Optical-electric hybrid ultra-low delay communication method and system

Abstract

The invention provides an optical-electric hybrid ultra-low time delay communication method and system, and relates to the technical field of communication, and the method comprises the steps: outputting a global synchronous clock signal with clock jitter less than 1ppm, dynamically dividing micro time slot resources in an orthogonal frequency division multiplexing symbol period, and generating a dynamically adjusted micro time slot resource distribution result; calculating an optimal phase offset matrix of the metasurface intelligent reflecting surface through a depth deterministic strategy gradient algorithm to obtain an optimized electromagnetic wave propagation path; generating a global optimization check matrix by aggregating the locally trained lightweight error correction model gradient of each node to obtain a compensated data stream; and constructing a causal graph model dynamic pruning high-entropy path to minimize causal entropy, through multi-agent reinforcement learning, taking time delay-energy efficiency as a game target to decide an optimal modulation order and a subcarrier switching strategy, and obtaining an optimized stable communication link. According to the invention, high-reliability and low-delay communication basic support is provided for high-precision intelligent manufacturing.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an optical-electrical hybrid ultra-low latency communication method and system. Background Art

[0002] With the rapid development of industrial automation and intelligent manufacturing, hybrid optical-electronic communication networks face stringent latency and reliability requirements in scenarios such as collaborative control of factory equipment and real-time data transmission. Conventional hybrid optical-electronic systems often employ fixed spectrum allocation, static equalizers, and half-duplex communication modes, making them difficult to cope with dynamic electromagnetic interference and multipath effects in complex industrial environments. For example, broadband interference sources such as arc welders and frequency converters within smart factories generate transient noise at the optoelectronic converter, simultaneously causing co-channel interference on the subcarriers of the 37 GHz millimeter wave air interface, resulting in a sharp drop in the subcarrier signal-to-noise ratio and an increase in the bit error rate. The delay spread caused by millimeter wave multipath propagation further exacerbates timing chaos, causing the end-to-end latency of robot collaborative control commands to fluctuate by more than milliseconds, severely impacting machining accuracy and system stability.

[0003] To address the above issues, existing solutions mostly rely on adding redundant optical fiber wavelengths, expanding millimeter-wave bandwidth, or deploying static reflective surfaces. However, these methods have defects such as low spectrum utilization, high hardware costs, and poor dynamic adaptability. Static reflective surfaces cannot adjust the beam direction in real time, resulting in high path loss. Traditional redundant retransmission requires full retransmission of data, which significantly increases latency and energy consumption, making it difficult to meet the sub-millisecond real-time requirements of industrial scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide an optical-electrical hybrid ultra-low latency communication method and system to improve the above-mentioned problems. To achieve the above-mentioned purpose, the technical solution adopted by the present invention is as follows: In a first aspect, the present application provides an optical-electrical hybrid ultra-low latency communication method, comprising: A clock signal source is obtained and the clock synchronization channel is encrypted using a quantum key distribution protocol. The random phase noise generated by the quantum entropy source is used as the observation noise input to perform Kalman filtering correction on the phase estimate of the classical clock source. This outputs a globally synchronized clock signal with a clock jitter of less than 1 ppm, and a spatiotemporal hash grid is constructed based on the globally synchronized clock signal. The optical-electrical hybrid data stream is segmented into quantized data blocks according to the coordinates of the spatiotemporal hash grid. Each quantized data block is bound to a unique spatiotemporal hash value and a priority tag, with the robot joint control instructions marked as level L0, the highest priority. Based on the quantized data blocks of the spatiotemporal hash grid and real-time channel state information, a spatiotemporal convolutional neural network is used to predict the distribution of spectrum holes within the next 10ms. Micro-slot resources within the orthogonal frequency division multiplexing symbol period are dynamically allocated, L0-level instructions are preferentially mapped to low-interference subcarriers, and polar codes are used to perform asymmetric redundant encoding on micro-slot data units to generate dynamically adjusted micro-slot resource allocation results. Based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, the optimal phase offset matrix of the metasurface intelligent reflector is calculated through a deep deterministic policy gradient algorithm. The electromagnetic wave propagation path is dynamically reconstructed, and full-duplex self-interference cancellation technology is integrated to achieve concurrent uplink and downlink transmission in the same frequency band, thereby obtaining the optimized electromagnetic wave propagation path. Based on the optimized electromagnetic wave propagation path, the channel impulse response and bit error rate (BER) obtained at the receiving feedback end are then received. A time-frequency dual-domain joint equalizer is used to collaboratively compensate for multipath fading and narrowband interference. If the BER exceeds a preset threshold, an incremental redundancy retransmission mechanism based on federated learning is triggered. This mechanism generates a globally optimized parity check matrix by aggregating the gradients of the lightweight error correction model trained locally on each node. Only the sparse parity bits of the damaged mini-slot are retransmitted to obtain the compensated data stream. Based on the compensated data stream, real-time device temperature, and subcarrier load rate data obtained from transceiver statistics, a causal graph model is constructed to dynamically prune high-entropy paths to minimize causal entropy. Multi-agent reinforcement learning is used to determine the optimal modulation order and subcarrier switching strategy with delay-energy efficiency as the game objective. A phase change material heat dissipation module is embedded to regulate the optical module driving current, resulting in an optimized stable communication link.

[0005] Preferably, the clock signal source is obtained, and the clock synchronization channel is encrypted through a quantum key distribution protocol; the random phase noise generated by the quantum entropy source is used as the observation noise input, and the phase estimate of the classical clock source is corrected by Kalman filtering, and a global synchronized clock signal with a clock jitter of less than 1ppm is output, and a spatiotemporal hash grid is constructed based on the global synchronized clock signal; the optical-electrical hybrid data stream is divided into quantized data blocks according to the coordinates of the spatiotemporal hash grid, wherein each quantized data block is bound to a unique spatiotemporal hash value and a priority label, wherein the robot joint control instruction is marked as the highest priority level L0, including: Based on the factory local area network infrastructure, it acquires GPS pulse-second signals, receives IEEE 1588 protocol synchronization messages, and generates a quantum random number entropy source through a quantum random number generator based on an optical quantum noise chip. It then encrypts the clock synchronization channel through the quantum key distribution protocol to generate a quantum key stream. Based on the quantum key stream, the random phase noise of the quantum entropy source is integrated with the phase of the classical clock source, and the quantum Kalman filter algorithm is used to eliminate clock drift and generate a globally synchronized clock signal with a clock jitter of less than 1ppm. A spatiotemporal hash grid is constructed based on the global synchronous clock signal to obtain the coordinates of the spatiotemporal hash grid. The optical-electrical hybrid data stream is then divided into quantized data blocks. Each data block is bound to a unique spatiotemporal hash value and a priority label, where the robot joint control instructions are marked as the highest priority level L0.

[0006] Preferably, the quantized data blocks and real-time channel state information of the space-time hash grid are used to predict the spectrum hole distribution within the next 10ms through a space-time convolutional neural network, dynamically divide the micro-slot resources within the orthogonal frequency division multiplexing symbol period, preferentially map the L0 level instructions to low-interference subcarriers, and use polar codes to perform asymmetric redundant encoding on the micro-slot data units to generate a dynamically adjusted micro-slot resource allocation result, which includes: Based on the spatiotemporal hash values ​​and priority tags of quantized data blocks, combined with historical spectrum occupancy and real-time channel status information, a spatiotemporal convolutional neural network is used to predict the distribution of spectrum holes within the next 10ms. Dynamically divide OFDM symbol periods based on predicted spectrum hole distribution The duration of each mini-slot is Δ t =1μs, bandwidth is Δ f =4.3125kHz, preferentially mapping L0 level instructions to low-interference subcarriers, where the low-interference subcarriers meet the signal-to-noise ratio threshold greater than or equal to 25dB, and generating a mini-slot resource allocation table; Based on the mini-slot resource allocation table, polarization code encoding is performed on the mini-slot data unit.

[0007] Preferably, the real-time electromagnetic interference heat map generated by the micro-slot resource allocation results and distributed sensors calculates the optimal phase offset matrix of the metasurface intelligent reflector through a deep deterministic policy gradient algorithm, dynamically reconstructs the electromagnetic wave propagation path, and integrates full-duplex self-interference elimination technology to achieve concurrent uplink and downlink transmission in the same frequency band, thereby obtaining an optimized electromagnetic wave propagation path, which includes: A deep deterministic policy gradient model is established based on the micro-slot resource allocation results and a real-time electromagnetic interference heat map generated by distributed spectrum sensors based on electromagnetic interference intensity. The micro-slot resource allocation results and the real-time electromagnetic interference heat map are input into the deep deterministic policy gradient model, which outputs the phase offset matrix of the metasurface smart reflector. The training objective of the deep deterministic policy gradient model is to minimize path loss. By iteratively optimizing the phase offset of the reflector unit, the total path loss after the channel response of the reflected path and the direct path is minimized. Based on the phase offset matrix, the phase offset of the metasurface's intelligent reflector is dynamically adjusted to steer the electromagnetic wave propagation path around the interference area, resulting in an optimized path. Based on the optimized path, a minimum mean square error (MMSE) algorithm is used to cancel the self-interference of uplink and downlink signals in the same frequency band. This involves constructing the autocorrelation matrix of the self-interference signal based on the known signal waveform and channel impulse response at the transmitter, and solving the optimal interference cancellation weight matrix under the minimum mean square error criterion by maximizing the signal-to-interference-noise ratio. The optimal interference cancellation weight matrix is ​​applied to filter the received signal to suppress the self-interference component in the same frequency band, thereby outputting the full-duplex concurrent transmission configuration parameters and obtaining the optimized electromagnetic wave propagation path.

[0008] Preferably, the channel impulse response and bit error rate obtained by the receiving feedback end based on the optimized electromagnetic wave propagation path are received, and a time-frequency dual-domain joint equalizer is used to collaboratively compensate for multipath fading and narrowband interference. If the bit error rate exceeds a preset threshold, an incremental redundancy retransmission mechanism based on federated learning is triggered. A globally optimized check matrix is ​​generated by aggregating the gradients of the lightweight error correction model trained locally at each node, and only the sparse check bits of the damaged mini-time slot are retransmitted to obtain a compensated data stream, which includes: Based on the optimized phase offset matrix of the metasurface smart reflector in the electromagnetic wave propagation path, the channel impulse response and bit error rate fed back by the receiver, a time-frequency dual-domain joint equalizer is designed. This includes: calculating the time-domain equalization weights by minimizing the inter-symbol interference criterion to compensate for the delay caused by multipath fading; identifying the interference frequency set based on the narrowband interference power spectral density, and calculating the frequency-domain suppression coefficient; and superimposing the time-domain equalization weights with the frequency-domain suppression results to generate an equalized low-error signal stream. If the bit error rate BER>10 -6 , extract the damaged micro-time slot index, identify the fault frequency and time slice, where the communication node performs local training, where the local training includes: using the original data of the damaged micro-time slot and the equalized received signal as input data, training a lightweight error correction model to minimize the reconstruction error and generate a local gradient; Based on the local gradient, it is encrypted and uploaded to the central server through the secure multi-party computing protocol, aggregated to generate the global gradient, update the global check matrix, and only retransmit the sparse check bits of the damaged micro-time slot to obtain the globally optimized check matrix. The globally optimized check matrix and the sparse check bits are then error-corrected and decoded to restore the compensated data stream.

[0009] It should be noted that in this embodiment, the networking architecture of the optical-electrical hybrid communication terminal is as follows: the central office (CO) is located on the service provider side, connected to the backbone network via optical fiber and connected to the customer premises equipment (CPE) via a 37 GHz millimeter wave link, forming a "fiber + millimeter wave" hybrid backhaul. The CPE, deployed on the user side, features a built-in 37 GHz transceiver module and intelligent metasurface reflector, bridging millimeter wave signals to Ethernet / Wi-Fi, supporting line-of-sight communication up to 1 km. Optical fiber aggregates multiple card reader outstations to the CO, providing a high-bandwidth, low-latency backbone channel. The millimeter wave air interface carries data backhaul between the outstations and the CPE, ensuring high-speed access in the last mile. A video compression module can be integrated into the CPE or outstation to perform H.265 / H.266 encoding on the video stream captured by the front-end, significantly reducing air interface bandwidth requirements. Wi-Fi devices access the network through the CPE's LAN port or 2.4 / 5 GHz 2×2 MIMO wireless module, providing flexible coverage for mobile terminals (surveillance devices, handheld PADs, etc.). A star topology is used between the substations and the CPE. Each substation independently manages local access control card swipe and video surveillance data, which is then transmitted back to the CO via millimeter wave links, achieving stable transmission over long distances with low latency.

[0010] In summary, the characteristics of this network are: the fiber optic backbone ensures high bandwidth, the millimeter wave air interface ensures flexible access, and it takes into account both deployment cost and performance. It is suitable for scenarios such as security monitoring, smart buildings, and smart factories. The coverage range and terminal capacity can be flexibly expanded by adding CPE or card reader substations.

[0011] In terms of terminal specifications in this embodiment, the device provides one 37 GHz millimeter-wave air port (industrial connector, supporting beamforming and IRS phase control), four 10 / 100 / 1000Base-T(X) RJ45 network ports (auto-sensing full / half duplex and MDI / MDI-X), one 1000Base-X fiber port (compatible with 100Base-FX, SC / FC optional), and 2.4 / 5 GHz 2×2 MIMO Wi-Fi (802.11a / b / g / n / ac / ax). It operates from a DC 12 V power supply with power consumption of less than 10 W. The operating temperature range is -20°C to +75°C, the storage temperature range is -40°C to +85°C, and the relative humidity range is 5% to 95% (non-condensing). The software supports Chinese and English language switching, status query, basic and advanced network configuration, and device management (upgrade, reboot, and backup).

[0012] In a second aspect, the present application also provides an optical-electrical hybrid ultra-low latency communication system, comprising: Acquisition module: used to obtain the clock signal source and encrypt the clock synchronization channel through the quantum key distribution protocol. The random phase noise generated by the quantum entropy source is used as the observation noise input to perform Kalman filtering correction on the phase estimate of the classical clock source, outputting a globally synchronized clock signal with a clock jitter of less than 1ppm. A spatiotemporal hash grid is constructed based on the globally synchronized clock signal. The optical-electrical hybrid data stream is divided into quantized data blocks according to the coordinates of the spatiotemporal hash grid. Each quantized data block is bound to a unique spatiotemporal hash value and priority label, and the robot joint control instructions are marked as level L0, the highest priority. Prediction module: This module uses a spatiotemporal convolutional neural network to predict the distribution of spectrum holes within the next 10ms based on the quantized data blocks of the spatiotemporal hash grid and real-time channel state information. It dynamically allocates micro-slot resources within the orthogonal frequency division multiplexing symbol period, prioritizes mapping L0-level instructions to low-interference subcarriers, and uses polarization codes to perform asymmetric redundant encoding on micro-slot data units to generate dynamically adjusted micro-slot resource allocation results. Computing module: Based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, it calculates the optimal phase offset matrix of the metasurface intelligent reflector through a deep deterministic policy gradient algorithm, dynamically reconstructs the electromagnetic wave propagation path, and integrates full-duplex self-interference cancellation technology to achieve concurrent uplink and downlink transmission in the same frequency band, thereby obtaining the optimized electromagnetic wave propagation path; The generation module receives the channel impulse response and bit error rate (BER) obtained from the feedback end based on the optimized electromagnetic wave propagation path. It uses a time-frequency dual-domain joint equalizer to collaboratively compensate for multipath fading and narrowband interference. If the BER exceeds a preset threshold, it triggers an incremental redundancy retransmission mechanism based on federated learning. This mechanism generates a globally optimized parity check matrix by aggregating the gradients of the lightweight error correction model trained locally on each node. Only the sparse parity bits of the damaged mini-slot are retransmitted to obtain the compensated data stream. Build an optimization module: This module is used to construct a causal graph model to dynamically prune high-entropy paths to minimize causal entropy based on the compensated data stream, real-time device temperature, and subcarrier load rate data obtained from transceiver statistics. Through multi-agent reinforcement learning, the optimal modulation order and subcarrier switching strategy are determined with delay-energy efficiency as the game objective. A phase change material heat dissipation module is embedded in the module to adjust the optical module drive current, thereby obtaining an optimized stable communication link.

[0013] In a third aspect, the present application further provides an optical-electrical hybrid ultra-low latency communication device, comprising: memory for storing computer programs; A processor is configured to implement the steps of the optical-electrical hybrid ultra-low latency communication method when executing the computer program.

[0014] In a fourth aspect, the present application also provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned ultra-low latency communication method based on optical-electrical hybrid are implemented.

[0015] The beneficial effects of the present invention are: Based on the quantum Kalman filter algorithm, this invention integrates the quantum entropy source and the classical clock source to generate a globally synchronized clock signal with a jitter of ≤5 nanoseconds. It combines the space-time hash grid to dynamically segment the data stream, solving the timing confusion problem caused by traditional clock drift, reducing delay fluctuations, and effectively solving the phase drift problem caused by environmental interference in traditional clock synchronization, significantly reducing the risk of timing confusion, and providing a precise space-time benchmark for subsequent spectrum resource allocation and path optimization.

[0016] The present invention uses deep reinforcement learning to adjust the metasurface phase offset matrix in real time, and combines real-time electromagnetic interference heat maps to dynamically reconstruct the electromagnetic wave propagation path. It can actively avoid high-interference areas (such as strong radiation areas of arc welding machines and inverters), reduce the superposition of multipath effects and external interference, and significantly improve the stability and reliability of signal transmission.

[0017] The present invention is based on a sparse check bit and security gradient aggregation mechanism. Through local lightweight error correction model training and security gradient aggregation, a globally optimized sparse check matrix is ​​generated, and only the check bits of damaged micro-time slots are retransmitted, breaking through the delay and bandwidth limitations of traditional full retransmission. Combined with the interference compensation capability of the time-frequency dual-domain joint equalizer, efficient redundant repair is achieved, significantly improving the fault tolerance and real-time performance in complex channel environments.

[0018] The present invention constructs a causal graph model to analyze the timing dependencies of data transmission paths, dynamically prunes high-entropy paths to reduce the probability of timing chaos, and simultaneously balances latency and power consumption through multi-agent reinforcement learning. Combined with the thermal adaptive control of phase change materials, the optical module drive current and subcarrier switching state are intelligently adjusted under high temperature and high load scenarios, significantly extending the equipment life and maintaining optimal system energy efficiency.

[0019] The present invention realizes dynamic interference suppression and energy efficiency optimization through the integration of multiple technologies, specifically combining quantum enhanced clock synchronization, intelligent reflector beamforming, federated learning incremental retransmission and causal entropy energy efficiency game to break through the bottleneck of traditional technology; first, based on the quantum Kalman filter algorithm, high-precision synchronous clock signals are generated, and a space-time hash grid is constructed to dynamically segment data streams. Deep reinforcement learning is combined to optimize the phase of the intelligent reflector in real time to avoid high interference areas and reduce path loss; secondly, through the time-frequency dual-domain joint equalizer and the sparse check bit retransmission mechanism driven by federated learning, multi-modal interference suppression and microsecond-level redundant repair are achieved; in addition, multi-agent reinforcement learning and phase change material heat dissipation technology are introduced to dynamically balance delay and energy efficiency to ensure the stable operation of the system under high temperature and high load; compared with the existing technology, this application significantly improves the communication reliability, delay stability and energy efficiency indicators in complex electromagnetic environments, providing highly robust communication guarantees for intelligent manufacturing scenarios.

[0020] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the embodiments of the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 Schematic diagram of the process flow of the optical-electrical hybrid ultra-low latency communication method described in an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the optical-electrical hybrid ultra-low latency communication system described in an embodiment of the present invention; Figure 3 Schematic diagram of the structure of the optical-electrical hybrid ultra-low latency communication device described in an embodiment of the present invention.

[0023] In the figure: 701, acquisition module; 702, prediction module; 703, calculation module; 704, generation module; 705, construction optimization module; 800, optical-electrical hybrid ultra-low latency communication equipment; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are used only to distinguish the description and should not be understood as indicating or implying relative importance.

[0026] Embodiment 1:

[0027] This embodiment provides an optical-electrical hybrid ultra-low latency communication method.

[0028] See also Figure 1 , the figure shows that the method includes step S100, step S200, step S300, step S400 and step S500.

[0029] S100. Obtain a clock signal source and encrypt the clock synchronization channel through a quantum key distribution protocol. Use the random phase noise generated by the quantum entropy source as the observation noise input to perform Kalman filtering correction on the phase estimate of the classical clock source, output a global synchronized clock signal with a clock jitter less than 1ppm, and construct a spatiotemporal hash grid based on the global synchronized clock signal. Split the optical-electrical hybrid data stream into quantized data blocks according to the coordinates of the spatiotemporal hash grid, where each quantized data block is bound to a unique spatiotemporal hash value and a priority label, and the robot joint control instructions are marked as the highest priority level L0.

[0030] It is understood that in this step, data sources include GPS pulse-per-second signals from a satellite receiving module (such as a GPS antenna on the factory roof), IEEE 1588 PTP synchronization messages from a master clock server within the factory's local area network, a quantum random number entropy source generated by a quantum physics device (such as an entropy source chip based on optical quantum noise), and a clock synchronization channel for a dedicated communication link (such as an encrypted fiber channel) used to transmit clock synchronization data. The random phase noise (physical unpredictability) of the quantum entropy source is injected into the phase calibration module of the classical clock source. Clock drift is eliminated using a quantum Kalman filter algorithm. A spatiotemporal coordinate system (such as a timestamp + physical location hash value) generated based on the synchronized clock signal is used to provide spatiotemporal labels for data blocks. The optical-electrical hybrid data stream is segmented into minimum transmission units based on the spatiotemporal grid coordinates. Each data block is bound to a unique spatiotemporal label (for example, timestamp + factory robot coordinates + priority) using the SHA-3 hash algorithm. It should be noted that the spatiotemporal hash grid in this step provides a spatiotemporal reference for spectrum prediction, and the quantized data block labels drive micro-slot allocation.

[0031] It can be understood that step S100 includes steps S101, S102, and S103, wherein: S101. Based on the factory local area network infrastructure, obtain GPS pulse-second signals, receive IEEE 1588 protocol synchronization messages, generate a quantum random number entropy source through a quantum random number generator based on an optical quantum noise chip, encrypt the clock synchronization channel through the quantum key distribution protocol, and generate a quantum key stream; It should be noted that the quantum key stream is used to encrypt IEEE 1588 PTP synchronization messages, ensuring the security of the clock synchronization channel. The encrypted clock synchronization channel provides a secure clock signal for the subsequent quantum-classical clock source fusion.

[0032] S102, based on the quantum key stream, integrates the random phase noise of the quantum entropy source with the phase of the classical clock source, and uses the quantum Kalman filter algorithm to eliminate clock drift and generate a globally synchronized clock signal with a clock jitter of less than 1ppm. The jitter calculation formula is as follows:

[0033] Where Jitter is the clock jitter, N is the number of clock cycle samples measured, and t i is the measured value of the clock period, is the theoretical period, i For the i clock cycle samples; It should be noted that the random phase noise of the quantum entropy source is generated by a chip based on optical quantum noise. This chip can generate high-entropy random noise to enhance the stability and anti-interference ability of the clock signal. It can be understood that the jitter formula is to quantify the jitter degree of the clock signal by calculating the sum of the squares of the deviations between all measured clock cycle samples and the theoretical clock cycle, and then taking the average value. The smaller the jitter, the better the stability of the clock signal.

[0034] S103. Construct a spatiotemporal hash grid based on the global synchronous clock signal, obtain the coordinates of the spatiotemporal hash grid, and divide the optical-electrical hybrid data stream into quantized data blocks. Each data block is bound to a unique spatiotemporal hash value and a priority label, where the robot joint control instructions are marked as the highest priority level L0.

[0035] It should be noted that time is divided into fixed-length time segments (e.g., each segment is 1 millisecond) based on a global synchronous clock signal; space is divided into two-dimensional or three-dimensional grid cells based on the physical layout of factory equipment. Spatial coordinates ( x , y ) can be obtained through the UWB (Ultra Wideband) positioning system, where the location of each device or sensor is mapped to a specific grid cell. For each time-space cell, a unique hash value is generated H , the hash value can be generated using the SHA-3 algorithm, with the input being a timestamp t and spatial coordinates ( x , y ): H =SHA3( t , x , y ). In this way, each time-space unit has a unique hash value to identify the unit. All time-space units and their corresponding hash values ​​are organized into a grid structure to form a time-space hash grid G ( x , y , t). This grid can be viewed as a multidimensional array or hash table for fast data lookup and management. It is understandable that optical-electrical hybrid data streams are obtained from the optical-electrical hybrid access points of the factory local area network. These data streams may contain robot control instructions, sensor data or other information that needs to be transmitted; according to the coordinates of the spatiotemporal hash grid, the data block segmentation criteria are determined. Each data block corresponds to a time-space unit, and its size can be adjusted according to actual needs. The optical-electrical hybrid data stream is divided according to the time-space unit to generate quantized data blocks. Among them, each data block contains the following information: the spatiotemporal hash value generated by the SHA-3 algorithm and the priority label assigned according to the type and importance of the data; for example, the robot joint control instruction is marked as P = L 0 (highest priority).

[0036] S200. Based on the quantized data blocks and real-time channel status information of the space-time hash grid, the space-time convolutional neural network is used to predict the distribution of spectrum holes in the next 10ms, dynamically divide the micro-slot resources within the orthogonal frequency division multiplexing symbol period, preferentially map the L0 level instructions to the low-interference subcarriers, and use polarization codes to perform asymmetric redundant encoding on the micro-slot data units to generate dynamically adjusted micro-slot resource allocation results.

[0037] It is understood that in this step, the input data is historical spectrum occupancy (database records) and real-time channel state information (CSI, collected by the optical-electronic hybrid transceiver). A spatiotemporal convolutional neural network (ST-CNN) is used to output the interference intensity distribution map of each subcarrier within the next 10ms, marking low-interference areas as "spectrum holes." The robot joint control instructions (level 0) are assigned to the predicted low-interference subcarriers (e.g., subcarrier indices 1-10). The coding redundancy (level 0 redundancy ≥ 30%) is dynamically selected based on priority. Redundant information is injected through the frozen bits of the polarization code. The result is a dynamically divided micro-slot resource allocation table (including subcarrier mapping) and asymmetric redundantly encoded micro-slot data units, which is the dynamically adjusted micro-slot resource allocation result.

[0038] It should be noted that step S200 includes steps S201, S202, and S203, wherein: S201. Based on the spatiotemporal hash value and priority tag of the quantized data block, combined with historical spectrum occupancy and real-time channel status information, the spectrum hole distribution within the next 10 ms is predicted using a spatiotemporal convolutional neural network. The calculation formula is as follows:

[0039] In the formula, Mask priority(P) is the priority mask matrix, Holes(f,t) is the predicted spectrum hole distribution, STCNN(S hist ,SNR) is the spatiotemporal convolutional neural network; S202: Dynamically divide the OFDM symbol period according to the predicted spectrum hole distribution The duration of each mini-slot is Δ t =1μs, bandwidth is Δ f =4.3125kHz, preferentially mapping L0 level instructions to low-interference subcarriers, where the low-interference subcarriers meet the signal-to-noise ratio threshold greater than or equal to 25dB, and generating a mini-slot resource allocation table; It should be noted that spectrum hole prediction enables real-time optimization of micro-slot resource allocation, reduces the transmission latency of high-priority instructions, and dynamically identifies available spectrum resources, improving spectrum utilization and reducing spectrum waste. Micro-slot resources are divided within the available spectrum region based on the spectrum hole distribution Holes(f,t). The start time and frequency position of each micro-slot are determined by the spectrum hole distribution. Dynamic allocation of micro-slot resources allows flexible adjustment of resource allocation based on the real-time distribution of spectrum holes, improving spectrum resource utilization and reducing resource waste.

[0040] S203. Based on the mini-slot resource allocation table, polar code encoding is performed on the mini-slot data unit to generate a codeword as follows:

[0041] Where, For the encoded codeword, PolarEncode(D slot ,R) is the mini-slot data D slot Perform polarization coding and add redundant information with redundancy R; Among them, redundancy R The dynamic adjustment based on priority is as follows:

[0042] Where R is the redundancy, P is the priority label of the data block, L0 is the highest priority, L1 is the medium priority, and L2 is the lowest priority.

[0043] It should be noted that R =0.3 (when P = L 0): For the highest priority data block ( L 0), the redundancy is set to 0.3, which means that 30% of the encoded data is redundant information, which is used to improve error correction capability; R =0.5 (when P = L1 time): For medium priority data blocks ( L 1) Redundancy is set to 0.5, which means that 50% of the encoded data is redundant information; R =0.7 (when P = L 2): For the lowest priority data block ( L 2) Redundancy is set to 0.7, which means that 70% of the encoded data is redundant information to ensure reliable transmission even under poor channel conditions. Therefore, the L0 level instruction redundancy is 70%, and the bit error rate tolerance is improved to BER≤10 -8 Therefore, this method of dynamically adjusting redundancy can optimize transmission efficiency and reliability according to the importance of data and channel conditions, and the mini-slot resource allocation table also provides a scheduling basis for subsequent beamforming and equalization.

[0044] S300, based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, calculates the optimal phase offset matrix of the metasurface intelligent reflector through a deep deterministic policy gradient algorithm, dynamically reconstructs the electromagnetic wave propagation path, and integrates full-duplex self-interference elimination technology to achieve concurrent uplink and downlink transmission in the same frequency band, obtaining the optimized electromagnetic wave propagation path.

[0045] It can be understood that in this step, the input data source is the quantized data block label generated in the above step, and the distributed electromagnetic sensors in the factory (such as the spectrum analyzer array) generate the electromagnetic interference heat map in real time. The phase parameters of the IRS reflection unit are calculated by the deep deterministic policy gradient (DDPG) algorithm, so that the electromagnetic waves bypass the interference area (such as the arc welder). When the uplink and downlink are transmitted in the same frequency band, the digital domain interference cancellation technology is used to eliminate the echo interference, and finally the optimized electromagnetic wave propagation path with a delay of ≤0.15ms and the concurrent transmission configuration parameters (such as the phase matrix and frequency allocation) are obtained.

[0046] It should be noted that step S300 includes S301, S302 and S303, wherein: S301. Establish a deep deterministic policy gradient model based on the micro-slot resource allocation results and a real-time electromagnetic interference heat map generated by the distributed spectrum sensor based on the electromagnetic interference intensity. Input the micro-slot resource allocation results and the real-time electromagnetic interference heat map into the deep deterministic policy gradient model, and output the phase offset matrix of the metasurface intelligent reflector. The training objective of the deep deterministic policy gradient model is to minimize path loss, and by iteratively optimizing the phase offset of the reflector unit, the total path loss after the channel response of the reflected path and the direct path is superimposed is minimized. It should be noted that the micro-slot resource allocation result is the result of the dynamic division of the OFDM symbol period, which identifies the frequency point. f The real-time electromagnetic interference heat map is generated by using a distributed spectrum sensor network (such as a software-defined radio node) deployed throughout the factory to collect the electromagnetic radiation intensity (in dBm) at each coordinate point (x, y). Kriging interpolation is then used to generate a spatially continuous heat map with a resolution of 1 meter. Typical interference sources include arc welders (radiation intensity > −80 dBm) and inverters.

[0047] It can be understood that the construction of the deep deterministic policy gradient model includes the definition of the state space, the definition of the action space, and the design of the reward function. The model input state defined by the state space includes the micro-slot resource allocation results and the electromagnetic interference heat map, while the action space is defined as the output of the phase offset matrix of the metasurface intelligent reflective surface, with dimension N×N (N is the number of reflective units), and each element ϕn∈[0,2π) represents the phase offset of the nth unit. The reward function is designed with minimizing path loss as the core goal, and its reward function is defined as:

[0048] Where R is the reward function, is the weight coefficient, is the time delay of the signal in the transmission path, is the cumulative interference value of the path coverage area, L path To minimize path loss; The phase-shift matrix optimization mechanism involves summing the channel matrices of the reflected and direct paths to form the total channel response. The path loss is then calculated as the inverse squared Frobenius norm of the channel matrix. Phase-shift actions are then generated through the actor network, and the critic network evaluates the value of these actions. This process combines the experience replay pool with the target network update strategy until the path loss converges to a minimum. This step reduces the number of multipath reflections and path propagation distance, resulting in an end-to-end latency of less than or equal to 0.15 ms, meeting the sub-millisecond requirements of robot collaborative control and enhancing anti-interference capabilities.

[0049] S302. Dynamically adjust the phase offset of the metasurface smart reflector according to the phase offset matrix to guide the electromagnetic wave propagation path around the interference area to obtain an optimized path. Based on the optimized path, use the minimum mean square error algorithm to cancel the self-interference of the uplink and downlink signals in the same frequency band. This includes: constructing the autocorrelation matrix of the self-interference signal based on the known signal waveform and channel impulse response at the transmitting end, and solving the optimal interference cancellation weight matrix under the minimum mean square error criterion by maximizing the signal to interference and noise ratio. The calculation formula is as follows:

[0050] Where W MMSE is the optimal interference cancellation weight matrix, R xy is the cross-correlation matrix between the self-interference signal and the desired signal, is the inverse matrix of the autocorrelation matrix; S303 , applying the optimal interference cancellation weight matrix to filter the received signal, suppressing the self-interference component in the same frequency band, thereby outputting full-duplex concurrent transmission configuration parameters and obtaining an optimized electromagnetic wave propagation path.

[0051] It should be noted that based on the known signal at the transmitter and the channel impulse response, the autocorrelation matrix of the self-interference signal is constructed. This matrix is ​​used to describe the statistical characteristics of the self-interference signal and provide a basis for subsequent weight calculations. Then, by maximizing the signal-to-interference-plus-noise ratio (SINR), the MMSE interference cancellation weight matrix is ​​calculated. This weight matrix is ​​used to suppress the self-interference components within the same frequency band to ensure that the uplink and downlink signals do not interfere with each other in full-duplex transmission. The weight matrix is ​​applied to filter the received signal to reduce the residual interference power to no more than -30 dBm. Through this process, the spectrum efficiency is significantly improved to η = 8 bps / Hz, and the full-duplex concurrent transmission configuration parameters are generated. ,in, is the uplink frequency, is the downlink frequency, and η is the spectrum efficiency. This effectively reduces the self-interference of uplink and downlink signals in the same frequency band, significantly improving spectrum efficiency and achieving efficient full-duplex communication. It can be understood that the smaller the residual interference power after filtering, the better the interference suppression effect. The goal of maximizing SINR is to ensure that the signal maintains high quality despite interference and noise. By maximizing SINR, the weight matrix can be optimized to maximize the useful portion of the signal while minimizing the mean square error.

[0052] S400, based on the optimized electromagnetic wave propagation path, receives the channel impulse response and bit error rate obtained at the feedback end, and uses a time-frequency dual-domain joint equalizer to collaboratively compensate for multipath fading and narrowband interference. If the bit error rate exceeds the preset threshold, the incremental redundant retransmission mechanism based on federated learning is triggered. By aggregating the gradients of the lightweight error correction model trained locally at each node, a globally optimized check matrix is ​​generated. Only the sparse check bits of the damaged micro-time slot are retransmitted to obtain the compensated data stream.

[0053] It can be understood that the data of the optimized electromagnetic wave propagation path in this step is derived from the optimized path data in the above step (such as the channel impulse response) for equalizer compensation. If BER ≤ threshold: directly enter the energy efficiency optimization of the following steps; if BER > threshold (1e-6): trigger the federated learning retransmission mechanism, namely local training: each node trains the error correction model (such as a sparse check matrix generator) based on the historical error pattern, gradient aggregation: encrypt and upload the model gradient to the central server to generate a global optimized check matrix, and sparse check bit retransmission: only retransmit the check bits of the damaged micro-time slot (instead of all data), and compress the retransmission delay to ≤20μs. It only retransmits the sparse check bits of the damaged micro-time slot, which means that after detecting an error, the system will not retransmit the entire damaged micro-time slot data, but only retransmit those sparse check bits containing key check information. This method can significantly reduce the amount of retransmitted data, improve transmission efficiency, and reduce delay. At the same time, the data stream processed by the time-frequency dual-domain joint equalizer has a lower bit error rate and improved signal quality.

[0054] It should be noted that step S400 includes S401, S402 and S403, wherein: S401. Based on the optimized phase offset matrix of the metasurface intelligent reflector in the electromagnetic wave propagation path, the channel impulse response and bit error rate fed back by the receiving end, a time-frequency dual-domain joint equalizer is designed, including: calculating the time-domain equalization weight by minimizing the inter-symbol interference criterion to compensate for the delay caused by multipath fading, identifying the interference frequency set based on the narrowband interference power spectral density, and calculating the frequency-domain suppression coefficient; superimposing the time-domain equalization weight and the frequency-domain suppression result to generate an equalized low-error signal stream; It should be noted that by minimizing the inter-symbol interference criterion calculation, compensating for the time delay caused by multipath fading, and then obtaining the time domain equalization weight, compensating for the multipath effect in the time domain, reducing inter-symbol interference, and improving signal clarity; then based on the narrowband interference power spectral density, identifying the interference frequency set, calculating the frequency domain suppression coefficient, performing notch filtering on the interference frequency, suppressing narrowband interference in the frequency domain, reducing the impact of interference on the signal, and improving the signal-to-noise ratio of the signal; combining the time domain equalization weight and the frequency domain suppression coefficient, obtaining the equalized signal, outputting the equalized low-error signal stream, significantly reducing the bit error rate, improving signal quality, and achieving comprehensive optimization of the signal.

[0055] S402, if the bit error rate BER>10 -6 , extract the damaged micro-time slot index, identify the fault frequency and time slice, where the communication node performs local training, where the local training includes: using the original data of the damaged micro-time slot and the equalized received signal as input data, training a lightweight error correction model to minimize the reconstruction error and generate a local gradient; It should be noted that if the bit error rate BER≤10-6 , it is considered that the signal quality of the current transmission link meets the requirements and there is no need to trigger the incremental redundancy retransmission mechanism. The system will continue to transmit data normally, maintain the current communication status, and continuously monitor the bit error rate to ensure communication reliability. In this step, the time-frequency dual-domain joint equalizer is used to effectively compensate for multipath fading and narrowband interference, significantly reducing the bit error rate and improving signal quality. When the bit error rate exceeds the threshold, the lightweight error correction model is trained through the federated learning mechanism to generate local gradients and perform secure aggregation to generate a global optimized check matrix. The calculation formula for minimizing the reconstruction error is as follows:

[0056] Where, L local is the local reconstruction error, D slot is the original data of the damaged mini-slot, M local is a lightweight error correction model, S eq (t) is the received signal after being processed by the time-frequency dual-domain joint equalizer.

[0057] S403. Based on the local gradient, the data is encrypted and uploaded to the central server through the secure multi-party computing protocol, and the global gradient is aggregated to generate the global gradient. The global check matrix is ​​updated, and only the sparse check bits of the damaged micro-time slot are retransmitted to obtain the global optimized check matrix. The global optimized check matrix and the sparse check bits are error-corrected and decoded to restore the compensated data stream.

[0058] It should be noted that only the sparse check bits of the damaged micro-time slots are retransmitted, which reduces the amount of retransmitted data, reduces the retransmission delay, and improves the transmission efficiency. In addition, the secure upload and aggregation of gradient information is ensured through the secure multi-party computing protocol, protecting data privacy and enhancing the security of the system. Therefore, this method achieves efficient interference suppression and reliable incremental redundant retransmission in a multimodal interference environment, significantly improving the transmission efficiency and reliability of the system. The compensated data stream finally obtained is a low bit error rate data stream that has been error-corrected and meets the system's reliability requirements. Among them, the calculation formula for the global gradient generated by aggregation is as follows, which is encrypted and uploaded to the central server through the secure multi-party computing protocol:

[0059] Where, is the global gradient, M is the total number of communication nodes, For the i The local gradients generated by the communication nodes.

[0060] S500, based on the compensated data stream, real-time device temperature, and subcarrier load rate data obtained from transceiver statistics, constructs a causal graph model to dynamically prune high-entropy paths to minimize causal entropy. Through multi-agent reinforcement learning, the optimal modulation order and subcarrier switching strategy are determined with delay-energy efficiency as the game objective. A phase change material heat dissipation module is embedded to adjust the optical module drive current, resulting in an optimized stable communication link.

[0061] It can be understood that in this step, the device temperature and subcarrier load rate data come from the real-time monitoring system of the device. Based on the demodulated data block sequence, device temperature and subcarrier load rate data, the high entropy path is dynamically pruned to reduce the risk of timing chaos. At the same time, based on the entropy increase index and real-time temperature feedback, the modulation order and subcarrier switching state are dynamically adjusted to optimize energy efficiency. The optimized modulation order and subcarrier switching strategy improves the energy efficiency and reliability of the system, while reducing the risk of timing chaos and improving the stability of data transmission.

[0062] It should be noted that based on the balanced signal flow, device temperature (collected in real time by the temperature sensor) and subcarrier load rate (identifying subcarrier f Utilization percentage), build causal graph model G causal (V, E), where V represents the data transmission path (such as direct path, reflection path, redundant relay path); E represents the delay dependency between paths (such as the timing dependency introduced by multi-hop relay), and then calculate the path entropy increase , where H pre is the original path entropy value, H post is the entropy value after path adjustment, High entropy path greater than 0 ( θ =( ...)))

[0063] Next, define the agent, whose state space (temperature, load rate, bit error rate), action space A = {QAM order, subcarrier switch state}; the reward function is designed as follows:

[0064] Where R is the design reward function, is the weight coefficient related to the path delay, is the weight coefficient related to the total power consumption, is the end-to-end transmission delay, P totalis the total system power consumption. The Q-learning algorithm then updates the strategy and outputs the optimized modulation parameters. This step uses latency and energy efficiency as joint objectives, adaptively adjusting the QAM order and subcarrier switching, improving spectral efficiency by 20%.

[0065] Based on the real-time device temperature and subcarrier switching status, the optical module drive current is dynamically adjusted through the heat absorption characteristics of the phase change material. r When the temperature exceeds the preset threshold, the drive current is reduced linearly to suppress the power consumption of the optical module, so that the total power consumption of the system is stabilized below the target value while maintaining the delay. , thereby achieving an optimized, stable communication link. Phase-change materials achieve closed-loop regulation of temperature and power consumption, improving heat dissipation efficiency by 40% and extending device life. Therefore, this step, through causal entropy pruning, reinforcement learning game theory, and thermal adaptive control, achieves ultra-low latency and high energy efficiency for optical-electrical hybrid links in complex industrial environments, providing reliable communication support for intelligent manufacturing.

[0066] Therefore, the present invention breaks through the bottlenecks of delay, interference and energy efficiency of traditional optical-electrical hybrid technology in complex industrial environments through multi-level collaboration of quantum-classical fusion synchronization, dynamic optimization of intelligent reflecting surfaces, sparse retransmission of federated learning and causal entropy energy efficiency game, and provides a highly reliable and low-latency communication infrastructure support for high-precision intelligent manufacturing.

[0067] Example 2:

[0068] like Figure 2 As shown, this embodiment provides an optical-electrical hybrid ultra-low latency communication system, see Figure 2 The system comprises: Acquisition module 701: used to obtain a clock signal source and encrypt the clock synchronization channel through a quantum key distribution protocol; use the random phase noise generated by the quantum entropy source as the observation noise input, perform Kalman filtering correction on the phase estimate of the classical clock source, output a globally synchronized clock signal with a clock jitter of less than 1ppm, and construct a spatiotemporal hash grid based on the globally synchronized clock signal; divide the optical-electrical hybrid data stream into quantized data blocks according to the coordinates of the spatiotemporal hash grid, where each quantized data block is bound to a unique spatiotemporal hash value and priority label, and the robot joint control instructions are marked as level L0 with the highest priority; Prediction module 702: used to predict the distribution of spectrum holes within the next 10ms based on the quantized data blocks of the spatiotemporal hash grid and real-time channel state information through a spatiotemporal convolutional neural network, dynamically allocate mini-slot resources within the orthogonal frequency division multiplexing symbol period, preferentially map L0 level instructions to low-interference subcarriers, and use polarization codes to asymmetric redundantly encode mini-slot data units to generate dynamically adjusted mini-slot resource allocation results; Computation Module 703: Based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, it calculates the optimal phase offset matrix of the metasurface intelligent reflector using a deep deterministic policy gradient algorithm, dynamically reconstructs the electromagnetic wave propagation path, and integrates full-duplex self-interference cancellation technology to achieve concurrent uplink and downlink transmission in the same frequency band, thereby obtaining an optimized electromagnetic wave propagation path. Generation module 704: Based on the optimized electromagnetic wave propagation path, the channel impulse response and bit error rate obtained by the feedback end are received, and a time-frequency dual-domain joint equalizer is used to collaboratively compensate for multipath fading and narrowband interference. If the bit error rate exceeds a preset threshold, an incremental redundancy retransmission mechanism based on federated learning is triggered. The globally optimized check matrix is ​​generated by aggregating the gradients of the lightweight error correction model trained locally at each node. Only the sparse check bits of the damaged mini-slot are retransmitted to obtain a compensated data stream. Constructing an optimization module 705: It is used to construct a causal graph model to dynamically prune high-entropy paths to minimize causal entropy based on the compensated data stream, the real-time collected device temperature, and the subcarrier load rate data obtained from transceiver statistics. Through multi-agent reinforcement learning, the optimal modulation order and subcarrier switching strategy are determined with delay-energy efficiency as the game goal. A phase change material heat dissipation module is embedded to adjust the optical module drive current to obtain an optimized stable communication link.

[0069] Specifically, the acquisition module 701 includes: The first generation unit is used to obtain GPS pulse-per-second signals based on the factory local area network infrastructure, receive IEEE1588 protocol synchronization messages, generate a quantum random number entropy source through a quantum random number generator based on an optical quantum noise chip, encrypt the clock synchronization channel through the quantum key distribution protocol, and generate a quantum key stream; The second generation unit is used to combine the random phase noise of the quantum entropy source with the phase of the classical clock source based on the quantum key stream, and use the quantum Kalman filter algorithm to eliminate clock drift and generate a globally synchronized clock signal with a clock jitter of less than 1ppm. The jitter calculation formula is as follows:

[0070] Where Jitter is the clock jitter, N is the number of clock cycle samples measured, and t i is the measured value of the clock period, is the theoretical period, i For the i clock cycle samples; Segmentation unit: used to build a spatiotemporal hash grid based on the global synchronous clock signal, obtain the coordinates of the spatiotemporal hash grid, and segment the optical-electrical hybrid data stream into quantized data blocks. Each data block is bound to a unique spatiotemporal hash value and priority label, among which the robot joint control instructions are marked as the highest priority level L0.

[0071] Specifically, the prediction module 702 includes: Prediction unit: Based on the spatiotemporal hash value and priority tag of the quantized data block, combined with historical spectrum occupancy and real-time channel status information, the prediction unit uses a spatiotemporal convolutional neural network to predict the spectrum hole distribution within the next 10ms. The calculation formula is as follows:

[0072] In the formula, Mask priority (P) is the priority mask matrix, Holes(f,t) is the predicted spectrum hole distribution, STCNN(S hist ,SNR) is the spatiotemporal convolutional neural network; Division unit: used to dynamically divide the OFDM symbol period according to the predicted spectrum hole distribution The duration of each mini-slot is Δ t =1μs, bandwidth is Δ f =4.3125kHz, preferentially mapping L0 level instructions to low-interference subcarriers, where the low-interference subcarriers meet the signal-to-noise ratio threshold greater than or equal to 25dB, and generating a mini-slot resource allocation table; Coding unit: Used to perform polarization code encoding on the mini-slot data unit based on the mini-slot resource allocation table to generate the following codewords:

[0073] Where, For the encoded codeword, PolarEncode(D slot ,R) is the mini-slot data D slot Perform polarization coding and add redundant information with redundancy R; Among them, redundancy R The dynamic adjustment based on priority is as follows:

[0074] Where R is the redundancy, P is the priority label of the data block, L0 is the highest priority, L1 is the medium priority, and L2 is the lowest priority.

[0075] Specifically, the calculation module 703 includes: A model building unit is used to establish a deep deterministic policy gradient model based on the micro-slot resource allocation results and a real-time electromagnetic interference heat map generated by distributed spectrum sensors based on electromagnetic interference intensity. The micro-slot resource allocation results and the real-time electromagnetic interference heat map are input into the deep deterministic policy gradient model, which outputs a phase offset matrix for the metasurface smart reflector. The training objective of the deep deterministic policy gradient model is to minimize path loss. The phase offset of the reflector unit is iteratively optimized to minimize the total path loss after the channel responses of the reflected path and the direct path are superimposed. The optimization unit is used to dynamically adjust the phase offset of the metasurface intelligent reflector according to the phase offset matrix, guiding the electromagnetic wave propagation path to bypass the interference area to obtain an optimized path. Based on the optimized path, the minimum mean square error algorithm is used to cancel the self-interference of the uplink and downlink signals in the same frequency band. This includes: constructing the autocorrelation matrix of the self-interference signal based on the known signal waveform and channel impulse response at the transmitting end, and solving the optimal interference cancellation weight matrix under the minimum mean square error criterion by maximizing the signal-to-interference-noise ratio. The calculation formula is as follows:

[0076] Where W MMSE is the optimal interference cancellation weight matrix, R xy is the cross-correlation matrix between the self-interference signal and the desired signal, is the inverse matrix of the autocorrelation matrix; Processing unit: Used to apply the optimal interference cancellation weight matrix to filter the received signal, suppress the self-interference component in the same frequency band, and thus output the full-duplex concurrent transmission configuration parameters to obtain the optimized electromagnetic wave propagation path.

[0077] Specifically, the generating module 704 includes: Design unit: Used to design a time-frequency dual-domain joint equalizer based on the optimized phase offset matrix of the metasurface smart reflector in the electromagnetic wave propagation path, the channel impulse response and bit error rate fed back by the receiver. This includes: calculating the time-domain equalization weights by minimizing the inter-symbol interference criterion to compensate for the delay caused by multipath fading, identifying the interference frequency set based on the narrowband interference power spectral density, and calculating the frequency-domain suppression coefficient; superimposing the time-domain equalization weights with the frequency-domain suppression results to generate an equalized low-error signal stream; Extraction and recognition unit: used if the bit error rate BER>10 -6 , extract the damaged micro-time slot index, identify the fault frequency and time slice, where the communication node performs local training, where the local training includes: using the original data of the damaged micro-time slot and the equalized received signal as input data, training a lightweight error correction model to minimize the reconstruction error and generate a local gradient; Update error correction unit: Based on the local gradient, it is encrypted and uploaded to the central server through the secure multi-party computing protocol, aggregated to generate the global gradient, update the global check matrix, retransmit only the sparse check bits of the damaged micro-time slot, obtain the global optimized check matrix, perform error correction decoding on the global optimized check matrix and the sparse check bits, and restore the compensated data stream.

[0078] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0079] Example 3:

[0080] Corresponding to the above method embodiment, this embodiment also provides an optical-electrical hybrid ultra-low latency communication device. The optical-electrical hybrid ultra-low latency communication device described below and the optical-electrical hybrid ultra-low latency communication method described above can be referenced to each other.

[0081] Figure 3 FIG is a block diagram of an optical-electrical hybrid ultra-low latency communication device 800 according to an exemplary embodiment. Figure 3 As shown, the optical-electrical hybrid ultra-low latency communication device 800 includes: a processor 801 and a memory 802. The optical-electrical hybrid ultra-low latency communication device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0082] The processor 801 is used to control the overall operation of the optical-electrical hybrid ultra-low latency communication device 800 to complete all or part of the steps in the above-mentioned optical-electrical hybrid ultra-low latency communication method. The memory 802 is used to store various types of data to support the operation of the optical-electrical hybrid ultra-low latency communication device 800. For example, this data may include instructions for any application or method operating on the optical-electrical hybrid ultra-low latency communication device 800, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be a keyboard, a mouse or buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the optical-electrical hybrid ultra-low latency communication device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 may include: a Wi-Fi module, a Bluetooth module or an NFC module.

[0083] In an exemplary embodiment, the optical-electrical hybrid ultra-low latency communication device 800 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-mentioned optical-electrical hybrid ultra-low latency communication method.

[0084] In another exemplary embodiment, a computer-readable storage medium including program instructions is further provided. When executed by a processor, the program instructions implement the steps of the aforementioned optical-electrical hybrid ultra-low-latency communication method. For example, the computer-readable storage medium may be the aforementioned memory 802 including the program instructions. The program instructions may be executed by the processor 801 of the optical-electrical hybrid ultra-low-latency communication device 800 to implement the aforementioned optical-electrical hybrid ultra-low-latency communication method.

[0085] Embodiment 4:

[0086] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the optical-electrical hybrid ultra-low latency communication method described above can refer to each other.

[0087] A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, the steps of the optical-electrical hybrid ultra-low delay communication method of the above method embodiment are implemented.

[0088] The readable storage medium may specifically be any readable storage medium that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0089] In summary, the present invention realizes dynamic interference suppression and energy efficiency optimization through the integration of multiple technologies, specifically combining quantum enhanced clock synchronization, intelligent reflector beamforming, federated learning incremental retransmission and causal entropy energy efficiency game to break through the bottleneck of traditional technology; first, based on the quantum Kalman filter algorithm, a high-precision synchronous clock signal is generated, and a space-time hash grid is constructed to dynamically segment the data stream. The phase of the intelligent reflector is optimized in real time by combining deep reinforcement learning to avoid high interference areas and reduce path loss; secondly, through the time-frequency dual-domain joint equalizer and the sparse check bit retransmission mechanism driven by federated learning, multi-modal interference suppression and microsecond-level redundant repair are achieved; in addition, multi-agent reinforcement learning and phase change material heat dissipation technology are introduced to dynamically balance delay and energy efficiency to ensure the stable operation of the system under high temperature and high load; compared with the existing technology, the present application significantly improves the communication reliability, delay stability and energy efficiency indicators in complex electromagnetic environments, providing highly robust communication guarantees for intelligent manufacturing scenarios.

[0090] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

[0091] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An optical-electrical hybrid ultra-low latency communication method, characterized in that: include: Obtain a clock signal source and encrypt the clock synchronization channel through the quantum key distribution protocol; Using random phase noise generated by a quantum entropy source as the observation noise input, the phase estimate of the classical clock source is corrected by Kalman filtering, outputting a globally synchronized clock signal with a clock jitter of less than 1ppm. A spatiotemporal hash grid is then constructed based on the globally synchronized clock signal. The optical-electrical hybrid data stream is segmented into quantized data blocks according to the coordinates of the spatiotemporal hash grid, where each quantized data block is bound to a unique spatiotemporal hash value and a priority label, with the robot joint control instructions marked as level L0, the highest priority. Based on the quantized data blocks of the spatiotemporal hash grid and real-time channel state information, a spatiotemporal convolutional neural network is used to predict the distribution of spectrum holes within the next 10ms. Micro-slot resources within the orthogonal frequency division multiplexing symbol period are dynamically allocated, L0-level instructions are preferentially mapped to low-interference subcarriers, and polar codes are used to perform asymmetric redundant encoding on micro-slot data units to generate dynamically adjusted micro-slot resource allocation results. Based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, the optimal phase offset matrix of the metasurface intelligent reflector is calculated through a deep deterministic policy gradient algorithm. The electromagnetic wave propagation path is dynamically reconstructed, and full-duplex self-interference cancellation technology is integrated to achieve concurrent uplink and downlink transmission in the same frequency band, thereby obtaining the optimized electromagnetic wave propagation path. Based on the optimized electromagnetic wave propagation path, the channel impulse response and bit error rate (BER) obtained at the receiving feedback end are then received. A time-frequency dual-domain joint equalizer is used to collaboratively compensate for multipath fading and narrowband interference. If the BER exceeds a preset threshold, an incremental redundancy retransmission mechanism based on federated learning is triggered. This mechanism generates a globally optimized parity check matrix by aggregating the gradients of the lightweight error correction model trained locally on each node. Only the sparse parity bits of the damaged mini-slot are retransmitted to obtain the compensated data stream. Based on the compensated data stream, real-time device temperature, and subcarrier load rate data obtained from transceiver statistics, a causal graph model is constructed to dynamically prune high-entropy paths to minimize causal entropy. Multi-agent reinforcement learning is used to determine the optimal modulation order and subcarrier switching strategy with delay-energy efficiency as the game objective. A phase change material heat dissipation module is embedded to regulate the optical module driving current, resulting in an optimized stable communication link.

2. The optical-electrical hybrid ultra-low latency communication method according to claim 1, characterized in that: The method obtains a clock signal source and encrypts a clock synchronization channel through a quantum key distribution protocol; uses random phase noise generated by a quantum entropy source as an observation noise input, performs Kalman filtering correction on the phase estimate of the classical clock source, outputs a globally synchronized clock signal with a clock jitter less than 1 ppm, and constructs a spatiotemporal hash grid based on the globally synchronized clock signal; divides the optical-electrical hybrid data stream into quantized data blocks according to the coordinates of the spatiotemporal hash grid, wherein each quantized data block is bound to a unique spatiotemporal hash value and a priority label, wherein the robot joint control instructions are marked as the highest priority level L0, including: Based on the factory local area network infrastructure, it acquires GPS pulse-second signals, receives IEEE 1588 protocol synchronization messages, and generates a quantum random number entropy source through a quantum random number generator based on an optical quantum noise chip. It then encrypts the clock synchronization channel through the quantum key distribution protocol to generate a quantum key stream. Based on the quantum key stream, the random phase noise generated by the quantum entropy source is used as the observation noise input. The quantum Kalman filter algorithm is used to eliminate clock drift and generate a globally synchronized clock signal with a clock jitter of less than 1ppm. The jitter calculation formula is as follows: Where Jitter is the clock jitter, N is the number of clock cycle samples measured, and t i is the measured value of the clock period, is the theoretical period, i For the i clock cycle samples; A spatiotemporal hash grid is constructed based on the global synchronous clock signal to obtain the coordinates of the spatiotemporal hash grid. The VDSL data stream is then divided into quantized data blocks. Each data block is bound to a unique spatiotemporal hash value and a priority label, where the robot joint control instructions are marked as the highest priority level L0.

3. The optical-electrical hybrid ultra-low latency communication method according to claim 1, characterized in that: The method uses a spatiotemporal convolutional neural network to predict the distribution of spectrum holes within the next 10ms based on the quantized data blocks and real-time channel state information of the spatiotemporal hash grid, dynamically divides micro-slot resources within the orthogonal frequency division multiplexing symbol period, preferentially maps L0-level instructions to low-interference subcarriers, and uses polarization codes to perform asymmetric redundant encoding on micro-slot data units to generate dynamically adjusted micro-slot resource allocation results, including: Based on the spatiotemporal hash value and priority label of the quantized data block, combined with historical spectrum occupancy and real-time channel status information, the spatiotemporal convolutional neural network is used to predict the spectrum hole distribution within the next 10ms. The calculation formula is as follows: In the formula, Mask priority (P) is the priority mask matrix, Holes(f,t) is the predicted spectrum hole distribution, STCNN(S hist ,SNR) is the spatiotemporal convolutional neural network; Dynamically divide OFDM symbol periods based on predicted spectrum hole distribution The duration of each mini-slot is Δ t =1μs, bandwidth is Δ f =4.3125kHz, preferentially mapping L0 level instructions to low-interference subcarriers, where the low-interference subcarriers meet the signal-to-noise ratio threshold greater than or equal to 25dB, and generating a mini-slot resource allocation table; Based on the mini-slot resource allocation table, polar code encoding is performed on the mini-slot data unit to generate the following codeword: Where, For the encoded codeword, PolarEncode(D slot ,R) is the mini-slot data D slot Perform polarization coding and add redundant information with redundancy R; Among them, redundancy R The dynamic adjustment based on priority is as follows: Where R is the redundancy, P is the priority label of the data block, L0 is the highest priority, L1 is the medium priority, and L2 is the lowest priority.

4. The optical-electrical hybrid ultra-low latency communication method according to claim 1, characterized in that: Based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, the optimal phase offset matrix of the metasurface intelligent reflector is calculated through a deep deterministic policy gradient algorithm, the electromagnetic wave propagation path is dynamically reconstructed, and full-duplex self-interference cancellation technology is integrated to achieve concurrent uplink and downlink transmission in the same frequency band. The optimized electromagnetic wave propagation path is obtained, which includes: A deep deterministic policy gradient model is established based on the micro-slot resource allocation results and a real-time electromagnetic interference heat map generated by distributed spectrum sensors based on electromagnetic interference intensity. The micro-slot resource allocation results and the real-time electromagnetic interference heat map are input into the deep deterministic policy gradient model, which outputs the phase offset matrix of the metasurface smart reflector. The training objective of the deep deterministic policy gradient model is to minimize path loss. By iteratively optimizing the phase offset of the reflector unit, the total path loss after the channel response of the reflected path and the direct path is minimized. According to the phase offset matrix, the phase offset of the metasurface smart reflector is dynamically adjusted to guide the electromagnetic wave propagation path around the interference area to obtain an optimized path. Based on the optimized path, the minimum mean square error algorithm is used to cancel the self-interference of the uplink and downlink signals in the same frequency band. This includes: constructing the autocorrelation matrix of the self-interference signal based on the known signal waveform and channel impulse response at the transmitting end, and solving the optimal interference cancellation weight matrix under the minimum mean square error criterion by maximizing the signal-to-interference-noise ratio. The calculation formula is as follows: Where W MMSE is the optimal interference cancellation weight matrix, R xy is the cross-correlation matrix between the self-interference signal and the desired signal, is the inverse matrix of the autocorrelation matrix; The optimal interference cancellation weight matrix is ​​applied to filter the received signal to suppress the self-interference component in the same frequency band, thereby outputting the full-duplex concurrent transmission configuration parameters and obtaining the optimized electromagnetic wave propagation path.

5. The optical-electrical hybrid ultra-low latency communication method according to claim 1, characterized in that: Based on the optimized electromagnetic wave propagation path, the channel impulse response and bit error rate obtained at the receiving feedback end are received, and a time-frequency dual-domain joint equalizer is used to collaboratively compensate for multipath fading and narrowband interference. If the bit error rate exceeds a preset threshold, an incremental redundancy retransmission mechanism based on federated learning is triggered. By aggregating the gradients of the lightweight error correction model trained locally at each node, a globally optimized check matrix is ​​generated. Only the sparse check bits of the damaged micro-timeslot are retransmitted to obtain a compensated data stream, which includes: Based on the optimized phase offset matrix of the metasurface smart reflector in the electromagnetic wave propagation path, the channel impulse response and bit error rate fed back by the receiver, a time-frequency dual-domain joint equalizer is designed. This includes: calculating the time-domain equalization weights by minimizing the inter-symbol interference criterion to compensate for the delay caused by multipath fading; identifying the interference frequency set based on the narrowband interference power spectral density, and calculating the frequency-domain suppression coefficient; and superimposing the time-domain equalization weights with the frequency-domain suppression results to generate an equalized low-error signal stream. If the bit error rate BER>10 -6 , extract the damaged micro-time slot index, identify the fault frequency and time slice, where the communication node performs local training, where the local training includes: using the original data of the damaged micro-time slot and the equalized received signal as input data, training a lightweight error correction model to minimize the reconstruction error and generate a local gradient; Based on the local gradient, it is encrypted and uploaded to the central server through the secure multi-party computing protocol, aggregated to generate the global gradient, update the global check matrix, and only retransmit the sparse check bits of the damaged micro-time slot to obtain the globally optimized check matrix. The globally optimized check matrix and the sparse check bits are then error-corrected and decoded to restore the compensated data stream.

6. An optical-electrical hybrid ultra-low latency communication system, based on the optical-electrical hybrid ultra-low latency communication method according to claim 1, characterized in that: include: Acquisition module: used to obtain the clock signal source and encrypt the clock synchronization channel through the quantum key distribution protocol; Using random phase noise generated by a quantum entropy source as the observation noise input, the phase estimate of the classical clock source is corrected by Kalman filtering, outputting a globally synchronized clock signal with a clock jitter of less than 1ppm. A spatiotemporal hash grid is then constructed based on the globally synchronized clock signal. The optical-electrical hybrid data stream is segmented into quantized data blocks according to the coordinates of the spatiotemporal hash grid, where each quantized data block is bound to a unique spatiotemporal hash value and a priority label, with the robot joint control instructions marked as level L0, the highest priority. Prediction module: This module uses a spatiotemporal convolutional neural network to predict the distribution of spectrum holes within the next 10ms based on the quantized data blocks of the spatiotemporal hash grid and real-time channel state information. It dynamically allocates micro-slot resources within the orthogonal frequency division multiplexing symbol period, prioritizes mapping L0-level instructions to low-interference subcarriers, and uses polarization codes to perform asymmetric redundant encoding on micro-slot data units to generate dynamically adjusted micro-slot resource allocation results. Computing module: Based on the micro-slot resource allocation results and the real-time electromagnetic interference heat map generated by distributed sensors, it calculates the optimal phase offset matrix of the metasurface intelligent reflector through a deep deterministic policy gradient algorithm, dynamically reconstructs the electromagnetic wave propagation path, and integrates full-duplex self-interference cancellation technology to achieve concurrent uplink and downlink transmission in the same frequency band, thereby obtaining the optimized electromagnetic wave propagation path; The generation module receives the channel impulse response and bit error rate (BER) obtained from the feedback end based on the optimized electromagnetic wave propagation path. It uses a time-frequency dual-domain joint equalizer to collaboratively compensate for multipath fading and narrowband interference. If the BER exceeds a preset threshold, it triggers an incremental redundancy retransmission mechanism based on federated learning. This mechanism generates a globally optimized parity check matrix by aggregating the gradients of the lightweight error correction model trained locally on each node. Only the sparse parity bits of the damaged mini-slot are retransmitted to obtain the compensated data stream. Build an optimization module: This module is used to construct a causal graph model to dynamically prune high-entropy paths to minimize causal entropy based on the compensated data stream, real-time device temperature, and subcarrier load rate data obtained from transceiver statistics. Through multi-agent reinforcement learning, the optimal modulation order and subcarrier switching strategy are determined with delay-energy efficiency as the game objective. A phase change material heat dissipation module is embedded in the module to adjust the optical module drive current, thereby obtaining an optimized stable communication link.

7. The optical-electrical hybrid ultra-low latency communication system according to claim 6, characterized in that: The acquisition module includes: The first generation unit is used to obtain GPS pulse-per-second signals based on the factory local area network infrastructure, receive IEEE 1588 protocol synchronization messages, generate a quantum random number entropy source through a quantum random number generator based on an optical quantum noise chip, encrypt the clock synchronization channel through the quantum key distribution protocol, and generate a quantum key stream; The second generation unit is used to combine the random phase noise of the quantum entropy source with the phase of the classical clock source based on the quantum key stream, and use the quantum Kalman filter algorithm to eliminate clock drift and generate a globally synchronized clock signal with a clock jitter of less than 1ppm. The jitter calculation formula is as follows: Where Jitter is the clock jitter, N is the number of clock cycle samples measured, and t i is the measured value of the clock period, is the theoretical period, i For the i clock cycle samples; Segmentation unit: used to build a spatiotemporal hash grid based on the global synchronous clock signal, obtain the coordinates of the spatiotemporal hash grid, and segment the optical-electrical hybrid data stream into quantized data blocks. Each data block is bound to a unique spatiotemporal hash value and priority label, among which the robot joint control instructions are marked as the highest priority level L0.

8. The optical-electrical hybrid ultra-low latency communication system according to claim 6, characterized in that: The prediction module includes: Prediction unit: Based on the spatiotemporal hash value and priority tag of the quantized data block, combined with historical spectrum occupancy and real-time channel status information, the prediction unit uses a spatiotemporal convolutional neural network to predict the spectrum hole distribution within the next 10ms. The calculation formula is as follows: In the formula, Mask priority (P) is the priority mask matrix, Holes(f,t) is the predicted spectrum hole distribution, STCNN(S hist ,SNR) is the spatiotemporal convolutional neural network; Division unit: used to dynamically divide the OFDM symbol period according to the predicted spectrum hole distribution The duration of each mini-slot is Δ t =1μs, bandwidth is Δ f =4.3125kHz, preferentially mapping L0 level instructions to low-interference subcarriers, where the low-interference subcarriers meet the signal-to-noise ratio threshold greater than or equal to 25dB, and generating a mini-slot resource allocation table; Coding unit: Used to perform polarization code encoding on the mini-slot data unit based on the mini-slot resource allocation table to generate the following codewords: Where, For the encoded codeword, PolarEncode(D slot ,R) is the mini-slot data D slot Perform polarization coding and add redundant information with redundancy R; Among them, redundancy R The dynamic adjustment based on priority is as follows: Where R is the redundancy, P is the priority label of the data block, L0 is the highest priority, L1 is the medium priority, and L2 is the lowest priority.

9. The optical-electrical hybrid ultra-low latency communication system according to claim 6, characterized in that: The computing module includes: A model building unit is used to establish a deep deterministic policy gradient model based on the micro-slot resource allocation results and a real-time electromagnetic interference heat map generated by distributed spectrum sensors based on electromagnetic interference intensity. The micro-slot resource allocation results and the real-time electromagnetic interference heat map are input into the deep deterministic policy gradient model, which outputs a phase offset matrix for the metasurface smart reflector. The training objective of the deep deterministic policy gradient model is to minimize path loss. The phase offset of the reflector unit is iteratively optimized to minimize the total path loss after the channel responses of the reflected path and the direct path are superimposed. The optimization unit is used to dynamically adjust the phase offset of the metasurface intelligent reflector according to the phase offset matrix, guiding the electromagnetic wave propagation path to bypass the interference area to obtain an optimized path. Based on the optimized path, the minimum mean square error algorithm is used to cancel the self-interference of the uplink and downlink signals in the same frequency band. This includes: constructing the autocorrelation matrix of the self-interference signal based on the known signal waveform and channel impulse response at the transmitting end, and solving the optimal interference cancellation weight matrix under the minimum mean square error criterion by maximizing the signal-to-interference-noise ratio. The calculation formula is as follows: Where W MMSE is the optimal interference cancellation weight matrix, R xy is the cross-correlation matrix between the self-interference signal and the desired signal, is the inverse matrix of the autocorrelation matrix; Processing unit: Used to apply the optimal interference cancellation weight matrix to filter the received signal, suppress the self-interference component in the same frequency band, and thus output the full-duplex concurrent transmission configuration parameters to obtain the optimized electromagnetic wave propagation path.

10. The optical-electrical hybrid ultra-low latency communication system according to claim 6, characterized in that: The generation module includes: Design unit: Used to design a time-frequency dual-domain joint equalizer based on the optimized phase offset matrix of the metasurface smart reflector in the electromagnetic wave propagation path, the channel impulse response and bit error rate fed back by the receiver. This includes: calculating the time-domain equalization weights by minimizing the inter-symbol interference criterion to compensate for the delay caused by multipath fading, identifying the interference frequency set based on the narrowband interference power spectral density, and calculating the frequency-domain suppression coefficient; superimposing the time-domain equalization weights with the frequency-domain suppression results to generate an equalized low-error signal stream; Extraction and recognition unit: used if the bit error rate BER>10 -6 , extract the damaged micro-time slot index, identify the fault frequency and time slice, where the communication node performs local training, where the local training includes: using the original data of the damaged micro-time slot and the equalized received signal as input data, training a lightweight error correction model to minimize the reconstruction error and generate a local gradient; Update error correction unit: Based on the local gradient, it is encrypted and uploaded to the central server through the secure multi-party computing protocol, aggregated to generate the global gradient, update the global check matrix, retransmit only the sparse check bits of the damaged micro-time slot, obtain the global optimized check matrix, perform error correction decoding on the global optimized check matrix and the sparse check bits, and restore the compensated data stream.

Images