Method and system for generating and transmitting signals carrying multi-dimensional optical tags

Abstract

The present invention provides a method and system for generating and transmitting signals carrying multi-dimensional optical tags. The method includes: at the transmitting end of an optical network, generating an original optical signal, generating a modulation signal and converting it into a drive current signal; superimposing the drive current signal and the operating current signal of a semiconductor optical amplifier to obtain a modulation current signal; inputting the original optical signal into the semiconductor optical amplifier and modulating the original optical signal according to the modulation current signal to generate an optical signal carrying one-dimensional optical tag information; repeatedly executing the steps of generating, converting, and current superimposing the modulation signal multiple times. Each time, input the optical signal carrying the optical tag information of the corresponding dimension generated last time into the semiconductor optical amplifier, and modulate the optical signal carrying the optical tag information of the corresponding dimension according to the modulation current signal obtained each time to generate an optical signal carrying multi-dimensional optical tag information and send it to an exchange node. The present invention can generate optical signals carrying multi-dimensional optical tags without disturbing the original optical signal.

Description

[0001] This application is a divisional application of a Chinese patent application with an application date of July 31, 2024, an application number of 202411043309.9, and an invention title of "Optical Network Operation and Maintenance Method, Signal Generation and Transmission Method and System Carrying Multi-Dimensional Optical Tags". Technical Field

[0002] The present invention relates to the field of optical communication technologies, and in particular, to a network operation and maintenance method for multi-dimensional optical tags, a signal generation and transmission method and system carrying multi-dimensional optical tags. Background Art

[0003] Currently, with the continuous expansion of the scale of optical switches in optical networks, in-band optical tags based on tone mapping signals have attracted much attention due to their flexible spectral characteristics and no additional bandwidth consumption. Especially in passive elastic optical networks, the increasing spectral granularity has brought challenges to operation and maintenance. Using optical tags to indicate bandwidth, central frequency, and destination can greatly reduce operation and maintenance costs. Although progress has been made in the autonomous maintenance of optical networks, there are still challenges in the rapid generation and identification of optical tags and the autonomous setting of optical paths, which limits the application scope of optical tags.

[0004] Traditional optical tag signal generation methods rely on digital signal processors (DSPs) for driving, which requires modifications to the signal transmission devices in optical networks, increasing the complexity and cost of optical network systems. Moreover, in identifying optical tags, they rely on DSP algorithms, and complex algorithms have a long processing time, which may introduce additional signal processing delays and affect the real-time performance of the system. In addition, due to the low-frequency characteristics of tone mapping signals, the time for signal capture and analysis may be very long, making it difficult for switching nodes to quickly respond to changes in optical paths. To achieve the active operation and maintenance functions of optical transmission networks, network controllers need to automatically identify optical tags.

[0005] Furthermore, in the field of optical networks, the communication tasks between nodes are becoming increasingly heavy, the data transmission volume is huge, and the timeliness requirements of data packets vary. Therefore, without changing the existing data transmission protocol, how to improve the operation and maintenance efficiency of optical networks, reduce operating costs, improve the utilization efficiency of network resources, and at the same time improve service quality is particularly crucial and challenging in current network optimization practices. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a network operation and maintenance method for multi-dimensional optical tags, a signal generation and transmission method and system carrying multi-dimensional optical tags, so as to eliminate or improve one or more defects existing in the prior art.

[0007] The first aspect of the present invention provides a method for network operation and maintenance of multi-dimensional optical tags, which is applied to each switching node of an optical network and includes the following steps:

[0008] The switching node of the optical network receives an optical signal carrying multi-dimensional optical tag information generated by a sending end or forwarded by other switching nodes, and splits the optical signal carrying multi-dimensional optical tag information into a higher-power optical signal carrying multi-dimensional optical tag information and a lower-power optical signal carrying multi-dimensional optical tag information;

[0009] Perform restoration preprocessing on the lower-power optical signal carrying multi-dimensional optical tag information to obtain multiple restored one-dimensional optical tag signals;

[0010] Input the two-dimensional matrix representation of the multiple restored one-dimensional optical tag signals into a multi-dimensional optical tag recognition model deployed in the switching node for recognition, so that the multi-dimensional optical tag recognition model outputs multi-dimensional optical tag information. The multi-dimensional optical tag recognition model includes a spatial module, an activation function layer, a time module, a splicing module, a fully connected layer, and a normalized exponential function output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module;

[0011] Construct a multi-level dynamic feedback queue according to the output multi-dimensional optical tag information, and forward the higher-power optical signal carrying multi-dimensional optical tag information in a wavelength-division multiplexing manner according to the multi-level dynamic feedback queue.

[0012] In some embodiments of the present invention, the performing restoration preprocessing on the lower-power optical signal carrying multi-dimensional optical tag information to obtain multiple restored one-dimensional optical tag signals includes:

[0013] Sequentially pass the lower-power optical signal carrying multi-dimensional optical tag information through photoelectric conversion, analog-to-digital conversion, and fast Fourier transform to obtain multiple digital electrical signals carrying one-dimensional optical tag information;

[0014] Obtain corresponding multiple carrier signals from the multiple digital electrical signals carrying one-dimensional optical tag information through the Costas loop method;

[0015] Sequentially pass the multiple carrier signals through carrier synchronization, clock synchronization, and demodulation to obtain multiple restored one-dimensional optical tag signals.

[0016] In some embodiments of the present invention, the spatial module includes a multi-layer graph convolutional neural network for extracting spatial features of the multiple restored one-dimensional optical tag signals and outputting multiple two-dimensional matrices and multiple one-dimensional vectors containing corresponding spatial features;

[0017] The activation function layer is used to filter the element values in the multiple two-dimensional matrices containing corresponding spatial features;

[0018] The time module is used to extract the time features of multiple two-dimensional matrices containing corresponding spatial features after filtering, and output multiple one-dimensional vectors containing corresponding time features;

[0019] The splicing module is used to splice multiple one-dimensional vectors containing corresponding spatial features and multiple one-dimensional vectors containing corresponding time features respectively;

[0020] The fully connected layer is used to integrate the spatial features and time features of each of the multiple spliced one-dimensional vectors, and output multiple global features;

[0021] The softmax output layer is used to classify the multiple global features and output multi-dimensional optical label information.

[0022] In some embodiments of the present invention, before inputting the two-dimensional matrix representation of the restored multiple one-dimensional optical label signals into the multi-dimensional optical label recognition model deployed in the switching node, the method further includes:

[0023] Regarding the restored multiple one-dimensional optical label signals as multiple two-dimensional matrix graphs composed of multiple pixel points respectively, and dividing each two-dimensional matrix graph into multiple pixel blocks;

[0024] Constructing a feature matrix based on the multiple pixel blocks, where the element value in the feature matrix is the average value of the RGB values of each pixel point in the corresponding pixel block, and constructing an adjacency matrix based on the adjacent relationship between the multiple pixel blocks;

[0025] Performing normalization processing on the multiple feature matrices to obtain multiple normalized feature matrices, and the two-dimensional matrix representation includes the normalized feature matrices and the corresponding adjacency matrices.

[0026] In some embodiments of the present invention, the multi-dimensional optical label information includes wavelength information and priority information corresponding to the original optical signal generated by the sending end.

[0027] In some embodiments of the present invention, constructing a multi-level dynamic feedback queue according to the output multi-dimensional optical label information includes:

[0028] Dividing multiple optical signals with different wavelengths into multiple first groups according to the wavelength information;

[0029] In each first group, constructing a first-level queue in the order of priority from high to low according to the priority information carried by different optical signals with the same wavelength, where different optical signals with the same priority form a second group in the first-level queue;

[0030] In each second group, constructing a first-in-first-out queue as the second-level queue according to the order of arrival of different optical signals with the same priority at the switching node;

[0031] If the length, waiting time, and quality of service of the queue all reach the corresponding preset thresholds, dynamically adjust and increase the priority of the queue;

[0032] If an optical signal with a higher priority than the optical signal being transmitted is received during the forwarding process, stop the transmission of the current queue until the forwarding of the optical signal with a higher priority than the optical signal being transmitted is completed.

[0033] The second aspect of the present invention provides a method for generating and transmitting a signal carrying multi-dimensional optical tags, which is applied to the sending end of an optical network and includes the following steps:

[0034] The sending end of the optical network generates an original optical signal through a high-speed transceiver, and drives a digital-to-analog conversion module through a programmable logic array to generate a modulation signal, and the frequency of the modulation signal is lower than the frequency of the original optical signal;

[0035] Convert the modulation signal into a driving current signal through a voltage-controlled current source circuit;

[0036] Superimpose the driving current signal and the operating current signal of the semiconductor optical amplifier through the semiconductor optical amplifier to obtain a modulation current signal;

[0037] Input the original optical signal into the semiconductor optical amplifier, and fine-tune the original optical signal according to the modulation current signal through the semiconductor optical amplifier to generate an optical signal carrying one-dimensional optical tag information;

[0038] Repeat the steps of generating, converting, and current superimposing the modulation signal multiple times. Each time, input the optical signal carrying the corresponding dimension optical tag information generated last time into the semiconductor optical amplifier, and fine-tune the optical signal carrying the corresponding dimension optical tag information according to the modulation current signal obtained each time through the semiconductor optical amplifier to generate an optical signal carrying multi-dimensional optical tag information, and the frequencies of the modulation signals generated each time are different;

[0039] Send the optical signal carrying multi-dimensional optical tag information to the switching node of the optical network.

[0040] In some embodiments of the present invention, the method further includes: connecting a temperature controller through the semiconductor optical amplifier to suppress the temperature drift of the current signal.

[0041] In some embodiments of the present invention, after converting the modulation signal into a driving current signal through the voltage-controlled current source circuit, the method further includes:

[0042] Convert the driving current signal into a differential enable signal through a driver / receiver chip;

[0043] The differential enable signal and the operating current signal of the semiconductor optical amplifier are superimposed through the semiconductor optical amplifier to obtain a modulation current signal.

[0044] In some embodiments of the present invention, the multi-dimensional optical label information includes wavelength information and priority information corresponding to the original optical signal generated by the transmitting end.

[0045] The third aspect of the present invention provides a network operation and maintenance system for multi-dimensional optical labels. The system is deployed at each switching node in the optical network and includes: an optical coupler, an optical label signal recovery module, a processor, a memory, and computer instructions stored on the memory.

[0046] The optical coupler is configured to receive an optical signal carrying multi-dimensional optical label information generated by the transmitting end or forwarded by other switching nodes, and split the optical signal carrying multi-dimensional optical label information into a higher-power optical signal carrying multi-dimensional optical label information and a lower-power optical signal carrying multi-dimensional optical label information.

[0047] The optical label signal recovery module includes a photodiode, a transimpedance amplifier, an analog-to-digital converter, a fast Fourier transform circuit, a Costas loop circuit, a carrier synchronization module, a clock synchronization module, and a demodulation module connected in sequence, and is configured to perform recovery preprocessing on the lower-power optical signal carrying multi-dimensional optical label information to obtain multiple recovered one-dimensional optical label signals.

[0048] The processor is configured to execute the computer instructions. When the computer instructions are executed, the system implements the following steps in the network operation and maintenance method for multi-dimensional optical labels described in the foregoing first aspect:

[0049] The two-dimensional matrix representation of the multiple recovered one-dimensional optical label signals is input into a multi-dimensional optical label recognition model deployed in the switching node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a time module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence, and the output of the spatial module is also connected to the input of the splicing module.

[0050] A multi-level dynamic feedback queue is constructed according to the output multi-dimensional optical label information, and the higher-power optical signal carrying multi-dimensional optical label information is forwarded in a wavelength-division multiplexing manner according to the multi-level dynamic feedback queue.

[0051] The fourth aspect of the present invention provides a signal generation and transmission system carrying multi-dimensional optical labels. The system is deployed at the transmitting end in the optical network and includes: a high-speed transceiver, a programmable logic array, a digital-to-analog conversion module, a voltage-controlled current source circuit, and a semiconductor optical amplifier.

[0052] The high-speed transceiver is used to generate an original optical signal;

[0053] The programmable logic array is used to drive the digital-to-analog conversion module to generate modulation signals multiple times. The frequency of the modulation signal is lower than the frequency of the original optical signal, and the frequency of each generated modulation signal is different;

[0054] The voltage-controlled current source circuit is used to convert the modulation signal generated each time into a drive current signal;

[0055] The semiconductor optical amplifier is used to superimpose the drive current signal obtained each time and the working current signal of the semiconductor optical amplifier to obtain modulation current signals multiple times; receive the original optical signal, modulate the original optical signal according to the modulation current signal obtained for the first time to generate an optical signal carrying one-dimensional optical label information; and is also used to receive the optical signal carrying the corresponding-dimensional optical label information generated in each previous time multiple times, modulate the optical signal carrying the corresponding-dimensional optical label information according to the modulation current signal obtained each time to generate an optical signal carrying multi-dimensional optical label information, and send the optical signal carrying multi-dimensional optical label information to the switching node of the optical network.

[0056] In some embodiments of the present invention, the system further includes: a driver / receiver chip, and the driver / receiver chip is used to convert the drive current signal obtained each time into a differential enable signal;

[0057] The semiconductor optical amplifier is further used to superimpose the differential enable signal obtained each time and the working current signal of the semiconductor optical amplifier to obtain modulation current signals multiple times.

[0058] In some embodiments of the present invention, the system further includes: a temperature controller, and the temperature controller is connected to the semiconductor optical amplifier and is used to suppress the temperature drift of the current signal.

[0059] In some embodiments of the present invention, the multi-dimensional optical label information includes wavelength information and priority information corresponding to the original optical signal generated by the sending end.

[0060] The fifth aspect of the present invention provides an electronic device, which is deployed at each switching node in the optical network and includes: a processor, a memory, and computer instructions stored on the memory. The processor is used to execute the computer instructions, and when the computer instructions are executed, the system implements the following steps of the multi-dimensional optical label network operation and maintenance method described in the foregoing first aspect:

[0061] The two-dimensional matrix representation of the restored multiple one-dimensional optical label signals is input into a multi-dimensional optical label recognition model deployed in an exchange node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a temporal module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module; wherein, the restored multiple one-dimensional optical label signals are obtained by performing restoration preprocessing on optical signals with lower power carrying multi-dimensional optical label information. The optical signals with lower power carrying multi-dimensional optical label information are received by the exchange node of the optical network from the optical signals generated by the sending end or forwarded by other exchange nodes, and the optical signals with lower power carrying multi-dimensional optical label information are shunted to obtain optical signals with higher power carrying multi-dimensional optical label information and optical signals with lower power carrying multi-dimensional optical label information;

[0062] Construct a multi-level dynamic feedback queue according to the output multi-dimensional optical label information, and forward the optical signals with higher power carrying multi-dimensional optical label information in a wavelength division multiplexing manner according to the multi-level dynamic feedback queue.

[0063] The sixth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps of the network operation and maintenance method for multi-dimensional optical labels described in the first aspect are implemented:

[0064] The two-dimensional matrix representation of the restored multiple one-dimensional optical label signals is input into a multi-dimensional optical label recognition model deployed in an exchange node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a temporal module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module; wherein, the restored multiple one-dimensional optical label signals are obtained by performing restoration preprocessing on optical signals with lower power carrying multi-dimensional optical label information. The optical signals with lower power carrying multi-dimensional optical label information are received by the exchange node of the optical network from the optical signals generated by the sending end or forwarded by other exchange nodes, and the optical signals with lower power carrying multi-dimensional optical label information are shunted to obtain optical signals with higher power carrying multi-dimensional optical label information and optical signals with lower power carrying multi-dimensional optical label information;

[0065] Construct a multi-level dynamic feedback queue according to the output multi-dimensional optical label information, and forward the optical signals with higher power carrying multi-dimensional optical label information in a wavelength division multiplexing manner according to the multi-level dynamic feedback queue.

[0066] The seventh aspect of the present invention provides a computer program product, which includes computer instructions that, when executed by a processor, implement the following steps of the network operation and maintenance method of the multi-dimensional optical tag described in the foregoing first aspect:

[0067] Input the two-dimensional matrix representation of the restored multiple one-dimensional optical tag signals into the multi-dimensional optical tag recognition model deployed in the switching node for recognition, so that the multi-dimensional optical tag recognition model outputs multi-dimensional optical tag information. The multi-dimensional optical tag recognition model includes a spatial module, an activation function layer, a temporal module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module; wherein, the restored multiple one-dimensional optical tag signals are obtained by performing restoration preprocessing on the optical signal carrying multi-dimensional optical tag information with lower power. By receiving the optical signal carrying multi-dimensional optical tag information generated by the sending end or forwarded by other switching nodes by the switching node of the optical network, and splitting the optical signal carrying multi-dimensional optical tag information to obtain a high-power optical signal carrying multi-dimensional optical tag information and a low-power optical signal carrying multi-dimensional optical tag information;

[0068] Construct a multi-level dynamic feedback queue according to the output multi-dimensional optical tag information, and forward the high-power optical signal carrying multi-dimensional optical tag information in a wavelength division multiplexing manner according to the multi-level dynamic feedback queue.

[0069] The network operation and maintenance method of the multi-dimensional optical tag, the signal generation and transmission method and system carrying the multi-dimensional optical tag of the present invention. This network operation and maintenance method can receive the optical signal carrying the multi-dimensional optical tag through each switching node in the optical network and split it, separate the multi-dimensional optical tag signal carried from the optical signal with lower power after splitting, and input the separated multi-dimensional optical tag signal into the improved GraphFormer model for fast and accurate recognition. Finally, construct a multi-level dynamic feedback queue according to the recognized multi-dimensional optical tag result, and efficiently forward the high-power optical signal carrying the multi-dimensional optical tag after splitting according to this queue and the fiber wavelength division multiplexing technology. This solution not only improves the optical path utilization efficiency of the entire optical network, saves link resources, ensures the timeliness of data, but also greatly reduces the network operation and maintenance cost without occupying additional bandwidth of the optical network and without changing the original network transmission protocol. This method can also be extended to the transmission of optical signals carrying other dimensions or more dimensions of useful information.

[0070] The additional advantages, objects, and features of the present invention will be partially described below and will become partially apparent to those of ordinary skill in the art after studying the following. The objects and other advantages of the present invention can be achieved and obtained by the structure specifically pointed out in the specification and the drawings.

[0071] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and the above and other objectives achievable with the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and do not limit the present invention.

[0073] Figure 1 It is a schematic flowchart of a method for signal generation and transmission carrying a multi-dimensional optical tag in an embodiment of the present invention;

[0074] Figure 2 It is a schematic diagram of an optical signal carrying a multi-dimensional optical tag generated in an embodiment of the present invention;

[0075] Figure 3 It is a schematic flowchart of a method for network operation and maintenance of a multi-dimensional optical tag in an embodiment of the present invention;

[0076] Figure 4 It is a schematic structural diagram of an improved GraphFormer model in an embodiment of the present invention;

[0077] Figure 5 It is a schematic diagram of a multi-level dynamic feedback queue based on wavelength and priority constructed in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] To make the objectives, technical solutions, and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with the embodiments and the drawings. Herein, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but do not limit the present invention.

[0079] Herein, it should also be noted that in order to avoid obscuring the present invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the present invention are shown in the drawings, while other details less related to the present invention are omitted.

[0080] It should be emphasized that the term "including / comprising" when used herein refers to the presence of features, elements, steps, or components, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0081] Herein, it should also be noted that if not otherwise specified, the term "connection" in this document can not only refer to direct connection, but also represent indirect connection with an intermediate.

[0082] In the following, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0083] Regarding the operation and maintenance efficiency and cost issues brought about by the continuous expansion of the scale of optical switches in optical networks, as well as the challenges caused by the increase in spectral granularity in passive elastic optical networks, the purpose of the present invention is to provide a method for network operation and maintenance of multi-dimensional optical tags, a method for generating and transmitting signals carrying multi-dimensional optical tags, and a system corresponding to the method. Without changing the existing network transmission protocol, the method and system mainly deploy a signal generation system or an optical tag modulation system carrying multi-dimensional optical tags at the sending end of the optical network, perform pre-multiple modulation processing on the original optical signal to generate multi-dimensional optical tags, and superimpose the generated multi-dimensional optical tags on the original optical signal. A deep recognition model of multi-dimensional optical tags is deployed at each switching node of the optical network, which can quickly and intelligently identify and process multi-dimensional optical tag information, thereby realizing data transmission based on optical tag information and wavelength division multiplexing. For example, the utilization rate of the optical fiber link and the priority of the optical signal can be combined to complete differential processing of data packets with different importance levels.

[0084] Since the traditional method for generating signals carrying optical tags has problems such as making the optical network system more complex and increasing network maintenance costs when superimposing multi-dimensional optical tags, and the multi-dimensional optical tags superimposed by this method cause the dependent DSP algorithm to be unable to quickly and automatically identify the multi-dimensional optical tags, and may also introduce additional signal processing delays, affecting the real-time performance of the system, etc., the method and system for generating and transmitting signals carrying multi-dimensional optical tags provided by the embodiments of the present invention do not require any modification to the sending-end device of the optical network and completely overcome the above defects.

[0085] Figure 1 It is a schematic flow chart of the method for generating and transmitting signals carrying multi-dimensional optical tags in an embodiment of the present invention. As Figure 1 shown, the embodiments of the present invention provide a method for generating and transmitting signals carrying multi-dimensional optical tags, which is applied to the sending end of an optical network and includes the following steps:

[0086] Step S110, the sending end of the optical network generates an original optical signal through a high-speed transceiver, and drives a digital-to-analog conversion module through a programmable logic array to generate a modulation signal, and the frequency of the modulation signal is lower than the frequency of the original optical signal.

[0087] In this embodiment, the original data stream transmitted in the optical network is generated by modulating a high-speed transceiver, and a raw high-speed optical signal is generated by the high-speed transceiver. Therefore, when superimposing the optical label, it is a key consideration to avoid any interference with the normal forwarding of the superimposed signal. To solve this problem, since a field-programmable gate array (FPGA) can achieve high-precision and low-latency signal generation and processing, ensuring that the generation of the signal carrying the multi-dimensional optical label does not introduce additional latency or noise, and thus does not interfere with the normal transmission of the high-speed data stream, this embodiment uses a programmable logic array to drive a digital-to-analog conversion module to generate a low-speed modulation signal.

[0088] It should be noted that in the fields of communication and the like, the optical label is also called a pilot tone, a low-frequency perturbation signal, or an overmodulation signal. It refers to generating a low-speed optical in-band signal by means of peak clipping and loading it on the main signal of the wavelength channel for wavelength monitoring and transmitting optical in-band overhead, etc. Therefore, it can also be called a peak-clipping signal. If wavelength information is to be carried in the original optical signal, the peak-clipping signal can be loaded on the wavelength signal at the optical transmitting end in a wavelength-division multiplexing system, and different-frequency peak-clipping signals are superimposed on the original optical signals of each different wavelength. The receiving end can obtain the wavelength information by detecting the frequency of the peak-clipping signal. It can be seen that the low-speed modulation signal generated in the embodiment of the present invention is the optical label that the transmitting end wants to carry, and it can be a low-speed sine signal with adjustable frequency and peak-to-peak value, and the frequency range is 500 KHz to 100 MHz. The high-speed optical signal refers to an information signal that uses an optical fiber as a transmission medium and transmits at an extremely high speed through light waves (usually lasers). The high-speed optical signal generated by the high-speed transceiver utilizes the characteristics of light, such as extremely high speed (close to the speed of light in a vacuum), high frequency, and large bandwidth, to achieve efficient, fast, and long-distance data transmission. The frequency of the low-speed modulation signal is lower than that of the original high-speed optical signal.

[0089] Step S120: Convert the modulation signal into a drive current signal through a voltage-controlled current source circuit.

[0090] In this embodiment, since the SOA allows an upper limit value for the input current value, it is necessary to convert the low-speed modulation signal generated in step S110 into a drive current suitable for the SOA. The voltage-controlled current source circuit uses a voltage-controlled current source circuit built with TL431P, which can convert the sine voltage signal with a peak value of 5 volts output by the digital-to-analog conversion module into a sine current signal fluctuating up and down by about 2 mA. In other embodiments, voltage-controlled current source circuits with other chip models and structures can also be used to achieve this function.

[0091] Step S130: Superimpose the drive current signal and the working current signal of the semiconductor optical amplifier through the semiconductor optical amplifier to obtain a modulation current signal.

[0092] In this step, by superimposing the current signal output by the voltage-controlled current source on the DC operating current of the SOA, the operating current of the SOA for modulation is obtained, which can enable the SOA to modulate the intensity of the original high-speed optical signal without affecting the transmission of the original high-speed optical signal subsequently.

[0093] Step S140: Input the original optical signal into the semiconductor optical amplifier, and the semiconductor optical amplifier finely adjusts the original optical signal according to the modulation current signal to generate an optical signal carrying one-dimensional optical label information.

[0094] In this step, the semiconductor optical amplifier (SOA) receives the original high-speed optical signal generated by the high-speed transceiver at the input end, and slightly modulates the intensity of the input original high-speed optical signal by modulating the operating current of the SOA. This kind of modulation will not significantly change the characteristics of the original optical signal, but can introduce a low-speed modulation signal, that is, an optical label, into the original optical signal. Due to the optical gain characteristic of the SOA, modulating the operating current of the SOA generates an optical signal containing the low-speed modulation signal at the output end of the SOA, that is, this optical signal contains the information of the first-dimensional optical label and at the same time maintains the main characteristics of the original high-speed optical signal.

[0095] The information of the first-dimensional optical label can be the wavelength information of the original optical signal or any other label information.

[0096] Step S150: Repeatedly execute the steps of generating, converting, and current superposition of the modulation signal multiple times. Each time, input the optical signal carrying the optical label information of the corresponding dimension generated last time into the semiconductor optical amplifier, and the semiconductor optical amplifier finely adjusts the optical signal carrying the optical label information of the corresponding dimension according to the modulation current signal obtained each time to generate an optical signal carrying multi-dimensional optical label information, and the frequencies of the modulation signals generated each time are different.

[0097] In this step, after the second execution of the steps of generating the low-speed modulation signal in step S110 and steps S120 - S130, where the frequency of the low-speed modulation signal generated the second time is different from that generated the first time, input the optical signal carrying the one-dimensional optical label information generated the first time into the input end of the SOA, and finely adjust the intensity of the optical signal carrying the one-dimensional optical label information through the modulation current signal generated after the second execution of steps S120 - S130 to generate an optical signal carrying two-dimensional optical label information, realizing the superposition of the two-dimensional optical label information on the original optical signal. The same principle applies to the third, fourth, and even more executions. Through multiple executions, an optical signal carrying multi-dimensional optical label information is finally generated. In this process, multiple frequencies from low frequency to high frequency are considered for the modulation signal to achieve multiple modulations of the original optical signal.

[0098] In a specific example, the optical signal carrying two-dimensional optical label information is obtained by superimposing, on the optical signal carrying the optical label information representing the wavelength of the original optical signal, the optical label information representing the priority level of the original optical signal (the service priority level in the optical network). That is, the superimposed second-dimensional optical label information is the service priority level information of the original optical signal in the optical network, and can also be represented as other information according to specific requirements, such as the service type of the original optical signal transmission, etc. In the optical signal carrying two-dimensional optical label information obtained in this specific example, the formula of the two-dimensional optical label signal is as follows:

[0099]

[0100] Wherein, f 1 and f 2 respectively represent the frequencies of the first-dimensional optical label signal (representing the carried wavelength) and the second-dimensional optical label signal (representing the carried service priority level), and both frequencies are in the range of 500 KHz to 100 MHz. In order for the receiving end to correctly receive and identify these two optical label signals, the interval between f 1 and f 2 should generally be greater than 30 MHz, and f 1 > f 2 . The optical signal carrying two-dimensional optical label information obtained through two modulations is as Figure 2 shown.

[0101] Step S160: Send the optical signal carrying multi-dimensional optical label information to the switching node of the optical network, so that each switching node in the optical network receives the optical signal carrying multi-dimensional optical label information and processes the signal.

[0102] In another embodiment of the present invention, after converting the modulation signal into a drive current signal through a voltage-controlled current source circuit, the method for generating and transmitting the signal carrying multi-dimensional optical label further includes the following steps:

[0103] Step S230: Convert the drive current signal into a differential enable signal through a driver / receiver chip.

[0104] In this embodiment, the driver / receiver chip selects the DS90LV019TMLVDS chip. The input of the DS90LV019TMLVDS chip is at a high level, and the output is a pair of differential signals with the LVDS level standard. Converting the drive current signal into a differential enable signal through this chip can improve the anti-interference ability of the signal to ensure the stability of signal transmission.

[0105] Step S240: Superimpose the differential enable signal and the operating current signal of the semiconductor optical amplifier through the semiconductor optical amplifier to obtain a modulation current signal.

[0106] The other steps in this embodiment are the same as steps S110, S120, S140, and S150, and will not be described in detail. In step S240, a differential enable signal is added to the operating current of the SOA, enabling the SOA to perform appropriate modulation.

[0107] In another embodiment of the present invention, the method for generating and transmitting a signal carrying a multi-dimensional optical label further includes the following steps: connecting a thermostatic controller to a semiconductor optical amplifier to suppress the temperature drift of the current signal.

[0108] In this embodiment, using the HTC1500 chip as the thermostatic controller can prevent the drift of the operating current of the SOA, ensuring the stability of the system and the accuracy of the optical signal carrying the optical label.

[0109] Through the above steps, different from the traditional method, the method for generating and transmitting a signal carrying a multi-dimensional optical label in each embodiment of the present invention uses a semiconductor optical amplifier (SOA) as the optical label modulation device, and realizes the multi-dimensional superposition of optical labels. It can generate and load an optical signal carrying multi-dimensional optical label information without interfering with the original high-speed optical signal. And by transmitting this optical signal, the receiving end (switching node in the optical network) can quickly and automatically identify the multi-dimensional optical label information superimposed here. The wavelength of the transmitted signal can also be used as the first-dimensional optical label and superimposed on the transmitted signal, and corresponding priorities are set in advance for different signals at the transmitting end, and the priority is also superimposed on the transmitted signal as the second-dimensional optical label, that is, the wavelength and task priority information of the transmitted signal are embedded in the data packet transmitted in the optical network, so as to realize the differential processing of data packets with different importance levels.

[0110] Figure 3 It is a schematic flowchart of the network operation and maintenance method of multi-dimensional optical labels in an embodiment of the present invention. As Figure 3 shown, the network operation and maintenance method of multi-dimensional optical labels in this embodiment is applied to each switching node of the optical network, and includes the following steps:

[0111] Step S310, the switching node of the optical network receives the optical signal carrying multi-dimensional optical label information generated by the transmitting end or forwarded by other switching nodes, and splits the optical signal carrying multi-dimensional optical label information into a higher-power optical signal carrying multi-dimensional optical label information and a lower-power optical signal carrying multi-dimensional optical label information.

[0112] In this step, when the optical signal carrying multi-dimensional optical label information sent by the sending end or forwarded by the switching node arrives at the switching node, the optical coupler deployed at the switching node divides the optical signal into two parts according to the signal power. The power ratio of the two parts of the optical signal is approximately 1:9. And the optical signal with higher power in the two parts of the optical signal is sent to the server deployed at the switching node to wait for being forwarded by the switching node after the switching node receives and identifies the multi-dimensional optical label information. The optical signal with lower power is sent to the host of the deep recognition model of the multi-dimensional optical label deployed at the switching node and is used for the recognition of the multi-dimensional optical label information, so as to realize the signal shunt. If the above power ratio is too large, it will affect the reception of the optical signal. On the contrary, if the ratio is too small, it will affect the automatic recognition of the multi-dimensional optical label.

[0113] In an embodiment of the present invention, the multi-dimensional optical label information includes the wavelength information and priority information corresponding to the original optical signal generated by the sending end.

[0114] Step S320, perform restoration preprocessing on the optical signal with lower power carrying multi-dimensional optical label information to obtain multiple restored one-dimensional optical label signals.

[0115] In this step, before being recognized by the deep recognition model of the multi-dimensional optical label, it is necessary to restore the carried multi-dimensional optical label information from the optical signal with lower power carrying multi-dimensional optical label information, so as to realize the accurate recognition of the multi-dimensional optical label. In an embodiment, the multiple restored one-dimensional optical label signals are signals respectively representing the wavelength information and priority information to be carried.

[0116] In an embodiment of the present invention, performing restoration preprocessing on the optical signal with lower power carrying multi-dimensional optical label information to obtain multiple restored one-dimensional optical label signals includes the following steps:

[0117] Step S321, sequentially perform photoelectric conversion, analog-to-digital conversion, and fast Fourier transform on the optical signal with lower power carrying multi-dimensional optical label information to obtain multiple digital electrical signals carrying one-dimensional optical label information.

[0118] In this embodiment, a low-speed avalanche photodiode (APD) and a transimpedance amplifier (TIA) deployed at the switch point convert an optical signal with low power carrying multi-dimensional optical tag information into a chopped-top analog electrical signal. An analog-to-digital converter with a sampling frequency of 250 MHz converts the analog electrical signal into a digital electrical signal. The sampling output bit width of the analog-to-digital converter is 14 bits. Then, a fast Fourier transform circuit performs a fast Fourier transform on the digital electrical signal to obtain multiple digital electrical signals carrying one-dimensional optical tag information. In a specific example, the fast Fourier transform circuit is formed by using internal resources such as logic gates and look-up tables of an FPGA. The input of the circuit is the digital electrical signal after analog-to-digital conversion, and the output is the discrete spectrum of the digital electrical signal. By finding the coordinates n 1 、n 2 ,n 1 >n 2 , the signal frequencies of the two-dimensional optical tag containing the carried wavelength information and priority information can be obtained as follows:

[0119]

[0120] where f 3 、f 4 are the two signal frequencies of the two-dimensional optical tag. The two-dimensional optical tag can be understood as the first-dimensional optical tag in the two-dimensional optical tag × the second-dimensional optical tag in the two-dimensional optical tag. f s =250·M(Hz), where M is the number of sampling points of the fast Fourier transform. According to f 3 =f 1 +f 2 , f 4 =f 1 -f 2 , the signal frequencies of the first-dimensional optical tag (wavelength) and the second-dimensional optical tag (priority) in the two-dimensional optical tag can be obtained as

[0121] Step S322: Obtain corresponding multiple carrier signals from multiple digital electrical signals carrying one-dimensional optical tag information through the Costas loop method.

[0122] This step is implemented through a Costas loop circuit. Specifically, internal resources such as look-up tables and logic gates in a field-programmable gate array (FPGA) are converted into a corresponding hardware circuit through operations such as hardware description (RTL) language programming, synthesis, placement and routing, and timing optimization. The above-mentioned Costas loop circuit is formed by building this hardware circuit.

[0123] Step S323: Successively subject the multiple carrier signals to carrier synchronization, clock synchronization, and demodulation to obtain multiple restored one-dimensional optical tag signals.

[0124] Through the above steps, the recovery of multi-dimensional optical label information can be achieved from optical signals with lower power carrying multi-dimensional optical label information, enabling the accurate recognition and classification of multi-dimensional optical labels by the deep recognition model of multi-dimensional optical labels.

[0125] Step S330: Input the two-dimensional matrix representation of the recovered multiple one-dimensional optical label signals into the multi-dimensional optical label recognition model deployed in the switching node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a temporal module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module.

[0126] In this step, the multi-dimensional optical label recognition model is an improved GraphFormer model. The improved GraphFormer model is used to recognize the recovered multi-dimensional optical label information, and the improved GraphFormer model is deployed in the host at the switching node.

[0127] Specifically, as Figure 4 shown, the spatial module includes multiple layers of graph convolutional neural networks (GCNs) for extracting the spatial features of the recovered multiple one-dimensional optical label signals and outputting multiple two-dimensional matrices and multiple one-dimensional vectors containing the corresponding spatial features; the activation function layer is used to filter the element values in the multiple two-dimensional matrices containing the corresponding spatial features; the temporal module includes multiple Transformer layers, each Transformer layer including an embedding layer, a first linear layer, a multi-head attention layer, an attention splicing layer, and a second linear layer connected in sequence, for extracting the temporal features of the multiple two-dimensional matrices containing the corresponding spatial features after filtering and outputting multiple one-dimensional vectors containing the corresponding temporal features; the splicing module is used to splice the multiple one-dimensional vectors containing the corresponding spatial features and the multiple one-dimensional vectors containing the corresponding temporal features respectively; the fully connected layer is used to integrate the respective spatial and temporal features in the multiple spliced one-dimensional vectors and output multiple global features; the softmax output layer is used to classify the multiple global features and output multi-dimensional optical label information.

[0128] After the model automatically recognizes the final multi-dimensional optical label result, the host deployed at the switching node sends the final multi-dimensional optical label result to the server deployed at the switching node, enabling the switching node to efficiently forward the optical signal with higher power carrying multi-dimensional optical label information located in the server according to the automatically recognized multi-dimensional optical label result, and saving optical fiber link resources. The multi-dimensional optical label result can be specific wavelength information of the original optical signal, specific levels of priorities, information such as high or low priorities, etc.

[0129] Step S340: Construct a multi-level dynamic feedback queue based on the output multi-dimensional optical label information, and forward the optical signals with higher power carrying the multi-dimensional optical label information in a wavelength-division multiplexing manner according to the multi-level dynamic feedback queue.

[0130] In an embodiment of the present invention, constructing a multi-level dynamic feedback queue based on the output multi-dimensional optical label information includes the following steps:

[0131] Step S341: Divide the optical signals with multiple different wavelengths into multiple first groups according to the wavelength information;

[0132] Step S342: In each first group, construct a first-level queue in the order of decreasing priority according to the priority information carried by the optical signals with the same wavelength, where the optical signals with the same priority form a second group in the first-level queue;

[0133] Step S343: In each second group, construct a first-in-first-out queue as the second-level queue according to the order of arrival of the optical signals with the same priority at the switching node;

[0134] Step S344: If the length, waiting time, and quality of service of the queue all reach the corresponding preset thresholds, dynamically adjust and increase the priority of the queue;

[0135] Step S345: If an optical signal with a higher priority than the currently transmitted optical signal is received during the forwarding process, stop the transmission of the current queue until the optical signal with a higher priority than the currently transmitted optical signal is forwarded completely.

[0136] As Figure 5 shown, the forwarding mechanism of the multi-level dynamic feedback queue constructed by the above steps is specifically: construct n first groups according to different wavelengths λ 1 ,λ 2 ,…,λ n ; under each first group, construct first-level queues P 1 ,P 2 ,…,P m ,P 1 with the highest priority, P m with the lowest priority, where the optical signals with the same priority form a second group in the first-level queue. For example, P 1etc. Under each second-level group, a second-level queue is constructed according to the order of arrival at the switching node, that is, a first-in-first-out (FIFO) queue, forming a total of n×m queues, where n is the total number of signal wavelength types sent by the sending end, and m is the total number of preset priority levels at the sending end. After each optical signal arrives at the switching node and undergoes automatic recognition of the multi-dimensional optical label, the switching node adds the corresponding optical signal to the above corresponding queue according to the recognized multi-dimensional optical label - wavelength and priority. The length L of each queue 1 , L 2 , …, L p is inversely proportional to the priority of the queue. Based on the fiber optic wavelength division multiplexing technology, each switching node sends signals of multiple wavelengths in parallel and preferentially forwards the signals in the queue with a higher priority under each wavelength group. For example, only when all the signals in P 1 under the second group of wavelength λ 1 are forwarded, the signals in P 1 under the second group of wavelength λ 2 are forwarded. In this embodiment, in order to ensure that high-priority signals can be forwarded in time, during the forwarding process, if the switching node receives a signal with a higher priority, it will discard the currently transmitted signal with a lower priority and stop the transmission of the current queue, and instead transmit the just-received signal with a higher priority.

[0137] Moreover, in order to prevent extreme situations such as queue overflow or starvation of lower-priority members from occurring, and the resulting large number of packet losses, which affect the service quality and reduce the network throughput, a method of dynamically adjusting the queue priority is also adopted to solve this problem. Specifically, the priority is dynamically adjusted through the following formula:

[0138]

[0139] where P i1 represents the priority, i1 = 1, 2, …, m, L i2，th , T wait，th and QoS th are the preset thresholds of the expected queue length, waiting time, and service quality respectively. i2 = 1, 2, …, p. When the actual queue length L i2 reaches the preset threshold L i2，th of the expected queue length, the actual waiting time T wait reaches the preset threshold T wait，th of the waiting time, and the actual service quality QoS reaches the preset threshold QoS thWhen it is, the priority of the corresponding queue is increased, so as to preferentially process the data forwarding of this queue. α, β, and γ are influencing factors, and they are taken from random numbers generated between 0 and 0.5. The quality of service QoS is affected by data such as packet loss rate, delay, jitter, and bandwidth.

[0140] It can be seen from this that according to the above multi-level dynamic feedback queue, the optical signal superimposed with multi-dimensional optical labels in the optical network can be more efficiently forwarded based on wavelength division multiplexing and priority, and differential forwarding processing is performed on data packets with different task priorities and importance, thereby significantly improving the utilization efficiency of fiber link resources and the quality of service in the entire optical network, and realizing the time-sensitive guarantee of data.

[0141] In another embodiment of the present invention, in order to more conveniently input the restored multi-dimensional optical label information into the model and be accurately recognized, before inputting the two-dimensional matrix representation of the restored multiple one-dimensional optical label signals into the multi-dimensional optical label recognition model deployed in the switching node, the network operation and maintenance method of this multi-dimensional optical label further includes the following steps:

[0142] Step S326, respectively regard the restored multiple one-dimensional optical label signals as multiple two-dimensional matrix diagrams composed of multiple pixel points, and divide each two-dimensional matrix diagram into multiple pixel blocks.

[0143] Step S327, construct a feature matrix based on multiple pixel blocks. The element value in the feature matrix is the average value of the RGB values of each pixel point in the corresponding pixel block, and construct an adjacency matrix based on the adjacent relationship between multiple pixel blocks.

[0144] Step S328, perform normalization processing on multiple feature matrices to obtain multiple normalized feature matrices. The two-dimensional matrix representation includes the normalized feature matrices and the corresponding adjacency matrices. That is to say, the multiple adjacency matrices and multiple normalized feature matrices obtained by preprocessing the restored multiple one-dimensional optical label signals are used as the input of the improved GraphFormer model.

[0145] Since there are countless pixel points that make up the two-dimensional matrix diagram, in order to avoid the problem of soaring computational complexity caused by an overly large matrix due to excessive pixel points in the picture, in this embodiment, multiple two-dimensional matrix diagrams are each sliced into several 16×16 pixel blocks. Each pixel block is regarded as a node, and each picture can be regarded as being composed of several pixel blocks spliced together. The adjacent relationship between pixel blocks is regarded as the edge connecting nodes, and an adjacency matrix is constructed according to this connection relationship. In this way, two types of matrices are obtained for inputting into the GCN layer of the spatial module of the improved GraphFormer model, namely the feature matrix X composed of each pixel block and the adjacency matrix A used to represent the connection relationship between each pixel block in each picture. There are multiple feature matrices X and adjacency matrices A. The value of each matrix element in the feature matrix is the result obtained by averaging the RGB values of each pixel point in the corresponding pixel block, that is, the average value of the RGB values of each pixel point in the corresponding pixel block.

[0146] Before inputting the above-mentioned feature matrix and adjacency matrix into the model, each feature matrix is preprocessed by normalization. The specific formula for feature matrix normalization is as follows:

[0147]

[0148] where, X n represents the normalized feature matrix, X i，j represents the element in the i-th row and j-th column of the feature matrix, X min represents the minimum value among the elements of the feature matrix, X max represents the maximum value among the elements of the feature matrix.

[0149] Input multiple normalized feature matrices and the corresponding multiple adjacency matrices into the improved GraphFormer model. First, the spatial features of the multiple normalized feature matrices and multiple adjacency matrices corresponding to representing multiple one-dimensional optical label signals are extracted through multiple layers of GCN in the spatial module, that is, the spatial correlation relationships of each signal waveform are obtained. The output of each layer of GCN is as follows:

[0150]

[0151]

[0152] where, H (l+1) represents the two-dimensional matrix output after activation of the (l + 1)-th layer of GCN, H (l) represents the two-dimensional matrix output after activation of the l-th layer of GCN, H (l) ∈R M*D , M represents the number of nodes in each picture, D represents the dimension of the output features of each node, H (0) = X n Xn represents the normalized feature matrix of the input; σ represents the activation function of each layer of GCN; represents the normalization matrix, and the normalization matrix is the result of adding the out-degree matrix and the identity matrix. The out-degree matrix represents the matrix described by the edges that each pixel block in the graph can extend to; represents the self-connection adjacency matrix, A represents the adjacency matrix; W (l) represents the self-learning weight matrix of the i-th layer of GCN; I represents the identity matrix, N represents the dimension of the identity matrix, and in a specific example, N is one-dimensional.

[0153] It can be seen from this that the output of the spatial module is multiple two-dimensional spatial feature matrices that contain the spatial correlation relationships and spatial features of the corresponding one-dimensional optical label signals. To prevent the spatial features from being insignificant after the two-dimensional matrix passes through the time module, the spatial module in this embodiment also uses the TopK algorithm to retain the spatial features learned by the spatial module. Specifically, for each pixel block in the two-dimensional spatial feature matrix, the score of each pixel block is calculated through the following formula:

[0154] Score i，j =H i，j +∑A i，j H i，j

[0155] where H i，j represents the eigenvalue of the pixel block at the i-th row and j-th column in the two-dimensional spatial feature matrix, that is, the matrix element value, and ∑A i，j H i，j represents the sum of the features of the other pixel blocks in the matrix connected to the pixel block at the i-th row and j-th column in the two-dimensional spatial feature matrix. Sort the element values of each two-dimensional spatial feature matrix output by the last layer of GCN in descending order according to the calculated scores of each element value, and the top 50% of the sorted values can be selected. Expand this part to form a one-dimensional spatial feature vector containing spatial features, denoted as Spatial output , thus forming multiple one-dimensional spatial feature vectors, which are used to be concatenated with the one-dimensional vectors output by the subsequent time module to complete the recognition of multi-dimensional optical labels.

[0156] Take the multiple two-dimensional spatial feature matrices H output by the last layer of GCN of the spatial module as the input of the time module, and extract the time features of the matrix data through the time module. To enable the improved GraphFormer model to better understand the sequential relationship of the input matrix data in the time series, the time module uses the method of position embedding (embedding layer) to distinguish the elements at different positions in the time series, so as to understand the relationship between the elements in the input matrix. The calculation formula of the position embedding is as follows:

[0157]

[0158]

[0159] Among them, pos represents the data position in the time series, k represents the time series index, and PE (pos，2k) and PE (pos，2k+1) represent the data of the odd and even dimensions respectively, and d model represents the embedding dimension.

[0160] Through the above position embedding formula, a new feature matrix expression can be obtained, that is, the input matrix data of the time module - the vector representing the position of the two-dimensional spatial feature matrix in the time series. And, in order to better obtain the correlation between data, a multi-head attention mechanism is added to the time module. The purpose of the attention mechanism is to enable the model to dynamically focus on different parts of the sequence when processing sequence data. The traditional standard attention mechanism calculates a single weighted sum, while the multi-head attention mechanism calculates through multiple independent attention heads. Each attention head can focus on different features or positions, so as to capture richer context information. The multi-head attention will perform attention calculations on the input data respectively, and finally splice the calculation results of all attention heads together to obtain the final output result. The key lies in generating multiple groups of QKV values, decomposing the input two-dimensional spatial feature matrix into multiple different subspaces, enabling the model to focus on different positions in each space. The specific process is as follows:

[0161] First is the linear transformation. For each attention head, the embedding of the input sequence will be linearly transformed into query (Query), key (Key), and value (Value) matrices, as shown in the following formula:

[0162] Q = YW Q , K = YW K , V = YW V

[0163] Among them, W Q , W K , W V are learnable weight matrices, and Y represents the output of the embedding.

[0164] Then calculate the attention of each attention head through the following formula:

[0165]

[0166] Among them, d K is the dimension of the key vector, T represents the transpose operation, and the normalization exponential function (SoftMax) can convert a real number vector into a probability vector, as shown in the following formula: Among them, zj denotes the input vector, where \(j = 1, 2, \ldots, q\), and \(q\) represents the number of elements of the input vector, which is the length of the input sequence here.

[0167] Finally, the attentions output by each attention head are concatenated together through the following formula:

[0168] MultiHead(Q, K, V) = Concat(head 1 , \ldots, head h )W o

[0169] where, MultiHead() represents the multi - head attention function, Concat() represents the concatenation function, head h denotes the output of each attention head, \(h\) represents the number of attention heads, and W o represents the weight matrix of the linear transformation.

[0170] The time module finally outputs a one - dimensional time feature vector, which is similar to the one - dimensional spatial feature vector output by the spatial module, denoted as Time output , and finally the vectors Spatial output and Time output are concatenated and then passed through a fully - connected layer and a SoftMax output layer in sequence to obtain the final optical label recognition result. In a specific example, the optical label recognition result is the wavelength information and priority information of the original optical signal sent by the sending end.

[0171] For the network operation and maintenance method of the multi - dimensional optical label in each of the above embodiments of the present invention, the optical signals carrying the multi - dimensional optical label are received by each switching node in the optical network and split. The multi - dimensional optical label signals carried are separated from the optical signals with lower power after splitting, and the separated multi - dimensional optical label signals are input into the improved GraphFormer model for fast and accurate recognition. Finally, a multi - level dynamic feedback queue is constructed according to the recognized multi - dimensional optical label results, and the optical signals with higher power carrying the multi - dimensional optical label after splitting are efficiently forwarded according to this queue and the fiber wavelength division multiplexing technology. This solution realizes the effective combination of the optical fiber link and signal priority, etc. without occupying additional bandwidth of the optical network and without changing the original transmission protocol of the network. It not only improves the optical path utilization efficiency of the entire optical network, saves link resources, ensures the timeliness of data, but also greatly reduces the network operation and maintenance cost, and this method can also be extended to the transmission of optical signals carrying other dimensions or more dimensions of useful information.

[0172] The present invention aims to optimize the operation and maintenance process of optical networks, improve network performance, reduce operating costs, and meet the growing demand for optical network communication through innovative multi-dimensional optical label generation technology and a deep recognition model for multi-dimensional optical labels.

[0173] Correspondingly, an embodiment of the present invention also provides a network operation and maintenance system for multi-dimensional optical labels. This system is deployed at each switching node of the optical network and includes an optical coupler, an optical label signal recovery module, and a computer device.

[0174] Among them, the optical coupler is used to receive the optical signal carrying multi-dimensional optical label information generated by the sending end or forwarded by other switching nodes, and split the optical signal carrying multi-dimensional optical label information into a higher-power optical signal carrying multi-dimensional optical label information and a lower-power optical signal carrying multi-dimensional optical label information.

[0175] The optical label signal recovery module includes a photodiode, a transimpedance amplifier, an analog-to-digital converter, a fast Fourier transform circuit, a Costas loop circuit, a carrier synchronization module, a clock synchronization module, and a demodulation module connected in sequence. It is used to perform recovery preprocessing on the lower-power optical signal carrying multi-dimensional optical label information to obtain multiple recovered one-dimensional optical label signals.

[0176] The computer device includes a processor and a memory. Computer instructions are stored in the memory, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the following steps in the aforementioned network operation and maintenance method for multi-dimensional optical labels:

[0177] Input the two-dimensional matrix representation of the multiple recovered one-dimensional optical label signals into the multi-dimensional optical label recognition model deployed in the switching node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a time module, a splicing module, a fully connected layer, and a normalized exponential function output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module.

[0178] Construct a multi-level dynamic feedback queue according to the output multi-dimensional optical label information, and forward the higher-power optical signal carrying multi-dimensional optical label information in a wavelength-division multiplexing manner according to the multi-level dynamic feedback queue.

[0179] An embodiment of the present invention also provides a signal generation and transmission system carrying multi-dimensional optical labels. This system is deployed at the sending end of the optical network and includes a high-speed transceiver, a programmable logic array, a digital-to-analog conversion module, a voltage-controlled current source circuit, a driver / receiver chip, a semiconductor optical amplifier, and a thermostat.

[0180] Among them, the high-speed transceiver is used to generate an original optical signal.

[0181] The programmable logic array is used to drive the digital-to-analog conversion module to generate modulation signals multiple times. The frequency of the modulation signal is lower than that of the original optical signal, and the frequency of each generated modulation signal is different;

[0182] The voltage-controlled current source circuit is used to convert the modulation signal generated each time into a drive current signal;

[0183] The driver / receiver chip is used to convert the drive current signal obtained each time into a differential enable signal;

[0184] The semiconductor optical amplifier is used to superimpose the differential enable signal obtained each time and the working current signal of the semiconductor optical amplifier to obtain modulation current signals multiple times; receive the original optical signal, fine-tune the original optical signal according to the modulation current signal obtained for the first time, and generate an optical signal carrying one-dimensional optical label information; it is also used to receive the optical signals carrying the corresponding-dimensional optical label information generated in each previous time multiple times, fine-tune the optical signals carrying the corresponding-dimensional optical label information according to the modulation current signal obtained each time, generate an optical signal carrying multi-dimensional optical label information, and send the optical signal carrying multi-dimensional optical label information to the switching node of the optical network;

[0185] The thermostat is connected to the semiconductor optical amplifier and is used to suppress the temperature drift of the current signal.

[0186] An embodiment of the present invention further provides an electronic device, which is deployed at each switching node in the optical network, including: a processor, a memory, and computer instructions stored on the memory. The processor is used to execute the computer instructions. When the computer instructions are executed, the system implements the following steps of the foregoing multi-dimensional optical label network operation and maintenance method:

[0187] Input the two-dimensional matrix representation of the restored multiple one-dimensional optical label signals into the multi-dimensional optical label recognition model deployed in the switching node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a time module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module; among them, the restored multiple one-dimensional optical label signals are obtained by performing restoration preprocessing on the optical signal carrying multi-dimensional optical label information with lower power. By receiving the optical signal carrying multi-dimensional optical label information generated by the sending end or forwarded by other switching nodes by the switching node of the optical network, and splitting the optical signal carrying multi-dimensional optical label information to obtain the optical signal carrying multi-dimensional optical label information with higher power and the optical signal carrying multi-dimensional optical label information with lower power;

[0188] Construct a multi-level dynamic feedback queue based on the output multi-dimensional optical label information, and forward the optical signal with higher power carrying the multi-dimensional optical label information in a wavelength division multiplexing manner according to the multi-level dynamic feedback queue. The electronic device can be various types of mobile terminals, desktop computers, servers, etc.

[0189] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps of the foregoing network operation and maintenance method of the multi-dimensional optical label are implemented:

[0190] Input the two-dimensional matrix representation of the restored multiple one-dimensional optical label signals into the multi-dimensional optical label recognition model deployed in the switching node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a time module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module; among them, the restored multiple one-dimensional optical label signals are obtained by performing restoration preprocessing on the optical signal with lower power carrying the multi-dimensional optical label information. The optical signal with lower power carrying the multi-dimensional optical label information is obtained by receiving the optical signal with multi-dimensional optical label information generated by the sending end or forwarded by other switching nodes by the switching node of the optical network, and splitting the optical signal with multi-dimensional optical label information to obtain the optical signal with higher power carrying the multi-dimensional optical label information and the optical signal with lower power carrying the multi-dimensional optical label information;

[0191] Construct a multi-level dynamic feedback queue based on the output multi-dimensional optical label information, and forward the optical signal with higher power carrying the multi-dimensional optical label information in a wavelength division multiplexing manner according to the multi-level dynamic feedback queue. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0192] An embodiment of the present invention further provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the following steps of the foregoing network operation and maintenance method of the multi-dimensional optical label are implemented:

[0193] The two-dimensional matrix representation of the restored multiple one-dimensional optical label signals is input into the multi-dimensional optical label recognition model deployed in the switching node for recognition, so that the multi-dimensional optical label recognition model outputs multi-dimensional optical label information. The multi-dimensional optical label recognition model includes a spatial module, an activation function layer, a temporal module, a splicing module, a fully connected layer, and a softmax output layer connected in sequence. The output of the spatial module is also connected to the input of the splicing module. Among them, the restored multiple one-dimensional optical label signals are obtained by performing restoration preprocessing on the optical signal carrying multi-dimensional optical label information with lower power. The optical signal carrying multi-dimensional optical label information generated by the sending end or forwarded by other switching nodes is received by the switching node of the optical network, and the optical signal carrying multi-dimensional optical label information is split into a higher-power optical signal carrying multi-dimensional optical label information and a lower-power optical signal carrying multi-dimensional optical label information.

[0194] A multi-level dynamic feedback queue is constructed according to the output multi-dimensional optical label information, and the optical signal with higher power carrying multi-dimensional optical label information is forwarded in a wavelength-division multiplexing manner according to the multi-level dynamic feedback queue.

[0195] Those of ordinary skill in the art should understand that the various exemplary components, systems, and methods described in connection with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Specifically, whether to execute in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium or transmitted through a data signal carried in a carrier wave on a transmission medium or a communication link.

[0196] It should be clear that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present invention.

[0197] In the present invention, the features described and / or illustrated for one embodiment can be used in the same or a similar manner in one or more other embodiments, and / or combined with the features of other embodiments or replace the features of other embodiments.

[0198] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for generating and transmitting a signal carrying a multi-dimensional optical tag, characterized in that: The method comprises: The transmitting end of the optical network generates an original optical signal through a high-speed transceiver, and drives a digital-to-analog conversion module through a programmable logic array to generate a modulated signal, wherein the frequency of the modulated signal is lower than the frequency of the original optical signal; Converting the modulation signal into a driving current signal through a voltage-controlled current source circuit; The driving current signal and the working current signal of the semiconductor optical amplifier are superimposed by a semiconductor optical amplifier to obtain a modulated current signal; Inputting the original optical signal into the semiconductor optical amplifier, modulating the original optical signal according to the modulation current signal through the semiconductor optical amplifier to generate an optical signal carrying one-dimensional optical label information; Repeat the steps of generating, converting and superimposing the modulation signal multiple times, inputting the optical signal carrying the corresponding dimension optical label information generated last time into the semiconductor optical amplifier each time, modulating the optical signal carrying the corresponding dimension optical label information by the semiconductor optical amplifier according to the modulation current signal obtained each time, and generating an optical signal carrying multi-dimensional optical label information, wherein the frequency of the modulation signal generated each time is different; The optical signal carrying the multi-dimensional optical label information is sent to a switching node of the optical network.

2. The method according to claim 1, characterized in that The method further includes: connecting a temperature controller via the semiconductor optical amplifier to suppress temperature drift of the current signal.

3. The method according to claim 1, characterized in that After converting the modulation signal into a driving current signal through a voltage-controlled current source circuit, the method further includes: Converting the drive current signal into a differential enable signal through a driver / receiver chip; The differential enable signal and the working current signal of the semiconductor optical amplifier are superimposed through the semiconductor optical amplifier to obtain a modulated current signal.

4. The method according to any one of claims 1 to 3, characterized in that The multi-dimensional optical label information includes wavelength information and priority information corresponding to the original optical signal generated by the transmitting end.

5. A signal generation and transmission system carrying a multi-dimensional optical tag, characterized in that: The system is deployed at the transmitting end of the optical network, and includes: a high-speed transceiver, a programmable logic array, a digital-to-analog conversion module, a voltage-controlled current source circuit, and a semiconductor optical amplifier. The high-speed transceiver is used to generate an original optical signal; The programmable logic array is used to drive the digital-to-analog conversion module to generate a modulation signal multiple times, the frequency of the modulation signal is lower than the frequency of the original optical signal, and the frequency of the modulation signal generated each time is different; The voltage-controlled current source circuit is used to convert the modulation signal generated each time into a driving current signal; The semiconductor optical amplifier is used to superimpose the driving current signal obtained each time and the working current signal of the semiconductor optical amplifier to obtain the modulated current signal multiple times; receive the original optical signal, modulate the original optical signal according to the modulated current signal obtained for the first time, and generate an optical signal carrying one-dimensional optical label information; and is also used to receive the optical signal carrying the corresponding dimensional optical label information generated each time multiple times, modulate the optical signal carrying the corresponding dimensional optical label information according to the modulated current signal obtained each time, generate an optical signal carrying multi-dimensional optical label information, and send the optical signal carrying the multi-dimensional optical label information to the switching node of the optical network.

6. The system according to claim 5, characterized in that The system further comprises: a driver / receiver chip, the driver / receiver chip being used to convert the driving current signal obtained each time into a differential enable signal; The semiconductor optical amplifier is also used to superimpose the differential enable signal obtained each time and the working current signal of the semiconductor optical amplifier to obtain the modulated current signal multiple times.

7. The system according to claim 5, characterized in that The system further comprises: a temperature controller, which is connected to the semiconductor optical amplifier and is used to suppress the temperature drift of the current signal.

8. The system according to any one of claims 5 to 7, characterized in that The multi-dimensional optical label information includes wavelength information and priority information corresponding to the original optical signal generated by the transmitting end.

Images