Weak network dynamic adaptive transmission method and device based on two-dimensional link entropy

Abstract

This invention discloses a dynamic adaptive transmission method and apparatus for weak networks based on dual-dimensional link entropy, relating to the fields of communication network transmission and data communication technology. The method includes: starting the transmitting end system and configuring core parameters; collecting and preprocessing multi-dimensional link features within a fixed time window; normalizing the delay difference sequence; calculating the delay difference entropy and packet loss location entropy; combining weights to obtain and calibrate the comprehensive link feature entropy; generating adjustment instructions according to the entropy value range; adjusting the service flow transmission parameters; allocating bandwidth according to service priority and feature entropy; adjusting the transmission rate and fine-tuning parameters in real time to complete the transmission. This invention can accurately distinguish between random packet loss and congestion packet loss in weak networks, improve bandwidth utilization and transmission stability, reduce retransmission rate and stuttering rate, and ensure the transmission quality of high-priority services.

Description

Technical Field

[0001] This invention relates to the field of communication network transmission and data communication technology, and in particular to a dynamic adaptive transmission method and apparatus for weak networks based on dual-dimensional link entropy. Background Technology

[0002] With the rapid popularization of mobile internet and the Internet of Things, the demand for weak network transmission scenarios such as subways, remote areas, high-speed mobile networks, and industrial plants continues to grow. These scenarios generally suffer from narrow bandwidth and high packet loss rates (3%). 20%), large latency jitter (20ms) With characteristics such as 200ms and frequent dynamic fluctuations in channel status, traditional transmission schemes are no longer able to meet the high-quality transmission requirements of high-definition video, real-time control, and remote interaction.

[0003] Currently, weak network transmission often employs congestion control schemes centered on RTT and packet loss rate, with typical examples being the BBR and CUBIC algorithms, widely used in various terminals, routers, and servers. Existing technologies use a binary approach of adjusting the sending window and rate through fixed thresholds, which is only suitable for relatively stable network scenarios. In weak network environments, they cannot distinguish between random packet loss and congestion-induced packet loss, resulting in incomplete link state representation, rigid adjustment strategies, and delayed response. They are ill-suited to dynamically fluctuating channels and cannot meet the requirements for high real-time and high-reliability transmission. Summary of the Invention

[0004] The main objective of this invention is to provide a dynamic adaptive transmission method for weak networks based on two-dimensional link entropy.

[0005] Another objective of this invention is to propose a dynamic adaptive transmission device for weak networks based on two-dimensional link entropy.

[0006] The third objective of this invention is to provide an electronic device.

[0007] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0008] To achieve the above objectives, a first aspect of the present invention proposes a dynamic adaptive transmission method for weak networks based on dual-dimensional link entropy, comprising:

[0009] Start the sending system and configure the core parameters, collect multi-dimensional link characteristic parameters within a fixed time window and perform preprocessing; Based on the preprocessed feature sequence, the delay difference sequence is normalized, the delay difference entropy and packet loss location entropy are calculated, and the link comprehensive feature entropy is calculated and calibrated in combination with the weight coefficient. The calibrated link comprehensive characteristic entropy is divided into different numerical ranges, corresponding adjustment instructions are generated, and the sending parameters of each service flow are adjusted according to the bandwidth allocation ratio. Obtain the preset priority of each service flow, and calculate the bandwidth allocation weight of each service flow and allocate the link bandwidth based on the calibrated link comprehensive feature entropy and the transmission parameters of each service flow. The system extracts packets from each service flow and sends them to intermediate routing nodes. It adjusts the packet sending rate in real time through the rate control interface, fine-tunes parameters based on the actual sending status feedback, and finally determines whether the transmission task is complete.

[0010] Optionally, start the sending system and configure core parameters, collect multi-dimensional link characteristic parameters within a fixed time window and perform preprocessing, including: Start the sending system and configure the following core parameters: set the time window T and weight coefficient. Calibration coefficient k, number of packet loss location intervals m, acquisition frequency, priority of each service flow and minimum bandwidth guarantee value, initial value of sending window W, initial value of sending rate V; Initialize the transmit buffer and ACK receive buffer, and clear the historical data and entropy calculation results; The control link feature acquisition unit collects the delay difference sequence D, packet loss location sequence L, and message interval sequence I within the current time window T at a preset acquisition frequency. Among them, the delay difference sequence D is used to characterize the fluctuation of link delay, the packet loss location sequence L is used to characterize the randomness of packet loss, and the message interval sequence I is used to assist in correcting the accuracy of entropy value calculation. The collected sequence data is denoised, and missing data is supplemented using linear interpolation to obtain the preprocessed feature sequence.

[0011] Optionally, based on the preprocessed feature sequence, the delay difference sequence is normalized, the delay difference entropy and packet loss location entropy are calculated, and combined with weighting coefficients, the link comprehensive feature entropy is calculated and calibrated, including: After receiving the preprocessed feature sequence through the link feature entropy calculation unit, the delay difference sequence D is normalized, and each element in D is normalized to the interval [0,1]. The normalization formula is as follows:

[0012] in, Let i be the i-th element in the time delay difference sequence D. Let be the minimum value of sequence D. The maximum value of sequence D; Perform probability distribution statistics on the normalized time delay difference sequence, calculate the probability of each value, and obtain the time delay difference probability distribution:

[0013] Calculate the probability distribution of time delay difference and the entropy of time delay difference. :

[0014] Where i ranges from 1 to n, and n is the length of sequence D, when When, define ; Perform location probability statistics on the packet loss location sequence L, divide the transmission sequence within the time window T into m equal intervals, and calculate the probability of packet loss in each interval to obtain the packet loss location probability distribution:

[0015] Calculate the entropy of packet loss location :

[0016] Among them, when When, define ; The formula for calculating the comprehensive feature entropy H is:

[0017] in, These are the weighting coefficients. The larger the value of , the higher the weight of the impact of latency fluctuations on the link status; Based on the message interval sequence I, the overall entropy value H is calibrated using the following formula:

[0018] in, For message interval sequence The standard deviation of , where k is the calibration coefficient.

[0019] Optionally, the calibrated link composite characteristic entropy is divided into different numerical ranges, and corresponding adjustment instructions are generated. Based on the bandwidth allocation ratio, the transmission parameters of each service flow are adjusted, including: The calibrated comprehensive entropy value is received through the entropy weight adaptive adjustment unit. Based on a preset entropy range, the current link status is determined, and a corresponding transmission command is generated. like If ∈ [0, 0.3), then the corresponding link state is determined to be a stable link. At this time, the link has no obvious fluctuations, the random packet loss rate is <3%, and the latency jitter is <20ms. The sending window increases by 1.25 times the time window, and the sending rate increases by 10% of the link bandwidth limit per time window until the sending rate reaches the link bandwidth limit. The retransmission strategy adopts fast retransmission, and the timeout retransmission time is set to 1.5 times the smooth RTT. like If ∈ [0.3, 0.7), then the corresponding link status is determined to be a weakly fluctuating link. At this time, the link fluctuates slightly, with a random packet loss rate of 3%-8% and a latency jitter of 20ms-80ms. The sending window remains unchanged at its current size, and the sending rate adopts a pacing smoothing strategy. The retransmission strategy adopts a combination of fast retransmission and selective retransmission, only retransmitting lost packets, and setting the timeout retransmission time to twice the smoothing RTT. like If ∈ [0.7, 1.0], the corresponding link state is determined to be a strongly random link. At this time, the link fluctuates violently, with a random packet loss rate >8% and latency jitter >80ms. The sending window is reduced to 70% of the current size, the sending rate is reduced to 60% of the current rate, and the rate anti-jitter mechanism is enabled. The retransmission strategy adopts a combination of fast retransmission and redundant retransmission. High-priority packets are retransmitted once more, and the timeout retransmission time is set to 2.5 times the smoothed RTT. It receives the bandwidth allocation ratio from the multi-stream entropy weight scheduling unit and coordinates the adjustment of the transmission parameters of each service stream.

[0020] Optionally, obtain the preset priority of each service flow, calculate the bandwidth allocation weight of each service flow and allocate link bandwidth based on the calibrated link comprehensive characteristic entropy and the transmission parameters of each service flow, including: The comprehensive entropy value is received through the multi-stream entropy weight scheduling unit. Obtain the priority and bandwidth requirements of all currently active service flows, and calculate the bandwidth allocation weight for each service flow. The system generates a bandwidth allocation instruction, transmits it to the sending execution unit, and feeds back the bandwidth allocation ratio to the entropy weight adaptive adjustment unit. The formula for calculating the bandwidth allocation weights for each service flow is as follows:

[0021] in, Assign a weight to the bandwidth of the i-th service flow. Let be the priority coefficient for the i-th business flow. Let be the bandwidth requirement coefficient for the i-th service flow. This is the sum of the priority coefficients for all business flows.

[0022] Optionally, extract packets from each service flow and send them to intermediate routing nodes. Adjust the packet sending rate in real time through the rate control interface, fine-tune parameters based on the actual sending status feedback, and finally determine whether the transmission task is complete, including: The sending execution unit receives sending parameter adjustment instructions and bandwidth allocation instructions, extracts messages from each service flow from the sending buffer, sends messages according to the adjusted sending window and sending rate, and simultaneously receives ACK messages from the ACK receiving unit, identifies lost messages, and performs retransmission operations according to the adjusted retransmission strategy. The control and transmission execution unit feeds back the actual transmission status of this iteration to the entropy weight adaptive adjustment unit, and the link feature acquisition unit feeds back the feature sequence acquired in this iteration to the link feature entropy calculation unit. The entropy-weighted adaptive adjustment unit fine-tunes the weighting coefficients based on the actual transmission status feedback. Based on the entropy value range threshold, the link feature entropy calculation unit adjusts the acquisition frequency according to the feedback feature sequence; Determine if the transmission task is complete. If not, proceed to the next time window T and repeat the above steps. If complete, terminate the process and shut down the system unit.

[0023] To achieve the above objectives, a second aspect of the present invention provides a dynamic adaptive transmission device for weak networks based on dual-dimensional link entropy, comprising: The preprocessing module is used to start the sending system and configure core parameters, collect multi-dimensional link characteristic parameters within a fixed time window and perform preprocessing. The calculation module is used to normalize the delay difference sequence based on the preprocessed feature sequence, calculate the delay difference entropy and packet loss location entropy, and calculate the link comprehensive feature entropy and perform calibration in combination with the weight coefficient. The adjustment module is used to divide the calibrated link comprehensive characteristic entropy into different numerical ranges, generate corresponding adjustment instructions, and adjust the sending parameters of each service flow according to the bandwidth allocation ratio. The allocation module is used to obtain the preset priority of each service flow, calculate the bandwidth allocation weight of each service flow and allocate the link bandwidth based on the calibrated link comprehensive feature entropy and the transmission parameters of each service flow. The judgment module is used to extract the packets of each service flow and send them to the intermediate routing node. It adjusts the packet sending rate in real time through the rate control interface, fine-tunes the parameters based on the actual sending status feedback, and finally determines whether the transmission task is completed.

[0024] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0025] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing the weak network dynamic adaptive transmission method based on dual-dimensional link entropy as described in the first aspect embodiment.

[0026] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements the weak network dynamic adaptive transmission method based on two-dimensional link entropy as described in the first aspect embodiment.

[0027] The embodiments of the present invention have the following beneficial effects: This invention can accurately distinguish between random packet loss and congestion packet loss in weak network scenarios, avoiding misjudgment of congestion; through a dynamic adjustment strategy based on link feature entropy, it significantly improves link bandwidth utilization and transmission stability, and reduces packet retransmission rate and service interruption rate; at the same time, through an entropy weighted bandwidth allocation mechanism, it ensures the transmission quality of high-priority services, solving the shortcomings of existing technologies in weak network scenarios. Attached Figure Description

[0028] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart of a dynamic adaptive transmission method for weak networks based on dual-dimensional link entropy provided in an embodiment of the present invention; Figure 2 A schematic diagram of a dynamic adaptive transmission system for weak networks based on dual-dimensional link entropy, provided for an embodiment of the present invention; Figure 3 This is a structural diagram of a dynamic adaptive transmission device for weak networks based on dual-dimensional link entropy, provided in an embodiment of the present invention. Detailed Implementation

[0029] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] The following description, with reference to the accompanying drawings, describes a dynamic adaptive transmission method and apparatus for weak networks based on dual-dimensional link entropy, according to embodiments of the present invention.

[0032] Example 1 This invention provides a dynamic adaptive transmission method for weak networks based on dual-dimensional link entropy, such as... Figure 1 As shown, the method includes the following steps: S1, start the sending end system and configure the core parameters, collect multi-dimensional link characteristic parameters within a fixed time window and perform preprocessing.

[0033] To achieve stable and reliable link transmission control and accurate status awareness, this application completes the initialization of the transmitting end system, parameter configuration, link feature acquisition, and data preprocessing.

[0034] In this embodiment, the transmitting system is first started, entering the initialization configuration phase. Specifically, this application sets the time window T to 200 milliseconds, which serves as the basic time unit for link feature acquisition and status analysis; weighting coefficients are configured. The value is set to 0.6, which is used for weighted balancing of features or indicators of different dimensions in subsequent multi-factor decision-making or fusion calculations; the calibration coefficient k is set to 0.1, which is used for real-time calibration and error correction of collected data or calculation results; at the same time, the number of packet loss location intervals m is set to 10, which divides the possible packet loss range into equal intervals, making it easier for refined analysis and location.

[0035] In this embodiment, the link data acquisition frequency is preset to 100Hz to ensure that the dynamic changes in the link status can be captured at a sufficiently high frequency. For different service data streams, this application sets their corresponding priority levels and minimum bandwidth guarantees to ensure the quality of service for critical services. The core transmission parameters of the sending end are also initialized, including setting the initial value W of the sending window to 128 kilobytes and the initial value V of the sending rate to 50% of the link bandwidth limit, thereby establishing a reasonable initial transmission baseline for data transmission.

[0036] In this embodiment, after configuring the core parameters, the initialization of the relevant data storage areas is completed simultaneously. This application initializes the sending buffer to prepare for receiving and temporarily storing service data to be sent; simultaneously, it initializes the ACK receiving buffer to store acknowledgment information returned by the peer device. To ensure the accuracy and independence of this communication process, this application clears all historically collected link data and previous entropy calculation results to avoid interference or deviation of historical data from the current link status analysis.

[0037] In this embodiment, after the system is ready, the data acquisition task is started in real time by the link feature acquisition unit according to the aforementioned preset acquisition frequency of 100Hz. Within the currently set time window T, three types of key link feature sequences are continuously acquired.

[0038] In this embodiment, the specifically collected sequences include a delay difference sequence D, a packet loss location sequence L, and a message interval sequence I. Each of these three sequences plays a unique role in representing the link state. The delay difference sequence D primarily reflects the delay fluctuations between data packets during transmission and is a core indicator for evaluating link delay stability and jitter. The packet loss location sequence L records the specific location information of data packets lost along the transmission path; analyzing this sequence can effectively characterize the random distribution characteristics of network packet loss. The message interval sequence I is mainly used for auxiliary calibration, improving the accuracy of subsequent link state quantification calculations, especially entropy calculations, and providing supplementary data transmission timing analysis.

[0039] In this embodiment, the acquired raw sequence data is inevitably affected by factors such as network noise and device interference, and some data may be missing. To ensure the reliability of subsequent analysis and calculation, this application requires preprocessing of the raw acquired data.

[0040] First, the collected sequence data undergoes denoising processing. Abnormal noise points are removed using filtering or smoothing algorithms to restore the true trend of the data. For missing values ​​encountered during data acquisition, this application employs linear interpolation to fill in the gaps. By linearly fitting adjacent valid data points, the missing data is estimated and filled in, thus obtaining a complete and continuous preprocessed feature sequence. This provides high-quality, highly reliable data support for subsequent link analysis and transmission control decisions.

[0041] In this embodiment, a reliable operating environment and high-quality feature data are provided for subsequent transmission control through system startup, parameter configuration, buffer initialization, link feature acquisition and data preprocessing.

[0042] S2, based on the preprocessed feature sequence, normalizes the delay difference sequence, calculates the delay difference entropy and packet loss location entropy, and combines the weight coefficients to calculate the link comprehensive feature entropy and perform calibration.

[0043] In order to accurately quantify the uncertainty of the link state, this application performs entropy calculation and calibration on the preprocessed feature sequence to obtain the calibrated comprehensive feature entropy.

[0044] In this embodiment of the application, the link feature entropy calculation unit calculates and calibrates the entropy value of the preprocessed feature sequence to quantify the disorder and uncertainty of the link state, thereby providing a reliable basis for subsequent transmission control.

[0045] In this embodiment, after receiving the preprocessed delay difference sequence D, packet loss location sequence L, and message interval sequence I, the link feature entropy calculation unit first performs normalization processing on the delay difference sequence D, mapping each element in the sequence to a numerical range of 0 to 1, eliminating calculation biases caused by different units and numerical ranges. The specific formula for normalization is:

[0046] in, Let D represent the i-th time delay difference element after normalization, min(D) be the minimum value of the time delay difference sequence D, and max(D) be the maximum value of the time delay difference sequence D.

[0047] In this embodiment of the application, after normalization, the application performs probability distribution statistics on the processed delay difference sequence, counts the frequency of each value, and then calculates the corresponding delay difference probability distribution. The calculation formula is as follows:

[0048] In this embodiment of the application, based on the obtained delay difference probability distribution, the delay difference entropy is further calculated. The specific calculation formula is as follows:

[0049] Where i ranges from 1 to n, and n represents the length of the time delay difference sequence D, when the probability When equal to 0, the definition is... The value of is set to 0 to avoid meaningless calculation results.

[0050] In this embodiment, the application simultaneously performs packet loss location probability statistics on the packet loss location sequence L. The overall transmission sequence within the current time window T is evenly divided into m intervals, and the number of lost packets and the total number of transmitted packets are counted in each interval. The packet loss probability corresponding to each interval is then calculated to obtain the packet loss location probability distribution. The calculation formula is as follows:

[0051] In this embodiment, based on the packet loss location probability distribution, the packet loss location entropy is calculated. The specific calculation formula is as follows:

[0052] When probability When it equals 0, the same applies to this item. The value of is defined as 0.

[0053] In this embodiment of the application, the delay difference entropy is... Entropy of packet loss location We perform weighted fusion to obtain the comprehensive feature entropy H, calculated using the following formula:

[0054] in, The pre-set weighting coefficient has a default value of 0.6. The larger the value, the higher the weight of latency fluctuations in the overall link state assessment. In this embodiment, to further improve the accuracy and robustness of the comprehensive feature entropy, this application calibrates the comprehensive entropy value H in conjunction with the message interval sequence I. The specific calibration formula is as follows:

[0055] in, is the standard deviation of the message interval sequence I, used to reflect the degree of fluctuation of the message interval, and k is a pre-configured calibration coefficient. This calibration process can effectively reduce the interference of abnormal fluctuations in the message interval on the entropy calculation, and obtain a calibrated comprehensive feature entropy that is more in line with the real link state.

[0056] In this embodiment of the application, by calculating the entropy values ​​of the delay difference sequence and the packet loss location sequence respectively and weighting and fusing them, and combining them with the message interval sequence for calibration, a comprehensive feature entropy that can accurately reflect the link status is obtained, which lays the foundation for subsequent adjustment of the transmission parameters of each service flow.

[0057] S3 divides the calibrated link comprehensive characteristic entropy into different numerical ranges, generates corresponding adjustment instructions, and adjusts the sending parameters of each service flow according to the bandwidth allocation ratio.

[0058] In order to adaptively optimize transmission performance based on link status, this application determines the link status based on the comprehensive entropy value range and dynamically adjusts the sending and retransmission parameters.

[0059] In this embodiment, after receiving the calibrated comprehensive entropy value H_cal, the entropy weight adaptive adjustment unit classifies the current link state according to the preset entropy value range, and generates a corresponding transmission control command based on the different link states to ensure stable and efficient data transmission in different network environments.

[0060] In the embodiments of this application, when When the link is in the range [0, 0.3), this application determines that the current link is a stable link. In this state, the link has no obvious fluctuations, the random packet loss rate is less than 3%, and the latency jitter is less than 20ms. At this time, this application increases the sending window by 1.25 times per time window and increases the sending rate by 10% of the upper limit of the link bandwidth per time window until the sending rate reaches the upper limit of the link bandwidth. At the same time, the retransmission strategy adopts fast retransmission and the timeout retransmission time is set to 1.5 times the smooth RTT.

[0061] In the embodiments of this application, when When the value is in the range of [0.3, 0.7), this application determines that the current link is a weakly fluctuating link. In this state, the link has slight fluctuations, the random packet loss rate is between 3% and 8%, and the latency jitter is between 20ms and 80ms. At this time, this application keeps the sending window unchanged, adopts a pacing smoothing strategy for the sending rate, and adopts a combination of fast retransmission and selective retransmission for the retransmission strategy. Only lost packets are retransmitted, and the timeout retransmission time is set to twice the smoothing RTT.

[0062] In the embodiments of this application, when When the link is in the range of [0.7, 1.0], this application determines that the current link is a strongly random link. In this state, the link fluctuates violently, with a random packet loss rate greater than 8% and a latency jitter greater than 80ms. At this time, this application reduces the sending window to 70% of the current size, reduces the sending rate to 60% of the current rate, and enables the rate anti-jitter mechanism. The retransmission strategy adopts a combination of fast retransmission and redundant retransmission. High-priority packets are retransmitted once for redundancy. At the same time, the timeout retransmission time is set to 2.5 times the smoothed RTT.

[0063] In this embodiment, the bandwidth allocation ratio fed back from the multi-stream entropy weight scheduling unit is also received, and the transmission parameters of each service stream are adjusted collaboratively based on the feedback information to achieve overall transmission performance optimization in multi-service stream scenarios.

[0064] In this embodiment, the link status is determined by the comprehensive entropy value range, and the sending window, rate and retransmission strategy are adaptively adjusted to achieve stable and efficient transmission under multiple service flows. This lays the foundation for subsequent calculation of bandwidth allocation weights for each service flow and allocation of link bandwidth.

[0065] S4: Obtain the preset priority of each service flow, calculate the bandwidth allocation weight of each service flow and allocate the link bandwidth based on the calibrated link comprehensive characteristic entropy and the transmission parameters of each service flow.

[0066] To ensure the transmission quality of different services under complex links, this application divides multiple service flows according to priority and adaptively allocates bandwidth based on the comprehensive entropy value.

[0067] In this application embodiment, multiple service flows are prioritized and adaptively allocated bandwidth to achieve hierarchical guarantee of transmission quality for different services under complex link conditions. This application divides the multiple service flows in the system into three configurable priorities, ranked from highest to lowest: high-priority services, medium-priority services, and low-priority services. High-priority services include real-time control, high-definition live streaming, and emergency data transmission; medium-priority services include ordinary video and voice calls; and low-priority services include file downloads and data synchronization.

[0068] In this embodiment of the application, corresponding minimum bandwidth guarantee values ​​are set for service flows of different priorities. The minimum bandwidth guarantee value for high-priority service flows is not less than 50% of the link bandwidth, the minimum bandwidth guarantee value for medium-priority service flows is not less than 30% of the link bandwidth, and the minimum bandwidth guarantee value for low-priority service flows is not less than 20% of the link bandwidth.

[0069] In this embodiment of the application, the bandwidth allocation weight of each service flow is calculated based on the aforementioned link comprehensive feature entropy. The specific allocation formula is as follows:

[0070] in, Assign a weight to the bandwidth corresponding to the i-th service flow. Let be the priority coefficient of the i-th service flow, and let be the priority coefficient of higher priority service flows. A value of 0.8 indicates a medium-priority service flow. A value of 0.5 indicates a low-priority service flow. The value is 0.2. This is the bandwidth requirement coefficient for the i-th service flow, set according to the pre-defined bandwidth requirement for that service flow, with a value ranging from 0.1 to 1.0. This is the sum of the priority coefficients for all business flows.

[0071] In this embodiment of the application, the bandwidth allocation weight is calculated. The application allocates actual link bandwidth to each service flow, prioritizing the minimum bandwidth requirement of each service flow during the allocation process. In this embodiment, when there is remaining link bandwidth, the remaining bandwidth is redistributed according to the weight ratio of each service flow; when link bandwidth resources are insufficient, the application prioritizes reducing the bandwidth occupied by low-priority service flows, thereby ensuring the transmission stability and service quality of high-priority service flows.

[0072] In this embodiment, by classifying multiple service flows, setting minimum bandwidth guarantees, and adaptively calculating the weight allocation of bandwidth based on the comprehensive entropy value, a reasonable scheduling of high-priority service priority guarantee is achieved, laying the foundation for subsequent judgment on whether the transmission task has been completed.

[0073] S5 extracts packets from each service flow and sends them to intermediate routing nodes. It adjusts the packet sending rate in real time through the rate control interface, fine-tunes parameters based on the actual sending status feedback, and finally determines whether the transmission task is completed.

[0074] To achieve closed-loop adaptive optimization of the transmission process, this application completes continuous transmission until the task ends through message sending, retransmission, status feedback and parameter iteration.

[0075] In this embodiment of the application, the sending execution unit receives a sending parameter adjustment instruction from the entropy weight adaptive adjustment unit and a bandwidth allocation instruction from the multi-stream entropy weight scheduling unit. According to the instruction content, it extracts the messages to be sent corresponding to each service stream from the sending buffer and completes the message sending operation according to the adjusted sending window size and sending rate.

[0076] In this embodiment of the application, the sending execution unit simultaneously receives the ACK confirmation message fed back by the ACK receiving unit, identifies the lost messages during the transmission process based on the ACK message, and performs the corresponding retransmission operation on the lost messages according to the previously determined retransmission strategy to ensure the reliability of data transmission.

[0077] In this embodiment, the sending execution unit feeds back the actual sending status during the current time window iteration process to the entropy weight adaptive adjustment unit, and the link feature acquisition unit feeds back the feature sequence acquired within the current time window to the link feature entropy calculation unit, forming a closed-loop feedback mechanism.

[0078] In this embodiment, the entropy weight adaptive adjustment unit adaptively fine-tunes the weight coefficient α and the entropy value range threshold according to the received actual transmission status, thereby improving the accuracy of subsequent link status judgment. The link feature entropy calculation unit dynamically adjusts the acquisition frequency according to the feedback feature sequence, so that the data acquisition is more in line with the actual link changes.

[0079] In this embodiment, it is determined in real time whether the overall transmission task has been completed. If it has not been completed, the next time window T is entered, and the complete process of parameter configuration, link acquisition, entropy calculation, status determination, bandwidth allocation and message sending is repeated. If the transmission task has been completed, the entire transmission process is terminated and the relevant system units are shut down.

[0080] In the embodiments of this application, a closed-loop control is formed by message sending, retransmission, status feedback and parameter iterative adjustment, and the transmission task is continuously executed until completion.

[0081] To fully verify the effectiveness and practicality of the technical solution of this invention, and to specifically address the shortcomings of existing technologies in weak network scenarios, three typical weak network scenarios were selected for comparative testing. Using existing technologies (CUBIC algorithm) as a control, the advantages of this invention in key indicators such as link bandwidth utilization, latency jitter, and packet retransmission rate were verified. Specific implementation details are as follows: The implementation scenario parameters are as follows: Test scenarios (3 typical weak network scenarios, covering mainstream application scenarios): Scenario type: Weak 5G network in the subway (mixed coverage of China Mobile and China Unicom 5G), simulating the link status during peak hours (morning peak 7:30-9:00).

[0082] Scenario 1: Weak 5G network in the subway (peak hours), link bandwidth 1Mbps-5Mbps, random packet loss rate 3%-8%, latency jitter 20ms-80ms, baseline RTT 150ms; Scenario 2: Weak 4G network in remote areas, link bandwidth 500Kbps-2Mbps, random packet loss rate 5%-12%, latency jitter 50ms-120ms, baseline RTT 200ms; Scenario 3: Vehicle-mounted high-speed mobile weak network, link bandwidth 2Mbps-6Mbps, random packet loss rate 4%-10%, latency jitter 40ms-100ms, baseline RTT 180ms.

[0083] Test equipment: Sending end: Huawei Mate 60 Pro (equipped with Linux kernel 4.19) and industrial gateway (ARM Cortex-A53), both deployed with the system unit of this invention; Receiver: Server (CPU Intel i7-12700H, 16GB RAM), deploys standard receiving programs to record transmitted data; Auxiliary equipment: 5G / 4G signal simulator (for stabilizing weak network environments), network monitor (for real-time collection of link parameters).

[0084] Test service: The test service adopts "high-definition live streaming (1080P, bitrate 2Mbps) + real-time control commands (100 bytes / command, 10Hz transmission frequency)", with dual service multiplexing links, both set to high priority, to simulate real application scenarios.

[0085] Test duration: 30 minutes for each scenario, repeated 3 times, and the average of the 3 tests is taken as the final result to ensure data reliability.

[0086] The system unit parameters of this invention are uniformly configured (and can be fine-tuned according to the scenario), as follows: Key parameters: Time window T=200ms, link feature acquisition frequency 100Hz, weight coefficient α=0.6, calibration coefficient k=0.1, number of packet loss location intervals m=10; Entropy range thresholds: [0, 0.3) (stable link), [0.3, 0.7) (weakly fluctuating link), [0.7, 1.0] (strongly random link); Service configuration: Minimum bandwidth guarantee of 50% for high-priority services; multi-stream entropy weight allocation weights are based on the formula. calculate; Adjustment strategy: Strictly follow the window, rate, and retransmission adjustment rules corresponding to each entropy value interval, with closed-loop fine-tuning amplitude ≤10%. Test results and comparisons are shown in Table 1: Table 1

[0087] The tests conducted through the above embodiments demonstrate that the solution of the present invention can accurately distinguish between random packet loss and congestion packet loss in weak network scenarios, avoiding misjudgment of congestion. Through a dynamic adjustment strategy based on link characteristic entropy, the link bandwidth utilization and transmission stability are significantly improved, while reducing packet retransmission rate and service interruption rate. At the same time, through the entropy weight bandwidth allocation mechanism, the transmission quality of high-priority services is guaranteed, completely solving the defects of existing technologies in weak network scenarios and achieving the purpose of the present invention.

[0088] Example 2 This invention relates to a dynamic adaptive transmission system for weak networks based on dual-dimensional link entropy, such as... Figure 2 As shown, the system is deployed at the sending end (which can be integrated into terminal devices, servers, gateways, etc.), forming a complete transmission link with the receiving end and intermediate routing nodes. The system mainly consists of 5 functional units, which are interconnected and work together. The specific structure and connection relationships are as follows: The link feature acquisition unit is deployed in the kernel layer of the sending end (deeply integrated with the TCP / UDP protocol stack). It captures the link transmission data between the sending end and the receiving end through hook functions and collects multi-dimensional link feature parameters within a fixed time window. This unit is bidirectionally connected to the entropy calculation unit and transmits the collected feature parameters to the link feature entropy calculation unit in real time. At the same time, it receives the acquisition frequency adjustment instructions fed back by the entropy calculation unit.

[0089] The link feature entropy calculation unit is deployed between the application layer and the kernel layer of the sending end, serving as the core decision support unit of the system. This unit is bidirectionally connected to the link feature acquisition unit and the entropy weight adaptive adjustment unit, respectively. It receives multi-dimensional feature parameters transmitted by the link feature acquisition unit, calculates the comprehensive link feature entropy value, and transmits the entropy value to the entropy weight adaptive adjustment unit in real time. At the same time, according to the rate of change of the entropy value, it sends an acquisition frequency adjustment command to the link feature acquisition unit (increase the acquisition frequency when the entropy value fluctuates greatly; decrease the acquisition frequency when the entropy value is stable).

[0090] The entropy weight adaptive adjustment unit is deployed at the kernel layer of the sending end and is integrated with the congestion control unit and rate control unit of the TCP / UDP protocol stack. This unit is bidirectionally connected to the link feature entropy calculation unit, the multi-stream entropy weight scheduling unit, and the sending execution unit. It receives the comprehensive entropy value transmitted by the link feature entropy calculation unit, and generates corresponding sending window adjustment instructions, sending rate adjustment instructions, and retransmission strategy adjustment instructions according to the numerical range of the entropy value. At the same time, it receives the bandwidth allocation ratio fed back by the multi-stream entropy weight scheduling unit and coordinates the adjustment of the sending parameters of each service flow.

[0091] The multi-stream entropy weight scheduling unit is deployed at the application layer of the sending end and is responsible for bandwidth allocation and scheduling of multiple service streams. This unit is bidirectionally connected to the entropy weight adaptive adjustment unit and the sending execution unit. It receives the comprehensive entropy value transmitted by the entropy weight adaptive adjustment unit, obtains the preset priority of each service stream, calculates the entropy weight allocation ratio of each service stream, transmits the bandwidth allocation instruction to the sending execution unit, and feeds back the bandwidth allocation ratio to the entropy weight adaptive adjustment unit to collaboratively optimize the sending parameters.

[0092] Sending execution unit: Deployed in the kernel layer of the sending end, it is integrated with the sending buffer and ACK receiving unit. This unit is connected to the entropy weight adaptive adjustment unit and the multi-stream entropy weight scheduling unit respectively. It receives the adjustment instructions and bandwidth allocation instructions transmitted by the two, executes message sending, rate control and retransmission operations, and feeds back the actual sending status (such as sending rate and number of retransmissions) to the entropy weight adaptive adjustment unit to form a closed-loop control.

[0093] Detailed functions of each unit: Link feature acquisition unit: The collected parameters are: delay difference sequence D (the RTT difference between two adjacent ACK packets) within a fixed time window T, packet loss location sequence L (the position index of the lost packet in the transmission sequence), and packet interval sequence I (the time interval between two adjacent transmitted packets); among them, delay difference sequence D is used to characterize the fluctuation of link delay, packet loss location sequence L is used to characterize the randomness of packet loss, and packet interval sequence I is used to assist in correcting the accuracy of entropy value calculation.

[0094] Acquisition frequency: The default acquisition frequency is 100Hz (i.e., data is acquired once every 10ms), which can be dynamically adjusted according to the instructions of the link feature entropy calculation unit (adjustment range is 50Hz-200Hz); the time window T can be configured according to the actual scenario, with a default value of 200ms and a configuration range of 100ms-500ms.

[0095] Data preprocessing: The collected sequence data is denoised (outliers are removed; outliers are defined as data that deviate from the sequence mean by more than three standard deviations). Missing data is supplemented using linear interpolation to ensure the accuracy of feature parameters.

[0096] Link feature entropy calculation unit: Core computational logic: Based on the entropy calculation principle in information theory, the link comprehensive feature entropy H is obtained by calculating the delay difference entropy H1 and the packet loss location entropy H2, combined with the weighting coefficient α. The entropy value ranges from [0,1]. The smaller the entropy value, the more stable the link state and the lower the random uncertainty. The larger the entropy value, the more unstable the link state and the higher the random uncertainty.

[0097] Detailed calculation steps: (1) Normalize the time delay difference sequence D by normalizing each element in D to the interval [0,1]. The normalization formula is as follows: ,in Let be the i-th element in the time delay difference sequence D, min(D) be the minimum value of sequence D, and max(D) be the maximum value of sequence D.

[0098] (2) For the normalized time delay difference sequence Perform probability distribution statistics and calculate the probability of each value. Thus, the probability distribution of the time delay difference is obtained.

[0099] (3) Calculate the time delay difference entropy The calculation formula is: i takes values ​​from 1 to n, where n is the length of sequence D; when When, define

[0100] (4) Perform location probability statistics on the packet loss location sequence L, divide the transmission sequence within the time window T into m equal intervals (m defaults to 10), and calculate the probability of packet loss in each interval. We obtain the probability distribution of packet loss locations (j from 1 to m).

[0101] (5) Calculate the entropy of the packet loss location The calculation formula is: 1 to m); when When, define

[0102] (6) Calculate the comprehensive feature entropy H. The calculation formula is as follows: ;in This is the weighting coefficient, with a default value of 0.6 (which can be configured according to the actual scenario, with a configuration range of 0.5-0.7). The larger the value of α, the higher the weight of the impact of latency fluctuations on the link status.

[0103] Entropy calibration: The overall entropy value H is calibrated based on the message interval sequence I. The calibration formula is as follows: Where σ(I) is the standard deviation of the message interval sequence I, and k is the calibration coefficient (default value is 0.1); the purpose of calibration is to eliminate the influence of message transmission interval fluctuations on entropy calculation and improve the accuracy of entropy representation.

[0104] Entropy weight adaptive adjustment unit: Entropy range division: the calibrated comprehensive entropy value The system is divided into three intervals, each corresponding to a different link status and adjustment strategy, as detailed below: (1) Interval 1: ∈ [0, 0.3), corresponding to a "stable link" status (no significant fluctuations, random packet loss rate <3%, latency jitter <20ms); (2) Interval 2: ∈ [0.3, 0.7), corresponding to the link status of "weak fluctuation link" (slight fluctuation, random packet loss rate of 3%-8%, and latency jitter of 20ms-80ms); (3) Interval 3: ∈ [0.7, 1.0], corresponding to the link state of "strong random link" (severe fluctuations, random packet loss rate >8%, latency jitter >80ms).

[0105] Sending parameter adjustment strategy: (1) Interval 1 (stable link): The sending window increases by 1.25 times the time window, and the sending rate increases by 10% of the link bandwidth limit per time window until the sending rate reaches the link bandwidth limit; the retransmission strategy adopts fast retransmission (retransmission is triggered when 3 duplicate ACKs are received), and the timeout retransmission time is set to 1.5 times the smooth RTT to reduce retransmission latency. (2) Interval 2 (weak fluctuation link): The sending window remains unchanged at its current size, and the sending rate adopts the pacing smoothing strategy (rate fluctuation is controlled within 5%). The retransmission strategy adopts a combination of "fast retransmission + selective retransmission", only retransmitting lost packets to avoid unnecessary retransmissions. At the same time, the timeout retransmission time is set to twice the smoothing RTT to balance the timeliness and accuracy of retransmission.

[0106] (3) Interval 3 (Strong random link): The sending window is reduced to 70% of the current size, the sending rate is reduced to 60% of the current rate, and the rate anti-jitter mechanism is enabled (rate fluctuation is controlled within 3%). The retransmission strategy adopts a combination of "fast retransmission + redundant retransmission", and adds one redundant retransmission for high priority packets to reduce the probability of packet loss. The timeout retransmission time is set to 2.5 times the smooth RTT to avoid frequent false triggering of retransmission.

[0107] Closed-loop feedback adjustment: After each time window T, the weight coefficient α and the entropy range threshold are fine-tuned based on the actual transmission status (transmission rate, number of retransmissions, packet loss rate) fed back by the transmission execution unit (the fine-tuning range does not exceed 10%) to ensure that the adjustment strategy matches the link status in real time.

[0108] Multi-stream entropy weight scheduling unit: Service flow priority division: Multiple service flows are divided into 3 priority levels (configurable), with priorities from high to low as follows: high priority (real-time control, high-definition live streaming, emergency data transmission), medium priority (normal video, voice calls), and low priority (file download, data synchronization); each priority corresponds to a different minimum bandwidth guarantee value (high priority minimum bandwidth guarantee value ≥ 50% of link bandwidth, medium priority ≥ 30%, low priority ≥ 20%).

[0109] Entropy-weighted bandwidth allocation calculation: based on the comprehensive characteristic entropy of the link. Based on this, the bandwidth allocation weight for each service flow is calculated using the following formula: ,in Assign a weight to the bandwidth of the i-th service flow. The priority coefficient of the i-th service flow (high priority) =0.8, medium priority =0.5, low priority =0.2), This is the bandwidth requirement coefficient for the i-th service flow (set according to the preset bandwidth requirement of the service flow, ranging from 0.1 to 1.0). This is the sum of the priority coefficients for all business flows.

[0110] Scheduling logic: Weight allocation based on bandwidth Allocate link bandwidth to each service flow, prioritizing the minimum bandwidth requirements of each service flow; when there is remaining link bandwidth, allocate the remaining bandwidth to each service flow according to the weight ratio; when link bandwidth is insufficient, prioritize reducing the bandwidth of low-priority service flows to ensure the transmission quality of high-priority service flows.

[0111] Send execution unit: Message transmission: Based on the transmission window and transmission rate instructions of the entropy weight adaptive adjustment unit, and the bandwidth allocation instructions of the multi-stream entropy weight scheduling unit, the corresponding service flow messages are extracted from the transmission buffer and sent to the intermediate routing nodes in sequence; at the same time, the transmission timestamp and transmission sequence number of each message are recorded.

[0112] Rate control: Through the rate control interface at the kernel layer, the message sending rate is adjusted in real time to ensure that rate fluctuations meet the requirements of the entropy weight adaptive adjustment unit; when the link bandwidth changes abruptly, the adjustment command is responded to quickly to avoid excessive rate fluctuations.

[0113] Retransmission execution: According to the retransmission strategy instructions of the entropy weight adaptive adjustment unit, receive the ACK message fed back by the ACK receiving unit, identify the lost message, and execute the corresponding retransmission operation; at the same time, record the number of retransmissions and the retransmission time, and feed them back to the entropy weight adaptive adjustment unit for fine-tuning the adjustment strategy.

[0114] Example 3 This invention provides a dynamic adaptive transmission device for weak networks based on dual-dimensional link entropy, such as... Figure 3 As shown, the device includes: The preprocessing module 100 is used to start the transmitting end system, configure core parameters, collect multi-dimensional link characteristic parameters within a fixed time window, and perform preprocessing. The calculation module 200 is used to normalize the delay difference sequence based on the preprocessed feature sequence, calculate the delay difference entropy and packet loss location entropy, and calculate the link comprehensive feature entropy and perform calibration in combination with the weight coefficient. The adjustment module 300 is used to divide the calibrated link comprehensive characteristic entropy into different numerical ranges, generate corresponding adjustment instructions, and adjust the sending parameters of each service flow according to the bandwidth allocation ratio. The allocation module 400 is used to obtain the preset priority of each service flow, calculate the bandwidth allocation weight of each service flow and allocate the link bandwidth based on the calibrated link comprehensive feature entropy and the transmission parameters of each service flow. The judgment module 500 is used to extract the packets of each service flow and send them to the intermediate routing node. It adjusts the packet sending rate in real time through the rate control interface, fine-tunes the parameters based on the actual sending status feedback, and finally determines whether the transmission task is completed.

[0115] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0116] Example 4 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0117] Example 5 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0118] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0119] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0120] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A dynamic adaptive transmission method for weak networks based on dual-dimensional link entropy, characterized in that, include: Start the sending system and configure the core parameters, collect multi-dimensional link characteristic parameters within a fixed time window and perform preprocessing; Based on the preprocessed feature sequence, the delay difference sequence is normalized, the delay difference entropy and packet loss location entropy are calculated, and the link comprehensive feature entropy is calculated and calibrated in combination with the weight coefficient. The calibrated link comprehensive characteristic entropy is divided into different numerical ranges, corresponding adjustment instructions are generated, and the sending parameters of each service flow are adjusted according to the bandwidth allocation ratio. Obtain the preset priority of each service flow, and calculate the bandwidth allocation weight of each service flow and allocate the link bandwidth based on the calibrated link comprehensive feature entropy and the transmission parameters of each service flow. The system extracts packets from each service flow and sends them to intermediate routing nodes. It adjusts the packet sending rate in real time through the rate control interface, fine-tunes parameters based on the actual sending status feedback, and finally determines whether the transmission task is complete.

2. The method according to claim 1, characterized in that, The process of starting the transmitting system and configuring core parameters, collecting multi-dimensional link characteristic parameters within a fixed time window and performing preprocessing includes: Start the sending system and configure the following core parameters: set the time window T and weight coefficient. Calibration coefficient k, number of packet loss location intervals m, acquisition frequency, priority of each service flow and minimum bandwidth guarantee value, initial value of sending window W, initial value of sending rate V; Initialize the transmit buffer and ACK receive buffer, and clear the historical data and entropy calculation results; The control link feature acquisition unit collects the delay difference sequence D, packet loss location sequence L, and message interval sequence I within the current time window T at a preset acquisition frequency. Among them, the delay difference sequence D is used to characterize the fluctuation of link delay, the packet loss location sequence L is used to characterize the randomness of packet loss, and the message interval sequence I is used to assist in correcting the accuracy of entropy value calculation. The collected sequence data is denoised, and missing data is supplemented using linear interpolation to obtain the preprocessed feature sequence.

3. The method according to claim 1, characterized in that, The process involves normalizing the delay difference sequence based on the preprocessed feature sequence, calculating the delay difference entropy and packet loss location entropy, and then, in conjunction with weighting coefficients, calculating and calibrating the link comprehensive feature entropy. This includes: After receiving the preprocessed feature sequence through the link feature entropy calculation unit, the delay difference sequence D is normalized, and each element in D is normalized to the interval [0,1]. The normalization formula is as follows: in, Let i be the i-th element in the time delay difference sequence D. Let be the minimum value of sequence D. The maximum value of sequence D; Perform probability distribution statistics on the normalized time delay difference sequence, calculate the probability of each value, and obtain the time delay difference probability distribution: Calculate the probability distribution of time delay difference and the entropy of time delay difference. : Where i ranges from 1 to n, and n is the length of sequence D, when When, define ; Perform location probability statistics on the packet loss location sequence L, divide the transmission sequence within the time window T into m equal intervals, and calculate the probability of packet loss in each interval to obtain the packet loss location probability distribution: Calculate the entropy of packet loss location : Among them, when When, define ; The formula for calculating the comprehensive feature entropy H is: in, These are the weighting coefficients. The larger the value of , the higher the weight of the impact of latency fluctuations on the link status; Based on the message interval sequence I, the overall entropy value H is calibrated using the following formula: in, For message interval sequence The standard deviation of , where k is the calibration coefficient.

4. The method according to claim 1, characterized in that, The process of dividing the calibrated link comprehensive characteristic entropy into different numerical ranges, generating corresponding adjustment instructions, and adjusting the transmission parameters of each service flow according to the bandwidth allocation ratio includes: The calibrated comprehensive entropy value is received through the entropy weight adaptive adjustment unit. Based on a preset entropy range, the current link status is determined, and a corresponding transmission command is generated. like If ∈ [0, 0.3), then the corresponding link state is determined to be a stable link. At this time, the link has no obvious fluctuations, the random packet loss rate is <3%, and the latency jitter is <20ms. The sending window increases by 1.25 times the time window, and the sending rate increases by 10% of the link bandwidth limit per time window until the sending rate reaches the link bandwidth limit. The retransmission strategy adopts fast retransmission, and the timeout retransmission time is set to 1.5 times the smooth RTT. like If ∈ [0.3, 0.7), then the corresponding link status is determined to be a weakly fluctuating link. At this time, the link fluctuates slightly, with a random packet loss rate of 3%-8% and a latency jitter of 20ms-80ms. The sending window remains unchanged at its current size, and the sending rate adopts a pacing smoothing strategy. The retransmission strategy adopts a combination of fast retransmission and selective retransmission, only retransmitting lost packets, and setting the timeout retransmission time to twice the smoothing RTT. like If ∈ [0.7, 1.0], the corresponding link state is determined to be a strongly random link. At this time, the link fluctuates violently, with a random packet loss rate >8% and latency jitter >80ms. The sending window is reduced to 70% of the current size, the sending rate is reduced to 60% of the current rate, and the rate anti-jitter mechanism is enabled. The retransmission strategy adopts a combination of fast retransmission and redundant retransmission. High-priority packets are retransmitted once more, and the timeout retransmission time is set to 2.5 times the smoothed RTT. It receives the bandwidth allocation ratio from the multi-stream entropy weight scheduling unit and coordinates the adjustment of the transmission parameters of each service stream.

5. The method according to claim 1, characterized in that, The process of obtaining the preset priority of each service flow, calculating the bandwidth allocation weight of each service flow and allocating link bandwidth based on the calibrated link comprehensive feature entropy and the transmission parameters of each service flow, includes: The comprehensive entropy value is received through the multi-stream entropy weight scheduling unit. Obtain the priority and bandwidth requirements of all currently active service flows, and calculate the bandwidth allocation weight for each service flow. The system generates a bandwidth allocation instruction, transmits it to the sending execution unit, and feeds back the bandwidth allocation ratio to the entropy weight adaptive adjustment unit. The formula for calculating the bandwidth allocation weights for each service flow is as follows: in, Assign a weight to the bandwidth of the i-th service flow. Let be the priority coefficient for the i-th business flow. Let be the bandwidth requirement coefficient for the i-th service flow. This is the sum of the priority coefficients for all business flows.

6. The method according to claim 1, characterized in that, The process of extracting packets from each service flow and sending them to intermediate routing nodes, adjusting the packet sending rate in real time through a rate control interface, fine-tuning parameters based on the actual sending status feedback, and finally determining whether the transmission task is complete includes: The sending execution unit receives sending parameter adjustment instructions and bandwidth allocation instructions, extracts messages from each service flow from the sending buffer, sends messages according to the adjusted sending window and sending rate, and simultaneously receives ACK messages from the ACK receiving unit, identifies lost messages, and performs retransmission operations according to the adjusted retransmission strategy. The control and transmission execution unit feeds back the actual transmission status of this iteration to the entropy weight adaptive adjustment unit, and the link feature acquisition unit feeds back the feature sequence acquired in this iteration to the link feature entropy calculation unit. The entropy-weighted adaptive adjustment unit fine-tunes the weighting coefficients based on the actual transmission status feedback. Based on the entropy value range threshold, the link feature entropy calculation unit adjusts the acquisition frequency according to the feedback feature sequence; Determine if the transmission task is complete. If not, proceed to the next time window T and repeat the above steps. If complete, terminate the process and shut down the system unit.

7. A dynamic adaptive transmission device for weak networks based on dual-dimensional link entropy, characterized in that, include: The preprocessing module is used to start the sending system and configure core parameters, collect multi-dimensional link characteristic parameters within a fixed time window and perform preprocessing. The calculation module is used to normalize the delay difference sequence based on the preprocessed feature sequence, calculate the delay difference entropy and packet loss location entropy, and calculate the link comprehensive feature entropy and perform calibration in combination with the weight coefficient. The adjustment module is used to divide the calibrated link comprehensive characteristic entropy into different numerical ranges, generate corresponding adjustment instructions, and adjust the sending parameters of each service flow according to the bandwidth allocation ratio. The allocation module is used to obtain the preset priority of each service flow, calculate the bandwidth allocation weight of each service flow and allocate the link bandwidth based on the calibrated link comprehensive feature entropy and the transmission parameters of each service flow. The judgment module is used to extract the packets of each service flow and send them to the intermediate routing node. It adjusts the packet sending rate in real time through the rate control interface, fine-tunes the parameters based on the actual sending status feedback, and finally determines whether the transmission task is completed.

8. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images