Multi-mode optical signal transmission method and device, electronic equipment and storage medium

Abstract

The invention discloses a multi-mode optical signal transmission method and device, electronic equipment and a storage medium, and relates to the technical field of computers.The transmitted data stream is divided into a control signal and a data signal, the control signal is a visible light wave band, the data signal is an infrared wave band, and the control signal and the data signal are combined to a single multi-core optical fiber through a combiner; the optical network controller allocates subcarrier resources as required according to the change of network flow in a spectrum dynamic allocation mode to perform data transmission, so that the subcarrier resource allocation adapted to the change of network flow improves the dynamic adaptability, and the allocation of the subcarrier resources as required improves the equipment cooperation efficiency, and in the data transmission process, the data transmission efficiency is improved. Clock synchronization and signal encryption are realized through a quantum key distribution mode, and the security is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a multimodal optical signal transmission method, device, electronic device, and storage medium. Background Art

[0002] Significant progress has been made in PCIe (PCIe) optical interconnect technology, such as increasing the PCIe Gen5 signal transmission distance to 30 meters through mixed-rate linear optical transmission solutions, and achieving efficient long-distance transmission through fiber extension cables. However, existing technologies still have the following limitations: insufficient dynamic adaptability, with traditional solutions relying on fixed modulation formats (such as 16QAM) and unable to optimize transmission efficiency in real time based on channel quality; low multi-device collaboration efficiency, with existing multi-channel multiplexing technologies (such as dense wavelength division multiplexing) lacking intelligent scheduling mechanisms, which can easily lead to resource contention and delays; and weak security, with optical signal transmission susceptible to physical layer eavesdropping and existing technologies lacking quantum-level encryption. Summary of the Invention

[0003] The present application provides a multimodal optical signal transmission method, device, electronic device and storage medium to at least solve the problems of insufficient dynamic adaptability, low equipment coordination efficiency and weak security in related technologies.

[0004] The present application provides a multimodal optical signal transmission method, comprising:

[0005] Build and train deep reinforcement learning models;

[0006] Monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate the channel quality index;

[0007] Inputting the channel quality index into the deep reinforcement learning model, and selecting an optimal modulation format for transmitting data streams based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate;

[0008] The data stream is divided into a control signal and a data signal, wherein the control signal is in the visible light band and the data signal is in the infrared band. The control signal is combined into a single multi-core optical fiber through a combiner, and an optical network controller is used to dynamically allocate subcarrier resources on demand according to changes in network traffic for data transmission;

[0009] During data transmission, clock synchronization and signal encryption are achieved through quantum key distribution.

[0010] The present application also provides a multimodal optical signal transmission device, comprising:

[0011] AI dynamic modulation engine for building and training deep reinforcement learning models;

[0012] Channel quality index monitoring module, used to monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate the channel quality index;

[0013] a modulation format selection module, configured to input the channel quality index into the deep reinforcement learning model, and select an optimal modulation format for transmitting a data stream based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate;

[0014] A multimodal optical signal processor is used to split the data stream into a control signal and a data signal, where the control signal is in the visible light band and the data signal is in the infrared band. The control signal and the data signal are combined into a single multi-core optical fiber through a combiner. The optical network controller is used to dynamically allocate subcarrier resources on demand for data transmission based on changes in network traffic through spectrum dynamic allocation.

[0015] The quantum synchronization unit is used to achieve clock synchronization and signal encryption through quantum key distribution during data transmission.

[0016] The present application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for implementing the steps of any of the above-mentioned multimodal optical signal transmission methods when executing the computer program:

[0017] Build and train deep reinforcement learning models;

[0018] Monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate the channel quality index;

[0019] Inputting the channel quality index into the deep reinforcement learning model, and selecting an optimal modulation format for transmitting data streams based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate;

[0020] The data stream is divided into a control signal and a data signal, wherein the control signal is in the visible light band and the data signal is in the infrared band. The control signal is combined into a single multi-core optical fiber through a combiner, and an optical network controller is used to dynamically allocate subcarrier resources on demand according to changes in network traffic for data transmission;

[0021] During data transmission, clock synchronization and signal encryption are achieved through quantum key distribution.

[0022] The present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of any of the above-mentioned multimodal optical signal transmission methods are implemented:

[0023] Build and train deep reinforcement learning models;

[0024] Monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate the channel quality index;

[0025] Inputting the channel quality index into the deep reinforcement learning model, and selecting an optimal modulation format for transmitting data streams based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate;

[0026] The data stream is divided into a control signal and a data signal, wherein the control signal is in the visible light band and the data signal is in the infrared band. The control signal is combined into a single multi-core optical fiber through a combiner, and an optical network controller is used to dynamically allocate subcarrier resources on demand according to changes in network traffic for data transmission;

[0027] During data transmission, clock synchronization and signal encryption are achieved through quantum key distribution.

[0028] Through this application, the transmitted data stream is divided into a control signal and a data signal, the control signal is in the visible light band, and the data signal is in the infrared band. They are combined into a single multi-core optical fiber through a combiner, and an optical network controller is used to allocate subcarrier resources on demand for data transmission in a spectrum dynamic allocation manner according to changes in network traffic. In this way, subcarrier resource allocation that adapts to changes in network traffic improves dynamic adaptability, and the on-demand allocation of subcarrier resources improves the equipment coordination efficiency. In the process of data transmission, clock synchronization and signal encryption are achieved through quantum key distribution, thereby improving security. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0030] Figure 1 This is a diagram of an application environment of a multimodal optical signal transmission method according to an embodiment of the present application;

[0031] Figure 2 This is a flow chart of a multi-modal optical signal transmission method according to an embodiment of the present application;

[0032] Figure 3 This is a flow chart of dividing a data stream into a control signal and a data signal in one embodiment of the present application;

[0033] Figure 4 This is a schematic diagram of a process for adjusting the gain of an optical amplifier according to a channel attenuation model in one embodiment of the present application;

[0034] Figure 5This is a structural block diagram of a multi-modal optical signal transmission device in one embodiment of the present application;

[0035] Figure 6 This is a diagram of the internal structure of a computer device in one embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0037] It should be noted that, in the description of this application, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. The terms "first," "second," etc., in this application are used to distinguish similar objects, and are not used to describe a particular order or sequence.

[0038] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0039] The multimodal optical signal transmission method provided in this application can be applied to Figure 1 The system significantly improves transmission efficiency, security, and scalability by integrating AI-driven adaptive modulation, visible light / infrared coordinated transmission, and quantum encryption technologies. The system includes an intelligent resource scheduler and modular optical interfaces, making it suitable for next-generation data centers and edge computing scenarios.

[0040] like Figure 2 As shown, an embodiment of the present application provides a multimodal optical signal transmission method, comprising the following steps:

[0041] Step S1, building and training a deep reinforcement learning model;

[0042] Step S2, real-time monitoring of the eye diagram parameters, noise level and bit error rate of the optical fiber link to generate a channel quality index;

[0043] Step S3: inputting the channel quality index into the deep reinforcement learning model, and selecting the optimal modulation format for transmitting the data stream based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate;

[0044] Step S4: Split the data stream into a control signal and a data signal. The control signal is in the visible light band, and the data signal is in the infrared band. The signals are combined into a single multi-core optical fiber through a combiner. The optical network controller is used to dynamically allocate subcarrier resources on demand for data transmission based on changes in network traffic.

[0045] Step S5: During data transmission, clock synchronization and signal encryption are achieved through quantum key distribution.

[0046] This embodiment divides the transmitted data stream into control signals and data signals, where the control signal is in the visible light band and the data signal is in the infrared band. The signals are combined into a single multi-core optical fiber through a combiner, and an optical network controller is used to allocate subcarrier resources on demand for data transmission in a spectrum dynamic allocation manner according to changes in network traffic. This improves dynamic adaptability by allocating subcarrier resources in accordance with changes in network traffic, and the on-demand allocation of subcarrier resources improves device collaboration efficiency. In the process of data transmission, clock synchronization and signal encryption are achieved through quantum key distribution, thereby improving security.

[0047] Among them, the eye diagram parameters include eye height and eye width, eye height ≥ 0.8UI, eye width ≥ 0.6UI, and noise level SNR ≥ 20dB.

[0048] Among them, real-time monitoring of the eye diagram parameters, noise level and bit error rate of the optical fiber link and generation of the channel quality index include:

[0049] The channel quality index is calculated as follows: CQI = a*(Eye_Height / Eye_Height_max)+b*(Eye_Width / Eye_Width_max)+c*log2(1+SNR)+d*Historical_CQI_avg;

[0050] Where Eye_Height is the eye height, Eye_Width is the eye width, Eye_Height / Eye_Width is the currently measured vertical / horizontal eye opening, Eye_Height_max is the maximum eye height under ideal conditions, Eye_Width_max is the maximum eye width under ideal conditions, Eye_Height_max / Eye_Width_max is the maximum eye opening under ideal conditions, SNR is the noise level, Historical_CQI_avg is the sliding average of the historical channel quality index (such as the average of the past 10 measurements), and a, b, c, and d are weighting coefficients, where a+b+c+d=1.

[0051] This application introduces a sliding average of historical channel quality indexes in combination with eye diagram parameters, noise level, and bit error rate to obtain the current channel quality index, which can improve the accuracy of the channel quality index.

[0052] like Figure 3 As shown, a multimodal optical signal processor is used to split the data stream into control and data signals. This supports the coordinated transmission of visible light (380-750nm) and infrared light (850-1550nm), leveraging the characteristics of optical signals in different wavelength bands to achieve more efficient data transmission. This multimodal transmission method improves the system's transmission capacity and anti-interference capabilities. The visible light band is used to transmit low-latency control commands (such as PREST# and CLK), while the infrared band carries high-speed data streams (such as communication between AI accelerators and GPUs). This signal segmentation method fully utilizes the characteristics of optical signals in different wavelength bands to achieve efficient transmission of control and data signals. Dynamic spectrum allocation utilizes software-defined optical networking (SDON) technology to allocate subcarrier resources on demand. For example, in bursty traffic scenarios, a backup wavelength (such as 1550nm) is automatically activated to expand bandwidth, supporting single-channel rates up to 128GT / s (PCIe Gen7 standard). By flexibly allocating spectrum resources, SDON technology can quickly respond to changes in network traffic and improve system resource utilization.

[0053] In this embodiment, building and training a deep reinforcement learning model (DRL) includes:

[0054] Construct a deep reinforcement learning model (DRL) through state space, action space and reward function;

[0055] The state space settings include the channel quality index (CQI), payload, and environmental parameters. The payload is the current data rate (e.g., 128 GT / s) and link utilization. The environmental parameters include temperature, electromagnetic interference (EMI), and fiber length (L).

[0056] Setting the action space includes modulation format selection, gain coefficient adjustment, and spectrum allocation strategy. Modulation format selection involves dynamically switching the modulation format from a set of modulation formats. Gain coefficient adjustment involves adjusting the optical amplifier gain based on a channel attenuation model.

[0057] The reward function is set to the sum of the throughput reward and the bit error rate penalty as the comprehensive reward. The throughput reward is proportional to the total throughput. The bit error rate penalty is a negative reward if the bit error rate (BER) of any subcarrier exceeds a preset threshold.

[0058] Monte Carlo Tree Search (MCTS) is used to explore the action space, the policy parameters of the deep reinforcement learning model (DRL) are optimized by gradient ascent, and the deep reinforcement learning model (DRL) is trained using historical state-action-reward tuples.

[0059] Based on a deep reinforcement learning (DRL) model, it can perceive channel conditions in real time and optimize modulation methods and gain coefficients to adapt to changing network environments. The DRL model continuously learns the optimal modulation strategy through interaction with the environment, improving overall system performance.

[0060] Among them, a deep reinforcement learning model is constructed by jointly constructing the state space, action space and reward function, so that the model can adjust the action through the state, and adjust the reward based on training to optimize the state-action-reward relationship, thereby improving the overall performance of the system.

[0061] like Figure 4 As shown, in this embodiment, when setting the action space including modulation format selection, gain coefficient adjustment and spectrum allocation strategy, it also includes:

[0062] The gain coefficient is adjusted to adjust the optical amplifier gain according to the channel attenuation model. The formula is A(t) = α*e -β*L +γ*CQI(t), where α, β, and γ are parameters trained by the deep reinforcement learning model (DRL), L is the fiber length, and CQI(t) is the channel quality index at time t;

[0063] The way to set the reward function as the sum of throughput reward and bit error rate penalty as the comprehensive reward is: set the throughput reward to Among them, B i is the bandwidth of the ith subcarrier (unit: Hz), M i is the modulation order of the ith subcarrier, indicating the number of bits carried by each symbol; the bit error rate penalty is set to impose a negative reward R when the bit error rate (BER) of any subcarrier exceeds the preset threshold. penalty ; The total reward is R total =R throughput +R penalty .

[0064] Here, α, β, and γ are parameters of the deep reinforcement learning (DRL) model optimized and trained through 100,000 iterations. The reward function is set as the sum of the throughput reward and the bit error rate penalty. With the dual objectives of increasing throughput and reducing bit error rate, the optimization strategy is implemented using Monte Carlo Tree Search (MCTS).

[0065] In this embodiment, a channel quality index (CQI) is input into a deep reinforcement learning model (DRL), and an optimal modulation format is selected based on the channel quality index CQI to transmit a data stream with the goal of maximizing throughput and minimizing bit error rate, including:

[0066] Get the modulation format set in the deep reinforcement learning model (DRL). The modulation format set is {16QAM, 64QAM, OOK}, where 16QAM is M i =16, corresponding to log2(16)=4 bits / symbol; 64QAM is M i =64, corresponding to log2(64)=6 bits / symbol; OOK is on-off keying, which is M i =2, corresponding to log2(2)=1 bit / symbol;

[0067] Set up an objective function to maximize throughput and minimize bit error rate: Set the constraint condition to the bit error rate BER of the i-th subcarrier i Meet BER i ≤10 -12 , where i = 1, 2, 3, ..., n;

[0068] Based on the objective function, the total throughput of all subcarriers is maximized, that is, the sum of the product of spectral efficiency and bandwidth, to obtain the optimal modulation format corresponding to the channel quality index (CQI). When the channel quality is good, the high-order modulation 64QAM is used, and in low signal-to-noise ratio environments, the order is automatically downgraded to 16QAM or OOK.

[0069] Determine one by one whether the bit error rate of each subcarrier meets the constraints. If so, maintain the modulation order and calculate the gain coefficient. Otherwise, adjust the modulation order and use an adjustable gain amplifier to dynamically adjust the gain coefficient according to the predicted channel attenuation to dynamically compensate for the transmission loss.

[0070] The objective function is to maximize the total throughput of all subcarriers, which is the sum of the product of spectrum efficiency and bandwidth. The constraint condition is to set the bit error rate of each subcarrier to be less than or equal to 10. -12 , ensuring high-reliability transmission (such as financial transactions and supercomputing scenarios).

[0071] In this embodiment, using an adjustable gain amplifier to dynamically adjust a gain coefficient according to predicted channel attenuation to dynamically compensate for transmission loss includes:

[0072] The compensation formula for setting the adjustable gain amplifier to dynamically adjust the gain coefficient according to the predicted channel attenuation is P out =A(t)*P in ; where P inis the gain factor of the input adjustable gain amplifier, P out is the compensated gain factor output from the adjustable gain amplifier.

[0073] In this embodiment, during the data transmission process, clock synchronization and signal encryption are achieved through quantum key distribution, including:

[0074] A quantum source generates entangled photon pairs, and a synchronous clock signal is generated through the quantum entangled photon pairs;

[0075] Distribute the synchronous clock signal to the mainboard and backplane through the beam splitter;

[0076] The mainboard and backplane achieve phase synchronization based on timestamp comparison, and the least squares method is used for error compensation during phase synchronization;

[0077] Deploy quantum key distribution (QKD) units between backplanes to generate quantum keys, set the quantum key update frequency, and distribute quantum keys using quantum entangled photon pairs;

[0078] The control signal is quantum-encoded according to the quantum key to achieve signal encryption.

[0079] The wavelength of the entangled photon pair is 1310nm / 1550nm. The BB84 protocol is used to generate quantum keys, with a key update frequency of ≥1MHz and an eavesdropping detection rate of ≥99.99%. When distributing quantum keys, the QKD unit uses the characteristics of quantum entangled photon pairs to securely distribute keys and provide encryption protection for data transmission. Signal encryption involves quantum state encoding of PCIE control signals (such as PERST# and CLK). For example, the CLK signal is mapped to the polarization state of the photon (H / V or +45° / -45°), which can effectively prevent tampering and ensure the security of data transmission.

[0080] In this embodiment, the multimodal optical signal transmission method further includes:

[0081] Optimize bandwidth allocation and path selection between multiple devices through federated learning;

[0082] Set the path selection formula to Among them, Delay i is the delay of the i-th subcarrier, ∑ i Delay i is the path delay, BER i is the bit error rate of the i-th subcarrier, λ is the weight coefficient, Path k The optimal path selected is the kth path;

[0083] Based on the graph neural network (GNN), the topology and traffic characteristics of the link are learned, the link congestion is predicted, the optimal path is dynamically selected according to the path selection formula, and the switching transmission path is controlled in real time.

[0084] Each backplane node trains its own bandwidth allocation model, while a central server aggregates the global strategy to achieve low-latency collaboration. Federated learning, while protecting data privacy, aggregates the local models of each node to generate a global optimization strategy, improving overall system performance.

[0085] The above-mentioned multimodal optical signal transmission method has the following technical effects:

[0086] 1. Transmission efficiency and spectrum utilization are significantly improved

[0087] Dynamic modulation optimization:

[0088] The deep reinforcement learning (DRL) model adjusts the modulation format (16QAM / 64QAM / OOK) in real time, enabling higher-order modulation (such as 64QAM) when channel quality is good, improving spectral efficiency by 50%.

[0089] Automatically downgrade to 16QAM or OOK in low signal-to-noise ratio environments, ensuring a stable bit error rate (BER) of ≤10 -12 , meeting the stringent requirements of supercomputing centers and financial transaction scenarios.

[0090] Multimodal collaborative transmission:

[0091] Visible light (VLC band) and infrared light (850-1550nm) share control signals and data streams. A single multi-core optical fiber supports 128-channel PCIE Gen6 signal transmission, and the spectrum efficiency is 2 times higher than that of traditional solutions.

[0092] 2. Comprehensively enhanced anti-interference capabilities and security

[0093] Optical signal anti-electromagnetic interference (EMI):

[0094] The all-optical link design completely avoids electromagnetic interference from metal cables, and the bit error rate fluctuation is less than 0.1% in strong electromagnetic environments (such as industrial control scenarios);

[0095] Quantum-level encryption and synchronization:

[0096] The BB84 protocol is used to encode the quantum state of PCIE control signals (such as CLK, PERST#), with an eavesdropping detection rate of ≥99.99%, a key update frequency of 1MHz, and financial-grade security (risk ≤10 -9 );

[0097] Quantum entangled photon pairs achieve picosecond clock synchronization (error ≤ 1ps), eliminating the cumulative error of traditional PLLs and are suitable for high-frequency trading and real-time control systems.

[0098] 3. Achieve precise clock synchronization

[0099] The quantum synchronization unit generates a synchronous clock signal through quantum entangled photon pairs, eliminating the cumulative error of traditional PLL, and achieving synchronization accuracy at the picosecond level, providing high-precision guarantee for the system's clock synchronization.

[0100] 4. Adapt to sudden traffic

[0101] It adopts software-defined optical network (SDON) technology to allocate subcarrier resources on demand, supports elastic expansion under burst traffic, can quickly respond to changes in network traffic, and improve system resource utilization and flexibility.

[0102] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method.

[0103] This application can not only be used for PCIE backplane connections inside servers, but can also be extended to high-speed interconnection between server racks in data centers, achieving a more efficient data center network architecture, improving the overall data transmission efficiency of the data center, reducing network latency, and meeting the needs of cloud computing, big data and other applications for high-performance computing and storage in data centers.

[0104] As energy consumption in data centers and communication networks continues to grow, reducing energy consumption has become an important research direction. Based on this invention, the system's energy management strategy can be further optimized, such as by adopting dynamic power control technology to automatically adjust the output power of the optical transmitter and the gain of the optical amplifier according to actual data transmission requirements to avoid energy waste; or by developing efficient optoelectronic conversion chips and optical signal processing algorithms to reduce the system's energy consumption during signal conversion and processing. Furthermore, energy-saving designs at the hardware level can be combined, such as using low-power lasers, detectors, and digital signal processors, as well as optimizing circuit board layout and heat dissipation design, to improve the system's energy efficiency, reduce operating costs, and achieve green communication and computing.

[0105] The software-defined optical transport network (SDON) architecture is adopted to separate the control plane and data plane of the optical transport network, and the entire optical transport system is software-defined and dynamically managed through a centralized controller. Users can flexibly configure and manage the parameters of the optical transmission link, such as modulation format, bandwidth allocation, routing selection, etc. through the software interface, realizing the rapid deployment, dynamic adjustment and automated operation and maintenance of the optical transmission network. For example, when the business needs of the data center change, the network administrator only needs to perform simple operations on the controller to reallocate the bandwidth resources of the optical transmission link and adjust the modulation format to adapt to the new transmission needs, without having to manually adjust the hardware equipment. This greatly improves the flexibility and scalability of the system and can quickly respond to changing business needs and technological development trends.

[0106] In one embodiment, Figure 1 、 Figure 5 As shown, a multimodal optical signal transmission device 10 is provided, including: an AI dynamic modulation engine 1, a channel quality index monitoring module 2, a modulation format selection module 3, a multimodal optical signal processor 4, and a quantum synchronization unit 5.

[0107] AI dynamic modulation engine 1 is used to build and train deep reinforcement learning models.

[0108] The channel quality index monitoring module 2, namely the CQI monitoring module, is used to monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate a channel quality index.

[0109] The modulation format selection module 3 is used to input the channel quality index into the deep reinforcement learning model, and select the optimal modulation format to transmit the data stream based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate.

[0110] The multimodal optical signal processor 4 is used to split the data stream into control signals and data signals. The control signal is in the visible light band, and the data signal is in the infrared band. They are combined into a single multi-core optical fiber through a combiner. The optical network controller is used to allocate subcarrier resources on demand for data transmission in a dynamic spectrum allocation manner according to changes in network traffic.

[0111] The quantum synchronization unit 5 is used to achieve clock synchronization and signal encryption through quantum key distribution during data transmission.

[0112] In this embodiment, building and training a deep reinforcement learning model (DRL) includes:

[0113] Construct a deep reinforcement learning model (DRL) through state space, action space and reward function;

[0114] The state space settings include the channel quality index (CQI), payload, and environmental parameters. The payload is the current data rate (e.g., 128 GT / s) and link utilization. The environmental parameters include temperature, electromagnetic interference (EMI), and fiber length (L).

[0115] Setting the action space includes modulation format selection, gain coefficient adjustment, and spectrum allocation strategy. Modulation format selection involves dynamically switching the modulation format from a set of modulation formats. Gain coefficient adjustment involves adjusting the optical amplifier gain based on a channel attenuation model.

[0116] The reward function is set to the sum of the throughput reward and the bit error rate penalty as the comprehensive reward. The throughput reward is proportional to the total throughput. The bit error rate penalty is a negative reward if the bit error rate (BER) of any subcarrier exceeds a preset threshold.

[0117] Monte Carlo Tree Search (MCTS) is used to explore the action space, the policy parameters of the deep reinforcement learning model (DRL) are optimized by gradient ascent, and the deep reinforcement learning model (DRL) is trained using historical state-action-reward tuples.

[0118] In this embodiment, when setting the action space including modulation format selection, gain coefficient adjustment, and spectrum allocation strategy, it also includes:

[0119] The gain coefficient is adjusted to adjust the optical amplifier gain according to the channel attenuation model. The formula is A(t) = α*e -β*L +γ*CQI(t), where α, β, and γ are parameters trained by the deep reinforcement learning model (DRL), L is the fiber length, and CQI(t) is the channel quality index at time t;

[0120] The way to set the reward function as the sum of throughput reward and bit error rate penalty as the comprehensive reward is: set the throughput reward to Among them, B i is the bandwidth of the ith subcarrier (unit: Hz), M i is the modulation order of the ith subcarrier, indicating the number of bits carried by each symbol; the bit error rate penalty is set to impose a negative reward R when the bit error rate (BER) of any subcarrier exceeds the preset threshold. penalty ; The total reward is R total =R throughput +R penalty .

[0121] In this embodiment, a channel quality index (CQI) is input into a deep reinforcement learning model (DRL). Based on the channel quality index CQI, the optimal modulation format (16QAM / 64QAM / OOK) is selected to transmit the data stream with the goal of maximizing throughput and minimizing bit error rate. The method includes:

[0122] Get the modulation format set in the deep reinforcement learning model (DRL). The modulation format set is {16QAM, 64QAM, OOK}, where 16QAM is M i =16, corresponding to log2(16)=4 bits / symbol; 64QAM is M i =64, corresponding to log2(64)=6 bits / symbol; OOK is on-off keying, which is M i =2, corresponding to log2(2)=1 bit / symbol;

[0123] Set up an objective function to maximize throughput and minimize bit error rate: Set the constraint condition to the bit error rate BER of the i-th subcarrier i Meet BER i ≤10 -12 , where i = 1, 2, 3, ..., n;

[0124] According to the objective function, the total throughput of all subcarriers is maximized (i.e., the sum of the product of spectral efficiency and bandwidth), and the optimal modulation format corresponding to the channel quality index (CQI) is obtained (high-order modulation (such as 64QAM) is enabled when the channel quality is good, and the order is automatically downgraded to 16QAM or OOK in low signal-to-noise ratio environments);

[0125] Determine one by one whether the bit error rate of each subcarrier meets the constraints. If so, maintain the modulation order and calculate the gain coefficient. Otherwise, adjust the modulation order and use an adjustable gain amplifier to dynamically adjust the gain coefficient according to the predicted channel attenuation to dynamically compensate for the transmission loss.

[0126] In this embodiment, using an adjustable gain amplifier to dynamically adjust a gain coefficient according to predicted channel attenuation to dynamically compensate for transmission loss includes:

[0127] The compensation formula for setting the adjustable gain amplifier to dynamically adjust the gain coefficient according to the predicted channel attenuation is P out =A(t)*P in ; where P in is the gain factor of the input adjustable gain amplifier, P out is the compensated gain factor output from the adjustable gain amplifier.

[0128] In this embodiment, clock synchronization and signal encryption are achieved through quantum key distribution during data transmission, including:

[0129] A quantum source generates entangled photon pairs, and a synchronous clock signal is generated through the quantum entangled photon pairs;

[0130] Distribute the synchronous clock signal to the mainboard and backplane through the beam splitter;

[0131] The mainboard and backplane achieve phase synchronization based on timestamp comparison, and the least squares method is used for error compensation during phase synchronization;

[0132] Deploy quantum key distribution (QKD) units between backplanes to generate quantum keys (using the BB84 protocol), set the quantum key update frequency, and distribute the quantum keys using quantum entangled photon pairs;

[0133] The control signal is quantum-encoded according to the quantum key to achieve signal encryption.

[0134] In this embodiment, if Figure 5 As shown, the multi-mode optical signal transmission device 10 further includes: an intelligent resource scheduler 6. The intelligent resource scheduler 6 is used to:

[0135] Optimize bandwidth allocation and path selection between multiple devices through federated learning;

[0136] Set the path selection formula to Among them, Delay i is the delay of the i-th subcarrier, ∑ i Delay i is the path delay, BER i is the bit error rate of the i-th subcarrier, λ is the weight coefficient, Path k The optimal path selected is the kth path;

[0137] Based on the graph neural network (GNN), the topology and traffic characteristics of the link are learned, the link congestion is predicted, the optimal path is dynamically selected according to the path selection formula, and the switching transmission path is controlled in real time.

[0138] In the above-mentioned multimodal optical signal transmission device, technical effect.

[0139] For the description of the features in the embodiments corresponding to the multimodal optical signal transmission apparatus, reference may be made to the relevant description of the embodiments corresponding to the multimodal optical signal transmission method, and will not be repeated here.

[0140] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any of the above-mentioned multi-modal optical signal transmission method embodiments.

[0141] In one embodiment, the electronic device may be a server, and its internal structure diagram may be as follows: Figure 6As shown. The electronic device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the electronic device is used to store multimodal optical signal transmission data. The network interface of the electronic device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a multimodal optical signal transmission method is implemented.

[0142] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above-mentioned multi-modal optical signal transmission method embodiments when running.

[0143] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0144] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the computer program implements the steps in any of the above-mentioned multi-modal optical signal transmission method embodiments.

[0145] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above-mentioned multi-modal optical signal transmission method embodiments are implemented.

[0146] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0147] The above is a detailed introduction to a multimodal optical signal transmission method, device, electronic device and storage medium provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. A multimodal optical signal transmission method, characterized in that: include: Build and train deep reinforcement learning models; Monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate the channel quality index; Inputting the channel quality index into the deep reinforcement learning model, and selecting an optimal modulation format for transmitting data streams based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate; The data stream is divided into a control signal and a data signal, wherein the control signal is in the visible light band and the data signal is in the infrared band. The control signal is combined into a single multi-core optical fiber through a combiner, and an optical network controller is used to dynamically allocate subcarrier resources on demand according to changes in network traffic for data transmission; During data transmission, clock synchronization and signal encryption are achieved through quantum key distribution.

2. The multimodal optical signal transmission method according to claim 1, wherein: The construction and training of the deep reinforcement learning model includes: Build a deep reinforcement learning model through state space, action space and reward function; Setting the state space to include a channel quality index, a load, and environmental parameters, wherein the load is the current data transmission rate and link utilization, and the environmental parameters include temperature, electromagnetic interference level, and optical fiber length; Setting the action space includes modulation format selection, gain coefficient adjustment, and spectrum allocation strategy, wherein the modulation format selection is to dynamically switch the modulation format from a modulation format set, and the gain coefficient adjustment is to adjust the optical amplifier gain according to a channel attenuation model; The reward function is set to be the sum of a throughput reward and a bit error rate penalty as a comprehensive reward, wherein the throughput reward is proportional to the total throughput, and the bit error rate penalty is a negative reward imposed if the bit error rate of any subcarrier exceeds a preset threshold; Monte Carlo tree search is used to explore the action space, the policy parameters of the deep reinforcement learning model are optimized by gradient ascent, and the deep reinforcement learning model is trained using historical state-action-reward tuples.

3. The multimodal optical signal transmission method according to claim 2, wherein: When setting the action space including modulation format selection, gain coefficient adjustment and spectrum allocation strategy, it also includes: The gain coefficient is adjusted to adjust the optical amplifier gain according to the channel attenuation model. The formula is A(t)=α*e -β*L +γ*CQI(t), where α, β, and γ are the parameters trained by the deep reinforcement learning model, L is the fiber length, and CQI(t) is the channel quality index at time t; The reward function is set to the sum of the throughput reward and the bit error rate penalty as the comprehensive reward: the throughput reward is set to Among them, B i is the bandwidth of the i-th subcarrier, M i is the modulation order of the ith subcarrier, indicating the number of bits carried by each symbol; the bit error rate penalty is set to impose a negative reward R when the bit error rate of any subcarrier exceeds a preset threshold. penalty ; The comprehensive reward is R total =R throughput +R penalty .

4. The multimodal optical signal transmission method according to claim 3, wherein: Inputting the channel quality index into the deep reinforcement learning model, and selecting an optimal modulation format for transmitting the data stream based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate includes: Get the modulation format set in the deep reinforcement learning model, the modulation format set is {16QAM, 64QAM, OOK}, where 16QAM is M i =16, corresponding to log2(16)=4 bits / symbol; 64QAM is M i =64, corresponding to log2(64)=6 bits / symbol; OOK is on-off keying, which is M i =2, corresponding to log2(2)=1 bit / symbol; Set up an objective function to maximize throughput and minimize bit error rate: Set the constraint condition to the bit error rate BER of the i-th subcarrier i Meet BER i ≤10 -12 , where i = 1, 2, 3, ..., n; Maximizing the total throughput of all subcarriers according to the objective function to obtain an optimal modulation format corresponding to the channel quality index; Determine one by one whether the bit error rate of each subcarrier meets the constraints. If so, maintain the modulation order and calculate the gain coefficient. Otherwise, adjust the modulation order and use an adjustable gain amplifier to dynamically adjust the gain coefficient according to the predicted channel attenuation to dynamically compensate for the transmission loss.

5. The multimodal optical signal transmission method according to claim 4, wherein: The method of dynamically adjusting the gain coefficient according to the predicted channel attenuation using the adjustable gain amplifier to dynamically compensate for the transmission loss includes: The compensation formula for setting the adjustable gain amplifier to dynamically adjust the gain coefficient according to the predicted channel attenuation is P out =A(t)*P in ; where P in is the gain factor of the input adjustable gain amplifier, P out is the compensated gain factor output from the adjustable gain amplifier.

6. The multimodal optical signal transmission method according to claim 1, wherein: The method of implementing clock synchronization and signal encryption by quantum key distribution during data transmission includes: A quantum source generates an entangled photon pair, and a synchronous clock signal is generated by the quantum entangled photon pair; Distributing the synchronous clock signal to the main board and the back board through a beam splitter; The mainboard and the backboard achieve phase synchronization based on a timestamp comparison method, and adopt the least square method to perform error compensation during phase synchronization; deploying a quantum key distribution unit between the backplanes, generating a quantum key, setting an update frequency of the quantum key, and distributing the quantum key using the quantum entangled photon pair; The control signal is quantum-encoded according to the quantum key to achieve signal encryption.

7. The multimodal optical signal transmission method according to claim 1, wherein: Also includes: Optimize bandwidth allocation and path selection between multiple devices through federated learning; Set the path selection formula to Among them, Delay i is the delay of the i-th subcarrier, ∑ i Delay i is the path delay, BER i is the bit error rate of the i-th subcarrier, λ is the weight coefficient, Path k The optimal path selected is the kth path; Based on the graph neural network, the topology and traffic characteristics of the link are learned, the congestion of the link is predicted, and the optimal path is dynamically selected according to the path selection formula, and the switching transmission path is controlled in real time.

8. A multimodal optical signal transmission device, characterized in that: include: AI dynamic modulation engine for building and training deep reinforcement learning models; Channel quality index monitoring module, used to monitor the eye diagram parameters, noise level and bit error rate of the optical fiber link in real time and generate the channel quality index; a modulation format selection module, configured to input the channel quality index into the deep reinforcement learning model, and select an optimal modulation format for transmitting a data stream based on the channel quality index with the goal of maximizing throughput and minimizing bit error rate; A multimodal optical signal processor is used to split the data stream into a control signal and a data signal, where the control signal is in the visible light band and the data signal is in the infrared band. The control signal and the data signal are combined into a single multi-core optical fiber through a combiner. The optical network controller is used to dynamically allocate subcarrier resources on demand for data transmission based on changes in network traffic through spectrum dynamic allocation. The quantum synchronization unit is used to achieve clock synchronization and signal encryption through quantum key distribution during data transmission.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the multimodal optical signal transmission method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the multimodal optical signal transmission method according to any one of claims 1 to 7 are implemented.

Images