A decoding method and related apparatus

Abstract

The embodiment of the application provides a decoding method and a related device, wherein the scheme can send service demand information to a receiving end, and the receiving end obtains a fine tuning factor through nonlinear mapping according to the service demand information and channel state information. The fine tuning factor is used to modify the reference weight of a neural decoding network, and the modified weight is used to scale the message amplitude of a check node transmitted to a variable node. The scheme only needs low-dimensional features (SRI and channel state information) to obtain the fine tuning factor for the reference weight, realizes real-time modification of offline weights at a very small calculation cost, and takes into account the adaptability of dynamic channels and the calculation power limit of terminal equipment. Meanwhile, the scheme can actively adjust the weight of the neural decoding network according to the service demand, realizes active matching of physical layer resources and upper layer service demand, and avoids resource mismatch caused by passive and lagging network configuration adjustment.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a decoding method and related apparatus. Background Technology

[0002] As mobile communication technology evolves from 5G to 6G, communication systems are facing unprecedented challenges: they must not only support extremely high throughput, but also simultaneously meet highly differentiated service requirements such as hyper-reliable low latency communication (HRLLC) and massive machine-type communications (mMTC), with target bit error rates even needing to be as low as... .

[0003] Low-density parity-check (LDPC) codes play a crucial role in communication systems as a standard coding scheme for data channels. LDPC codes are constructed from sparse parity-check matrices, and decoding iteration information can be passed between check nodes and variable nodes to correct errors. In practical applications, normalized min-sum (NMS) and layered min-sum (LMS) decoding algorithms are widely used to balance performance and complexity. To further explore the potential of LDPC codes, neural network decoders based on "deep unrolling" (neural NMS / LMS, NNMS / NLMS) are widely adopted. This involves mapping the iterative structure of traditional NMS or LMS decoding algorithms into a neural network, using weight parameters obtained through offline training to replace the fixed normalization factor in the traditional algorithm, thereby achieving superior error correction performance.

[0004] Although neural network decoding technology performs well in theory, it suffers from architectural rigidity and parameter mismatch when faced with complex and ever-changing real-world scenarios. Summary of the Invention

[0005] In view of this, this application provides a decoding method and related apparatus to solve at least some of the above-mentioned problems, and the disclosed technical solution is as follows:

[0006] In a first aspect, this application provides a decoding method executed by a receiving end. The method includes: receiving a first message from a sending end, the first message including service requirement information, the service requirement information indicating the current service's requirement for at least one of power consumption, reliability, and throughput; determining a matching fine-tuning factor based on the service requirement information and channel state information; receiving a service signal from the sending end; inputting the fine-tuning factor and the decoding information corresponding to the service signal into a neural decoding network to obtain a decoding result, wherein the fine-tuning factor is used to correct the baseline weights within the neural decoding network, and the corrected weights are used to scale the message amplitude passed from the verification node to the variable node.

[0007] In the above scheme, the sending end can send the service requirement information of the current transmission service to the receiving end. The receiving end inputs the service requirement and the local channel state into the fine-tuning network and outputs a fine-tuning factor. The fine-tuning factor is used to adjust the baseline weights of the neural decoding network. The corrected weights are used to scale the message amplitude passed from the check node to the variable node in the neural decoding network. It can be seen that this scheme only requires low-dimensional features (SRI and channel state information) to obtain the fine-tuning factor for the baseline weights, achieving real-time correction of offline weights with minimal computational cost, while also considering the adaptability to dynamic channels and the computational limitations of terminal devices. Furthermore, this scheme can proactively adjust the weights of the neural decoding network according to service requirements, achieving proactive matching of physical layer resources with upper-layer service requirements, avoiding resource mismatch caused by passive and lagging adjustments to network configuration.

[0008] In one possible implementation, the service requirement information includes power consumption requirement fields, reliability requirement fields, and throughput requirement fields. Thus, the power consumption requirement field indicates the service's power requirements, the reliability requirement field indicates the service's reliability requirements, and the throughput requirement field indicates the service's throughput requirements. The corresponding fields can be filled in according to the actual transmission service requirements, thereby allowing the receiving end to clearly understand the actual needs of the current transmission service.

[0009] In one possible implementation, the power consumption requirement field can occupy 2 bits; a first value (e.g., 00) indicates that the current service has low power consumption requirements; a second value (e.g., 11) indicates that the current service is not sensitive to power consumption; and a third value (e.g., 01) or a fourth value (e.g., 10) indicates that the current service has moderate power consumption sensitivity. In this way, the transmitting end can indicate different levels of power consumption sensitivity of the service through the 2-bit power consumption requirement field, allowing the receiving end to understand the service's power consumption requirements in detail.

[0010] In one possible implementation, the reliability requirement field can occupy 2 bits. A first value (e.g., 00) indicates that the current service is an ultra-high reliability service; a second value (e.g., 11) indicates that the current service is a low reliability service; a third value (e.g., 01) indicates that the current service is a high reliability service; and a fourth value (e.g., 10) indicates that the current service is a standard reliability service. Thus, the transmitting end uses the 2-bit reliability requirement field to indicate different levels of reliability requirements for the current service, allowing the receiving end to accurately understand the different reliability requirements of the service.

[0011] In one possible implementation, the throughput requirement field occupies 1 bit; a first value (e.g., 1) indicates that the current service is a high-throughput service; a second value (e.g., 0) indicates that the current service is a low-throughput service. Thus, the sending end uses the 1-bit throughput requirement field to indicate the different levels of throughput requirements of the transmitted service, thereby facilitating the receiving end to accurately determine the reliability requirements of the current service.

[0012] In one possible implementation, before inputting the decoding information corresponding to the fine-tuning factor and the service signal into the neural decoding network to obtain the decoding result, the method further includes: determining a configuration strategy matching the service requirement information based on a first mapping relationship; and adjusting the configuration parameters of the neural decoding network based on the configuration strategy. The configuration parameters include at least one of hierarchical scheduling, iterative control, and precision control. In this scheme, the receiving end, based on adjusting the baseline weights of the neural decoding network according to the fine-tuning factor, further outputs corresponding decoding strategy instructions according to the service requirements. The fine-tuning factor and decoding strategy instructions are transmitted to the neural decoding network for network configuration adjustment, realizing proactive matching of physical layer resources and upper-layer service requirements. Moreover, this scheme introduces a multi-dimensional decoding control strategy to dynamically adjust the network configuration, achieving a fine-grained trade-off between power consumption, latency, and throughput while maintaining essentially lossless performance.

[0013] In one possible implementation, the configuration parameters of the neural decoding network are adjusted based on a configuration strategy, including: adjusting the critical layer update ratio of the neural decoding network based on hierarchical scheduling instructions. The critical layer update ratio is used to determine that in each iteration, only the top critical layer update ratios, arranged from largest to smallest error correction contribution, are selected for message updates.

[0014] As can be seen, in this scheme, the receiver evaluates the importance of each layer in the neural decoding network and updates messages only for a certain proportion (e.g., K layers) of the top key layers ranked from high to low importance according to the hierarchical scheduling instructions. This is no longer a fixed full-layer update mode, which reduces computational overhead and also reduces the power consumption of the receiver.

[0015] In one possible implementation, the configuration parameters of the neural decoding network are adjusted based on a configuration strategy, including adjusting the maximum number of iterations and / or the early stopping threshold of the neural decoding network based on iteration control instructions. This allows for early termination of iterations by dynamically setting the maximum number of iterations and / or the early stopping threshold, thus reducing computational overhead.

[0016] In one possible implementation, adjusting the configuration parameters of the neural decoding network based on the configuration strategy includes: adjusting the information quantization bit width during node updates in the neural decoding network based on precision control instructions. This allows the receiver to select the appropriate information bit width configuration according to the actual needs of the current service. For example, in mMTC scenarios where the decoder needs to process a large amount of low-rate data simultaneously, reducing the information quantization bit width (e.g., from 6 bits to 4 bits) can reduce the circuit flip-flop rate of the terminal device. This reduces the overall system power consumption without sacrificing throughput, achieving a flexible trade-off between performance and energy efficiency, and is particularly suitable for power-sensitive IoT terminals.

[0017] Secondly, this application also provides another decoding method, executed by the receiving end, the method comprising: receiving a first message from the sending end, the first message including service requirement information, the service requirement information indicating the current service's requirement for at least one of power consumption, reliability, and throughput; determining a configuration strategy matching the service requirement information based on a first mapping relationship; adjusting the configuration parameters of the neural decoding network based on the configuration strategy, the configuration parameters including at least one of key layer update ratio, maximum number of iterations, early stopping threshold, and precision control parameters; receiving a service signal from the sending end; inputting the decoding information corresponding to the service signal into the configured and adjusted neural decoding network to obtain a decoding result.

[0018] As can be seen, in this scheme, after the receiving end receives the service requirement information sent by the sending end, the receiving end determines the configuration strategy that matches the service requirement information based on the first mapping relationship and transmits it to the neural decoding network. The neural decoding network adjusts its configuration parameters based on the decoding strategy instructions. The configuration parameters include at least one of hierarchical scheduling, iterative control, and precision control. This scheme can achieve proactive matching between physical layer resources and service requirements, avoiding resource mismatch caused by passive and lagging strategy adjustments. Moreover, this scheme introduces a multi-dimensional decoding control strategy to dynamically adjust the network configuration, achieving a fine-grained trade-off between power consumption, latency, and throughput while maintaining essentially lossless performance.

[0019] Thirdly, this application also provides a decoding method executed by a sending end. The method includes: sending a first message to a receiving end, the first message including service requirement information, the service requirement information indicating the current service's requirement for at least one of power consumption, reliability, and throughput; the service requirement information enabling the receiving end to obtain a fine-tuning factor matching the service requirement information and channel state information, the fine-tuning factor being used to correct the baseline weights within the neural decoding network, and the corrected weights being used to scale the message amplitude passed from the verification node to the variable node. It is evident that this scheme only requires low-dimensional features (SRI and channel state information) to obtain the fine-tuning factor for the baseline weights, achieving real-time correction of offline weights with minimal computational cost, while also considering the adaptability of dynamic channels and the computational limitations of terminal devices. Simultaneously, this scheme can proactively adjust the weights of the neural decoding network according to service requirements, achieving proactive matching of physical layer resources with upper-layer service requirements, avoiding resource mismatch caused by passive and delayed adjustments to network configuration.

[0020] Fourthly, this application also provides a communication device including at least one processor coupled to a memory storing a program or instructions, wherein the processor executes the program or instructions to cause the device to perform the method as described in any of the possible implementations of the first to third aspects.

[0021] Fifthly, this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed, cause a computer to perform the method as described in any of the possible implementations of the first to third aspects.

[0022] Sixthly, this application also provides a communication system, including the communication device as described in the fourth aspect.

[0023] In a seventh aspect, this application also provides a chip system including one or more processors, the one or more processors being configured to call and execute instructions stored in memory, such that the method described in any of the possible implementations of the first to third aspects is executed. Attached Figure Description

[0024] Figure 1 A schematic diagram of a communication network architecture provided in an embodiment of this application;

[0025] Figure 2 A schematic diagram of another communication network architecture provided in an embodiment of this application;

[0026] Figure 3 A schematic diagram of a neural decoding network provided for an embodiment of this application;

[0027] Figure 4 A flowchart of a decoding method provided in an embodiment of this application;

[0028] Figure 5 This is a schematic diagram of the message structure of a first message provided in an embodiment of this application;

[0029] Figure 6 For use Figure 4 A schematic diagram of the communication network architecture for the decoding method shown.

[0030] Figure 7 A flowchart illustrating another decoding method provided in this application embodiment;

[0031] Figure 8 For use Figure 7 A schematic diagram of the communication network architecture for the decoding method shown.

[0032] Figure 9 A flowchart illustrating yet another decoding method provided in this application embodiment;

[0033] Figure 10 For use Figure 9 A schematic diagram of the communication network architecture for the decoding method shown.

[0034] Figure 11 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0035] Figure 12 This is a schematic diagram of another communication device provided in an embodiment of this application. Detailed Implementation

[0036] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to be a limitation of this application. As used in the specification and appended claims of this application, the singular expressions "a," "an," "the," "the," "the," and "this" are intended to also include expressions such as "one or more," unless the context clearly indicates otherwise. It should also be understood that in the embodiments of this application, "one or more" refers to one, two, or more; "and / or" describes the relationship between related objects, indicating that three relationships may 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.

[0037] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0038] In this application, "multiple" refers to two or more embodiments. It should be noted that in the description of the embodiments of this application, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance, nor as indicating or implying order.

[0039] The technical solutions provided in this application can be applied to various communication systems, such as: Global System for Mobile Communications (GSM) systems, General Packet Radio Service (GPRS), Wireless Local Area Network (WLAN), Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD) systems, sidelink communication systems, Universal Mobile Telecommunication System (UMTS), Worldwide Interoperability for Microwave Access (WiMAX) communication systems, non-terrestrial network (NTN) communication systems, 5th generation (5G) mobile communication systems, or new radio access technology (NR). Among these, 5G mobile communication systems can include non-standalone (NSA) and / or standalone (SA) networking. The technical solutions provided in this application can also be applied to future communication systems. This application does not limit the scope of these applications.

[0040] Figure 1 This is a schematic diagram of a communication system 100 used in an embodiment of this application. The communication system 100 may include network devices, such as... Figure 1 The network device 110 is shown. The communication system 100 may also include terminal devices, such as... Figure 1 The terminal device 120 is shown. The network device 110 and the terminal device 120 can communicate via a wireless link.

[0041] Figure 1 An exemplary network device 110 and a terminal device 120 are shown. Optionally, the communication system 100 may also include multiple network devices and / or multiple terminal devices.

[0042] The network equipment in this application can be network-side equipment such as access network equipment and core network equipment. Access network equipment is sometimes also called access node. Access network equipment has wireless transceiver capabilities and is used to communicate with terminals. Access network equipment includes, but is not limited to, base stations, evolved NodeBs (eNodeBs), transmission reception points (TRPs) in the above-mentioned communication systems, next-generation NodeBs (gNBs) in 5G mobile communication systems, access network equipment or modules of access network equipment in open RAN (ORAN) systems, satellites in NTN communication systems, base stations in future mobile communication systems, or access nodes in WiFi systems. Access network equipment can also be modules or units that can implement some of the functions of a base station. Access network equipment can be macro base stations, micro base stations or indoor stations, relay nodes or donor nodes, or wireless controllers in cloud radioaccess network (CRAN) scenarios. Optionally, access network equipment can also be servers, wearable devices, or vehicle-mounted equipment, etc. For example, the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU). Multiple access network devices in a communication system can be base stations of the same type or different types. Base stations can communicate with terminals directly or via relay stations. Terminals can communicate with multiple base stations using different access technologies. The embodiments of this application do not limit the specific technology or device form used in the access network equipment. In this application, the access network equipment is referred to as a network device.

[0043] In this application, the means for implementing the functions of a network device can be a network device itself, or a means capable of supporting the network device in implementing those functions, such as a processor, circuit, chip, or chip system. This means can be installed in or connected to the network device. In the technical solutions provided in this application, the example of a network device being used to implement the functions of a network device is used to describe the technical solutions provided in this application.

[0044] The terminal device in this application can be a wireless terminal device capable of receiving network device scheduling and instruction information. The wireless terminal device can be a device providing voice and / or data connectivity to a user, a handheld device with wireless connectivity, or other processing devices connected to a wireless modem. For example, the terminal device can communicate with one or more core networks or the Internet via a radio access network (RAN). The terminal device can also be referred to as a terminal, user equipment (UE), mobile station, mobile terminal, etc. Terminal devices can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), Internet of Things (IoT), ultra-reliable low-latency communication (URLLC), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, smart cities, or satellite communication, etc. The terminal can be a mobile phone, tablet computer, computer with wireless transceiver capabilities, wearable device, vehicle, aircraft (such as drone, helicopter, airplane), hot air balloon, ship, robot, robotic arm, or smart home device, etc. The embodiments of this application do not limit the form of the terminal device.

[0045] In this application, the apparatus for implementing the functions of a terminal device can be the terminal device itself, or any apparatus capable of supporting the terminal device in implementing those functions, such as a processor, circuit, chip, or chip system. This apparatus can be installed in or connected to the terminal device. In the technical solutions provided in this application, the example of a terminal device being used to implement the functions of a terminal device is used to describe the technical solutions provided in this application.

[0046] Access network equipment and / or terminal equipment can be fixed or mobile. They can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; on water; or in the air on aircraft, balloons, and satellites. This application does not limit the application scenarios of the access network equipment and terminal equipment. They can be deployed in the same or different scenarios; for example, both can be deployed on land simultaneously; or the access network equipment can be deployed on land while the terminal equipment is deployed on water, etc., and so on.

[0047] In practical applications, multiple network devices can collaborate to assist terminals in achieving wireless access, with different network devices each implementing a portion of the base station's functions. For example, network devices can be central units (CUs), distributed units (DUs), CUs (control planes, CPs), CUs (user planes, UPs), or radio units (RUs), etc. CUs and DUs can be set up separately or included in the same network element, such as a baseband unit (BBU). RUs can be included in radio frequency equipment or radio frequency units, such as remote radio units (RRUs), active antenna units (AAUs), or remote radio heads (RRHs).

[0048] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called O-CU (Open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software and hardware modules. CU (or CU-CP and CU-UP), DU, and RU can implement different protocol layer functions.

[0049] like Figure 2As shown in the embodiments of this application, another communication network architecture is also provided, which includes a transmitter, a channel, and a receiver. The transmitter sends a signal to the channel, and the receiver receives the signal from the channel. The transmitter includes an encoder, a modulator, and a transceiver; the receiver includes a transceiver, a decoder, and a demodulator. The transceiver is used to receive analog signals, the modulator is used to convert digital signals into analog signals, and the demodulator is used to convert analog signals into digital signals so that the decoder can decode the digital signals.

[0050] The transmitting and receiving ends can be the network devices and terminal devices mentioned above. The channel mentioned above can be a wireless link.

[0051] by Figure 2 Taking the network architecture shown as an example, since the signal received by the receiver can be affected by channel fading, interference, noise, etc., leading to information transmission failure, coding techniques are usually used to improve the reliability of information transmission. The transmitter can first encode the information sequence (i.e., user data) using an encoder, then the modulator can modulate the encoded data, and finally the transceiver transmits the modulated data. The receiver can receive the data transmitted through the channel using a transceiver, demodulate the received data using a demodulator, and then send the demodulated data to a decoder. Based on the encoding method used by the transmitter, the decoder can decode the received data, thereby reliably recovering the original information sequence and storing the recovered original information sequence in memory. This encoding method must be visible to both the transmitter and receiver. Generally, the encoding method is based on forward error correction coding, where forward error correction coding adds some redundant information (i.e., check data obtained through encoding) to the information sequence. The receiver can use this redundant information to reliably recover the original information sequence.

[0052] Since LDPC codes in forward error correction coding are linear block codes that can be defined using a very sparse parity check matrix, it is precisely by utilizing the sparsity of its parity check matrix that low-complexity encoding and decoding can be achieved. LDPC codes are currently the highest-performing channel coding, with performance very close to the Shannon limit. Therefore, the encoder described above can be an LDPC code encoder, and the decoder can be an LDPC code decoder. Thus, when the transmitter encodes the information sequence, the encoder at the transmitter can encode the information sequence according to the LDPC code's parity check matrix to obtain the LDPC codeword. The LDPC codeword includes the information sequence and the parity check data. When the receiver decodes the received LDPC codeword, it can decode the received LDPC codeword according to the LDPC code's parity check matrix.

[0053] Neural network decoding techniques (such as NNMS or NLMS) unfold the iterative process of traditional decoding algorithms (NMS or LMS) into a neural network structure. The network weights correspond to the normalization factors in traditional algorithms. Through deep learning training, these parameters can be adaptively optimized, rather than using fixed values ​​as in traditional methods. Figure 3 The diagram shown is a schematic of a neural decoding network that iterates twice.

[0054] in, This represents prior probability information, which is initialized by the channel received value y and output as the initial prior probability information. , These represent the weights of the variable nodes calculated in the first and second iterations, respectively. This represents the baseline weight of the verification node calculated in the first iteration; , These represent the variable nodes calculated in the 1st and 2nd iterations, respectively. v To the verification node c The message; , These represent the verification nodes calculated in the first and second iterations, respectively. c To variable node v The news.

[0055] However, future business scenarios will place drastically different demands on decoders. For example, HRLLC services prioritize extreme reliability and low latency; mMTC services are constrained by device power consumption and are extremely sensitive to power consumption; and eMBB services focus on throughput. Existing hardware decoders typically employ a fixed architecture, using fixed parameters (such as maximum number of iterations, layer update strategy, and granularity). This inflexible design leads to severe resource mismatch. For instance, using high-configuration resources for high-reliability services to decode low-power services results in significant energy waste; conversely, using low-configuration resources for low-reliability services to decode high-reliability services fails to meet the performance metrics of critical business applications.

[0056] Meanwhile, most existing neural network decoders adopt an "offline training, online fixing" model. Weight training is often based on an ideal white Gaussian noise channel, while real-world channel environments are extremely complex, including Doppler frequency shifts caused by high-speed movement or sudden interference. When the actual channel characteristics encountered by the receiver are inconsistent with the training set distribution, these fixed weight parameters become unsuitable, causing the decoder to get stuck in local optima in high signal-to-noise ratio regions, producing an unavoidable "error plane," severely limiting the final performance of the system.

[0057] Current decoding schemes mainly rely on passive, blind estimation of channel conditions by the receiver. While the receiver can perceive the "channel quality," it cannot know the "urgency" of the transmitter's traffic. The decoder can only passively and laggingly adjust its strategy, unable to actively match according to the transmitter's service requirements.

[0058] In view of this, this application provides a decoding method in which the transmitting end can provide the receiving end with the service requirement information of the current transmission service. The receiving end obtains a fine-tuning factor through nonlinear mapping based on the service requirement information and channel state information (CSI). This fine-tuning factor is then used to correct the baseline weights of the neural decoding network. The corrected weights are used to scale the message amplitude passed from the check node to the variable node. A fine-tuning factor greater than 1 amplifies the baseline weights, while a fine-tuning factor less than 1 shrinks the baseline weights. Therefore, this method can proactively adjust the weights of the neural decoding network according to service requirements, achieving proactive matching between physical layer resources and upper-layer service requirements, and avoiding resource mismatch caused by passive and lagging network configuration adjustments.

[0059] The solution provided in this application will be described in detail below with reference to the corresponding flowcharts. It is understood that the illustrative flowcharts provided in this application primarily use different devices (e.g., terminal devices, network devices) as examples of the execution subjects of this interactive illustration to illustrate the method, but this application does not limit the execution subjects of the interactive illustrations. For example, the devices (e.g., terminal devices, network devices) in the illustrative flowcharts can also be chips, chip systems, or processors that support the implementation of this method on the device, or logic modules or software that can implement all or part of the functions of the device.

[0060] As a general statement, the message or signaling interactions involved in the interaction process of this application embodiment can be standard messages or signaling or newly introduced messages or signaling. This application embodiment does not make specific limitations on this.

[0061] Figure 4 This is a schematic diagram illustrating a communication method according to an embodiment of this application. It can be understood that... Figure 4 The receiving end in the middle can be Figure 1 The term "terminal device" can refer to any terminal device, or a component within that terminal device (such as a processor, chip, or chip system). The sending end can be... Figure 1 Any access network device, or a component within an access network device (such as a processor, chip, or chip system). Figure 4 As shown, the method includes the following steps:

[0062] S101, the sending end sends the first message. The corresponding receiving end receives the first message.

[0063] In this embodiment of the application, the first message includes the service requirement information of the current service, which is used to indicate the requirements of the current service for at least one of the indicators such as power consumption, reliability and throughput.

[0064] The physical layer at the sending end can identify the current service requirements from the information mapped by the MAC layer and add information indicating the service requirements to the downlink messages of the physical layer.

[0065] In one possible implementation, the first message can be downlink control information (DCI). For example, the sender can add a service requirement indicator (SRI) field to the DCI used for dynamically scheduling UE uplink transmission resources, and can use reserved bits in the DCI as the SRI field.

[0066] like Figure 5 As shown, the main fields in DCI include: ResourceAlloc, Modulation and Coding Scheme (MCS), Hybrid Automatic Repeat Request (HARQ), SRI, etc.

[0067] ResourceAlloc indicates the location and range of the specific physical resource block allocated to the UE, supporting continuous or non-contiguous allocation. MCS is used to specify the data modulation scheme (such as QPSK, 16QAM) and coding rate, determining the transmission rate and reliability. HARQ is used to indicate key parameters for HARQ retransmission.

[0068] In one possible implementation, such as Figure 5 As shown, the SRI field can include a power consumption requirement (PR) field, a reliability requirement (RR) field, and a throughput requirement (TT) field.

[0069] For example, the SRI field can occupy 5 bits, of which the PR field is 2 bits, the RR field is 2 bits, and the TT field is 1 bit.

[0070] In one possible implementation, PR=00 indicates energy saving priority or low power consumption; PR=11 indicates insensitivity to power consumption, and power consumption can be sacrificed for performance; PR=01 / 10 indicates moderate power consumption sensitivity.

[0071] RR=00 indicates ultra-high reliability; RR=01 indicates high reliability; RR=10 indicates standard reliability; RR=11 indicates best effort, with the lowest reliability requirements.

[0072] TT=1 indicates high throughput, and TT=0 indicates normal throughput.

[0073] S102, the receiving end determines a matching fine-tuning factor based on service requirement information and local channel state information.

[0074] The fine-tuning factor is used to adjust the baseline weights within the neural network decoder. The adjusted weights are then used to scale the magnitude of the message passed from the check node to the variable node.

[0075] In one possible implementation, such as Figure 6 As shown, an auxiliary lightweight fine-tuning network is deployed before the neural decoding network. The input feature vector of this fine-tuning network includes... Wherein, SRI is the service requirement information sent by the sender. The channel features (i.e., channel state information) estimated for the receiver include, but are not limited to, signal-to-noise ratio (SNR) and Doppler shift estimates. The fine-tuning network outputs a fine-tuning factor that matches the input feature vector in real time through a nonlinear mapping. .

[0076] like Figure 6 As shown, the fine-tuning factor of the fine-tuning network output. The signal is passed to the neural decoding network and fine-tuned. Baseline weights for neural decoding networks Make corrections to generate the corrected weights for the current time step. Its expression is shown in Formula 1:

[0077] (1)

[0078] in, This indicates element-wise multiplication. This indicates the neural decoding network's first... l The baseline weights for the next iteration include the baseline weights of the variable nodes. Baseline weights of verification nodes Exemplary It can be obtained by offline training on a large amount of data under standard channels.

[0079] The corrected weights are substituted into the check node update formula of the NNMS or NLMS algorithm in the neural decoding network to scale the magnitude of the message passed from the check node to the variable node, as shown in the following formula:

[0080] (2)

[0081] in, Indicates the firstl Sub-iteration verification node c To variable node v The message; Indicates the first l Sub-iteration verification node c To variable node v The corrected weights on this edge; Indicates the first l Next iteration variable node v` To the verification node c The message; N ( c ) is related to the verification node c The set of all connected variable nodes; It is with the verification node c Connected, except for variable nodes v The set of variable nodes afterwards; The sign represents a series of multiplication operations, used to represent a compressed form of a long sequence of multiplication operations; the Sign function returns the corresponding value based on the sign of the parameter: 1 for a positive number, -1 for a negative number, and 0 for 0.

[0082] Fine-tuning factor It changes according to service requirements and current channel conditions. For example, if the SRI indicates a high-reliability service and the channel quality is good, the fine-tuning factor of the network output is adjusted. This increases the baseline weights, thereby increasing the neural decoding network's confidence in the decision and accelerating convergence to reduce latency. Conversely, SRI indicates high-reliability services, and if sudden interference or high-speed movement causes a large channel state error, the fine-tuning factor output by the fine-tuning network is adjusted. By reducing the baseline weight, the spread of unreliable messages (such as information transmitted by variable nodes affected by sudden interference) can be suppressed, preventing the propagation of erroneous information and thus eliminating the error plane under high signal-to-noise ratio.

[0083] The fine-tuning network is a lightweight neural network. In one possible implementation, the fine-tuning network can be trained offline using a training set. The fine-tuning network and the neural decoding network can be trained jointly, or the parameters of the neural decoding network can be obtained by training the neural decoding network first, and then the fine-tuning network can be trained. This application does not limit the training method.

[0084] Taking the example of first training a neural decoding network and then training a fine-tuning network, the SRI (Signal Response Indicator) of the current service and the channel characteristics of the current channel can be input into the fine-tuning network. Simultaneously, the received service signal vector of the current channel can be input into the trained neural decoding network. The fine-tuning network outputs a corresponding fine-tuning factor based on the SRI and channel characteristics and passes it to the neural decoding network. The neural decoding network decodes the input service signal vector and outputs the decoding result. Then, a loss function is used to calculate the loss value between the decoding result output by the neural decoding network and the actual decoding result of the service signal. If the loss value meets a preset condition, the training process ends; if the loss value does not meet the preset condition, the network parameters in the fine-tuning network are adjusted according to the loss value, and training continues until the loss value meets the preset condition, ending the training process.

[0085] S103, the transmitting end sends a service signal to the receiving end. The corresponding receiving end receives the service signal.

[0086] Combination Figure 2 The system architecture shown involves the transmitter encoding the service data using an encoder, the modulator modulating the encoded data, and finally the transceiver transmitting the modulated service signal.

[0087] S104, the receiver inputs the information to be decoded and the fine-tuning vector into the neural decoding network to obtain the decoding result.

[0088] The receiving end uses a transceiver to receive the service signals transmitted through the channel, and a demodulator to demodulate the received service signals to obtain the demodulated data (or information to be decoded). The fine-tuning factor output by the fine-tuning network is input into the neural decoding network, and the information to be decoded is sent to the neural decoding network at the same time, finally outputting the decoding result.

[0089] The fine-tuning factor is used to adjust the baseline weights of the neural decoding network. The adjusted weights are then used to scale the magnitude of the messages passed from the check nodes to the variable nodes. Amplify the benchmark weights, Reduce the baseline weight.

[0090] The decoding method provided in this embodiment allows the sending end to send service requirement information of the current transmission service to the receiving end. The receiving end inputs the service requirement and local channel state into the fine-tuning network and outputs a fine-tuning factor. The fine-tuning factor is used to adjust the baseline weights of the neural decoding network. The corrected weights are used to scale the message amplitude passed from the check node to the variable node in the neural decoding network. This scheme only requires low-dimensional features (SRI and channel state information) to obtain the fine-tuning factor for the baseline weights, achieving real-time correction of offline weights with minimal computational cost, while also considering the adaptability to dynamic channels and the computing power limitations of terminal devices. Furthermore, this scheme can proactively adjust the weights of the neural decoding network according to service requirements, achieving proactive matching of physical layer resources with upper-layer service requirements, avoiding resource mismatch caused by passive and lagging adjustments to network configuration.

[0091] To address the issue of achieving a balance between performance, power consumption, and throughput when using a fixed decoder architecture, this application also provides an alternative decoding method embodiment. In this embodiment, the receiving end pre-configures a lookup table (LUT) to map service requirement information into specific decoding strategy configuration instructions. For example, the neural decoding network can be dynamically adjusted from three dimensions: hierarchical scheduling, iterative control, and precision control. Figure 7 As shown, the method may include the following steps:

[0092] S201, the sending end sends the first message. The corresponding receiving end receives the first message.

[0093] The first message includes the current service requirements. This first message can be a DCI message, and the service requirements information may include power requirements, reliability requirements, and throughput requirements. (See also...) Figure 4 The relevant description of S101 in the illustrated embodiment.

[0094] S202, the receiving end determines the decoding strategy instruction that matches the business requirement information based on the rule lookup table, and adjusts the configuration of the neural decoding network according to the decoding strategy instruction.

[0095] In one possible implementation, the decoding strategy instructions include at least one of hierarchical scheduling instructions, iterative control instructions, and precision control instructions.

[0096] like Figure 8 As shown, the sending end sends Service Request Information (SRI) to the receiving end. The receiving end has a pre-built rule lookup table containing the mapping relationship between different SRIs and decoding strategy instructions. For example, this rule lookup table can be obtained by offline training using a lightweight neural network, which can be deployed inside the receiving end.

[0097] After parsing the received SRI to obtain the power requirements, reliability requirements, and throughput requirements of the current service, the receiving end looks up the rule lookup table to obtain the decoding strategy instruction that matches the received SRI. The decoding strategy instruction includes at least one of the following: hierarchical scheduling instruction, iterative control instruction, and precision control instruction.

[0098] (1) Layer scheduling strategy

[0099] The hierarchical scheduling instruction is used to indicate the update ratio K of the critical layer. For example... Figure 8 As shown, the receiver evaluates the importance of each layer in the neural decoding network. Based on the hierarchical scheduling instructions, it only updates messages for the top K layers ranked from highest to lowest importance, instead of a fixed full-layer update pattern, thus reducing computational overhead. For example, when channel quality is good or the traffic load is heavy, updates to layers that contribute minimally to error correction can be ignored, reducing the receiver's power consumption.

[0100] The importance of each layer in a neural decoding network can be assessed using the log likelihood ratio (LLR) residual values ​​of each layer. The LLR residual measures which layers in a neural decoding network need parameter updates. A complete decoding process involves multiple iterations; the LLR residual of a layer represents the average change in the LLR of all nodes within that layer after a particular iteration compared to before the iteration. The LLR residual measures how much the confidence of all variable nodes within a layer changes after a complete message passing. A larger LLR residual indicates a greater contribution of that layer to the convergence of the decoding process; information is being rapidly exchanged and corrected, and a stable state has not yet been reached. Updating this layer will significantly contribute to the convergence of the entire decoding process. Conversely, a smaller LLR residual indicates that the iteration of that layer has essentially converged, and continuing to update that layer is inefficient or even ineffective.

[0101] (2) Iteration control strategy

[0102] The maximum number of iterations for a neural decoding network can be dynamically set using an iterative control strategy. I max Early stop threshold S min The neural decoding network has reached [number] iterations. I max Then the iteration stops. The decoding metric reaches the preset early stopping threshold. S min This indicates that the decoding process has converged to an acceptable level of semantic quality, at which point the iteration can be terminated early to reduce computational overhead.

[0103] For HRLLC business, the adjustment can be increased. Imax and raise the early stop threshold S min This is to ensure decoding accuracy, thus guaranteeing the ultra-high reliability requirements of the service. For eMBB services, the requirements can be appropriately reduced. I max An efficiency-first strategy should be adopted. For semantic communication-related services, costs can be appropriately reduced. S min The decoding index reached S min Iteration can be stopped early without waiting for the number of iterations to be reached. I max Then stop, which not only ensures the correct recovery of semantic information, but also significantly improves decoding efficiency and the real-time performance of semantic communication.

[0104] (3) Precision control strategy

[0105] In an exemplary embodiment, the precision control strategy refers to the receiving end adjusting the information quantization bit width during node updates based on the SRI.

[0106] like Figure 8 As shown, the receiving end can select the appropriate information bit width configuration according to the actual needs of the current business. For example, in mMTC scenarios, devices typically transmit small data packets (such as sensor data), have low requirements for quantization accuracy, and support massive device access, requiring high throughput. In this scenario, the decoder needs to process a large amount of low-rate data simultaneously. Reducing the information quantization bit width (e.g., from 6 bits to 4 bits) can reduce the circuit flip-flop rate of the terminal device, thereby reducing the overall system power consumption without sacrificing throughput. This achieves a flexible trade-off between performance and energy efficiency, and is especially suitable for power-sensitive IoT terminals.

[0107] In one possible implementation, the rule lookup table can be obtained by offline training on the training set data using a lightweight neural network. For example, the lightweight neural network and the decoding network can be trained jointly, or the decoding network can be trained first to obtain its parameters before training the lightweight neural network. This application does not limit the training method.

[0108] The example of training a neural decoding network first to determine the network parameters and then training a lightweight neural network is used for illustration. The training set data includes business requirement information (such as power requirements, reliability requirements, and throughput requirements) and the corresponding information to be decoded.

[0109] The training set data is input into a lightweight neural network, while the information to be decoded is input into a neural decoding network. The lightweight neural network outputs corresponding decoding strategy instructions based on the input SRI and passes them to the neural decoding network. The neural decoding network adjusts its decoding strategy configuration according to the decoding strategy instructions, then decodes the input information to be decoded and outputs the decoding prediction result. The decoding index is calculated based on the decoding prediction result and the actual decoding result, and it is determined whether the index meets the business requirements corresponding to the current service. If it does, the trained lightweight neural network is obtained. If the calculated decoding index does not meet the business requirements, the network parameters of the lightweight neural network are adjusted, and the above process is repeated until the decoding index meets the business requirements corresponding to the current service, resulting in a trained lightweight neural network. Further, the SRIs corresponding to each service are input into the trained lightweight neural network to obtain a rule lookup table.

[0110] S203, the transmitting end sends a service signal to the receiving end. The corresponding receiving end receives the service signal.

[0111] The transmitting end encodes the service data from the upper layer using an encoder, the modulator modulates the encoded data, and finally the transceiver sends the modulated service signal.

[0112] S204, the receiving end inputs the decoding information corresponding to the service signal into the adjusted and configured neural decoding network to obtain the decoding result.

[0113] The receiving end uses a transceiver to receive the service signals transmitted through the channel, and a demodulator to demodulate the received service signals to obtain demodulated data (or information to be decoded). Further, the information to be decoded is input into a neural decoding network with adjusted configuration for decoding, and finally, the decoding result is output.

[0114] The decoding method provided in this embodiment allows the sending end to send service requirement information of the current transmission service to the receiving end. The receiving end searches for a matching decoding strategy instruction based on a preset rule lookup table and transmits it to the neural decoding network. The neural decoding network adjusts its network configuration based on the decoding strategy instruction; for example, it ignores information updates of non-critical layers when the channel quality is good, and reduces the information quantization bit width in low-power scenarios. Then, it processes the input information to be decoded and outputs the final decoding result. It is evident that this scheme can achieve proactive matching between physical layer resources and service requirements, avoiding resource mismatch caused by passive and lagging strategy adjustments. Moreover, this scheme introduces a multi-dimensional decoding control strategy to dynamically adjust the network configuration, achieving a fine-grained trade-off between power consumption, latency, and throughput while maintaining essentially lossless performance.

[0115] like Figure 9As shown, this application embodiment also provides another decoding method embodiment. In this embodiment, the receiving end is deployed with a lightweight neural network, through which the fine-tuning factor and rule lookup table corresponding to the business requirement information can be obtained. Figure 9 As shown, the method may include the following steps:

[0116] S301, the sending end sends the first message. The corresponding receiving end receives the first message.

[0117] The first message includes the current service requirements. This first message can be a DCI message, and the service requirements information may include power requirements, reliability requirements, and throughput requirements. (See also...) Figure 4 The relevant description of S101 in the illustrated embodiment.

[0118] S302, the receiver obtains the fine-tuning factor based on the service requirement information and the local channel state information, and uses the fine-tuning factor to correct the base weight of the neural network decoder.

[0119] The revised weights are used to scale the magnitude of the messages passed from the verification node to the variable node.

[0120] The implementation process for this step can be found in [link to relevant documentation]. Figure 4 The relevant description of S102 in the illustrated embodiment will not be repeated here.

[0121] S303, the receiving end determines the decoding strategy instruction that matches the business requirement information based on the rule lookup table, and adjusts the configuration of the neural decoding network according to the decoding strategy instruction.

[0122] In one possible implementation, such as Figure 10 As shown, the fine-tuning factor and decoding strategy instructions can be obtained through the same lightweight neural network.

[0123] For example, the lightweight neural network may include a fine-tuning subnetwork and a policy adjustment subnetwork. The fine-tuning subnetwork outputs a corresponding fine-tuning factor based on service requirement information and channel state information; the policy adjustment subnetwork outputs a corresponding decoding policy instruction based on the service requirement information. This lightweight neural network can be trained offline. It can be jointly trained with a neural decoding network, or the neural decoding network can be trained first, and then the lightweight neural network can be trained using the neural decoding network; this application does not limit this approach.

[0124] The service requirement information sent by the transmitter is input into a lightweight neural network. The fine-tuning sub-network outputs a fine-tuning factor based on the service requirement information and the channel state information measured by the receiver. The policy adjustment sub-network outputs a corresponding decoding policy instruction based on the service requirement information. The lightweight neural network then transmits the output decoding policy instruction and fine-tuning factor to the neural decoding network, enabling the neural decoding network to adjust its decoding policy according to the corresponding decoding policy instruction.

[0125] In one possible implementation, the decoding strategy instructions include at least one of hierarchical scheduling instructions, iterative control instructions, and precision control instructions. The precision control parameters include the information quantization bit width configuration; furthermore, a fine-tuning factor can be passed as a precision control parameter to the neural decoding network.

[0126] Neural decoding networks use fine-tuning factors to adjust baseline weights. These adjusted weights are then used to scale the magnitude of the messages passed from the check nodes to the variable nodes. (See also...) Figure 4 The relevant descriptions in S102 of the illustrated embodiment will not be repeated here.

[0127] The process by which neural decoding networks adjust their decoding configuration using hierarchical scheduling instructions, iterative control instructions, and precision control instructions can be found in [link to relevant documentation]. Figure 7 The relevant descriptions in S202 of the illustrated embodiment will not be repeated here.

[0128] S304, the transmitting end sends a service signal to the receiving end. The corresponding receiving end receives the service signal.

[0129] The transmitting end encodes the service data using an encoder, the modulator modulates the encoded data, and finally the transceiver transmits the modulated service signal.

[0130] S305, the receiver inputs the information to be decoded, the fine-tuning factor, and the decoding strategy instructions into the neural decoding network to obtain the decoding result.

[0131] The receiving end uses a transceiver to receive the service signals transmitted through the channel, and a demodulator to demodulate the received service signals to obtain the demodulated data (or information to be decoded). The information to be decoded, the fine-tuning factor, and the decoding strategy instructions are input into a neural decoding network for processing, and finally the decoding result is output.

[0132] The decoding method provided in this embodiment allows the sending end to send service requirement information of the current transmission service to the receiving end. The receiving end inputs the service requirements and local channel state into the fine-tuning network to output a fine-tuning factor, and outputs corresponding decoding strategy instructions based on the service requirements. The fine-tuning factor and decoding strategy instructions are passed to the neural decoding network for network configuration adjustment, realizing proactive matching between physical layer resources and upper layer service requirements. Moreover, this scheme introduces a multi-dimensional decoding control strategy to dynamically adjust the network configuration, achieving a fine-grained trade-off between power consumption, latency, and throughput while maintaining essentially lossless performance.

[0133] It should be understood that Figures 1 to 10 The flowcharts or scene diagrams shown are for illustrative purposes only and are not intended to limit the embodiments of this application to the examples illustrated. In fact, those skilled in the art can interpret the embodiments based on... Figures 1 to 10 The examples in the document can be transformed into equivalent ways to obtain more implementations.

[0134] The above text combined Figures 1 to 10 This document describes in detail the communication method provided in the embodiments of this application. The following will combine... Figures 11 to 12 The device embodiments of this application are described in detail below. It should be understood that the communication device of this application embodiment can execute the various communication methods of the foregoing embodiments of this application, that is, the specific working processes of the various products below can be referred to the corresponding processes in the foregoing method embodiments.

[0135] In the embodiments described above, the terminal device may execute some or all of the steps in each embodiment; the network device may execute some or all of the steps in each embodiment. These steps or operations are merely examples, and the embodiments of this application may also perform other operations or variations thereof. Furthermore, the steps may be executed in different orders as presented in the embodiments, and it is not necessary to execute all the operations in the embodiments of this application. Moreover, the sequence number of each step does not imply the order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0136] Figure 11 This is a schematic block diagram of a communication device provided in an embodiment of this application. Figure 11 As shown, the communication device may include a communication module 102. The communication module 102 can implement corresponding communication functions, which can be internal communication functions of the communication device or communication functions between the communication device and other devices. Optionally, the communication module 102 may also be referred to as a communication interface or transceiver module. Optionally, the communication device further includes a processing module 101. The processing module 101 can implement corresponding processing functions.

[0137] Optionally, the communication device further includes a storage module, which can be used to store instructions and / or data; the processing module 101 can read the instructions and / or data in the storage module so that the communication device can implement the aforementioned method embodiments.

[0138] In one possible design, the communication device may correspond to the receiving end in the above method embodiments. The receiving end may be a terminal device or a component (such as a circuit, chip, or chip system) configured in the terminal device. The communication device may be used to perform the steps or processes executed by the sending end in any of the above method embodiments.

[0139] For example, the communication module 102 is configured to receive a first message from the transmitter, the first message including service requirement information, the service requirement information being used to indicate the current service's requirement for at least one of power consumption, reliability, and throughput; and to receive a service signal from the transmitter.

[0140] The processing module 101 is used to determine a matching fine-tuning factor based on service requirement information and channel state information; and to input the fine-tuning factor and the decoding information corresponding to the service signal into the neural decoding network to obtain the decoding result. The fine-tuning factor is used to correct the baseline weights in the neural decoding network, and the corrected weights are used to scale the message amplitude passed from the verification node to the variable node.

[0141] In one possible implementation, the business requirement information includes power consumption requirement field, reliability requirement field, and throughput requirement field.

[0142] In one possible implementation, the power consumption requirement field occupies 2 bits, the reliability requirement field occupies 2 bits, and the throughput requirement field occupies 1 bit.

[0143] In one possible implementation, the power consumption requirement field has a first value to indicate that the current service has low power consumption requirements; the power consumption requirement field has a second value to indicate that the current service is not sensitive to power consumption; and the power consumption requirement field has a third or fourth value to indicate that the current service has moderate power consumption sensitivity.

[0144] In one possible implementation, the reliability requirement field has a first value to indicate that the current service is an ultra-high reliability service; a second value to indicate that the current service is a low reliability service; a third value to indicate that the current service is a high reliability service; and a fourth value to indicate that the current service is a standard reliability service.

[0145] In one possible implementation, the throughput requirement field has a first value to indicate that the current service is a high-throughput service; the throughput requirement field has a second value to indicate that the current service is a low-throughput service.

[0146] In one possible implementation, before inputting the decoding information corresponding to the fine-tuning factor and the service signal into the neural decoding network to obtain the decoding result, the processor is further configured to: determine a configuration strategy that matches the service requirement information based on a first mapping relationship; and adjust the configuration parameters of the neural decoding network based on the configuration strategy, wherein the configuration parameters include at least one of hierarchical scheduling, iterative control, and precision control.

[0147] In one possible implementation, when the processor adjusts the configuration parameters of the neural decoding network based on a configuration strategy, it specifically adjusts the key layer update ratio of the neural decoding network based on a hierarchical scheduling instruction. The key layer update ratio is used to determine that in each iteration, only the first few layers with the largest error correction contribution, arranged in descending order, are selected for message updates.

[0148] In one possible implementation, when the processor adjusts the configuration parameters of the neural decoding network based on a configuration strategy, it specifically adjusts the maximum number of iterations and / or the early stopping threshold of the neural decoding network based on iteration control instructions.

[0149] In one possible implementation, when the processor adjusts the configuration parameters of the neural decoding network based on a configuration strategy, it specifically adjusts the information quantization bit width during node updates in the neural decoding network based on precision control instructions.

[0150] The above are merely examples; for detailed steps or procedures, please refer to the descriptions in the foregoing embodiments.

[0151] In one possible design, the communication device may correspond to the receiving end in the above method embodiments. The receiving end may be a terminal device, or a component (such as a circuit, chip, or chip system) configured in the terminal device. The communication device may be used to perform the steps or processes performed by the terminal device in any of the above method embodiments.

[0152] For example, the communication module 102 is configured to receive a first message from the sending end, the first message including service requirement information, the service requirement information being used to indicate the current service's requirement for at least one of power consumption, reliability, and throughput; and to receive a service signal from the sending end.

[0153] The processing module 101 is used to determine a configuration strategy that matches the business requirement information based on the first mapping relationship; adjust the configuration parameters of the neural decoding network based on the configuration strategy, the configuration parameters including at least one of the following: key layer update ratio, maximum number of iterations and early stop threshold, and precision control parameters; input the information to be decoded corresponding to the business signal into the neural decoding network after configuration adjustment, and obtain the decoding result.

[0154] In one possible implementation, the business requirement information includes power consumption requirement field, reliability requirement field, and throughput requirement field.

[0155] In one possible implementation, the power consumption requirement field occupies 2 bits, the reliability requirement field occupies 2 bits, and the throughput requirement field occupies 1 bit.

[0156] In one possible implementation, the configuration strategy includes at least one of hierarchical scheduling instructions, iterative control instructions, and precision control instructions.

[0157] In one possible implementation, the configuration parameters of the neural decoding network are adjusted based on a configuration strategy, including: adjusting the critical layer update ratio of the neural decoding network based on hierarchical scheduling instructions. The critical layer update ratio is used to determine that in each iteration, only the top critical layer update ratios, arranged from largest to smallest error correction contribution, are selected for message updates.

[0158] In one possible implementation, the configuration parameters of the neural decoding network are adjusted based on a configuration strategy, including adjusting the maximum number of iterations and / or the early stopping threshold of the neural decoding network based on iteration control instructions.

[0159] In one possible implementation, the configuration parameters of the neural decoding network are adjusted based on a configuration strategy, including adjusting the information quantization bit width when nodes in the neural decoding network are updated based on precision control instructions.

[0160] The above are merely examples; for detailed steps or procedures, please refer to the descriptions in the foregoing embodiments.

[0161] In one possible design, the communication device may correspond to the transmitting end in the above method embodiments. The transmitting end may be a network device, or a component (such as a circuit, chip, or chip system) configured in a network device. The communication device may be used to perform the steps or processes executed by the network device in any of the above method embodiments.

[0162] For example, the communication module 102 is used to send a first message to the receiving end. The first message includes service requirement information, which indicates the current service's requirement for at least one of power consumption, reliability, and throughput. The service requirement information is used to enable the receiving end to obtain a fine-tuning factor that matches the service requirement information and channel state information. The fine-tuning factor is used to correct the baseline weights in the neural decoding network. The corrected weights are used to scale the message amplitude passed from the verification node to the variable node.

[0163] In one possible implementation, the business requirement information includes power consumption requirement field, reliability requirement field, and throughput requirement field.

[0164] In one possible implementation, the power consumption requirement field occupies 2 bits, the reliability requirement field occupies 2 bits, and the throughput requirement field occupies 1 bit.

[0165] In one possible implementation, the power consumption requirement field is a first value to indicate that the current service has low power consumption requirements; the power consumption requirement field is a second value to indicate that the current service is not sensitive to power consumption; and the power consumption requirement field is a third or fourth value to indicate that the current service has moderate power consumption sensitivity.

[0166] In one possible implementation, the reliability requirement field has a first value to indicate that the current service is an ultra-high reliability service; a second value to indicate that the current service is a low reliability service; a third value to indicate that the current service is a high reliability service; and a fourth value to indicate that the current service is a standard reliability service.

[0167] In one possible implementation, the throughput requirement field has a first value to indicate that the current service is a high-throughput service; the throughput requirement field has a second value to indicate that the current service is a low-throughput service.

[0168] In one possible implementation, the service requirement information is also used to enable the receiver to obtain configuration instructions for the neural decoding network, the configuration instructions including at least one of hierarchical scheduling instructions, iterative control instructions, and precision control instructions.

[0169] In one possible implementation, the hierarchical scheduling instruction is used to enable the receiver to adjust the critical layer update ratio of the neural decoding network. The critical layer update ratio is used to determine that in each iteration, only the top critical layer update ratios, arranged from largest to smallest error correction contribution, are selected for message updates. The iteration control instruction is used to enable the receiver to adjust the maximum number of iterations and / or the early stopping threshold of the neural decoding network. The precision control instruction is used to enable the receiver to adjust the information quantization bit width when updating nodes in the neural decoding network.

[0170] The above are merely examples; for detailed steps or procedures, please refer to the descriptions in the foregoing embodiments.

[0171] Figure 12 This is another schematic block diagram of the communication device provided in the embodiments of this application. The communication device may be a chip, chip system, or processor, etc., in a terminal device or network device that implements the above-described methods. This communication device can be used to implement the methods described in the above-described method embodiments; for details, please refer to the descriptions in the above-described method embodiments.

[0172] like Figure 12As shown, the communication device may include one or more processors 201, which may also be referred to as processing units or processing modules, and can implement certain control functions. The processor 201 may be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, while the central processing unit can be used to control the communication device (e.g., base station, baseband chip, user, user chip), execute software programs, and process data from the software programs.

[0173] In an alternative design, the processor 201 may also store instructions and / or data that can be executed by the processor 201 to cause the communication device to perform the methods described in the above method embodiments.

[0174] In another alternative design, the communication device may include a communication interface 202 for implementing receiving and transmitting functions. For example, the communication interface 202 may be a transceiver circuit, interface, interface circuit, or transceiver. The transceiver circuit, interface, interface circuit, or transceiver for implementing receiving and transmitting functions may be separate or integrated. The aforementioned transceiver circuit, interface, interface circuit, or transceiver may be used for reading and writing code / data, or it may be used for transmitting or relaying signals.

[0175] Optionally, the communication device may include one or more memories 203, which may store instructions that can be executed on the processor 201, causing the communication device to perform the methods described in the above method embodiments. Optionally, the memories 203 may also store data. Optionally, the processor 201 may also store instructions and / or data. The processor 201 and the memories 203 may be provided separately or integrated together.

[0176] It should be understood that, in one possible design, the steps in the method embodiments provided in this application can be implemented by integrated logic circuits in the processor's hardware or by instructions in software form. The steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are not provided here.

[0177] In one implementation, the communication device may correspond to the terminal device in the above method embodiments and may be used to execute the various steps and / or processes executed by the terminal device in the above method embodiments. The processor 201 may be used to execute instructions stored in the memory 203, and when the processor 201 executes the instructions stored in the memory, the processor 201 is used to execute the various steps and / or processes of the above method embodiments corresponding to the terminal device.

[0178] In another implementation, the communication device may correspond to the network device in the above method embodiments and may be used to execute the various steps and / or processes executed by the network device in the above method embodiments. The processor 201 may be used to execute instructions stored in the memory 203, and when the processor 201 executes the instructions stored in the memory, the processor 201 is used to execute the various steps and / or processes of the above method embodiments corresponding to the network device.

[0179] It should be understood that the aforementioned processing device can be one or more chips. For example, the processing device can be a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a system-on-chip (SoC), a central processor unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips.

[0180] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. 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 linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0181] According to the method provided in the embodiments of this application, this application also provides a chip system, which includes one or more processors for calling and executing instructions stored in memory, thereby causing the method described in the embodiments of this application to be executed. The chip system may be composed of chips or may include chips and other discrete devices.

[0182] The chip system may include input circuits or interfaces for transmitting information or data, and output circuits or interfaces for receiving information or data.

[0183] According to the method provided in the embodiments of this application, this application also provides a communication system, which includes the aforementioned network device and terminal device.

[0184] According to the method provided in the embodiments of this application, this application also provides a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute the various steps or processes executed by the network device or terminal device in any of the foregoing method embodiments.

[0185] According to the method provided in the embodiments of this application, this application also provides a computer-readable storage medium storing program code, which, when run on a computer, causes the computer to execute the various steps or processes executed by the network device or terminal device in any of the foregoing method embodiments.

[0186] The computer-readable storage medium may be the aforementioned volatile memory or non-volatile memory, or it may include both volatile memory and non-volatile memory.

[0187] In the embodiments of this application, the terms and English abbreviations are exemplary examples given for ease of description and should not be construed as limiting the application in any way. This application does not preclude the possibility of defining other terms that can achieve the same or similar functions in existing or future agreements.

[0188] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated.

[0189] 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 through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0190] It should be understood that in the various embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0191] In summary, the above description is merely a preferred embodiment of the technical solution of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A decoding method, characterized in that, The method, executed by the receiving end, includes: Receive a first message from the sender, the first message including service requirement information, the service requirement information being used to indicate the current service’s requirement for at least one of power consumption, reliability and throughput, the service requirement information including a power consumption requirement field, a reliability requirement field and a throughput requirement field; The service requirement information and channel state information are input into the fine-tuning network to obtain a matching fine-tuning factor, which is obtained through offline training. Receive service signals from the transmitting end; The fine-tuning factor and the decoding information corresponding to the service signal are input into the neural decoding network to obtain the decoding result. The fine-tuning factor is used to correct the baseline weights corresponding to the variable nodes and verification nodes in the neural decoding network. The corrected weights of the verification nodes are substituted into the verification node update formula to scale the message amplitude passed from the verification node to the variable node.

2. The method according to claim 1, characterized in that, The power consumption requirement field occupies 2 bits, the reliability requirement field occupies 2 bits, and the throughput requirement field occupies 1 bit.

3. The method according to claim 2, characterized in that: The power consumption requirement field is set to a first value to indicate that the current service has a low power consumption requirement. The power consumption requirement field is a second value used to indicate that the current service is not sensitive to power consumption. The power consumption requirement field is a third or fourth value used to indicate that the current service has a moderate sensitivity to power consumption; The reliability requirement field is a first value used to indicate that the current service is an ultra-high reliability service; The reliability requirement field is a second value used to indicate that the current service is a low-reliability service; The reliability requirement field is a third value used to indicate that the current service is a high-reliability service; The reliability requirement field is the fourth value used to indicate that the current service is a standard reliable service; The throughput requirement field is a first value used to indicate that the current service is a high-throughput service; The throughput requirement field is a second value used to indicate that the current service is a non-high throughput service.

4. The method according to any one of claims 1-3, characterized in that, Before inputting the fine-tuning factor and the decoding information corresponding to the service signal into the neural decoding network to obtain the decoding result, the method further includes: Based on the first mapping relationship, a configuration strategy matching the business requirement information is determined; The configuration parameters of the neural decoding network are adjusted based on the configuration strategy, and the configuration parameters include at least one of hierarchical scheduling, iterative control, and precision control.

5. The method according to claim 4, characterized in that, The adjustment of the configuration parameters of the neural decoding network based on the configuration strategy includes: The key layer update ratio of the neural decoding network is adjusted based on the hierarchical scheduling instruction. The key layer update ratio is used to determine that in each iteration, only the first key layer update ratios, arranged from largest to smallest in error correction contribution, are selected for message updates. The maximum number of iterations and / or the early stopping threshold of the neural decoding network are adjusted based on iterative control instructions. The information quantization bit width during node updates in the neural decoding network is adjusted based on precision control instructions.

6. A decoding method, characterized in that, The method, executed by the receiving end, includes: Receive a first message from the sender, the first message including service requirement information, the service requirement information being used to indicate the current service’s requirement for at least one of power consumption, reliability and throughput, the service requirement information including a power consumption requirement field, a reliability requirement field and a throughput requirement field; Based on the first mapping relationship, a configuration strategy matching the business requirement information is determined. The configuration strategy includes at least one of hierarchical scheduling instructions, iterative control instructions, and precision control instructions. The first mapping relationship includes the mapping relationship between different business requirement information and the configuration strategy, and is obtained by offline training of a lightweight neural network. The configuration parameters of the neural decoding network are adjusted based on the configuration strategy. The configuration parameters include at least one of the following: critical layer update ratio, maximum number of iterations and early stopping threshold, and precision control parameters. Receive service signals from the transmitting end; The decoding information corresponding to the service signal is input into the configured and adjusted neural decoding network to obtain the decoding result.

7. The method according to claim 6, characterized in that, The power consumption requirement field occupies 2 bits, the reliability requirement field occupies 2 bits, and the throughput requirement field occupies 1 bit.

8. The method according to claim 6 or 7, characterized in that, The adjustment of the configuration parameters of the neural decoding network based on the configuration strategy includes: The key layer update ratio of the neural decoding network is adjusted based on the hierarchical scheduling instruction. The key layer update ratio is used to determine that in each iteration, only the first key layer update ratios, arranged from largest to smallest in error correction contribution, are selected for message updates. The maximum number of iterations and / or the early stopping threshold of the neural decoding network are adjusted based on iterative control instructions. The information quantization bit width during node updates in the neural decoding network is adjusted based on precision control instructions.

9. A decoding method, characterized in that, The method, executed by the sending end, includes: A first message is sent to the receiving end. The first message includes service requirement information, which indicates the current service's requirement for at least one of power consumption, reliability, and throughput. The service requirement information includes a power consumption requirement field, a reliability requirement field, and a throughput requirement field. The service requirement information enables the receiving end to obtain a fine-tuning factor that matches the service requirement information and channel state information. The fine-tuning factor is used to correct the baseline weights corresponding to the variable nodes and verification nodes in the neural decoding network. The corrected weights of the verification nodes are substituted into the verification node update formula to scale the message amplitude passed from the verification node to the variable nodes.

10. The method according to claim 9, characterized in that, The power consumption requirement field occupies 2 bits, the reliability requirement field occupies 2 bits, and the throughput requirement field occupies 1 bit.

11. The method according to claim 10, characterized in that: The power consumption requirement field is set to a first value to indicate that the current service has a low power consumption requirement. The power consumption requirement field is a second value used to indicate that the current service is not sensitive to power consumption. The power consumption requirement field is a third or fourth value used to indicate that the current service has a moderate sensitivity to power consumption; The reliability requirement field is a first value used to indicate that the current service is an ultra-high reliability service; The reliability requirement field is a second value used to indicate that the current service is a low-reliability service; The reliability requirement field is a third value used to indicate that the current service is a high-reliability service; The reliability requirement field is the fourth value used to indicate that the current service is a standard reliable service; The throughput requirement field is a first value used to indicate that the current service is a high-throughput service; The throughput requirement field is a second value used to indicate that the current service is a non-high throughput service.

12. The method according to any one of claims 9-11, characterized in that, The service requirement information is also used to enable the receiving end to obtain configuration instructions for the neural decoding network. The configuration instructions include at least one of hierarchical scheduling instructions, iterative control instructions, and precision control instructions.

13. The method according to claim 12, characterized in that: The hierarchical scheduling instruction is used to enable the receiver to adjust the key layer update ratio of the neural decoding network. The key layer update ratio is used to determine that in each iteration, only the first key layer update ratios, arranged from largest to smallest error correction contribution, are selected for message updates. The iteration control command is used to enable the receiving end to adjust the maximum number of iterations and / or the early stopping threshold of the neural decoding network; The precision control command is used to enable the receiver to adjust the information quantization bit width when updating nodes in the neural decoding network.

14. A communication device, characterized in that, The device includes at least one processor coupled to a memory storing a program or instructions, the processor executing the program or instructions to cause the device to perform the method as described in any one of claims 1 to 13.

15. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed, they cause the computer to perform the method as described in any one of claims 1 to 13.

Images