Artificial intelligence communication method and related apparatus

Abstract

The application provides an artificial intelligence communication method and related device for managing the combination of AI models associated with each other in a communication system, avoiding the management of AI models in the communication system separately, which makes the communication quality of the communication system fall into a local optimum, thereby causing the communication quality of the communication system to deteriorate. The method of the application runs a demodulation proxy model and a channel decoding proxy model in a network device, performs CRC on the decoding information output by the channel decoding proxy model, and determines that the performance of the demodulation proxy model and the performance of the channel decoding proxy model meet the standard when the decoding information passes the CRC. Compared with the conventional performance evaluation scheme of a single model, the scheme provided by the application performs performance evaluation on at least two proxy models in linkage, evaluates whether the multiple models in the AI communication system can work normally from the overall point of view, and avoids falling into a local optimum.

Description

Technical Field

[0001] This application relates to the field of wireless communication, and more particularly to an artificial intelligence communication method and related apparatus. Background Technology

[0002] Artificial intelligence (AI) is an important branch of computer science and a key area of ​​current development in the field. With the continuous advancement of AI technology, it is being applied to an increasing number of scenarios, including wireless communication systems. Wireless communication systems are increasingly using AI modules to replace functional modules, enabling real-time optimization of wireless communication performance.

[0003] Currently, when AI technology is applied to wireless communication scenarios, it is mostly used in single-module air interface AI applications, such as CSI feedback, positioning, and beam management. The AI ​​models are managed and operate independently, decoupled from each other. However, in a communication link, there may be interconnected modules. When setting up AI models within these interconnected modules, the challenge lies in managing and optimizing these interconnected AI model groups during their use. Summary of the Invention

[0004] This application provides an artificial intelligence communication method and related apparatus for managing the combination of interrelated AI models in a communication system, avoiding the communication quality of the communication system from falling into local optima due to the separate management of AI models in the communication system, which would lead to the degradation of the communication quality of the communication system.

[0005] In a first aspect, this application provides an AI communication method, which is executed by a network device, or by a component applied to the network device (e.g., a processor, chip, or chip system), or by a logic module or software capable of implementing all or part of the functions of the network device. In the first aspect and its possible implementations, the method is described using the execution by a network device as an example. The network device receives first channel information, which describes the channel characteristics of a target channel; the network device processes first information based on the first channel information to obtain second information, where the first information is information transmitted through the target channel; the second information and the first channel information are also input into a demodulation proxy model (PM) to obtain demodulated information; the demodulated information is also input into a channel decoding PM to obtain decoded information, which carries a first checksum; cyclic redundancy check (CRC) is performed on the decoded information based on the first checksum, and if the decoded information passes the CRC check, it is determined that the performance of the demodulation PM and the performance of the channel decoding PM are both up to standard.

[0006] In this embodiment, by running at least two Model Managers (PMs) in the network device, the performance of at least two models used in the terminal device is determined by whether the performance of the at least two PMs meets the standards. There is a one-to-one correspondence between the models in the at least two PMs and the models in the at least two models. For example, when the at least two PMs are a demodulation PM and a channel decoding PM, the decoding information output by the channel decoding PM is subjected to a CRC check. If the decoding information passes the CRC check, it is determined that the performance of both the demodulation PM and the channel decoding PM meets the standards. Compared to the conventional scheme of evaluating the performance of a single model, the scheme provided in this application evaluates the performance of at least two PMs in a coordinated manner, assessing the overall performance of multiple models in the AI ​​communication system to ensure they function correctly, thus avoiding getting trapped in local optima.

[0007] In one possible implementation of the first aspect, the method further includes:

[0008] If the decoded information fails the cyclic redundancy check, the performance of the demodulation PM or the channel decoding PM is confirmed to be substandard based on the performance curve of the network device.

[0009] In this embodiment of the application, when the decoded information fails the CRC test, it is proposed that the model for determining the substandard performance based on the bit error rate be demodulation PM or channel decoding PM. This helps to efficiently and accurately locate the fault point, shorten the time required for the abnormal recovery of the AI ​​communication system, and provide a guarantee for the reliable operation of the AI ​​communication system.

[0010] In one possible implementation of the first aspect, the first channel information includes a first SINR;

[0011] Based on the first channel information, the first information is processed to obtain the second information, including:

[0012] Based on the first SINR, a target channel is simulated, and a first simulated channel is obtained.

[0013] The first information is simulated and transmitted in a first simulated channel to obtain the second information.

[0014] In this embodiment, the network device simulates the transmission of the first information in the target channel to obtain the second information based on the first SINR fed back by the terminal device. Based on the simulated second information, the demodulation PM and channel decoding PM are verified. This can more effectively simulate the real operation of the demodulation model and channel decoding model in the terminal device, thereby improving the reliability of the network device in managing the AI ​​model.

[0015] In one possible implementation of the first aspect, the first channel information further includes channel information of the target channel, and the modulation model is a model trained based on the channel information of the target channel.

[0016] Before receiving the first channel information, the method also includes:

[0017] The encoded information is input into the modulation model to obtain the modulation information. The encoded information is the information obtained by channel coding the source information. The source information is the information generated by the network device and sent to the terminal device.

[0018] A first demodulation reference signal is added to the modulation information to obtain the first information.

[0019] In this embodiment, when the modulation is a non-standard constellation point, the modulation and demodulation are related to the target channel. The channel information of the target channel is introduced during the training phase of the modulation model and demodulation PM, making the modulation model, demodulation model and demodulation PM more in line with specific usage requirements, providing a more practical and effective solution, and ensuring that the solution provided by this application can be applied to more different multi-AI model communication systems.

[0020] In one possible implementation of the first aspect, the first channel information includes channel information of the target channel;

[0021] The method also includes:

[0022] The second information is input into the channel estimation PM to obtain the channel estimation information;

[0023] Based on the channel estimation information and the channel information of the target channel, we analyze whether the performance of the channel estimation PM meets the standards.

[0024] In this embodiment, the performance of the channel estimation PM is analyzed by analyzing the similarity between the channel estimation model and the first channel information (i.e., the channel information of the target channel) output by the channel estimation PM and the channel estimation information. This can help network devices to comprehensively and reliably manage the use of AI communication models by terminal devices, effectively improving the reliability and stability of the solution.

[0025] In one possible implementation of the first aspect, processing the first information based on the first channel information to obtain the second information includes:

[0026] The target channel is simulated based on the channel information of the target channel to obtain a second simulated channel;

[0027] The first information is simulated and transmitted in the second simulated channel to obtain the second information.

[0028] In one possible implementation of the first aspect, the second information includes first information transmitted via a second analog channel and subjected to resource demapping;

[0029] Before receiving the first channel message, the method also includes:

[0030] The modulation information and the first demodulation reference signal are pre-coded to obtain the third information. The modulation information is the information obtained by channel coding and modulation of the source information. The source information is the information generated by the network device and sent to the terminal device.

[0031] Perform resource mapping on the third message to obtain the first message;

[0032] The first information is transmitted through the target channel.

[0033] In this embodiment of the application, the modulation information and the first demodulation reference signal are pre-coded and resource-mapped to obtain the first information. When the first information is transmitted through the target channel, the influence of channel correlation on the transmitted signal can be effectively reduced by preprocessing the transmitted signal, for example, reducing the bit error rate and interference of the information.

[0034] Secondly, this application provides an AI communication method, including:

[0035] Receive the second information, which is obtained by transmitting the first information through the target channel;

[0036] Based on the analysis of the second information, the channel characteristics of the target channel are analyzed to obtain the first channel information;

[0037] Send first channel information, which is used to describe the channel characteristics of the target channel.

[0038] In one possible implementation of the second aspect, the method further includes:

[0039] The second information and the first channel information are input into the demodulation model to obtain the demodulation information;

[0040] The demodulated information is input into the channel decoding model to obtain the decoded information.

[0041] In one possible implementation of the second aspect, the first channel information includes a first SINR.

[0042] In one possible implementation of the second aspect, the first channel information further includes channel information of the target channel.

[0043] In one possible implementation of the second aspect, obtaining the first channel information by analyzing the channel characteristics of the target channel based on the second information includes:

[0044] The second information is input into the channel estimation model to obtain channel estimation information.

[0045] In one possible implementation of the second aspect, the second information is information that has been pre-encoded and resource-mapped;

[0046] Based on the analysis of the target channel's channel characteristics using the second information, the first channel information obtained includes:

[0047] Perform resource demapping on the second information;

[0048] Based on the second information analysis of resource access mapping, the channel characteristics of the target channel are obtained to acquire the first channel information.

[0049] The beneficial effects shown in this aspect are similar to those of the first aspect or any possible implementation of the first aspect, and will not be repeated here.

[0050] Thirdly, this application provides an AI communication device, comprising:

[0051] A receiving unit is used to receive first channel information, which is used to describe the channel characteristics of the target channel.

[0052] The processing unit is used to process the first information based on the first channel information to obtain the second information, wherein the first information is information transmitted through the target channel;

[0053] The processing unit is also used to input the second information and the first channel information into the demodulation proxy model PM to obtain demodulation information;

[0054] The processing unit is also used to input demodulated information into the channel decoder PM to obtain decoded information, which carries a first check code;

[0055] The verification unit is used to perform cyclic redundancy check on the decoded information based on the first check code. If the decoded information passes the cyclic redundancy check, it is determined that the performance of the demodulation PM meets the standard and the performance of the channel decoding PM meets the standard.

[0056] In one possible implementation of the third aspect, the verification unit is further configured to confirm, based on the performance curve of the network device, that the performance of the demodulation PM or the performance of the channel decoding proxy model is substandard if the decoded information fails the cyclic redundancy check.

[0057] In one possible implementation of the third aspect, the first channel information includes a first signal-to-interference-plus-noise ratio (SINR).

[0058] The processing unit is specifically used for:

[0059] Based on the first SINR, a target channel is simulated, and a first simulated channel is obtained.

[0060] The first information is simulated and transmitted in a first simulated channel to obtain the second information.

[0061] In one possible implementation of the third aspect, the first channel information further includes channel information of the target channel, and the modulation model is a model trained based on the channel information of the target channel.

[0062] The processing unit is also used for:

[0063] The encoded information is input into the modulation model to obtain the modulation information. The encoded information is the information obtained by channel coding the source information. The source information is the information generated by the network device and sent to the terminal device.

[0064] A first demodulation reference signal is added to the modulation information to obtain the first information.

[0065] In one possible implementation of the third aspect, the first channel information includes channel information of the target channel;

[0066] The processing unit is also used for:

[0067] The second information is input into the channel estimation PM to obtain the channel estimation information;

[0068] Based on the channel estimation information and the channel information of the target channel, we analyze whether the performance of the channel estimation PM meets the standards.

[0069] In one possible implementation of the third aspect, the processing unit is specifically used for:

[0070] The target channel is simulated based on the channel information of the target channel to obtain a second simulated channel;

[0071] The first information is simulated and transmitted in the second simulated channel to obtain the second information.

[0072] In one possible implementation of the third aspect, the second information includes first information transmitted via a second analog channel and demapped for resources;

[0073] The device also includes:

[0074] The precoding unit is used to precode the modulation information and the first demodulation reference signal to obtain the third information. The modulation information is the information obtained by channel coding and modulation of the source information. The source information is the information generated by the network device and sent to the terminal device.

[0075] The resource mapping unit is used to perform resource mapping on the third message to obtain the first information;

[0076] The transmitting unit is used to transmit the first information through the target channel.

[0077] Fourthly, this application provides an AI communication device, comprising:

[0078] A receiving unit is used to receive second information, which is obtained by transmitting first information through a target channel.

[0079] The analysis unit is used to analyze the channel characteristics of the target channel based on the second information to obtain the first channel information;

[0080] The transmitting unit is used to transmit first channel information, which describes the channel characteristics of the target channel.

[0081] In one possible implementation of the fourth aspect, the analysis unit is further configured to:

[0082] The second information and the first channel information are input into the demodulation model to obtain the demodulation information;

[0083] The demodulated information is input into the channel decoding model to obtain the decoded information.

[0084] In one possible implementation of the fourth aspect, the first channel information includes a first SINR.

[0085] In one possible implementation of the fourth aspect, the first channel information further includes channel information of the target channel.

[0086] In one possible implementation of the fourth aspect, the analysis unit is specifically used to input the second information into the channel estimation model to obtain channel estimation information.

[0087] In one possible implementation of the fourth aspect, the second information is information that has been pre-encoded and resource-mapped;

[0088] Analysis unit, specifically used for:

[0089] Perform resource demapping on the second information;

[0090] Based on the second information analysis of resource access mapping, the channel characteristics of the target channel are obtained to acquire the first channel information.

[0091] Fifthly, this application provides an AI communication system, including network equipment and terminal equipment;

[0092] A network device for implementing the method of the first aspect or any possible implementation thereof;

[0093] A terminal device for implementing the method in the second aspect or any possible implementation of the second aspect.

[0094] Sixthly, this application provides an AI communication device, including a processor coupled to a memory;

[0095] Instructions are stored in the memory, which, when executed on the processor, cause the communication device to implement the methods related to the first aspect, possible implementations of the first aspect, the second aspect, or any implementation of the second aspect.

[0096] The beneficial effects shown in this aspect are similar to those of the aforementioned first aspect, any possible implementation of the first aspect, the second aspect, or any possible implementation of the second aspect, and will not be repeated here.

[0097] Sixthly, this application provides a computer-readable storage medium storing instructions that, when executed on a processor, implement the methods shown in the first aspect, any possible implementation of the first aspect, the second aspect, or any possible implementation of the second aspect.

[0098] In a seventh aspect, this application provides a computer program product that, when executed on a processor, implements the method shown in the first aspect, any possible implementation of the first aspect, the second aspect, or any possible implementation of the second aspect.

[0099] The beneficial effects shown in any of the fifth to seventh aspects are similar to those of the first aspect, any possible implementation of the first aspect, the second aspect, or any possible implementation of the second aspect, and will not be repeated here. Attached Figure Description

[0100] Figure 1 A schematic diagram of an architecture of the wireless network system provided in this application;

[0101] Figure 2 A schematic diagram of the AI ​​communication system provided in this application;

[0102] Figure 3 A flowchart illustrating the AI ​​communication method provided in this application;

[0103] Figure 4 A schematic diagram of the performance curves provided in this application;

[0104] Figure 5 Another schematic diagram of the AI ​​communication system provided in this application;

[0105] Figure 6 Another flowchart illustrating the AI ​​communication method provided in this application;

[0106] Figure 7 Another schematic diagram of the AI ​​communication system provided in this application;

[0107] Figure 8 Another flowchart illustrating the AI ​​communication method provided in this application;

[0108] Figure 9 A schematic diagram of the AI ​​communication device provided in this application;

[0109] Figure 10 Another structural schematic diagram of the AI ​​communication device provided in this application;

[0110] Figure 11 A schematic diagram of the structure of the AI ​​communication system provided in this application;

[0111] Figure 12 Another structural schematic diagram of the AI ​​communication device provided in this application. Detailed Implementation

[0112] First, some terms used in the embodiments of this application will be explained to facilitate understanding by those skilled in the art.

[0113] (1) Demodulation reference signal (DMRS) is used for correlation demodulation of the physical uplink share channel (PUSCH) and physical uplink control channel (PUCCH) in Long Term Evolution (LTE).

[0114] (2) Channel state information (CSI) is the channel attribute of a communication link. It describes the fading factor of the signal on each transmission path, that is, the value of each element in the channel gain matrix H, such as signal scattering, environmental fading (multipath fading or shadowing fading), and power decay of distance. CSI enables the communication system to adapt to the current channel conditions, providing a guarantee for high-reliability and high-speed communication in multi-antenna systems.

[0115] (3) Signal to interference plus noise ratio (SINR): This refers to the ratio of the strength of the received useful signal to the strength of the received interference signal. It serves as a terminal device or channel quality indicator to feed back channel characteristics to network equipment, thereby adjusting the data rate of the transmitting antenna and achieving adaptive modulation.

[0116] (4) The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion so that a process, method, system, product, or device that comprises a series of units is not necessarily limited to those units, but may include other units not explicitly listed or inherent to those processes, methods, products, or devices. Additionally, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can be expressed as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0117] Artificial intelligence (AI) is an important branch of computer science and a key area of ​​current development in the field. With the continuous advancement of AI technology, it is being applied to an increasing number of scenarios, including wireless communication systems. Wireless communication systems are increasingly using AI modules to replace functional modules, thereby enabling real-time optimization of wireless communication performance.

[0118] In wireless communication, most applications of AI technology are single-module air interface AI use cases. For example, CSI feedback enhancement based on AI models can achieve spatial-frequency domain CSI compression through a bilateral AI model. A CSI generation model is deployed on the user equipment, and a corresponding CSI reconstruction model is deployed on the base station. These two models work together to complete the CSI compression, feedback, and reconstruction tasks, so that even when the CSI is significantly compressed on the user equipment, the base station can still obtain more accurate CSI data, thereby optimizing resource scheduling.

[0119] The application of single-module air interface AI use cases in the air interface also needs to ensure that the AI ​​model is trustworthy and controllable. Therefore, the model needs to be managed in the air interface to ensure that the AI ​​model is used when its performance meets the standards.

[0120] In actual pipelines, there are interconnected modules. When AI models are set up in these interconnected modules at the same time, how to manage this group of interconnected AI models during the use of the models, so as to ensure that they are used only when their performance meets the standards, becomes an urgent problem to be solved.

[0121] To address this issue, this application, when using a demodulation model and a channel decoding model in conjunction with the terminal device, maintains a model similar to the terminal device's demodulation model on the network device side. This model is called the demodulation PM, and the network device also maintains a model similar to the terminal device's channel decoding model, called the channel decoding PM. The network device receives first channel information, which describes the channel characteristics of the target channel, and processes the first information based on the first channel information to obtain second information (i.e., simulating the transmission of the first information in the target channel). This second information is then input into the demodulation PM to obtain demodulated information, and finally input into the channel decoding PM to obtain decoded information. Since the demodulation PM and the channel decoding PM are similar to the demodulation model maintained and used by the terminal device, the performance of both the demodulation PM and the channel decoding PM is simultaneously satisfied by verifying whether the decoded information passes the cyclic redundancy check (CRC). This allows the network device to determine whether the performance of the demodulation model and the channel decoding model maintained and used in the terminal device meets the required standards. This invention enables the management of multiple models used in terminal devices within network devices, and this management is coordinated. In AI communication systems using multiple AI models, the solution provided in this application evaluates the performance of multiple AI models from a global perspective, avoiding the pitfalls that might result from evaluating the performance of each AI model separately.

[0122] To facilitate understanding, let's first combine... Figure 1 The architecture of the wireless network system will be introduced.

[0123] like Figure 1 As shown, the wireless network system includes a source, a source coding module, a channel coding module, a modulation module, a communication channel, a demodulation module, a channel decoding module, a source decoding module, and a sink.

[0124] A source is used to generate information, which can be discrete or continuous data. For example, a source can be a network device.

[0125] The source coding module converts the information generated by the source into a digital sequence in a unified format. For example, continuous information requires analog-to-digital conversion, and then digital pulse combinations are used to identify the signal amplitude to ensure information validity. Discrete information is directly represented by digital pulse combinations. In this way, the information generated by the source is converted into symbols suitable for transmission in the channel.

[0126] The channel coding module adds redundant information to the encoded source information, enabling the receiver to perform forward error correction based on this redundancy. This helps counteract noise and attenuation in the channel and improves the receiver's error correction capabilities. Common channel coding methods include current block codes, convolutional codes, concatenated codes, turbo codes, and LDPC codes.

[0127] The modulation module is used to transform the spectrum of a signal to a higher frequency range.

[0128] The transmitting module is used to transmit wireless signals to the outside world to enable communication between devices.

[0129] A communication channel, as a medium for transmitting digital signals, is used to transmit signals from a source to a destination.

[0130] The receiving module is used to receive externally transmitted wireless signals to enable communication between devices.

[0131] The channel decoding module and the source decoding module correspond to the channel coding module and the source coding module, respectively, and are used to restore the encoded data.

[0132] A receiver is used to receive information and perform subsequent processing based on the received information. For example, the receiver can be a terminal device.

[0133] Based on the above system, the solution provided in this application will be described below with reference to the accompanying drawings.

[0134] Please see Figure 2 , Figure 2 A schematic diagram of the AI ​​communication system provided in this application.

[0135] Compared to the above Figure 1 The wireless network system shown in this application provides an AI communication system in which a simulated channel module is added to the base station (used to simulate the target channel in the method of this application), and a channel estimation module is added to the terminal device (used to estimate the target channel in the method of this application).

[0136] After receiving the second information sent by the network device, the terminal device uses the channel estimation module to estimate the channel characteristics of the target channel, obtains the first channel information, and reports the first channel information to the network device.

[0137] After receiving the first channel information reported by the terminal device, the network device simulates the loss of the first information in the target channel based on the first channel information to obtain the second information. The second information is then input into the demodulation phase (PM) to obtain demodulated information, and the demodulated information is input into the channel decoding phase (PM) to obtain decoded information. Finally, the performance of the demodulation and channel decoding phases is analyzed based on the decoded information to determine if they meet the required standards, thereby managing these two models. For example, if the performance of the demodulation or channel decoding phase is substandard, an alarm signal is sent to the terminal device, suspending the use of its corresponding demodulation or channel decoding model.

[0138] It should be understood that, due to the target channel (i.e., the aforementioned) Figure 1 Noise or attenuation may exist in the communication channel (in the target channel), which may affect the information transmitted in the target channel. Therefore, after the network device sends the first information through the target channel, the information received by the terminal device may be different from the first information. In order to distinguish the possible differences, the information received by the terminal device is referred to as the second information in this application.

[0139] In combination with the above Figure 2 The AI ​​communication system, combined with the following Figure 3 The methods that may be implemented in this scenario are described. Those skilled in the art will understand that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0140] S301. The network device receives the first channel information sent by the terminal device, that is, the terminal device sends the first channel information to the network device.

[0141] The first channel information is used to describe the channel characteristics of the target channel.

[0142] In one possible implementation, the first channel information is contained in a message sent by the terminal device to the network device via a radio resource control (RRC) message or downlink control information (DCI), the specific form and content of which are not limited here.

[0143] In one possible implementation, the first channel information may include a first SINR, wherein the first SINR is the SINR of the target channel. It should be understood that in some common scenarios, the signal-to-noise ratio (SNR) can be used instead of SINR, and this is not a limitation here.

[0144] In one possible implementation, the description of the channel characteristics of the target channel in the first channel information can be a description at the resource element (RE) granularity, a description at the resource block (RB) granularity, or a description at the subband granularity in the frequency domain. Moreover, the description of the channel characteristics of the target channel in the first channel information can be reported in the time domain on a unit of multiple frames, for example, analyzed and described in units of 100 frames, and then reported.

[0145] In this embodiment of the application, by adjusting the granularity and angle of the description of the channel characteristics of the target channel in the first channel information, it can adapt to more diverse application scenarios. For example, the target channel in the first channel information is described at the granularity of RB. Since 1 RB can be equivalent to 12 RE, describing the target channel at the granularity of RB can effectively reduce the length of the first channel information and reduce the reporting overhead of the first channel information compared to describing the target channel at the granularity of RE.

[0146] In one possible implementation, the first channel information may also include RE-related channel features on a portion of the transmission bandwidth occupied by the physical downlink shared channel (PUSCH) on the RB.

[0147] In this embodiment of the application, the first channel information can also be set as the channel features associated with REs on a portion of the bandwidth occupied by PUSCH. This can further reduce the length of the first channel information. By only reporting the channel features associated with REs that may be used, the length of the first channel information can be further reduced, which can also effectively reduce the reporting overhead of the first channel information.

[0148] S302. The network device processes the first information based on the first channel information to obtain the second information;

[0149] The first information is information sent through the target channel. In this embodiment, the first information is information sent by the network device to the terminal device through the target channel.

[0150] In one possible implementation, the first information may include a first demodulation reference signal (DMRS). When the first information includes the first DMRS, the first channel information is determined by the channel estimation module based on the content of the first DMRS field in the second information.

[0151] In one possible implementation, when the first channel information includes a first SINR, the analog channel module in the network device simulates noise and interference information in the target channel based on the first SINR to determine a first analog channel (the first analog channel is similar to the target channel, and therefore can also be understood as the channel obtained by simulating the target channel based on the first SINR). Furthermore, it simulates the transmission of the first information in the first analog channel to obtain second information; here, the information obtained by simulating the transmission of the first information in the first analog channel is also referred to as the second information, without considering any possible simulation errors.

[0152] In this embodiment of the application, a first SINR is proposed to represent the characteristics of the target channel, which can simply and efficiently complete the channel feature extraction and transmission of the target channel, and provide a certain guarantee for the efficiency and accuracy of the scheme implementation.

[0153] In this embodiment, the network device simulates the transmission of the first information in the target channel to obtain the second information based on the first SINR fed back by the terminal device. Based on the simulated second information, the demodulation PM and channel decoding PM are verified. This can more effectively simulate the real operation of the demodulation model and channel decoding model in the terminal device, thereby improving the reliability of the network device in managing the AI ​​model.

[0154] It should be understood that, due to the target channel (i.e., the aforementioned) Figure 1 The communication channel may contain noise and other interference factors. Therefore, after the network device sends the first information through the target channel, the first information received by the terminal device is called the second information to distinguish the possible differences between the two information. In this application, it is assumed that the first simulated channel can fully simulate the target channel. That is, the information obtained by inputting the first information into the first simulated channel and the target simulated channel respectively is the second information. No limitation is made here.

[0155] S303. The network device inputs the second information and the first channel information into the demodulation PM to obtain demodulated information;

[0156] In this setup, the demodulation PM and demodulation model exist in pairs. The demodulation PM is set on the network device, and the demodulation model is set on the terminal device.

[0157] In one possible implementation, when the first channel information is the first SINR, the network device inputs the second information and the first SINR into the demodulation PM to obtain demodulated information. This allows the demodulation PM to fully simulate the demodulation model in the terminal device demodulating the second information, fully considering possible interference in the target channel, and achieving better restoration results.

[0158] In one possible implementation, the network device can maintain a demodulation PM and send the demodulation PM to the terminal device, which uses it as a demodulation model. It should be understood that the names of the demodulation PM and demodulation model are only used to distinguish the storage and usage location of the model. The demodulation PM and demodulation model can be the same or different, and there is no restriction here.

[0159] In one possible implementation, the network device can receive a demodulation PM sent by the terminal device. The demodulation PM has the same or similar functions as the demodulation model used by the terminal device. It should be understood that the names of the demodulation PM and demodulation model here are only used to distinguish the storage and usage location of the model. The demodulation PM and demodulation model can be the same or different, and there is no restriction here.

[0160] S304. The network device inputs the demodulated information into the channel decoder PM to obtain the decoded information.

[0161] The decoded information carries a first check code, which is used to perform cyclic redundancy check (CRC) on the decoded information.

[0162] In one possible implementation, the channel decoding PM and the channel decoding model exist in pairs, with the channel decoding PM set in the network device and the channel decoding model set in the terminal device.

[0163] In one possible implementation, the network device can maintain a channel decoding PM and send the channel decoding PM to a terminal device, which uses it as a channel decoding model. It should be understood that the names of the channel decoding PM and the channel decoding model are only used to distinguish the storage and usage location of the model. The channel decoding PM and the channel decoding model can be the same or different, and there is no restriction here.

[0164] In one possible implementation, the network device can receive a channel decoding PM sent by the terminal device. The channel decoding PM has the same or similar functions as the channel decoding model used by the terminal device. It should be understood that the names of the channel decoding PM and the channel decoding model here are only used to distinguish the storage and usage location of the model. The channel decoding PM and the channel decoding model can be the same or different, and there is no restriction here.

[0165] S305. The network device performs cyclic redundancy check on the decoded information based on the first checksum.

[0166] In one possible implementation, the decoding information output by the channel decoder PM includes a check code field (i.e., the first check code). The network device uses a pre-agreed formula to calculate the non-check code field in the decoding information to obtain the second check code.

[0167] If the first check code and the second check code are the same, the decoded information is processed by CRC, and step S306 is executed.

[0168] If the first check code and the second check code are different, the decoded information fails the CRC test, and step S307 is executed.

[0169] S306. When the decoded information passes the cyclic redundancy check, the network device determines that the performance of the demodulated PM meets the standard and the performance of the channel decoded PM meets the standard.

[0170] In one possible implementation, when the decoded information passes through CRC, the network device determines that the performance of the demodulation PM and the channel decoding PM are up to standard. The network device can then send a first signal to the terminal device indicating that the performance of the demodulation PM and the channel decoding PM are up to standard, so that the terminal device can continue to use the demodulation model and the channel decoding model to analyze the communication data. No limitation is imposed here.

[0171] S307. If the decoded information fails the cyclic redundancy check, the network device determines that the performance of the demodulation PM or the performance of the channel decoding PM is substandard based on the network device's performance curve.

[0172] Among them, the performance curve of network equipment includes the bit error rate curve of network equipment.

[0173] In one possible implementation, the network device analyzes its first bit error rate (BER) based on a first SINR. If the BER is greater than a first threshold, the demodulated PM's performance is worse than expected, and the network device determines that the demodulated PM's performance is substandard. If the BER is less than or equal to the first threshold, the demodulated PM's performance meets the expected performance, and the network device determines that the channel decoding PM's performance is substandard. For example, the first threshold could be a value that deviates from the network device's expected performance curve by 20%, such as... Figure 4 As shown in the figure, the curves represent the performance curves of the network device. The horizontal axis represents the channel's SINR, and the vertical axis represents the network device's bit error rate. These performance curves are pre-set expected curves. Figure 4 As shown, if the first bit error rate of the network device is on or above the performance curve of the network device within a certain range (the first bit error rate is less than or equal to the first threshold), the demodulation PM performance of the network device can be considered to meet the standard, while the channel decoding PM performance is determined to be substandard.

[0174] In this embodiment of the application, when the decoded information fails the CRC test, it is proposed that the model for determining the substandard performance based on the bit error rate be demodulation PM or channel decoding PM. This helps to efficiently and accurately locate the fault point, shorten the time required for the abnormal recovery of the AI ​​communication system, and provide a guarantee for the reliable operation of the AI ​​communication system.

[0175] Optionally, the network device may also send a second signal to the terminal device to indicate that the demodulation PM performance is substandard if the demodulation PM performance is substandard. This signal may be used to alert the terminal device to check the performance of the demodulation model, retrain the demodulation model, or disable the demodulation model. No restrictions are imposed here.

[0176] Optionally, the network device may also send a third signal to the terminal device to indicate that the channel decoding PM's performance is substandard if the channel decoding PM's performance is substandard. This is to warn the terminal device that it needs to check the performance of the channel decoding model, or retrain the channel decoding model, or disable the channel decoding model, etc. There are no restrictions here.

[0177] In this embodiment, by running at least two Model Managers (PMs) in the network device, the performance of at least two models used in the terminal device is determined by whether the performance of the at least two PMs meets the standards. There is a one-to-one correspondence between the models in the at least two PMs and the models in the at least two models. For example, when the at least two PMs are a demodulation PM and a channel decoding PM, the decoding information output by the channel decoding PM is subjected to a CRC check. If the decoding information passes the CRC check, it is determined that the performance of both the demodulation PM and the channel decoding PM meets the standards. Compared to the conventional scheme of evaluating the performance of a single model, the scheme provided in this application evaluates the performance of at least two PMs in a coordinated manner, assessing the overall performance of multiple models in the AI ​​communication system to ensure they function correctly, thus avoiding getting trapped in local optima.

[0178] In one possible implementation, since the modulation may be at standard constellation points or non-standard constellation points, the aforementioned Figure 2 and Figure 3 The case of modulation to standard constellation points has been introduced, and the following section will combine... Figure 5 and Figure 6 The application describes the use of non-standard constellation points for modulation. The AI ​​communication system provided in this application can also, for example... Figure 5 As shown, to facilitate a clear understanding of the AI ​​communication system provided in this application, the following is combined with... Figure 2 The AI ​​communication system shown here is an example of another AI communication system provided in this application.

[0179] Compared to Figure 2In the AI ​​communication system shown in this application, during operation, the modulation module in the network device can also be a modulation model, and the first channel information sent by the terminal device to the network device can also include the channel information of the target channel. In this case, the modulation model is a model trained based on the channel information of the target channel, and the demodulation PM is also a model trained based on the channel information of the target channel.

[0180] In combination with the above Figure 5 The following is combined with Figure 6 Regarding the aforementioned Figure 5 The methods that may be implemented in the AI ​​communication system shown are introduced.

[0181] S601, Network equipment determines source information;

[0182] Among them, the source information is the information generated by the network device and needs to be sent to the terminal device.

[0183] S602. The network device performs channel coding on the source information to obtain coded information;

[0184] In one possible implementation, the network device performs channel coding on the source information by adding redundant information to the source information to obtain coded information. The method by which the network device adds redundant information to the source information can be to add redundant information to the source information according to a preset rule, or the network device can input the source information into a channel coding model to obtain coded information. No limitation is made here.

[0185] S603. The network device inputs the encoded information into the modulation model to obtain the modulation information;

[0186] The modulation model can be a model trained based on the channel information of the target channel, which is included in the first channel information.

[0187] In one possible implementation, the channel information of the target channel can be the equivalent channel of the target channel. For example, the equivalent channel can be described using W*H, where W describes the precoding or beamforming matrix and H describes the channel between the transmit antenna of the network device and the receive antenna of the terminal device.

[0188] S604. The network device adds a first demodulation reference signal to the modulation information to obtain the first information;

[0189] The first information includes a first DMRS, which is a high-density DMRS. Setting a high-density DMRS in the first information can effectively improve the accuracy of the first SINR reported by the terminal device.

[0190] S605. The network device sends the first information to the terminal device through the target channel, that is, the terminal device receives the first information sent by the network device through the target channel.

[0191] It should be understood that due to the presence of noise and other interference factors in the target channel, these interference factors may cause changes in the content of the first information. Therefore, in order to distinguish between the first information when inputting into the target channel and the first information when outputting from the target channel, the first information output from the target channel is called the second information. In practical applications, the first information and the second information can be the same or different, and no restrictions are imposed here.

[0192] S606. The terminal device performs channel estimation based on the received first information to obtain first channel information;

[0193] The first channel information includes the first SINR and the channel information of the target channel; the channel estimation can be a channel estimation model or a preset channel estimation method, which is not limited here.

[0194] It should be understood that the first information received is the aforementioned second information, and this application does not limit this in its description.

[0195] S607. The terminal device sends the first channel information to the network device, that is, the network device receives the first channel information sent by the terminal device.

[0196] In one possible implementation, the first channel information is included in the message sent by the terminal device to the network device via RRC message or DCI, and its specific form and content are not limited here.

[0197] In one possible implementation, the first channel information may include the first SINR and the channel information of the target channel. It should be understood that in some common scenarios, the signal-to-noise ratio can also be used instead of SINR, and this is not a limitation.

[0198] In one possible implementation, the first SINR can be a description at the resource element (RE) granularity, the resource block (RB) granularity, or the subband granularity in the frequency domain. Moreover, the description of the channel characteristics of the target channel in the first SINR can be reported in the time domain in units of multiple frames, for example, analyzed and described in units of 100 frames, and then reported.

[0199] S608. The network device processes the first information based on the first channel information to obtain the second information;

[0200] In one possible implementation, when the first channel information includes the first SINR and the channel information of the target channel, the simulated channel module in the network device simulates noise and interference information in the target channel based on the first SINR, and simulates the channel structure of the target channel based on the channel information of the target channel to determine the first simulated channel (the first simulated channel is similar to the target channel, so it can also be understood as the first simulated channel being the channel obtained by simulating the target channel based on the first SINR and the channel information of the target channel).

[0201] It should be understood that, due to the target channel (i.e., the aforementioned) Figure 1 The communication channel may contain noise and other interference factors. Therefore, after the network device sends the first information through the target channel, the first information received by the terminal device is called the second information to distinguish the possible differences between the two information. In this application, it is assumed that the first simulated channel can fully simulate the target channel. That is, the information obtained by inputting the first information into the first simulated channel and the target channel respectively is the second information. No limitation is made here.

[0202] S609. The network device inputs the second information and the first channel information into the demodulation PM to obtain demodulated information;

[0203] S610: The network device inputs the demodulated information into the channel decoder PM to obtain the decoded information;

[0204] S611. The network device performs cyclic redundancy check on the decoded information based on the first checksum.

[0205] S612. When the decoded information passes the cyclic redundancy check, the network device determines that the performance of the demodulation PM meets the standard and the performance of the channel decoding PM meets the standard.

[0206] S613. If the decoded information fails the cyclic redundancy check, the network device determines that the performance of the demodulation PM or the performance of the channel decoding PM is substandard based on the network device's performance curve.

[0207] It should be understood that steps S609 to S613 here are the same as those described above. Figure 3 The implementation methods for steps S303 to S307 are similar; please refer to the foregoing for details. Figure 3 The descriptions of steps S304 to S307 are omitted here.

[0208] In this embodiment, when the modulation is a non-standard constellation point, the modulation and demodulation are related to the target channel. The channel information of the target channel is introduced during the training phase of the modulation model and demodulation PM, making the modulation model, demodulation model and demodulation PM more in line with specific usage requirements, providing a more practical and effective solution, and ensuring that the solution provided by this application can be applied to more different multi-AI model communication systems.

[0209] In one possible implementation, the AI ​​communication system provided in this application can also be as follows: Figure 7 As shown, to facilitate a clear understanding of the AI ​​communication system provided in this application, the following is combined with... Figure 2 The AI ​​communication system shown here is an example of another AI communication system provided in this application.

[0210] Compared to Figure 2 The AI ​​communication system shown in this application allows for channel estimation (PM) to be set in the network device during operation, and the first channel information sent by the terminal device to the network device may include the channel information of the target channel.

[0211] In this scenario, after receiving the first channel information from the terminal device, the network device simulates the target channel based on the first channel information to obtain a second simulated channel. It then uses this simulated first information to simulate transmission within the second simulated channel, obtaining second information. This second information is input into the channel estimation process (PM) to obtain channel estimation information. The target channel information is then used as a label for the channel estimation PM to evaluate whether the PM's performance meets the requirements.

[0212] In combination with the above Figure 7 The following is combined with Figure 8 Regarding the aforementioned Figure 7 The methods that may be implemented in the AI ​​communication system shown are introduced.

[0213] S801, Network devices determine source information;

[0214] S802. The network device performs channel coding on the source information to obtain coded information;

[0215] It should be understood that steps S801 and S802 are the same as those described above. Figure 6 Steps S601 and S602 are similar, please refer to the foregoing for details. Figure 6 Steps S601 and S602 shown will not be repeated here.

[0216] S803: Network devices modulate the encoded information to obtain modulated information;

[0217] In one possible implementation, the process of the network device modulating the encoded information can be achieved by a pre-trained modulation model or by a pre-set modulation procedure, without any limitation.

[0218] S804, The network device pre-encodes the modulation information and the first demodulated reference signal to obtain the third information;

[0219] The first DMRS is a high-density DMRS, therefore the third information carries the first DMRS.

[0220] S805: The network device performs resource mapping on the third information to obtain the first information;

[0221] The first information contains content related to the first DMRS.

[0222] S806. The network device sends the first information to the terminal device through the target channel, that is, the terminal device receives the first information sent by the network device through the target channel.

[0223] It should be understood that due to the presence of noise and other interference factors in the target channel, these interference factors may cause changes in the content of the first information. Therefore, in order to distinguish between the first information when inputting into the target channel and the first information when outputting from the target channel, the first information output from the target channel is called the second information. In practical applications, the first information and the second information can be the same or different, and no restrictions are imposed here.

[0224] In this embodiment of the application, the modulation information and the first demodulation reference signal are pre-coded and resource-mapped to obtain the first information. When the first information is transmitted through the target channel, the influence of channel correlation on the transmitted signal can be effectively reduced by pre-processing the transmitted signal. For example, the bit error rate and interference of the information can be reduced.

[0225] S807. The terminal device performs resource demapping on the received first information to obtain the fourth information;

[0226] Since there may be interference, noise and other influencing factors in the target channel, the first information received by the terminal device can also be referred to as the second information in this application. That is, the terminal device performs resource demapping on the received second information to obtain the first information after resource demapping (in order to facilitate the distinction of the changes introduced by the third information through the target channel, the information obtained by performing resource demapping on the second information is referred to as the fourth information).

[0227] In one possible implementation, if there is no loss in the target channel, the fourth information is the same as the third information. In specific application scenarios, the third information and the fourth information can be the same or different, and there is no limitation here.

[0228] S808: The terminal device performs channel estimation based on the fourth information to obtain the first channel information;

[0229] The first channel information includes the channel information of the target channel; the terminal device performs channel estimation based on the fourth message, which is implemented based on the channel estimation model.

[0230] In one possible implementation, the channel estimation model set in the terminal device performs channel estimation based on the field related to the first DMRS in the fourth information to obtain the channel information of the target channel. It should be understood that in specific application scenarios, the channel information of the target channel can be the equivalent channel of the target channel. Therefore, the first channel information can also be referred to as the equivalent channel of the target channel, without limitation.

[0231] S809. The terminal device sends the first channel information to the network device, that is, the network device receives the first channel information sent by the terminal device.

[0232] In one possible implementation, the first channel information is included in the message sent by the terminal device to the network device via RRC message or DCI, and its specific form and content are not limited here.

[0233] S810: The network device simulates the target channel based on the first channel information to obtain the second simulated channel;

[0234] The first channel information includes the channel information of the target channel.

[0235] In one possible implementation, when the first channel information includes the channel information of the target channel, the simulated channel module in the network device simulates the target channel based on the channel information of the target channel to determine the second simulated channel (the second simulated channel is similar to the target channel, so it can also be understood as the second simulated channel being the channel obtained by simulating the target channel based on the first SINR and the channel information of the target channel).

[0236] S811, The network device simulates the transmission of the first information in the second analog channel to obtain the second information;

[0237] It should be understood that this application assumes that the second simulated channel can fully simulate the target channel. That is, the information obtained by inputting the first information into the second simulated channel and the target channel respectively is the second information, and no limitation is made here.

[0238] S812, The network device performs resource demapping on the second information to obtain the fourth information;

[0239] S813, The network device inputs the fourth information into the channel estimation PM to obtain the channel estimation information;

[0240] S814. The network device inputs the fourth information and channel estimation information into the demodulation PM to obtain demodulation information;

[0241] Since the channel estimation information is obtained by analyzing the relevant operations performed by the channel estimation model in the channel estimation PM simulation terminal device, the channel estimation information should be similar to the channel information of the target channel; that is, the channel estimation information should be similar to the first channel information. Furthermore, the fourth information is obtained by resource demapping the second information and is similar to the second information.

[0242] Therefore, step S814 is the same as described above. Figure 6 The implementation of step S609 shown is similar; please refer to [link / reference] for details. Figure 6 The details of step S609 will not be repeated here.

[0243] S815, the network device inputs the demodulated information into the channel decoder PM to obtain the decoded information;

[0244] S816. The network device performs cyclic redundancy check on the decoded information based on the first checksum.

[0245] S817. When the decoded information passes the cyclic redundancy check, the network device determines that the performance of the demodulated PM meets the standard and the performance of the channel decoded PM meets the standard.

[0246] S818. If the decoded information fails the cyclic redundancy check, the network device determines that the performance of the demodulation PM or the performance of the channel decoding PM is substandard based on the network device's performance curve.

[0247] It should be understood that steps S815 to S818 here are the same as those described above. Figure 6 The implementation methods for steps S610 to S613 are similar; please refer to the foregoing for details. Figure 6 The details of steps S610 to S613 are not repeated here.

[0248] S819. The network device analyzes whether the performance of the channel estimation PM meets the standard based on the channel estimation information and the channel information of the target channel.

[0249] In one possible implementation, the channel information of the target channel can be used as a label for the channel estimation PM to evaluate the performance of the channel estimation PM.

[0250] In this embodiment, the performance of the channel estimation PM is analyzed by analyzing the similarity between the channel estimation model and the first channel information (i.e., the channel information of the target channel) output by the channel estimation PM and the channel estimation information. This can help network devices to comprehensively and reliably manage the use of AI communication models by terminal devices, effectively improving the reliability and stability of the solution.

[0251] In one possible implementation, as described above Figure 2 , Figure 5 and Figure 7 In the communication system shown, CSI transmission can also be introduced to help network devices and terminal devices better utilize the target channel for data transmission.

[0252] exist Figure 2 , Figure 5 and Figure 7 In the communication system shown, when CSI transmission is introduced, the terminal device compresses the CSI corresponding to the target channel based on the channel estimation results to save channel overhead. The terminal device then sends the compressed CSI to the network device. The network device decompresses the compressed CSI to obtain the final CSI and uses it to adjust the information sent from the network device to the terminal device.

[0253] The AI ​​communication method provided in this application has been described above. The AI ​​communication device provided in this application will be described below with reference to the accompanying drawings.

[0254] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of the AI ​​communication device provided in this application.

[0255] The AI ​​communication device 90 includes:

[0256] The receiving unit 910 is used to receive first channel information, which is used to describe the channel characteristics of the target channel.

[0257] Processing unit 920 is used to process first information based on first channel information to obtain second information, wherein the first information is information transmitted through the target channel;

[0258] Processing unit 920 is also used to input the second information and the first channel information into demodulation proxy model PM to obtain demodulation information;

[0259] The processing unit 920 is also used to input demodulated information into the channel decoder PM to obtain decoded information, the decoded information carrying a first check code;

[0260] The verification unit 930 is used to perform cyclic redundancy check on the decoded information based on the first check code. If the decoded information passes the cyclic redundancy check, it is determined that the performance of the demodulation PM meets the standard and the performance of the channel decoding PM meets the standard.

[0261] Optionally, the verification unit 930 is also used to confirm, based on the performance curve of the network device, that the performance of the demodulation PM or the channel decoding proxy model is substandard if the decoded information fails the cyclic redundancy check.

[0262] Optionally, the first channel information includes the first signal-to-interference-plus-noise ratio (SINR);

[0263] Processing unit 920 is specifically used for:

[0264] Based on the first SINR, a target channel is simulated, and a first simulated channel is obtained.

[0265] The first information is simulated and transmitted in a first simulated channel to obtain the second information.

[0266] Optionally, the first channel information also includes the channel information of the target channel, and the modulation model is a model trained based on the channel information of the target channel;

[0267] Processing unit 920 is also used for:

[0268] The encoded information is input into the modulation model to obtain the modulation information. The encoded information is the information obtained by channel coding the source information. The source information is the information generated by the network device and sent to the terminal device.

[0269] A first demodulation reference signal is added to the modulation information to obtain the first information.

[0270] Optionally, the first channel information includes the channel information of the target channel;

[0271] Processing unit 920 is also used for:

[0272] The second information is input into the channel estimation PM to obtain the channel estimation information;

[0273] Based on the channel estimation information and the channel information of the target channel, we analyze whether the performance of the channel estimation PM meets the standards.

[0274] Optionally, the processing unit 920 is specifically used for:

[0275] The target channel is simulated based on the channel information of the target channel to obtain a second simulated channel;

[0276] The first information is simulated and transmitted in the second simulated channel to obtain the second information.

[0277] Optionally, the second information includes the first information transmitted via the second analog channel and used for resource demapping;

[0278] The device also includes:

[0279] The precoding unit 940 is used to precode the modulation information and the first demodulation reference signal to obtain the third information. The modulation information is the information obtained by channel coding and modulation of the source information. The source information is the information generated by the network device and sent to the terminal device.

[0280] Resource mapping unit 950 is used to perform resource mapping on the third message to obtain the first information;

[0281] The transmitting unit 960 is used to transmit first information through the target channel.

[0282] Please see Figure 10 This application provides an AI communication device 100, comprising:

[0283] The receiving unit 1010 is used to receive second information, which is obtained by transmitting first information through the target channel;

[0284] Analysis unit 1020 is used to analyze the channel characteristics of the target channel based on the second information to obtain the first channel information;

[0285] The transmitting unit 1030 is used to transmit first channel information, which is used to describe the channel characteristics of the target channel.

[0286] Optionally, the analysis unit 1020 is also used for:

[0287] The second information and the first channel information are input into the demodulation model to obtain the demodulation information;

[0288] The demodulated information is input into the channel decoding model to obtain the decoded information.

[0289] Optionally, the first channel information includes the first SINR.

[0290] Optionally, the first channel information may also include channel information of the target channel.

[0291] Optionally, the analysis unit 1020 is specifically used to input the second information into the channel estimation model to obtain channel estimation information.

[0292] Optionally, the second information is pre-encoded and resource-mapped information;

[0293] Analysis unit 1020 is specifically used for:

[0294] Perform resource demapping on the second information;

[0295] Based on the second information analysis of resource access mapping, the channel characteristics of the target channel are obtained to acquire the first channel information.

[0296] Please see Figure 11 , Figure 11 This is a schematic diagram of the structure of the AI ​​communication system provided in this application.

[0297] AI communication system 110 includes network equipment 1110 and terminal equipment 1120.

[0298] Among them, network device 1110 performs the aforementioned Figure 9 Similar functions and operations to AI communication devices.

[0299] Terminal device 1120 performs the aforementioned Figure 10 Similar functions and operations to AI communication devices.

[0300] Please see Figure 12 , Figure 12 Another structural schematic diagram of the AI ​​communication device provided in this application.

[0301] AI communication device 120 includes a transceiver 1220, a processor 1210, and a memory 1240. The processor 1210 and the transceiver 1220 can be interconnected via a bus 1230. The processor 1220 is coupled to the memory 1240 and is used to execute instructions stored in the memory 1240 to control the transceiver 1220 to send signals, and / or receive signals, and / or process signals.

[0302] Bus 1230 may include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the processing system. The bus communicatively couples various circuits together, including one or more processors (typically represented by a processor), memory, and computer-readable media (typically represented by a computer-readable media). The bus may also link various other circuits, such as timing sources, peripherals, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further. The bus interface provides the interface between the bus and transceivers, and between the bus and the interface.

[0303] Processor 1210 includes a microprocessor (e.g., x86, ARM), microcontroller, digital signal processor (DSP), field-programmable gate array (FPGA), GPU, programmable logic device (PLD), state machine, gated logic, discrete hardware circuitry, and other suitable hardware configured to perform various functions. The processor is responsible for managing the bus and general processing, including executing software stored on a computer-readable medium. When executed by the processor, the software causes the processing system to perform the various functions described below for any particular device.

[0304] The memory 1240 can be volatile memory or non-volatile memory, or may include both. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0305] The processor 1210, memory 1240, and computer-readable medium can perform functions such as: encoding, decoding, rate matching, rate dematching, scrambling, descrambling, modulation, demodulation, layer mapping, fast Fourier transform (FFT), inverse fast Fourier transform (IFFT), inverse discrete Fourier transform (IDFT), precoding, resource element (RE) mapping, channel equalization, RE demapping, digital beamforming (BF), adding cyclic prefix (CP), removing CP, etc.

[0306] Transceiver 1220 provides a communication interface or means for communicating with various other devices via a wireless transmission medium. The transceiver may be coupled to an antenna array, and the transceiver and antenna array may be used together for communication with a corresponding network type. At least one interface (e.g., a network interface and / or a user interface) provides a communication interface or means for communication via an internal bus or via an external transmission medium.

[0307] When the AI ​​communication device 120 performs the aforementioned Figure 3 , Figure 6 or Figure 8 When performing related operations, the processor 1210 needs to receive the first channel information, which is used to describe the channel characteristics of the target channel.

[0308] Based on the first channel information, the first information is processed to obtain the second information, where the first information is the information transmitted through the target channel.

[0309] The second information and the first channel information are input into the demodulation proxy model PM to obtain the demodulation information;

[0310] The demodulated information is input into the channel decoder PM to obtain the decoded information, which carries the first check code.

[0311] Cyclic redundancy check is performed on the decoded information based on the first check code. If the decoded information passes the cyclic redundancy check, it is determined that the performance of the demodulation PM and the performance of the channel decoding PM are both up to standard.

[0312] When the AI ​​communication device 120 performs the aforementioned Figure 3 , Figure 6 or Figure 8 When the network device performs related operations, the transceiver 1220 needs to receive the second information sent by the network device. The second information is obtained by transmitting the first information through the target channel.

[0313] Based on the analysis of the second information, the channel characteristics of the target channel are analyzed to obtain the first channel information;

[0314] Send first channel information, which is used to describe the channel characteristics of the target channel.

[0315] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0316] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0317] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0318] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0319] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An artificial intelligence (AI) communication method, characterized in that, include: Receive first channel information, which is used to describe the channel characteristics of the target channel; Based on the first channel information, the first information is processed to obtain the second information, wherein the first information is information transmitted through the target channel; The second information and the first channel information are input into the demodulation proxy model PM to obtain demodulation information; The demodulated information is input into the channel decoder PM to obtain decoded information, which carries a first check code. Cyclic redundancy check is performed on the decoded information based on the first check code. If the decoded information passes the cyclic redundancy check, it is determined that the performance of the demodulation PM meets the standard and the performance of the channel decoding PM meets the standard.

2. The method according to claim 1, characterized in that, The method further includes: If the decoded information fails the cyclic redundancy check, the performance of the demodulation PM or the channel decoding PM is confirmed to be substandard based on the performance curve of the network device.

3. The method according to claim 1 or 2, characterized in that, The first channel information includes the first signal-to-interference-plus-noise ratio (SINR); The process of processing the first information based on the first channel information to obtain the second information includes: Based on the first SINR, the target channel is simulated to obtain the first simulated channel; The second information is obtained by simulating the transmission of the first information in the first simulated channel.

4. The method according to any one of claims 1 to 3, characterized in that, The first channel information also includes the channel information of the target channel, and the modulation model is a model trained based on the channel information of the target channel; Before receiving the first channel information, the method further includes: The encoded information is input into the modulation model to obtain modulation information. The encoded information is information obtained by channel coding of the source information. The source information is information generated by the network device and sent to the terminal device. A first demodulation reference signal is added to the modulation information to obtain the first information.

5. The method according to claim 1 or 2, characterized in that, The first channel information includes the channel information of the target channel; The method further includes: The second information is input into the channel estimation PM to obtain channel estimation information; Based on the channel estimation information and the channel information of the target channel, analyze whether the performance of the channel estimation PM meets the standard.

6. The method according to claim 5, characterized in that, The process of processing the first information based on the first channel information to obtain the second information includes: Based on the channel information of the target channel, the target channel is simulated to obtain a second simulated channel; The transmission of the first information in the second simulated channel is simulated to obtain the second information.

7. The method according to claim 6, characterized in that, The second information includes the first information transmitted via the second analog channel and subjected to resource demapping; Before receiving the first channel message, the method further includes: The modulation information and the first demodulation reference signal are pre-coded to obtain the third information. The modulation information is the information obtained by channel coding and modulation of the source information. The source information is the information generated by the network device and sent to the terminal device. The third message is mapped to obtain the first information; The first information is transmitted through the target channel.

8. An artificial intelligence (AI) communication method, characterized in that, include: Receive second information, which is obtained by transmitting first information through a target channel; Based on the second information, the channel characteristics of the target channel are analyzed to obtain the first channel information; The first channel information is sent, which is used to describe the channel characteristics of the target channel.

9. The method according to claim 8, characterized in that, The method further includes: The second information and the first channel information are input into the demodulation model to obtain demodulation information; The demodulated information is input into the channel decoding model to obtain the decoded information.

10. The method according to claim 8 or 9, characterized in that, The first channel information includes the first signal-to-interference-plus-noise ratio (SINR).

11. The method according to claim 9, characterized in that, The first channel information also includes the channel information of the target channel.

12. The method according to any one of claims 8 to 11, characterized in that, The step of analyzing the channel characteristics of the target channel based on the second information to obtain the first channel information includes: The second information is input into the channel estimation model to obtain channel estimation information.

13. The method according to any one of claims 8 to 12, characterized in that, The second information is information that has undergone pre-coding and resource mapping; The step of analyzing the channel characteristics of the target channel based on the second information to obtain the first channel information includes: Perform resource demapping on the second information; Based on the second information of resource access mapping, the channel characteristics of the target channel are analyzed to obtain the first channel information.

14. An artificial intelligence (AI) communication device, characterized in that, include: A receiving unit is configured to receive first channel information, wherein the first channel information is used to describe the channel characteristics of the target channel; The processing unit is configured to process the first information based on the first channel information to obtain the second information, wherein the first information is information transmitted through the target channel; The processing unit is further configured to input the second information and the first channel information into the demodulation proxy model PM to obtain demodulation information; The processing unit is further configured to input the demodulation information into the channel decoding PM to obtain decoding information, wherein the decoding information carries a first check code; The verification unit is used to perform cyclic redundancy check on the decoded information based on the first check code, and if the decoded information passes the cyclic redundancy check, determine that the performance of the demodulation PM meets the standard and the performance of the channel decoding PM meets the standard.

15. An artificial intelligence (AI) communication device, characterized in that, include: The receiving unit is used to receive second information, which is obtained by transmitting first information through a target channel; An analysis unit is configured to analyze the channel characteristics of the target channel based on the second information to obtain first channel information; A transmitting unit is configured to transmit the first channel information, which describes the channel characteristics of the target channel.

16. An artificial intelligence (AI) communication system, characterized in that, This includes network equipment and terminal equipment; The network device is used to implement the method according to any one of claims 1 to 7; The terminal device is used to implement the method according to any one of claims 8 to 13.

17. An artificial intelligence (AI) communication device, characterized in that, Includes a processor, which is coupled to a memory; The memory stores instructions that, when executed on the processor, cause the communication device to perform the method of any one of claims 1 to 13.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a processor, cause the method of any one of claims 1 to 13 to be implemented.

19. A computer program product, characterized in that, When the computer program product is executed on a computer, the method of any one of claims 1 to 13 is implemented.

Images