron is assumed to be y=f(z)=z, and the output of the neuron is

[0080]

number

[0081] b can be of any possible type, such as a fractional number, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. The activation functions of different neurons in a neural network can be the same or different.

[0082] A neural network generally includes multiple layers, each of which may include one or more neurons. To improve the representation capability of a neural network, the depth and / or width of the neural network can be increased to provide more powerful information extraction and abstract modeling capabilities for complex systems. The depth of a neural network may refer to the number of layers included in the neural network, and the number of neurons included in each layer may be referred to as the layer width. In one implementation, a neural network includes an input layer and an output layer. The input layer of a neural network performs neuronal processing on received input information and transfers the processing results to the output layer. The output layer obtains the output result of the neural network. In another implementation, a neural network includes an input layer, a hidden layer, and an output layer. See FIG. 2B. The input layer of a neural network performs neuronal processing on received input information and transfers the processing results to an intermediate hidden layer. The hidden layer performs calculations on the received processing results to obtain calculation results. The hidden layer transfers the calculation results to the output layer or the next adjacent hidden layer. Finally, the output layer obtains the output result of the neural network. A neural network may include one hidden layer or multiple hidden layers connected in series, but this is not limited thereto.

[0083] The neural network in the present disclosure is, for example, a deep neural network (DNN). According to the network construction method, the DNN may include a feedforward neural network (FNN), a convolutional neural network (CNN), and a recurrent neural network (RNN).

[0084] A characteristic of an FNN network is that neurons in adjacent layers are fully connected to each other. Due to this characteristic, an FNN usually requires a large amount of storage space, which results in high computational complexity. Figure 2B shows an FNN network.

[0085] CNN is a neural network dedicated to processing data with a similar grid structure. For example, both time series data (timeline discrete sampling) and image data (two-dimensional discrete sampling) can be considered as data with a similar grid structure. CNN performs convolution operations by capturing partial information through a window with a fixed size, rather than using all input information at once, which significantly reduces the amount of calculation required for model parameters. Furthermore, different convolution kernel operations can be used for each window based on different types of information captured through the window (e.g., a person and an object in the same image are different types of information), so that CNN can better extract features from the input data.

[0086] RNN is a DNN network that uses feedback time-series information. The input of RNN includes the new input value at the current moment and the output value of the RNN at the previous moment. RNN is suitable for capturing time-correlated sequence features, and is particularly suitable for applications such as speech recognition and channel coding and decoding.

[0087] As described above, a loss function may be defined in the model training process. The loss function describes the gap or difference between the model's output value and an ideal target value. The specific form of the loss function is not limited by this disclosure. The model training process may be thought of as the following process: Some or all parameters of the model are adjusted so that the value of the loss function is less than a threshold or meets a target requirement.

[0088] The model may be referred to as an AI model, a rule, or by another name, without limitation. An AI model may be considered a specific method for implementing an AI function. An AI model represents a mapping relationship or function between the input and output of the model. The AI ​​function may include at least one of data collection, model training (or model learning), model information publication, model deduction (or called model inference, inference, prediction, etc.), model monitoring or model validation, inference result publication, etc. The AI ​​function may also be referred to as an AI (related) operation or an AI-related function.

[0089] In the present disclosure, an independent network element (e.g., referred to as an AI network element, AI node, or AI device) may be introduced into the communication system shown in FIG. 1 to implement some or all of the AI-related operations. The AI ​​network element may be directly connected to the access network device or indirectly connected to the access network device through a third-party network element. Optionally, the third-party network element may be a core network element. Alternatively, an AI entity may be configured or disposed in another network element in the communication system to implement the AI-related operations. The AI ​​entity may also be referred to as an AI module, AI unit, or by another name, and is primarily configured to implement some or all of the AI ​​functions. The specific name of the AI ​​entity is not limited in the present disclosure. Optionally, the other network element may be an access network device, a core network device, a cloud server, a network management system (operations, administration, and maintenance, OAM), etc. In this case, the network element that performs the AI-related operations is a network element in which the AI ​​functions are embedded. Because both AI network elements and AI entities implement AI-related functions, for ease of explanation, AI network elements and network elements with embedded AI functions are hereinafter collectively referred to as AI-function network elements.

[0090] In the present disclosure, an OAM is configured to operate, manage, and / or maintain a core network device (a network management system for the core network device) and / or an access network device (a network management system for the access network device). For example, the present disclosure includes a first OAM and a second OAM, where the first OAM is a network management system for the core network device and the second OAM is a network management system for the access network device. Optionally, the first OAM and / or the second OAM include an AI entity. In another example, the present disclosure includes a third OAM, where the third OAM is a network management system for both the core network device and the access network device. Optionally, the third OAM includes an AI entity.

[0091] Optionally, the AI ​​entity may be integrated into a terminal or a terminal chip to match and support the AI ​​function.

[0092] 3A-3D are exemplary diagrams of an application framework for AI in a communications system.

[0093] Optionally, as shown in FIG. 3A , an AI model is deployed in at least one of a core network device, an access network device, a terminal, or an OAM, and corresponding functions are implemented by using the AI ​​model. In the present disclosure, the AI ​​models deployed in different nodes may be the same or different. In the present disclosure, the models differ in at least one of the following: structural parameters of the model (e.g., at least one of the number of neural network layers, the neural network width, the connection relationships between layers, the neuron weights, the neuron activation functions, or the offsets in the activation functions); input parameters of the model (e.g., the type of the input parameters and / or the dimension of the input parameters); or output parameters of the model (e.g., the type of the output parameters and / or the dimension of the output parameters). Different input parameters of the model and / or different output parameters of the model may be described as different functions of the model. Unlike FIG. 3A , in FIG. 3B , the functions of the access network device are split between a CU and a DU. One or more AI models may be deployed in the CU, and / or one or more AI models may be deployed in the DU. Optionally, the CU in Figure 3B may be further split into a CU-CP and a CU-UP. Optionally, one or more AI models may be deployed in the CU-CP, and / or one or more AI models may be deployed in the CU-UP. Optionally, the OAM in Figure 3A or 3B may be further split into OAM of access network devices and OAM of core network devices.

[0094] Optionally, as shown in FIG. 3C , in a possible implementation, the access network device includes a near-real-time access network intelligent controller (RAN intelligent controller, RIC) module configured to perform model training and inference. For example, the near-real-time RIC may be configured to train an AI model and use the AI ​​model for inference. For example, the near-real-time RIC may obtain information about the network side and / or the terminal side from at least one of the CU, DU, RU, or terminal device, which may be used as training data or inference data. Optionally, the near-real-time RIC may deliver inference results to at least one of the CU, DU, RU, or terminal device. Optionally, the CU and DU may exchange inference results. Optionally, the DU and RU may exchange inference results. For example, the near-real-time RIC delivers the inference results to the DU, and the DU forwards the inference results to the RU.

[0095] Optionally, as shown in FIG. 3C , in another possible implementation, a non-real-time RIC configured to perform model training and inference is located outside the access network device (optionally, the non-real-time RIC may be located in an OAM, a cloud server, or a core network device). For example, the non-real-time RIC is configured to train an AI model and use the model for inference. For example, the non-real-time RIC may obtain information about the network side and / or the terminal side from at least one of a CU, a DU, a RU, or a terminal device, which may be used as training data or inference data. The inference result may be distributed to at least one of a CU, a DU, a RU, or a terminal device. Optionally, the CU and the DU may exchange the inference result. Optionally, the DU and the RU may exchange the inference result. For example, the non-real-time RIC distributes the inference result to the DU, and the DU forwards the inference result to the RU.

[0096] Optionally, as shown in FIG. 3C , in another possible implementation, the access network device includes a near-real-time RIC, and the non-real-time RIC is outside the access network device (optionally, the non-real-time RIC may be located in an OAM, a cloud server, or a core network device). As in the second possible implementation described above, the non-real-time RIC may be configured to perform model training and inference. Additionally / alternatively, as in the first possible implementation described above, the near-real-time RIC may be configured to perform model training and inference. Additionally / alternatively, the non-real-time RIC may perform model training, and the near-real-time RIC may acquire AI model information from the non-real-time RIC, acquire information about the network side and / or the terminal side from at least one of a CU, a DU, a RU, or a terminal device, and acquire an inference result based on the information and the AI ​​model information. Optionally, the near-real-time RIC may distribute the inference result to at least one of a CU, a DU, a RU, or a terminal device. Optionally, the CU and the DU may exchange the inference result. Optionally, the DU and the RU may exchange the inference result. For example, the near-real-time RIC delivers inference results to the DU, and the DU forwards the inference results to the RU. For example, the near-real-time RIC is configured to train model A and use model A for inference. For example, the non-real-time RIC is configured to train model B and use model B for inference. For example, the non-real-time RIC is configured to train model C and send information about model C to the near-real-time RIC, and the near-real-time RIC uses model C for inference.

[0097] In the present disclosure, one model may obtain one output through inference, and the output includes one or more parameters. The learning or training processes of different models may be deployed in different devices or nodes, or may be deployed in the same device or node. The inference processes of different models may be deployed in different devices or nodes, or may be deployed in the same device or node. This is not a limitation of the present disclosure.

[0098] In the present disclosure, the participating network elements may perform some or all of the steps or operations associated with the network elements. These steps or operations are merely examples. Other operations or variations of various operations may also be performed in the present disclosure. Furthermore, steps may be performed in a different sequence than presented in the present disclosure, and not all operations in the present disclosure may be performed.

[0099] In the examples of the present disclosure, unless otherwise specified or there is no logical contradiction, terms and / or descriptions in different examples may be mutually referenced, and technical features in different examples may be combined based on their internal logical relationships to form a new example.

[0100] In the present disclosure, "at least one (item)" may also be described as "one or more (items)," and "multiple (items)" may mean two (items), three (items), four (items), or more (items). This is not limited. " / " may represent an "or" relationship between associated objects. For example, A / B may represent A or B. "And / or" may indicate that there are three relationships between associated objects. For example, A and / or B may represent the following three cases: only A is present, both A and B are present, and only B is present, where A and B may be singular or plural. To facilitate the description of the technical solutions of the present disclosure, words such as "first," "second," "A," or "B" may be used to distinguish technical features with the same or similar functions. Words such as "first," "second," "A," or "B" do not limit the quantity and execution sequence. Furthermore, words such as "first," "second," "A," or "B" do not limit clear distinctions. Words such as "example" or "for example" are used to express an example, evidence, or illustration. Any design solution described as an "example" or "for example" should not be described as being preferred or having more advantages over another design solution. Words such as "example" or "for example" are intended to present related concepts in a particular manner for ease of understanding.

[0101] The network architectures and service scenarios described in this disclosure are intended to more clearly explain the technical solutions in this disclosure, and do not constitute limitations on the technical solutions provided in this disclosure. Those skilled in the art will know that with the development of network architectures and the emergence of new service scenarios, the technical solutions provided in this disclosure can also be applied to similar technical problems.

[0102] As described above, in a communication system supporting multi-antenna technology, signal transmission quality can be improved by using beamforming technology. When sending a signal to a terminal device, the access network device may perform beamforming at the access network device side. Specifically, the access network device aligns the beamforming direction of the signal's transmission beam with the exit angle of the main path of the channel so that the terminal device can obtain most of the signal transmission energy. Optionally, if the terminal device also supports multi-antenna technology, the terminal device may also perform beamforming. Specifically, the terminal device aligns the beamforming direction of the signal's reception beam with the incident angle of the main path of the channel. In another example, when receiving a signal from a terminal device, the access network device may perform beamforming at the access network device side. The access network device aligns the beamforming direction of the signal's reception beam with the incident angle of the main path of the channel so that the access network device can obtain most of the signal transmission energy. Optionally, if the terminal device also supports multi-antenna technology, the terminal device may also perform beamforming. The terminal device can align the beamforming direction of the signal transmission beam with the outgoing angle of the main path of the channel. By using the beamforming technique, the signal transmission can be completed with high quality and the signal energy received by the receiver can be improved.

[0103] To implement the transmitter beamforming technique, the signal transmitter may use a precoding technique so that the transmitted signal has a beamforming effect. Similarly, to implement the receiver beamforming technique, the signal receiver may use a precoding technique so that the received signal has a beamforming effect. For example, the signal propagation model may be expressed as Y=V*H*W*X+N (Equation 1).

[0104] N is noise, X is a transmitted signal, and W is a transmitter precoding matrix. The signal obtained by precoding X by using W is W*X. W*X is the final transmitted signal of the transmitter. W*X has a beamforming effect in space. The signal arriving at the receiver after channel propagation is H*W*X. V is the receiver precoding matrix. The signal obtained by precoding H*W*X by using V is V*H*W*X. V*H*W*X+N is the final received signal of the receiver, and V*H*W*X has a beamforming effect in space. The data format of the precoding matrix is ​​usually complex, and other data formats are not excluded in this disclosure. In this disclosure, the precoding matrix may also be referred to as a codebook, and one precoding matrix corresponds to one codebook.

[0105] In the above beamforming technique, W corresponds to the transmitter's transmitting beam, and V corresponds to the receiver's receiving beam. The angles of different main paths of the channel may be distributed over a wide range, for example, a horizontal range of 0 to 360 degrees and a vertical range of -90 to 90 degrees. However, one beam corresponding to one precoding matrix may only cover a limited angular range in space. Therefore, multiple precoding matrices may be supported in a system to support multiple beams to ensure good signal coverage. The process of determining W and / or V in a multi-beam system may be referred to as a beam management process, or the process of determining the beam corresponding to W and / or the beam corresponding to V may be referred to as a beam management process. The transmitter has a total of T candidate beams, in other words, the transmitter has a total of T candidate precoding matrices W, each of which is designated by W. i where i ranges from 0 to T−1, and the receiver has a total of R candidate beams. In other words, the receiver has R candidate precoding matrices V, each of which is V jwhere the value of j is assumed to range from 0 to R-1. T is a positive integer, e.g., a multiple or power of 2, such as 4, 8, 16, 32, or 64. Alternatively, the value of T is another possible integer, such as 5, 6, 10, or 12. This is not limited. R is a positive integer, such as 1, 2, or 4. In the beam management process, T precoding matrices W i and R precoding matrices V j From the top W and top V for communication between the transmitter and the receiver are determined, that is, the top beam pair is determined. In this process, the top beam pair can be determined through T*R rounds of beam polling. The T*R rounds of beam polling can be performed in a time-division manner.

[0106] For example, in the e-th round of beam polling, the transmitter sends a reference signal to the receiver device by using the i-th beam, and the receiver receives the reference signal by using the j-th beam. i Receive Y i =V j *H*W i *X+N (Equation 2).

[0107] X is the reference signal sent by the transmitter, and W i is the transmitter precoding matrix, H is the channel response, and V j is the receiver precoding matrix, and N is noise. e = i*R+j (corresponding to polling the receiver beam first, then the transmitter beam), or e = j*T+i (corresponding to polling the transmitter beam first, then the receiver beam), where the value of e ranges from 0 to T*R-1. iAfter receiving the e-th measurement, the receiver may obtain the e-th measurement through estimation. Through T*R rounds of beam polling, the receiver may obtain T*R measurements (also referred to as measurements, estimated values, etc.) of the reference signal. In this disclosure, the reference signal measurement is, for example, reference signal received power (RSRP), signal to interference plus noise ratio (SINR), or another possible estimated value obtained through measurement. The receiver selects one measurement with a top indicator from the T*R measurements and uses the beam corresponding to the measurement as the top beam. The receiver sends the transmitter a transmitter beam index (or precoding matrix index) corresponding to the measurement, and the transmitter may communicate with the receiver by using the beam (or precoding matrix). Furthermore, the receiver may communicate with the transmitter by using a receiver beam corresponding to the measurement.

[0108] In the present disclosure, for different receivers, such as terminal devices, the T candidate beams of a transmitter, such as an access network device, may be the same or different. In other words, for different terminal devices in a cell, the candidate beams used when an access network device communicates with the terminal device may be the same or different. This is not limited in the present disclosure. For one terminal device, the T candidate beams that can be used by the access network device to communicate with the terminal device may be referred to as the full beams of the terminal device. As described above, one beam corresponds to one precoding matrix, and the index of the beam may also be considered as the index of the precoding matrix.

[0109] In the method, the transmitter and receiver are related to a reference signal. In another signal, for example a signal sent by the receiver to the transmitter, the names of the receiver and transmitter may be interchanged.

[0110] The type of the reference signal is not limited. The value of the reference signal is notified to the receiver in advance so that the reference signal can be measured. For example, the value of the reference signal is agreed upon in a protocol or notified to the transmitter in advance by the receiver. This is not limited.

[0111] For example, when the transmitter is an access network device and the receiver is a terminal device, the reference signal may be a demodulation reference signal (DMRS) of a physical downlink control channel (PDCCH), a DMRS of a physical downlink shared channel (PDSCH), a channel state information reference signal (CSI-RS), a synchronization signal (e.g., a primary synchronization signal (PSS) and / or a secondary synchronization signal (SSS)), a DMRS of a synchronization signal, a phase tracking reference signal (PTRS), or another possible downlink signal.

[0112] For example, when the transmitter is a terminal device and the receiver is an access network device, the reference signal may be a DMRS on a physical uplink shared channel (PUSCH), a DMRS on a physical uplink control channel (PUCCH), a random access preamble (preamble), a sounding reference signal (SRS), or another possible uplink signal.

[0113] In the above method, the transmitter's top beam can be obtained from T candidate beams through T rounds of beam polling through measurements. The receiver's top beam can be obtained from R candidate beams through R rounds of beam polling through measurements. This sweeping process is sometimes called full beam sweeping, i.e., each candidate beam is swept. In this process, the transmitter and receiver implement angle alignment between the transmitter and receiver by sweeping all codebooks in a traversal manner. For example, if 64 precoding matrices in the transmitter's codebook correspond to 64 beamformed beams, respectively, and 4 precoding matrices in the receiver's codebook correspond to 4 beamformed beams, a total of 256 sweeps are required to determine the top receiver and top transmitter beamformed beam pair, resulting in very high sweeping overhead and latency.

[0114] For each transmitter beamformed beam, any receiver beamformed beam may form a transmit-receive beam pair with the transmitter beamformed beam. Therefore, the process of determining the top transmit-receive beam pair may be split into performing receiver beam sweeping on the transmitter beam to determine the top receiver beam that matches the transmitter beam. This process may then be repeated for each of the remaining T−1 transmitter beams to determine the overall top transmit-receive beam pair. Similarly, for each receiver beam, any transmitter beam may form a transmit-receive beam pair with the receiver beam. Therefore, the process of determining the top transmit-receive beam pair may be split into performing transmitter beam sweeping on the receiver beam to determine the top transmitter beam that matches the receiver beam. This process may then be repeated for each of the remaining R−1 receiver beams to determine the overall top transmit-receive beam pair. Because the principle of receiver beam sweeping is similar to that of transmitter beam sweeping, the beam management methods provided in this disclosure may be described herein by using transmitter beam sweeping as an example.

[0115] In the above full beam sweeping method, each candidate beam is swept, which causes high system overhead and latency. To reduce the overhead, AI technology is introduced in the present disclosure to implement beam management by using sparse beam sweeping and a beam prediction model. FIG. 4 shows an AI-based beam management method. By way of example and not limitation, T=64 in FIG. 4. In practice, the value of the number T of candidate beams can alternatively be another possible value. This is not a limitation.

[0116] S401: A transmitter sweeps a sparse beam by using a sparse beam pattern.

[0117] As shown in the sparse beam pattern in S401, among the 64 candidate beams, the transmitter sweeps a total of 16 beams. In the case of a full beam, i.e., 64 beams, the 16 beams are equivalent to several sparse beams in the full beam. For ease of understanding, in S401, horizontal and vertical directions are used to indicate the spatial direction of each beam. In practice, the beam division may be based on a two-dimensional planar direction, a three-dimensional spatial direction, or another possible scheme. This is not limited.

[0118] In S401, similar to the description related to Equation 1 above, the transmitter performs beam polling in a time-division manner by using each beam among the 16 beams shown in the black boxes. The transmitter separately performs beamforming on the reference signal by using 16 precoding matrices corresponding to the 16 beams, and sends the beamformed reference signal to the receiver in a time-division manner.

[0119] S402: The receiver receives a reference signal by using a sparse beam pattern and performs beam prediction.

[0120] The receiver may receive reference signals corresponding to 16 beams and obtain a total of 16 measurements. The 16 measurements are referred to as measurements corresponding to the sparse beam pattern in S401. The receiver inputs the 16 measurements into a beam prediction model and obtains the top K (Top-K) beams in the full beam pattern through inference. In other words, the receiver obtains the indices of the top K beams in the full beam pattern including 64 beams through prediction. K is a positive integer such as 1, 3, 4, 6, 8, or another possible value. In S402, K=3, as indicated by the shaded box.

[0121] The receiver may send the indices of the K beams or the indices of the K precoding matrices corresponding to the K beams to the transmitter. Optionally, the transmitter may communicate with the receiver by using any one beam in the K beams. Alternatively, when K is greater than 1, the transmitter may further determine the top beam from the K beams by using S403 and S404.

[0122] S403: The transmitter sweeps the top K beams.

[0123] K=3 is used as an example. Similar to Equation 1 above, the transmitter polls the reference signal by using each beam in the top three beams shown in the three shaded boxes. Specifically, the transmitter separately performs beamforming on the reference signal by using three precoding matrices (codebooks) corresponding to the three beams, and sends the beamformed reference signal to the receiver in a time-division manner.

[0124] S404: The receiver determines the top beam.

[0125] The receiver may receive three reference signals obtained through beamforming performed by using the top three beams and obtain a total of three measurements. The receiver may determine the top measurement among the three measurements. The beam corresponding to the top measurement is the top beam. The receiver may send an index of the top beam or an index of the precoding matrix corresponding to the top beam to the transmitter.

[0126] As shown in Figure 5, an example is used in which the transmitter is an access network device and the receiver is a terminal device. The top three beams are assumed to be beam 0, beam 1, and beam 2. The access network device sweeps the three beams, and the terminal device can learn through measurements that the top beam is beam 1.

[0127] After the top beam is determined, the transmitter and receiver may perform data channel transmission by using the beam. For example, the access network device may send downlink data, e.g., a PDSCH, to the terminal device by using the top beam, and / or the access network device may receive downlink data, e.g., a PUSCH, from the terminal device by using the top beam. In the present disclosure, sending data of a data channel by using a beam may be understood as precoding the data by using a precoding matrix corresponding to the beam.

[0128] According to the AI ​​model-based beam management method shown in Figure 4, the number of beam polls can be reduced from 64 time division polls to 16 time division polls or 16 + 3 = 19 time division polls, resulting in reduced system overhead and reduced latency.

[0129] The performance of the beam prediction model during inference is related to the sparse beam pattern used when the beam prediction model is trained. For example, when the beam prediction model is trained, the input of the beam prediction model is determined by using measurements corresponding to sparse beam pattern 1 so that the difference between the top K results output by the beam prediction model and the top K results obtained through full beam sweeping is smaller than a threshold. In this case, when inference is performed by using the beam prediction model obtained through training, measurements corresponding to sparse beam pattern 2 may be input to the beam prediction model. In this case, sparse beam pattern 2 does not correspond to or match the beam prediction model. The output result may be inaccurate, and even measurements corresponding to sparse beam pattern 2 may not be input to the model because the input dimension does not meet the requirements. For example, sparse beam pattern 1 indicates that eight beams are swept, and sparse beam pattern 2 indicates that 16 beams are swept.

[0130] In practice, channel environments are complex and subject to change. Different channel environments may use different sparse beam patterns, or even irregular sparse beam patterns. Therefore, in practice, a dozen, several dozen, or even hundreds of different sparse beam patterns may need to be used. FIG. 6 shows four different sparse beam patterns for a full beam including 64 beams. When the full beam patterns are different or the system supports multiple full beam patterns, for example, when the full beam pattern includes 128 beams, or when the system supports both a full beam pattern including 64 beams and a full beam pattern including 128 beams, there are more sparse beam patterns. Therefore, to meet practical application requirements, one beam prediction model may need to be configured for each sparse beam pattern. In this case, the receiver needs to consume a large amount of storage resources to store multiple beam prediction models or a large amount of air interface resources to exchange multiple beam prediction models with the transmitter.

[0131] To reduce resource consumption, the present disclosure proposes a beam management method. This method can reduce the number of beam prediction models in a system. In this method, an input adaptation layer (also called an input adaptation model) and / or an output adaptation layer (also called an output adaptation model) with a simple structure is introduced, so that a small number of beam prediction models can be used to support various system requirements and reduce system overhead. The beam prediction model may be referred to as a reference beam prediction model or a basic beam prediction model. The small number is not limited. For example, the small number can be 2, 3, 5, 6, 8, or another possible value. This is not limited.

[0132] As shown in FIG. 7, the methods provided in the present disclosure are respectively explained by using four scenarios.

[0133] Scenario 1: A sparse beam pattern matches the beam prediction model.

[0134] Similar to S401 and S402 in FIG. 4, the transmitter sends out a reference signal by using pattern A and sweeps the 16 beams in pattern A in a time-division manner. The receiver obtains the indices of the top K1 beams in the full beam through prediction based on the reference signal measurements and the beam prediction model. FIG. 7 shows an example where the full beam includes 64 beams and K1=5. In other words, the sparse beam pattern matching the beam prediction model in FIG. 7 is pattern A, and the output matching the beam prediction model is the indices of the top five beams in the full beam.

[0135] Scenario 2: The sparse beam pattern (input format) does not match the beam prediction model.

[0136] A method corresponding to Scenario 2 may be described as follows: The receiver maps reference signal measurements to input adaptation information by using an input adaptation layer, where a beam pattern corresponding to the reference signal measurements is a first beam pattern and the adaptation information corresponds to a second beam pattern, and the first beam pattern is different from the second beam pattern. The receiver obtains a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes the adaptation information. Alternatively, a method corresponding to Scenario 2 may be described as follows: The receiver maps reference signal measurements to input adaptation information by using an input adaptation layer, where a beam pattern corresponding to the reference signal measurements is the first beam pattern. The receiver obtains a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes the adaptation information, and the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern.

[0137] In this disclosure, when it is described that the input of a model includes one or more features, for example, that the input of a beam prediction model includes reference signal measurements or input adaptation information, it does not exclude that the input of the model may further include other features.

[0138] Unlike Scenario 1, in Scenario 2, the transmitter sends out a reference signal by using Pattern B. However, Pattern A matches the beam prediction model. Therefore, after obtaining the reference signal measurement corresponding to Pattern B, the transmitter maps the reference signal measurement to input adaptation information by using an input adaptation layer. The transmitter inputs the input adaptation information into the beam prediction model and outputs the indices of the top K1 beams in the full beam.

[0139] Scenario 3: The output format does not match the beam prediction model.

[0140] A method corresponding to Scenario 3 can be described as follows: A receiver inputs reference signal measurements into a beam prediction model to obtain a first beam prediction result, and the receiver maps the first beam prediction result to a second beam prediction result by using an output adaptation layer.

[0141] Unlike Scenario 1, the beam prediction results required in Scenario 3 are the indices of the top K2 beams, e.g., the indices of the top five beams. However, the beam prediction results that match the beam prediction model are the indices of the top K1 beams, e.g., the indices of the top three beams. Therefore, after obtaining the indices of the top five beams through inference by using the beam prediction model, the transmitter maps the indices of the top five beams to the indices of the top three beams by using an output adaptation layer.

[0142] Scenario 4: The sparse beam pattern (input format) and output format do not match the beam prediction model.

[0143] Scenario 4 may be understood as a combination of Scenario 2 and Scenario 3. A method corresponding to Scenario 4 may be described as follows: The receiver maps reference signal measurements to input adaptation information by using an input adaptation layer, where a beam pattern corresponding to the reference signal measurements is a first beam pattern and the adaptation information corresponds to a second beam pattern, and the first beam pattern is different from the second beam pattern. The receiver obtains a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes the adaptation information. The receiver maps the first beam prediction result to a second beam prediction result by using an output adaptation layer. Alternatively, a method corresponding to Scenario 4 may be further described as follows: The receiver maps reference signal measurements to input adaptation information by using an input adaptation layer, where a beam pattern corresponding to the reference signal measurements is the first beam pattern. The receiver obtains a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes the adaptation information, and the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. The receiver maps the first beam prediction result to the second beam prediction result by using an output adaptation layer.

[0144] In the present disclosure, the input adaptation layer is mainly configured to map the reference signal measurements to be in an input format that matches the beam prediction model. The name of the input adaptation layer is not limited in the present disclosure. For example, the input adaptation layer may be referred to as an input adaptation model, a first model, or by another name. The structure of the input adaptation layer is not limited in the present disclosure. Optionally, the structure of the input adaptation layer is a neural network. For example, the input adaptation layer includes at least one of one or more fully connected layers, one or more CNN layers, or one or more RNN layers. Figure 8 shows a possible structure of the input adaptation layer. For example, the input dimension of the input adaptation layer is [N in ,N sc] and the output dimension of the input adaptation layer can be expressed as [N out ,N sc ], where N in is the number of beamformed beams in the sparse beam pattern before mapping, and N out is the number of beamformed beams in the sparse beam pattern that matches the beam prediction model, and N sc is the number of resources occupied by the reference signal, e.g., the number of subcarriers or the number of resource elements (RE). in , N out , N sc is a positive integer. in and N out and can have the same value. For example, N in and N out correspond to Pattern 1 and Pattern 2, respectively, shown in Figure 6. Alternatively, N in and N out The values ​​of N and N can be different. in and N out and correspond to pattern 3 shown in Figure 6 and pattern A shown in Figure 7, respectively. in and N out corresponds to Pattern 3 shown in FIG. 6 and Pattern A shown in FIG. 7, respectively, the swept beam in Pattern 3 shown in FIG. 6 includes the swept beam in Pattern A shown in FIG. 7, so the input adaptation layer may perform adaptation by performing a downsampling operation.

[0145] In the present disclosure, the beam prediction model predicts a beam prediction result mainly based on reference signal measurement quantities. For the sake of explanation, the following uses an example in which the beam prediction result is the top beam of a full beam. As described above, the full beam pattern may be the same or different for different terminal devices in a cell. This is not limited. For example, if the full beam pattern is cell-level information, the full beam pattern is the same for different terminal devices in a cell. Alternatively, if the full beam pattern is terminal device-level information or terminal device group-level information, the full beam pattern may be different for different terminal devices in a cell. In an extensible manner, the beam prediction model in the present disclosure may further predict other types of results, and the principle is similar. For example, the beam prediction model may predict the top beam of a beam subset of a full beam, the top beam subset of a full beam, a beam in a full beam or a subset of a full beam whose corresponding measurement quantity is lower than a threshold, or a beam in a full beam or a subset of a full beam whose corresponding measurement quantity is higher than a threshold. This is not limited. The name of the beam prediction model is not limited in the present disclosure. For example, the beam prediction model may be referred to as a second model or by another name.

[0146] In the present disclosure, the output adaptation layer is mainly configured to map the first beam prediction result to the second beam prediction result, e.g., map the indices of the top K1 beams to the indices of the top K2 beams. K1 and K2 are positive integers, and the values ​​of K1 and K2 are different. The name of the output adaptation layer is not limited in the present disclosure. For example, the output adaptation layer may be referred to as an output adaptation model, a third model, or by another name. The structure of the output adaptation layer is not limited in the present disclosure. Optionally, the structure of the output adaptation layer is a neural network. For example, the output adaptation layer includes at least one of one or more fully connected layers, one or more CNN layers, or one or more RNN layers. FIG. 9 shows a possible structure of the output adaptation layer. The input of the output adaptation layer is the first beam prediction result, e.g., the indices of the top K1 beams, and the output of the output adaptation layer is the second beam prediction result, e.g., the indices of the top K2 beams.

[0147] 10 shows a specific beam prediction procedure according to the present disclosure. The procedure is described by using an example in which the transmitter is an access network device and the receiver is a terminal device. When the number of antennas or antenna ports on the network side is much larger than that on the terminal side, the number of beamformed beams on the network side is also much larger than the number of beams on the terminal side. In this way, the potential overhead and latency caused by implementing sparse beam sweeping on the network side are significantly reduced. As described above, in the present disclosure, in order to implement beam management on the terminal device side, it is clear that the transmitter can alternatively be a terminal device and the receiver is an access network device.

[0148] Optionally, in S1001, the terminal device reports capability information to the access network device.

[0149] The terminal device may report capability information to the access network device based on a query request of the access network device, or may actively report capability information to the access network device when accessing the network. A specific time or a specific trigger event for reporting capability information by the terminal device is not limited in the present disclosure.

[0150] The terminal device may report at least one of the following capability information to the access network device:

[0151] - Whether the end device supports running machine learning models or has AI capabilities;

[0152] - Machine learning model types supported by the terminal device, for example, the terminal device may report that the terminal device supports at least one of machine learning models such as CNN, RNN, and random forest models;

[0153] - the size of the memory space that can be used by the terminal device to store the machine learning model;

[0154] - computing power information of the terminal device, indicating at least one of the following:

[0155] ■ The computing capabilities of the terminal device for running the model, for example, at least one of the following information of the terminal device: the operating speed of the processor, the amount of data that can be processed by the processor, etc.; or

[0156] ■ Energy consumption information of the terminal device, for example, at least one of the operating power consumption of the chip or the battery capacity of the terminal device;

[0157] - hardware information of the terminal device, including, but not limited to, at least one of antenna information (e.g., the number and / or polarization direction of antennas) or radio frequency channels; and

[0158] - A beam prediction model stored in the terminal device.

[0159] When the terminal device supports the operation of a machine learning model or has AI capability, the access network device may configure an AI model for the terminal device to operate the method in the present disclosure. The access network device may configure an appropriate AI model for the terminal device based on the machine learning model type supported by the terminal device, the size of the memory space used to store the machine learning model, computing power information, hardware information, etc. Based on the beam prediction model stored in the terminal device, the access network device may know the existing beam prediction model on the terminal device side and / or determine whether a new beam prediction model needs to be further configured for the terminal device, etc.

[0160] S1001 is an optional step. For example, when the capabilities of a terminal device are agreed upon in a protocol, the terminal device does not need to report its capabilities by using S1001. Alternatively, when the terminal device has previously reported its capabilities to the access network device and the information about the capabilities is relatively fixed, in a particular beam prediction process, the terminal device may not need to report its capabilities again by using S1001.

[0161] Optionally, at S1002, the access network device sends information regarding the beam prediction model to the terminal device.

[0162] In a possible implementation, the beam prediction model of the terminal device is agreed upon in the protocol. In this case, the access network device does not need to configure the beam prediction model for the terminal device by using S1002. Optionally, for different terminal devices, the beam prediction models agreed upon in the protocol may be the same or different. This is not limited.

[0163] In a possible implementation, the access network device may send information about the beam prediction model to the terminal device. The beam prediction model is used by the terminal device to predict the beam. This information indicates specific structural information of the beam prediction model. Optionally, the beam prediction models configured by the access network device for different terminal devices may be the same or different. This is not limited.

[0164] In the present disclosure, when the structure of the beam prediction model is a neural network, the information about the beam prediction model may include at least one of the following information: a model index (or identifier), structural parameters of the model (e.g., at least one of the number of neural network layers, the neural network width, the connection relationships between layers, the neuron weights, the neuron activation functions, or the offsets in the activation functions), input parameters of the model (e.g., the types of the input parameters and / or the dimensions of the input parameters), or output parameters of the model (e.g., the types of the output parameters and / or the dimensions of the output parameters).

[0165] In the present disclosure, the access network device may send information to the terminal device in a broadcast or multicast manner, but this is not limited thereto, and different information may be sent in the same manner or in different manners, but this is not limited thereto.

[0166] In the present disclosure, the model presented to the terminal device by the access network device may be obtained by the access network device through training. For example, the model may be obtained through training by using a near-real-time RIC, CU, DU, or another model in the access network device, or may be downloaded by the access network device from a third-party website, or may be obtained through training by using a non-real-time RIC and sent to the access network device, or may be obtained through training by using OAM and sent to the access network device, or may be obtained through training by using a core network device and sent to the access network device. This is not limited to this.

[0167] In a possible implementation, the access network device may indicate one model from multiple candidate beam prediction models to the terminal device. The multiple candidate beam prediction models may also be referred to as a candidate beam prediction model set. For each model in the multiple candidate beam prediction models, information about the model may be agreed upon in a protocol or may be sent in advance by the access network device to the terminal device by using signaling. This is not limited thereto.

[0168] Method A1: E1 sparse beam patterns {F1, F2, ... F E1} and E2 candidate beam prediction models {S1,S2,...S E2} can be configured. This correspondence can be agreed upon in a protocol or can be notified in advance by the access network device to the terminal device by using signaling. E1 and E2 are positive integers, and E2 is less than or equal to E1. For example, E1 is equal to E2 and F i is S iwhere i ranges from 1 to E1. The E2 candidate beam prediction models may be referred to as reference beam prediction models, base beam prediction models, or by other names. This is not a limitation of the present disclosure. Information about each beam prediction model in the E2 candidate beam prediction models may be agreed upon in a protocol or notified in advance to the terminal device by the access network device using signaling. Each beam prediction model corresponds to one index. One or more sparse beam patterns may correspond to one candidate beam prediction model. The terminal device may select {F1, F2, ... F E1} and {S1,S2,...S E2} and {S1,S2,...S E2 When indicating one model A from the E2 beam prediction models to a terminal device, the access network device may indicate an index or identifier of model A, or the access network device may indicate a sparse beam pattern A, where pattern A is a set of {F1, F2, ... F E1}, where the terminal devices are {F1,F2,...F E1} and {S1,S2,...S E2}, a model A corresponding to pattern A can be determined.

[0169] Method A2: The access network device selects a plurality of candidate beam prediction models, for example, E2 candidate beam prediction models {S1, S2, ... S E2} to indicate to the terminal device the index of the beam prediction model configured for the terminal device. As described in Table 1, there are five candidate beam prediction models, and each candidate beam prediction model corresponds to one index. The access network device may indicate one index from the five indexes to the terminal device to indicate the model configured for the terminal device.

[0170] [Table 1]

[0171] Optionally, in S1003, the access network device sends information about the adaptation layer to the terminal device.

[0172] If the access network device is E1}, these beam patterns have corresponding basic beam prediction models, so the input adaptation layer is not required. In this case, S1003 does not need to be performed to configure the input adaptation layer. Alternatively, the access network device may use the beam patterns in {F1, F2, ... F E1 When it is expected to perform beam management by using a beam pattern other than the beam pattern in}, an input adaptation layer is required. The access network device may send information about the input adaptation layer to the terminal device by using S1003. For example, information about the input adaptation layer may be agreed upon in a protocol. In this case, S1003 does not need to be performed. Alternatively, the access network device may send information about the input adaptation layer to the terminal device. Alternatively, the access network device indicates one input adaptation layer from multiple candidate input adaptation layers to the terminal device. Information about each input adaptation layer in the multiple candidate input adaptation layers may be agreed upon in a protocol or may be sent in advance by the access network device to the terminal device by using signaling. This is not limited thereto.

[0173] When the output of the beam prediction model of the terminal device meets the requirements, an output adaptation layer is not required for result adaptation. In this case, the output adaptation layer does not need to be configured by using S1003. When the output of the beam prediction model of the terminal device does not meet the requirements, information about the output adaptation layer can be agreed upon in a protocol. In this case, S1003 does not need to be performed to configure the output adaptation layer. Alternatively, the access network device may send information about the output adaptation layer to the terminal device. Alternatively, the access network device indicates one output adaptation layer from multiple candidate output adaptation layers to the terminal device. Information about each output adaptation layer in the multiple candidate output adaptation layers can be agreed upon in a protocol or can be sent in advance by the access network device to the terminal device by using signaling. This is not limited thereto.

[0174] The input adaptation layer and the output adaptation layer may be indicated to the terminal device by the access network device using one message or two messages. This is not a limitation of the present disclosure. Optionally, the input adaptation layer and the output adaptation layer may be sent using two steps. For example, S1003 is split into S1003a and S1003b. S1003a is used to send the input adaptation layer, and S1003b is used to send the output adaptation layer.

[0175] Optionally, the structure of the input adaptation layer is a neural network, and the information about the input adaptation layer may include at least one of the following information: an index (or identifier) ​​of the model; structural parameters of the model (e.g., at least one of the number of neural network layers, the neural network width, the connection relationships between layers, the weights of neurons, the activation functions of neurons, or the offsets in the activation functions); input parameters of the model (e.g., the types of input parameters and / or the dimensions of the input parameters); or output parameters of the model (e.g., the types of output parameters and / or the dimensions of the output parameters).

[0176] Optionally, the structure of the output adaptation layer is a neural network, and the information about the output adaptation layer may include at least one of the following information: an index (or identifier) ​​of the model; structural parameters of the model (e.g., at least one of the number of neural network layers, the neural network width, the connection relationships between layers, the weights of neurons, the activation functions of neurons, or the offsets in the activation functions); input parameters of the model (e.g., the types of input parameters and / or the dimensions of the input parameters); or output parameters of the model (e.g., the types of output parameters and / or the dimensions of the output parameters).

[0177] S1004: The access network device sweeps the beam.

[0178] Possible Scenario 1:

[0179] The access network device sends a reference signal to the terminal device by using a second beam pattern. The terminal device receives the reference signal based on the second beam pattern and estimates a measurement quantity of the reference signal. The terminal device may predict indexes of the top K1 beams by using the beam prediction model configured by the access network device for the terminal device in S1002 and by using the method in Scenario 1 shown in FIG. 7. In this case, the second beam pattern is a pattern that matches the beam prediction model of the terminal device, and the required beam prediction result matches the output result of the beam prediction model.

[0180] Optionally, an index of the second beam pattern or an index of a precoding matrix (codebook) corresponding to the second beam pattern may be agreed upon in a protocol or configured for the terminal device by the access network device by using signaling. Based on the configuration, the terminal device may determine the second beam pattern. The terminal device receives a reference signal based on the second beam pattern.

[0181] The access network device configures a beam prediction model for the terminal device by using the method A1 in S1002, and the second beam pattern is {F1, F2, ... F E1}, the terminal device may generate a second beam pattern and E1 sparse beam patterns {F1, F2, ... F E1} and E2 candidate beam prediction models {S1,S2,...S E2} and the correspondence between the beam prediction model of the terminal device.

[0182] In the present disclosure, the required beam prediction result may be agreed upon in a protocol, determined by the terminal device (e.g., determined based on parameters such as channel quality), or configured by the access network device for the terminal device, without limitation.

[0183] Possible Scenario 2:

[0184] The access network device sends a reference signal to the terminal device by using a first beam pattern. The terminal device receives the reference signal based on the first beam pattern and estimates a measurement quantity of the reference signal. The terminal device may predict the indexes of the top K1 beams by using the beam prediction model configured by the access network device for the terminal device in S1002 and by using the method in Scenario 2 shown in FIG. 7. In this case, the second beam pattern is a pattern that matches the beam prediction model of the terminal device. The first beam pattern is different from the second beam pattern. The first beam pattern does not match the beam prediction model of the terminal device, but the required beam prediction result matches the output result of the beam prediction model.

[0185] Optionally, an index of the first beam pattern or an index of a precoding matrix (codebook) corresponding to the first beam pattern may be agreed upon in a protocol or configured for the terminal device by the access network device by using signaling. Based on the configuration, the terminal device may determine the first beam pattern. The terminal device receives a reference signal based on the first beam pattern.

[0186] For configuring a beam prediction model for a terminal device, see Possible Scenario 1 above, and the details will not be described again here.

[0187] Possible Scenario 3:

[0188] Unlike possible scenario 1 above, in possible scenario 3, the required beam prediction result does not match the output result of the beam prediction model. In this case, the terminal device maps the indices of the top K1 beams to the indices of the top K2 beams by using an output adaptation layer and by using the method shown in scenario 3 in Figure 7.

[0189] For the second beam pattern and for configuring a beam prediction model for the terminal device, please refer to Possible Scenario 1 above, and the details will not be described again here.

[0190] Possible Scenario 4:

[0191] Unlike possible scenario 2 above, in possible scenario 4, the required beam prediction result does not match the output result of the beam prediction model. In this case, the terminal device maps the indices of the top K1 beams to the indices of the top K2 beams by using an output adaptation layer and by using the method shown in scenario 4 in Figure 7.

[0192] For the first beam pattern and for configuring the input adaptation layer for the terminal device, please refer to possible scenario 2 above. Details will not be described again here. For configuring the output adaptation layer for the terminal device, please refer to possible scenario 3 above. Details will not be described again here.

[0193] S1005: The terminal device feeds back the index of the beam.

[0194] The terminal device reports to the access network device the beam prediction results obtained in S1003 that meet the requirements.

[0195] 11 illustrates another specific beam prediction procedure according to the present disclosure. In the procedure illustrated in FIG. 10, the adaptation layer is signaled to the terminal device by the access network device. However, in the procedure illustrated in FIG. 11, the adaptation layer is acquired by the terminal device through training.

[0196] Optionally, in S1101, the terminal device reports capability information to the access network device.

[0197] Same as S1001.

[0198] Optionally, at S1102, the access network device sends information regarding the beam prediction model to the terminal device.

[0199] Same as S1002.

[0200] Optionally, in S1103, the access network device sweeps the full beam. Optionally, in S1104, the terminal device trains an adaptation layer.

[0201] Possible Scenario 1:

[0202] Similar to possible scenario 1 in S1004, if the access network device prepares to communicate with the terminal device by using a second beam pattern, and the second beam pattern is a pattern that matches the beam prediction model of the terminal device, and if the required beam prediction result matches the output result of the beam prediction model, S1103 and S1104 do not need to be performed.

[0203] The method for configuring the second beam pattern is the same as that in Scenario 1 in S1004, and the details will not be described again here.

[0204] Possible Scenario 2:

[0205] Similar to possible scenario 2 in S1004, it is assumed that the access network device prepares to communicate with the terminal device by using a first beam pattern. The second beam pattern matches the beam prediction model of the terminal device, and the first beam pattern does not match the beam prediction model of the terminal device, but the required beam prediction result matches the output result of the beam prediction model. The method for configuring the first beam pattern is the same as that in scenario 2 in S1003. Details will not be described again here. In this case, the terminal device may obtain the input adaptation layer through training by using the method shown in FIG. 12A.

[0206] As shown in FIG. 12A, the access network device performs full beam sweeping by using S1103.

[0207] The terminal device trains the input adaptation layer by using S1104.

[0208] Operation 1: The terminal device obtains an ideal beam prediction result based on measurements corresponding to a full beam.

[0209] The terminal device receives a reference signal based on a full beam (a total of T beams) pattern and estimates T first measurements of the reference signal. The terminal device may obtain the top K1 first measurements among the T first measurements and use the K1 beams corresponding to the T first measurements as the ideal top K1 beams.

[0210] Operation 2: The terminal device maps the measurement quantity corresponding to the sparse beam of the first beam pattern to input adaptation information by using an input adaptation layer, and obtains an actual beam prediction result based on the input adaptation information and the beam prediction model.

[0211] The terminal device may extract t second measurements corresponding to the sparse beam pattern from the T second measurements based on the T second measurements corresponding to the full beams, the full beam pattern, and the sparse beam pattern, where t is the number of beams that need to be swept in the sparse beam pattern. Alternatively, the terminal device obtains the t second measurements corresponding to the sparse beams through estimation based on the full beam pattern and the sparse beam pattern. The type of the first measurement and the type of the second measurement may be the same or different. For example, the first measurement and the second measurement may be RSRP, SINR, estimated channel state information (CSI), or another possible measurement. This is not limited. For example, both the first measurement and the second measurement may be RSRP or CSI, or one may be RSRP and the other may be CSI. The terminal device uses the t second measurements as inputs of an input adaptation layer to obtain input adaptation information, similar to scenario 2 in FIG. 7. Then, the beam prediction model obtains indices of the top K1 beams based on the input adaptation information.

[0212] Operation 3: The terminal device trains an input adaptation layer based on the ideal beam prediction result and the actual beam prediction result. Specifically, the terminal device adjusts parameters of the input adaptation layer based on the ideal beam prediction result and the actual beam prediction result to enable the difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold.

[0213] The terminal device calculates a loss function based on the indices of the top K1 ideal beams obtained through measurement and the indices of the top K1 beams output by the beam prediction model. If the loss function meets the requirements, for example, if the loss function indicates that the difference between the indices (labels) of the top K1 ideal beams and the indices of the top K1 beams output by the beam prediction model is less than (or less than or equal to) a threshold, the training of the input adaptation layer is considered to be completed, the training process ends, and the input adaptation layer is then used as the input adaptation layer actually used to predict beams. If the loss function does not meet the requirements, the model parameters of the input adaptation layer are updated, and operation 2 above is performed.

[0214] The initial model information of the input adaptation layer may be agreed upon in a protocol or may be indicated to the terminal device by the access network device, but this is not limited thereto.

[0215] Possible Scenario 3:

[0216] Similar to possible scenario 3 in S1004, it is assumed that the access network device prepares to send a reference signal to the terminal device by using a second beam pattern. The second beam pattern matches the beam prediction model of the terminal device, and the required beam prediction result matches the output result of the beam prediction model. The method for configuring the second beam pattern is the same as that in scenario 3 in S1004. Details will not be described again here. In this case, the terminal device may acquire the input adaptation layer through training by using the method shown in FIG. 12B.

[0217] As shown in FIG. 12B, the access network device performs full beam sweeping by using S1103.

[0218] The terminal device trains the output adaptation layer by using S1104.

[0219] Operation 1: The terminal device obtains an ideal beam prediction result based on measurements corresponding to a full beam.

[0220] The terminal device receives a reference signal based on a full beam (a total of T beams) pattern and estimates T first measurement quantities of the reference signal. The terminal device may obtain the top K2 first measurement quantities in the T first measurement quantities and use the K2 beams corresponding to the T first measurement quantities as the ideal top K2 beams. The description of the first measurement quantity and the second measurement quantity is the same as that in FIG. 12A. The details will not be described again here.

[0221] Operation 2: The terminal device obtains a first beam prediction result based on a measurement corresponding to a sparse beam of the second beam pattern and a beam prediction model, and maps the first beam prediction result to an actual beam prediction result by using an output adaptation layer.

[0222] The terminal device may extract t second measurements corresponding to the sparse beam pattern from the T second measurements based on the T second measurements corresponding to the full beams, the full beam pattern, and the sparse beam pattern, where t is the number of beams that need to be swept in the sparse beam pattern. Alternatively, the terminal device obtains the t second measurements corresponding to the sparse beams through estimation based on the full beam pattern and the sparse beam pattern. The terminal device inputs the t second measurements into a beam prediction model to obtain indices of the top K1 beams. The terminal device inputs the indices of the top K1 beams into an output adaptation layer to obtain indices of the top K2 beams.

[0223] Operation 3: The terminal device trains an output adaptation layer based on the ideal beam prediction result and the actual beam prediction result. Specifically, the terminal device adjusts parameters of the output adaptation layer based on the ideal beam prediction result and the actual beam prediction result to enable the difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold.

[0224] The terminal device calculates a loss function based on the indices of the ideal top K2 beams obtained through measurement and the indices of the top K2 beams output by the output adaptation layer. If the loss function meets the requirement, for example, if the loss function indicates that the difference between the indices (labels) of the ideal top K2 beams and the indices of the top K2 beams output by the beam prediction model is less than (or less than or equal to) a threshold, the training of the output adaptation layer is considered to be completed, the training process ends, and the output adaptation layer is used as the output adaptation layer actually used to predict beams thereafter. If the loss function does not meet the requirement, the model parameters of the output adaptation layer are updated, and operation 2 above is performed.

[0225] The initial model information of the output adaptation layer may be agreed upon in a protocol or may be indicated to the terminal device by the access network device, without being limited thereto.

[0226] Possible Scenario 4:

[0227] Similar to possible scenario 4 in S1004, the access network device prepares to send a reference signal to the terminal device by using a first beam pattern. The first beam pattern does not match the beam prediction model of the terminal device, and the required beam prediction result matches the output result of the beam prediction model. The method for configuring the first beam pattern is the same as that in scenario 4 in S1004. Details will not be described again here. In this case, the terminal device may acquire the input adaptation layer through training by using the method shown in FIG. 12C.

[0228] As shown in FIG. 12C, the access network device performs full beam sweeping by using S1103.

[0229] The terminal device trains the input adaptation layer and / or the output adaptation layer by using S1104.

[0230] Operation 1: The terminal device obtains an ideal beam prediction result based on measurements corresponding to a full beam.

[0231] The terminal device receives a reference signal based on a full beam (a total of T beams) pattern and estimates T first measurements of the reference signal. The terminal device may obtain the top K2 first measurements in the T first measurements and use the K2 beams corresponding to the K2 first measurements as the ideal top K2 beams. The description of the first measurement and the second measurement is the same as that in 12A. Details will not be described again here.

[0232] Operation 2: The terminal device maps a measurement quantity corresponding to a sparse beam of the first beam pattern to the input adaptation information by using the input adaptation information, obtains a first beam prediction result based on the input adaptation information and the beam prediction model, and maps the first beam prediction result to an actual beam prediction result by using an output adaptation layer.

[0233] The terminal device may extract t measurements corresponding to the sparse beam pattern from the T second measurements based on the T second measurements corresponding to the full beams, the full beam pattern, and the sparse beam pattern, where t is the number of beams that need to be swept in the sparse beam pattern. Alternatively, the terminal device obtains the t second measurements corresponding to the sparse beams through estimation based on the full beam pattern and the sparse beam pattern. The terminal device uses the t measurements as inputs to an input adaptation layer to obtain input adaptation information. Then, the beam prediction model obtains indices of the top K1 beams based on the input adaptation information. The terminal device inputs the indices of the top K1 beams to an output adaptation layer to obtain indices of the top K2 beams.

[0234] Operation 3: The terminal device trains the input adaptation layer and / or the output adaptation layer based on the ideal beam prediction result and the actual prediction result.

[0235] The terminal device calculates a loss function based on the indices of the ideal top K2 beams obtained through measurements and the indices of the top K2 beams output by the beam prediction model. If the loss function satisfies the requirement—for example, if the loss function indicates that the difference between the indices (labels) of the ideal top K2 beams and the indices of the top K2 beams output by the beam prediction model is less than (or less than or equal to) a threshold—the training of the input adaptation layer and the output adaptation layer is considered complete, the training process ends, and the input adaptation layer is then used as the input adaptation layer actually used to predict beams. If the loss function does not satisfy the requirement, the model parameters of the input adaptation layer and / or the output adaptation layer are updated, and operation 2 above is performed. Updating the parameters of the input adaptation layer, updating the parameters of the output adaptation layer, or updating both the parameters of the input adaptation layer and the output adaptation layer may decrease the value of the loss function in the training process. Therefore, in a training round, one or both of the parameters of the input adaptation layer and the output adaptation layer may be updated, and operation 2 is then performed.

[0236] Optionally, at S1105, the terminal device sends a training completion indication to the access network device.

[0237] Based on the indication, the access network device may know that the terminal device has completed training, and beam management may be performed by using the trained input adaptation layer and / or output adaptation layer. This step is optional. For example, if the time period for the terminal device to complete training is agreed upon in the protocol, or if the access network device indicates the time period for the terminal device to complete training, S1105 does not need to be performed.

[0238] Optionally, in S1106, the access network device sweeps the beam. Optionally, in S1107, the terminal device feeds back the index of the beam.

[0239] Similar to S1004 and S1005, the access network device and the terminal device may perform sparse beam sweeping, and the terminal device feeds back the index of the beam. The method of determining the input adaptation layer and the output adaptation layer used in S1004 and S1005 is different from the method of determining the input adaptation layer and the output adaptation layer used in S1106 and S1107.

[0240] It can be understood that to implement the functions in the above methods, the access network device, the module of the access network device, the terminal device, the AI ​​function network element, etc. include corresponding hardware structures and / or software modules for implementing the functions. Those skilled in the art should easily recognize that the present disclosure can be implemented by hardware or a combination of hardware and computer software with reference to the units and method steps in the examples described in the present disclosure. Whether the functions are implemented by hardware or by hardware driven by computer software depends on the specific application scenario and design constraints of the technical solution.

[0241] 13 and 14 are diagrams of possible communication device structures according to the present disclosure. These communication devices may be configured to implement the functions of an access network device, a module of the access network device (e.g., a DU, a RU, and / or a near-real-time RIC), a terminal device, an AI function network element, etc. in the above-mentioned method. Therefore, these communication devices may also implement the beneficial effects of the above-mentioned method.

[0242] 13, a communication device 1300 includes a processing unit 1310 and a communication unit 1320. The communication device 1300 is configured to implement the methods presented above.

[0243] When the communication device 1300 is configured to implement the method performed by a reference signal receiver, the processing unit 1310 maps the reference signal measurements to input adaptation information by using an input adaptation layer, where the beam pattern corresponding to the reference signal measurements is a first beam pattern. The processing unit 1310 obtains a first beam prediction result by using a beam prediction model, where the input of the beam prediction model includes the adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. A reference signal is received by the communication unit 1320.

[0244] Alternatively, when the communication device 1300 is configured to implement the method performed by a reference signal receiver, the processing unit 1310 obtains a first beam prediction result by using a beam prediction model, where an input to the beam prediction model includes a reference signal measurement, and a beam pattern corresponding to the reference signal measurement is a second beam pattern. The processing unit 1310 maps the first beam prediction result to a second beam prediction result by using an output adaptation layer. A reference signal is received by the communication unit 1320.

[0245] Alternatively, when the communication device 1300 is configured to implement a method performed by a reference signal receiver, the processing unit 1310 maps the reference signal measurements to input adaptation information by using an input adaptation layer, where the beam pattern corresponding to the reference signal measurements is a first beam pattern. The processing unit 1310 obtains a first beam prediction result by using a beam prediction model, where the input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. The processing unit 1310 maps the first beam prediction result to a second beam prediction result by using an output adaptation layer. A reference signal is received by the communication unit 1320.

[0246] When the communication device 1300 is configured to implement the method performed by the reference signal transmitter, the communication unit 1320 is configured to send information related to an input adaptation layer, where the input adaptation layer is configured to adapt reference signal measurements to obtain inputs for a beam prediction model, a beam pattern corresponding to the reference signal measurements is a first beam pattern, the inputs for the beam prediction model match a second beam pattern, and the first beam pattern is different from the second beam pattern. The information related to the input adaptation layer is determined by the processing unit 1310.

[0247] Alternatively, when the communication device 1300 is configured to implement the method performed by the reference signal transmitter, the communication unit 1320 is configured to send information related to an output adaptation layer, where the output adaptation layer is configured to map a first beam prediction result output by the beam prediction model to a second beam prediction result, where the first beam prediction result is different from the second beam prediction result. The information related to the output adaptation layer is determined by the processing unit 1310.

[0248] Alternatively, when the communication device 1300 is configured to implement a method performed by a reference signal transmitter, the communication unit 1320 is configured to send information related to the input adaptation layer and information related to the output adaptation layer. The input adaptation layer is configured to adapt reference signal measurements to obtain inputs for a beam prediction model, where a beam pattern corresponding to the reference signal measurements is a first beam pattern, and the inputs for the beam prediction model match a second beam pattern, where the first beam pattern is different from the second beam pattern. The output adaptation layer is configured to map a first beam prediction result output by the beam prediction model to a second beam prediction result, where the first beam prediction result is different from the second beam prediction result. The information related to the input adaptation layer and the information related to the output adaptation layer are determined by the processing unit 1310.

[0249] For a more detailed functional description of the processing unit 1310 and the communication unit 1320, please refer to the relevant descriptions in the above method, and the details will not be described again here.

[0250] 14, a communications device 1400 includes a processor 1410 and an interface circuit 1420 configured to implement the method performed by the transmitter or receiver described above. The processor 1410 and the interface circuit 1420 are coupled to each other. It may be understood that the interface circuit 1420 may be a transceiver, a pin, an input / output interface, or another communications interface. Optionally, the communications device 1400 may further include a memory 1430 configured to store at least one of instructions executed by the processor 1410, input data required by the processor 1410 to execute the instructions, or data generated after the processor 1410 executes the instructions.

[0251] Optionally, the instructions executed by processor 1410 may be stored in processor 1410, stored in memory 1430, and / or downloaded by processor 1410 from a third-party website. The method for obtaining the instructions is not limited in this disclosure. The instructions in memory 1430 may be pre-stored, loaded later, or downloaded by processor 1410 from a third-party website and then stored in memory 1430.

[0252] When the communication device 1400 is configured to implement the above method, the processor 1410 is configured to implement the functions of the processing unit 1310 described above, and the interface circuit 1420 is configured to implement the functions of the communication unit 1320 described above.

[0253] When the communication device is a chip used in a terminal device, the chip in the terminal device implements the functions of the terminal device in the above method embodiments. The chip in the terminal device receives information from another module (e.g., a radio frequency module or an antenna) in the terminal device, and the information is sent to the terminal device by an access network device or the like. Alternatively, the chip in the terminal device sends information to another module (e.g., a radio frequency module or an antenna) in the terminal device, and the information is sent by the terminal device to an access network device or the like.

[0254] When the above communication apparatus is a module used in an access network device, the module in the access network device implements the functions of the access network device in the above method embodiments. The module in the access network device receives information from another module (e.g., a radio frequency module or an antenna) in the access network device, and the information is sent to the access network device by a terminal device or the like. Alternatively, the module in the access network device sends information to another module (e.g., a radio frequency module or an antenna) in the access network device, and the information is sent to a terminal device or the like by the access network device. The module in the access network device herein may be a baseband chip of the access network device, or may be a near-real-time RIC, CU, DU, or another module. The near-real-time RIC, CU, and DU in this specification may be the near-real-time RIC, CU, and DU in an O-RAN architecture.

[0255] In the present disclosure, a processor may include one or more processors to act as a combination of computing devices. The processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array, another programmable logic device, a discrete gate, a transistor logic device, a discrete hardware component, or the like, configured to implement or execute the methods, steps, and logic block diagrams of the present disclosure. The general-purpose processor may be a microprocessor, or any conventional processor, or the like. The steps of the methods disclosed in connection with the present disclosure may be performed and completed directly by a hardware processor, or may be performed and completed using a combination of hardware and software modules in the processor.

[0256] In the present disclosure, an interface circuit may include any suitable hardware or software for enabling communication with one or more computing devices (e.g., network elements in the present disclosure). For example, in some embodiments, the interface circuit may include wires for coupling a wired connection or terminals and / or pins of a wireless transceiver for coupling a wireless connection. In some embodiments, the interface circuit may include a transmitter, a receiver, a transceiver, and / or an antenna. The interface may be configured to enable communication between computing devices (e.g., network elements in the present disclosure) by using any available protocol (e.g., 3rd generation partnership project (3GPP) standard protocol).

[0257] In this disclosure, memory may be implemented by using any suitable storage technology. For example, memory may be any available storage medium accessible by a processor and / or computer. Non-limiting examples of storage media are random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM), removable media, compact disc memory, magnetic storage media, magnetic storage devices, flash memory, registers, state memory, remotely located memory, local or remote memory components, or any other medium capable of carrying or storing software, data, or information and accessible by a processor / computer.

[0258] The memory and processor in the present disclosure may be disposed separately or integrated together. The processor may read information from, store information in, and / or write information to the memory. The memory may be integrated into the processor. The processor and memory may be disposed in an integrated circuit (e.g., an application-specific integrated circuit (ASIC)). The integrated circuit may be disposed in a network element in the present disclosure or another network node.

[0259] Instructions in this disclosure may be referred to as programs, which are broadly defined as software. Software may be program code, a program, a subprogram, an instruction set, code, a code segment, a software module, an application program, a software application program, etc. The program may be run on a processor and / or computer to perform various functions and / or processes described in this disclosure.

[0260] All or part of the methods in this disclosure may be implemented by software, hardware, firmware, or any combination thereof. When software is used to implement the methods, all or part of the methods may be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, all or part of the procedures or functions according to the present application are performed. The computer may be a general-purpose computer, a special-purpose computer, a computer network, an access network device, a terminal device, a core network device, an AI function network element, or another programmable device. The computer program or instructions may be stored on a computer-readable storage medium or transmitted from a computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions may be transmitted from a website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless method. The computer-readable storage medium may be any available medium or data storage device that can be accessed by a computer, such as a server or data center that integrates one or more available media. The available media may be magnetic media such as floppy disks, hard disk drives, or magnetic tape; optical media such as digital video disks; or semiconductor media such as solid-state drives. The computer-readable storage media may be volatile or non-volatile storage media or may include both types of storage media: volatile and non-volatile storage media.

[0261] The above description is only a specific example of the present invention and does not limit the protection scope of the present invention. Any variations or replacements that can be easily thought of by those skilled in the art within the technical scope disclosed in the present invention shall fall within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. mapping reference signal measurements to input adaptation information by using an input adaptation layer, wherein a beam pattern corresponding to the reference signal measurements is a first beam pattern; obtaining a first beam prediction result by using a beam prediction model, wherein an input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern; A beam management method comprising:

2. obtaining a first beam prediction result by using a beam prediction model, wherein an input of the beam prediction model includes a reference signal measurement, and a beam pattern corresponding to the reference signal measurement is a second beam pattern; mapping the first beam prediction result to a second beam prediction result by using an output adaptation layer; A beam management method comprising:

3. mapping reference signal measurements to input adaptation information by using an input adaptation layer, wherein a beam pattern corresponding to the reference signal measurements is a first beam pattern; obtaining a first beam prediction result by using a beam prediction model, wherein an input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern; mapping the first beam prediction result to a second beam prediction result by using an output adaptation layer; A beam management method comprising:

4. The method of any one of claims 1 to 3, wherein the first beam prediction result includes the top K1 beams in a full beam corresponding to the first beam pattern or the second beam pattern, where K1 is a positive integer.

5. The method of any one of claims 2 to 4, wherein the second beam prediction result includes the top K2 beams in the full beam corresponding to the first beam pattern or the second beam pattern, where K2 is a positive integer.

6. The method of any one of claims 1 to 5, wherein the beam prediction model is included in a candidate beam prediction model set, each beam prediction model in the candidate beam prediction model set corresponds to one beam pattern, and the beam prediction model corresponds to the second beam pattern.

7. The method of claim 6 , wherein information indicative of the second beam pattern is received.

8. The method of claim 1 , wherein information about the input adaptation layer or information about the output adaptation layer is received.

9. The method of claim 1 , wherein the input adaptation layer or the output adaptation layer is obtained through training.

10. The method comprises: obtaining an ideal beam prediction result based on a measurement corresponding to a full beam; mapping a measurement corresponding to a sparse beam of the first beam pattern to the input adaptation information by using the input adaptation layer; obtaining an actual beam prediction result based on the input adaptation information and the beam prediction model; and adjusting parameters of the input adaptation layer based on the ideal beam prediction result and the actual beam prediction result to enable a difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold; or obtaining an ideal beam prediction result based on measurements corresponding to a full beam, obtaining an actual beam prediction result based on measurements of a sparse beam of the second beam pattern, the beam prediction model, and the output adaptation layer, and adjusting parameters of the output adaptation layer based on the ideal beam prediction result and the actual beam prediction result to enable a difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold; or obtaining an ideal beam prediction result based on a measurement corresponding to a full beam; mapping a measurement corresponding to a sparse beam of the first beam pattern to the input adaptation information by using the input adaptation layer; obtaining an actual beam prediction result based on the input adaptation information, the beam prediction model, and the output adaptation layer; and adjusting at least one of parameters of the input adaptation layer and parameters of the output adaptation layer based on the ideal beam prediction result and the actual beam prediction result to enable a difference between the ideal beam prediction result and the actual beam prediction result to be smaller than a threshold.

10. The method of claim 9, comprising:

11. sending information to an input adaptation layer, the input adaptation layer configured to perform adaptation based on reference signal measurements to obtain inputs for a beam prediction model, a beam pattern corresponding to the reference signal measurements being a first beam pattern, the inputs for the beam prediction model matching a second beam pattern, and the first beam pattern being different from the second beam pattern; A beam management method comprising:

12. sending information about an output adaptation layer, the output adaptation layer being configured to map a first beam prediction result output by a beam prediction model to a second beam prediction result, the first beam prediction result being different from the second beam prediction result; A beam management method comprising:

13. sending information about an input adaptation layer and information about an output adaptation layer, wherein the input adaptation layer is configured to perform adaptation based on reference signal measurements to obtain an input for a beam prediction model, a beam pattern corresponding to the reference signal measurements is a first beam pattern, the input for the beam prediction model matches a second beam pattern, the first beam pattern is different from the second beam pattern, and the output adaptation layer is configured to map a first beam prediction result output by the beam prediction model to a second beam prediction result, the first beam prediction result is different from the second beam prediction result; A beam management method comprising:

14. indicating the beam prediction model from a candidate beam prediction model set, each beam prediction model in the candidate beam prediction model set corresponding to one beam pattern, and the beam prediction model corresponding to the second beam pattern; The method of any one of claims 11 to 13, further comprising:

15. The method of claim 14, wherein information regarding each beam prediction model in the candidate beam prediction model set is agreed upon in a protocol, or the method includes a step of sending information regarding each beam prediction model in the candidate beam prediction model set.

16. The method of claim 14, wherein the correspondence between the beam prediction models in the candidate beam prediction model set and the beam patterns is agreed upon in a protocol, or the method includes a step of sending the correspondence between the beam prediction models in the candidate beam prediction model set and the beam patterns.

17. transmitting information indicating the second beam pattern; The method of claim 14 further comprising:

18. A communication device configured to implement the method of any one of claims 1 to 10.

19. A communication device comprising a processing circuit and a communication circuit, said processing circuit configured to perform a method according to any one of claims 1 to 10.

20. A communication device configured to implement the method of any one of claims 11 to 17.

21. A communications device comprising a processing circuit and a communications circuit, said processing circuit configured to perform a method according to any one of claims 11 to 17.

22. A communication system comprising a communication device according to claim 18 or 19 and a communication device according to claim 20 or 21.

23. 18. A computer-readable storage medium configured to store instructions that, when run on a computer, enable the computer to perform a method according to any one of claims 1 to 17.

24. 18. A computer program product comprising instructions, which when run on a computer, enable the computer to carry out a method according to any one of claims 1 to 17.

Images