Data transmission method and system for realizing low delay

Abstract

The invention discloses a low-delay data transmission method and system, and relates to the technical field of low-delay data transmission, and the method comprises the steps: analyzing the network state of a network communication protocol QUIC through employing a reinforcement learning model PPO, selecting an optimal path, and carrying out the data transmission according to a data transmission position requested by a user, according to a nearest neighbor principle, an optimal edge computing node is selected to process a user request, transmission data is compressed by using an adaptive Huffman coding technology through the characteristics of the optimal edge node, and when high delay and high packet loss rate occur in a data transmission process, the transmission data is compressed. And adjusting a transmission strategy by using a multi-dimensional action suggestion mechanism of a deep reinforcement learning model DDPG. According to the method, the data is ensured to be transmitted through the most appropriate path all the time, the data transmission delay can be greatly reduced while the data security is ensured, the transmission efficiency and the service response speed are improved, and the data are evolved continuously along with the change of the network environment and the user demand.

Description

Technical Field

[0001] The present invention relates to the technical field of reducing data transmission delay, and in particular to a method and system for achieving low-latency data transmission. Background Art

[0002] With the rapid development of information technology, network communication has become an indispensable part of modern society. From early analog signal transmission to modern digital communication technology, the speed and efficiency of data transmission have been greatly improved. Especially in recent years, with the development of cloud computing, big data analysis and emerging technologies of the Internet of Things (IoT), the demand for efficient and low-latency data transmission has been increasing. The network communication protocol QUIC is the core standard for realizing Internet data transmission. The network communication protocol QUIC performs well in ensuring the reliability and integrity of data transmission, making large-scale data transmission a reality.

[0003] Although there have been some attempts to solve the above problems, the network communication protocol QUIC still has some shortcomings. In terms of security, although traditional encryption algorithms and key exchange mechanisms can provide necessary protection for data, they add additional time overhead during actual deployment, especially in the handshake stage when the communication protocol establishes a connection, which undoubtedly exacerbates the overall delay problem. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for achieving low-latency data transmission to solve the problem of increased overall data transmission delay when a communication protocol establishes a connection.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for achieving low-latency data transmission, which comprises:

[0008] Use the network communication protocol QUIC and perform initial configuration;

[0009] Use the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path;

[0010] According to the data transmission location of the user request, the best edge computing node is selected based on the nearest neighbor principle to process the user request;

[0011] Adaptive Huffman coding technology is used to compress the transmitted data based on the characteristics of the optimal edge node;

[0012] Compressed transmission data is sent from the source to the destination according to the network communication protocol QUIC and the optimal path. When high latency and packet loss rate occur during data transmission, the multi-dimensional action suggestion mechanism of the deep reinforcement learning model DDPG is used to adjust the transmission strategy, and the link effectiveness is evaluated based on the deep reinforcement learning model DDPG feedback mechanism.

[0013] As a preferred solution of the method for realizing low-latency data transmission according to the present invention, the initial configuration is specifically as follows:

[0014] Use the network communication protocol QUIC, select the encryption algorithm AES-128-GCM and the key exchange mechanism X25519 to configure the encryption parameters in the QUIC protocol;

[0015] Adjust the QUIC handshake timeout based on the target network environment and use a congestion control algorithm.

[0016] Enable the connection multiplexing function of the network communication protocol QUIC, so that multiple HTTP requests share the same network communication protocol QUIC connection.

[0017] As a preferred solution for the method of realizing low-latency data transmission according to the present invention, the method of using the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path is specifically as follows:

[0018] Download Prometheus and Grafana to collect network information, and add logging capabilities to the receive buffer and send queue of the QUIC stack, a network communication protocol.

[0019] The bandwidth utilization is calculated based on the actual amount of data transmitted, the total available bandwidth, and the time interval. The packet loss rate is calculated by dividing the number of lost packets by the total number of sent packets. The comprehensive score S of each potential path is calculated based on the weight of the bandwidth utilization and the weight of the packet loss rate.

[0020] Create a PPO model instance and specify the multilayer perceptron (MLP) as the PPO model architecture. Input the comprehensive score S as the reward value into the PPO model. The PPO model outputs the expected reward value for each path and selects the path with the highest reward value as the optimal path.

[0021] As a preferred solution of the method for realizing low-latency data transmission according to the present invention, wherein: according to the data transmission location requested by the user, the best edge computing node is selected according to the nearest neighbor principle to process the user request, specifically:

[0022] Establish a database containing all edge computing nodes and their geographical locations, and update the database regularly;

[0023] When a user initiates a data transmission request, the user's geographic location is recorded, and a ping request is sent from the user device to all edge computing nodes to calculate the average round-trip time (RTT).

[0024] The nearest neighbor principle is used to select the edge node with the minimum distance value as the best edge node based on the distance between each edge computing node and the data transmission location.

[0025] As a preferred solution of the method for realizing low-latency data transmission according to the present invention, wherein: the transmission data is compressed using adaptive Huffman coding technology based on the characteristics of the optimal edge node, specifically:

[0026] Use APIs to collect characteristics of optimal edge computing nodes;

[0027] Initialize an adaptive Huffman encoder based on the characteristics of the optimal edge node and set the maximum resource consumption limit when the encoder is running;

[0028] Based on previous data transmission records, the frequency of characters in the transmitted data is collected and an initial character frequency table is created to construct a Huffman tree for data compression.

[0029] As a preferred solution of the method for achieving low-latency data transmission according to the present invention, wherein: the compressed transmission data is sent from the source end to the destination end according to the network communication protocol QUIC and the optimal path, specifically:

[0030] Assigning sequence numbers, checksums, marking path identifiers, and setting window sizes to the compressed transmission data to form data packets;

[0031] According to the network communication protocol QUIC and the optimal transmission path, the encapsulated data packet is sent from the source end to the destination end using the network communication protocol QUIC.

[0032] As a preferred solution of the method for realizing low-latency data transmission according to the present invention, the transmission strategy is adjusted as follows:

[0033] A multi-dimensional action suggestion mechanism based on the DRL model, setting the triggering conditions of the multi-dimensional action suggestion mechanism;

[0034] During data transmission, network delay and packet loss rate are collected in real time and input into the DDPG model. After triggering the DDPG multi-dimensional action suggestion mechanism of the deep reinforcement learning model, the transmission rate and retransmission threshold during data transmission are adjusted.

[0035] After adjusting the transmission rate and retransmission threshold, the link is evaluated and optimized based on the feedback of the adjustment results provided by the deep reinforcement learning model DDPG.

[0036] In a second aspect, the present invention provides a data transmission system for achieving low latency, comprising:

[0037] The initial configuration module uses the network communication protocol QUIC and performs initial configuration;

[0038] The path module uses the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path;

[0039] The processing module selects the best edge computing node to process the user request based on the data transmission location of the user request and the nearest neighbor principle;

[0040] The compression module compresses the transmitted data using adaptive Huffman coding technology based on the characteristics of the optimal edge node;

[0041] The adjustment module sends compressed transmission data from the source to the destination according to the network communication protocol QUIC and the optimal path. When high latency and packet loss rate occur during data transmission, the multi-dimensional action suggestion mechanism of the deep reinforcement learning model DDPG is used to adjust the transmission strategy, and the link effectiveness is evaluated based on the deep reinforcement learning model DDPG feedback mechanism.

[0042] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the method for achieving low-latency data transmission as described in the first aspect of the present invention.

[0043] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for achieving low-latency data transmission as described in the first aspect of the present invention.

[0044] The beneficial effects of the present invention are: through the reinforcement learning model PPO, the data transmission path can be automatically learned and optimized according to the real-time network conditions, and the adaptive Huffman coding technology is used to compress the transmission data, which not only improves the security and efficiency of data transmission, but also can dynamically adjust the transmission strategy according to the real-time network conditions to ensure that data is always transmitted through the most appropriate path. While ensuring data security, it can greatly reduce the delay of data transmission, improve transmission efficiency and service response speed, and continue to evolve with changes in network environment and user needs, continuously providing efficient and secure services, thereby improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 A flowchart of a method for achieving low-latency data transmission;

[0047] Figure 2 A schematic diagram of module interaction;

[0048] Figure 3 It is a schematic diagram of dynamically compressed data;

[0049] Figure 4 Schematic diagram of PPO path selection and network status. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0053] Reference Figures 1 to 4 This embodiment provides a method for achieving low-latency data transmission, comprising the following steps:

[0054] S1. Use the network communication protocol QUIC and perform initial configuration, specifically:

[0055] Log in to the server management interface, find the setting options of the network service software nginx, enable support for the network communication protocol QUIC, select AES-128-GCM as the default encryption algorithm in the encryption settings section of the server's configuration tool and management interface, and save the changes after completing the settings and restart the service to make the configuration take effect, ensuring that new connections use this AES-128-GCM encryption standard.

[0056] In the same configuration tool and management interface, set the key exchange algorithm to X25519 and use a QUIC handshake packet to find information related to key exchange, such as the cipher suite and algorithm identifiers and public key parameters, to verify that the new connection attempts to use the X25519 mechanism for key exchange. By enabling the network communication protocol QUIC protocol and selecting AES-128-GCM as the encryption algorithm and X25519 as the key exchange algorithm, the security of data transmission can be ensured. The network communication protocol QUIC itself is designed to be a more secure protocol and uses encryption by default, making man-in-the-middle attacks more difficult.

[0057] Use the ping command in the Windows CMD program to measure network latency and obtain the average round-trip time (RTT) of the network;

[0058] Check the configuration to confirm that the handshake timeout is correctly set to 4 times the RTT. Observe whether the actual connection establishment process meets expectations. A reasonable handshake timeout ensures that the connection can be successfully established even in network fluctuations.

[0059] According to the network communication protocol QUIC configuration interface, BBR is specified as the congestion control algorithm, which can effectively respond to network changes, optimize bandwidth utilization, and improve data transmission efficiency.

[0060] Configure HTTP / 3 support on the server side, allowing multiple HTTP requests to share the same QUIC connection. Use an HTTP client to send connection requests to the server, and use the Wireshark network traffic analysis tool to check whether multiple HTTP requests share the same QUIC connection to perform initial configuration. These configurations can adapt to different network environments and user needs, providing support and assurance for both highly secure enterprise applications and performance-demanding online services. They also lay the foundation for the early stages of data transmission, protect privacy and integrity during data transmission, and reduce the risk of man-in-the-middle attacks.

[0061] S2. Use the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path, specifically:

[0062] Visit the official website of the monitoring and alarm tool Prometheus, find the download page and select the Prometheus software package to download. After the download is complete, you will have a compressed file containing the Prometheus components. Select a directory on the server to store Prometheus, use the decompression tool to decompress the previously downloaded compressed file to the selected directory, open the Prometheus installation directory after decompression, find the configuration file and edit this file to add the network communication protocol QUIC address and port to be monitored. After the configuration is complete, Prometheus can capture the connection mode and layout between each node (such as computers, routers, switches) in the data transmission network according to the set goals. The ratio of lost data packets to the total number of sent data packets (packet loss rate) and the ratio of actually used bandwidth to available bandwidth (bandwidth utilization) are displayed. To visually view the network status information collected by Prometheus, you also need to download a Grafana (Grafana, often used to display time series data), select Prometheus as the type, fill in the URL address of the Prometheus instance, combine Prometheus and Grafana into a tool combination, create a custom dashboard, and select different visualization options for each custom dashboard, such as charts and tables, so that Grafana can display the network status information collected by Prometheus in real time. By using the combination of Prometheus and Grafana, network status information of the network communication protocol QUIC can be collected and displayed in real time, which provides network administrators with an intuitive data view, making it easier to discover and solve problems in a timely manner.

[0063] Modify the QUIC network communication protocol stack, add logging to the receive buffer and send queue, and regularly report the recorded network topology, bandwidth usage, and packet loss rate to Prometheus. Use Grafana to create a dashboard to display the network status information collected from Prometheus. Based on the collected network status information, calculate bandwidth utilization and packet loss rate. Specifically, calculate bandwidth utilization based on the actual amount of data transmitted, total available bandwidth, and time interval, and calculate packet loss rate by dividing the number of lost packets by the total number of packets sent. The collected bandwidth utilization and packet loss rate are used to calculate a comprehensive score for each potential path, expressed as:

[0064] S = w1·U-w2·L;

[0065] Where S represents the comprehensive score, w1 represents the bandwidth utilization weight, U represents the bandwidth utilization, w2 represents the packet loss rate weight, and L represents the packet loss rate.

[0066] The weighting can be adjusted based on specific application scenarios and preferences. For example, if you care more about transmission speed than data loss, you can increase the weight of bandwidth utilization and decrease the weight of packet loss rate. Conversely, if data integrity is more important, you can increase the weight of packet loss rate to emphasize the importance of low packet loss rate.

[0067] Use the StableBaselines3 framework to implement proximal policy optimization (PPO). Download the library and dependencies according to the official documentation of the selected StableBaselines3 framework, and configure the reinforcement learning environment, network state information, action space (that is, actions that can be taken, such as choosing different paths), and reward mechanism (based on the comprehensive score S calculated previously). The reinforcement learning environment can receive state input, perform actions, and return corresponding reward values.

[0068] Set model parameters. The learning rate determines the speed of model updates. Generally, start with a small value, such as 0.0003. The discount factor determines the importance of future rewards and is usually set to a value close to 1, such as 0.99. The reward decay coefficient, used in the generalized advantage estimate (GAE), helps improve learning efficiency. A common value is 0.95. Adjust these hyperparameters based on the specific data transmission scenario for optimal results.

[0069] Use the API provided by the selected StableBaselines3 framework and follow the above hyperparameter design method to create a new PPO model instance. Here, you need to specify the use of the multi-layer perceptron MLP and the reinforcement learning environment configured above.

[0070] The previously calculated comprehensive score S is used as the reward value. Higher comprehensive scores result in higher rewards, meaning that better network path performance corresponds to higher rewards. For each path, the comprehensive score S is input into the PPO model for evaluation. The PPO model outputs the expected reward value for each path and selects the path with the highest reward value as the optimal path.

[0071] In order to enable the PPO model to continuously adjust according to the latest network status information, a regular learning cycle is set, such as retraining the model every hour, to ensure that the PPO model adapts to the latest changes in the network environment. This mechanism ensures that the transmission path can continue to provide efficient and stable services in a dynamically changing network environment.

[0072] S3. Based on the data transmission location of the user request, the optimal edge computing node is selected according to the nearest neighbor principle to process the user request. Specifically:

[0073] Collect the IP addresses, longitude, and latitude coordinates of all edge computing nodes and create a relational database (RDBMS) to store their geographic locations. When creating the RDBMS structure, ensure that there are at least two fields for storing longitude and latitude coordinates. Use the SQL Server batch import tool to import all collected edge computing nodes into the RDBMS. Establish a regular update strategy, such as weekly checks on the status of all edge computing nodes. Newly added edge computing nodes should be added to the database using the SQL Server batch import tool as described above. Unused edge computing nodes should be deleted from the database.

[0074] Identify all available edge computing nodes and ensure they are online and accessible. When a user initiates a data transmission request, record the user's location information, namely the longitude and latitude coordinates. For each edge computing node, send a ping request from the user device to all edge computing nodes. Ping returns the time from sending the request to receiving the response. Send multiple ping requests for more accurate results; 3 to 5 ping requests are sufficient. Then take the average as the final RTT value.

[0075] The ping test results of all edge computing nodes are summarized to form a table containing the name, IP address, and corresponding RTT value of each edge computing node. The RTT value of each edge computing node is compared one by one, and the edge computing node with the lowest RTT value is found as the optimal edge computing node for processing user data transmission requests. The user's data transmission request is then redirected to this edge computing node. This dynamic selection of the closest and fastest-responding edge computing node to process requests based on the user's actual location not only optimizes the allocation of network resources, but also avoids unnecessary long-distance data transmission, reduces the burden on the backbone network, and thus improves overall network efficiency.

[0076] S4. Adaptive Huffman coding technology is used to compress the transmitted data based on the characteristics of the optimal edge node. Specifically:

[0077] Based on the optimal edge computing node, the API is used to collect relevant characteristics of the edge computing node, namely processing power (number of CPU cores and frequency), current load level (such as CPU utilization, memory occupancy), and response time (the time from receiving a request to starting processing).

[0078] Based on the characteristics of the optimal edge node, an adaptive Huffman encoder is initialized. Specifically, the maximum resource consumption limit of the encoder is determined based on the processing power, current load level, and response time of the optimal edge node. If the edge computing node is responsible for multiple tasks simultaneously, the data compression task should be prioritized to ensure that the data transmission task is not affected by the encoding task. Review past data transmission records, especially historical data similar to the type of data to be processed, to collect information about the frequency of character occurrence and create an initial character frequency table. Based on the current character frequency table, a Huffman tree is constructed. During this process, frequently occurring characters are assigned shorter codes, while uncommon characters are assigned longer codes for data compression. Throughout the compression process, the load of the edge computing node and the encoding progress are continuously monitored.

[0079] Perform decoding tests on the compressed data to ensure lossless restoration. At the same time, compare the file sizes before and after compression to confirm that the expected compression effect has been achieved.

[0080] This data compression solution takes into account the processing power and current load level of edge nodes, sets maximum resource consumption limits, and prioritizes data transmission tasks. This not only ensures that the data compression process does not excessively consume system resources and affect other critical tasks, but also ensures that data compression and transmission tasks can be completed efficiently even under high load conditions.

[0081] S5. Send the compressed transmission data from the source to the destination according to the network communication protocol QUIC and the optimal path. Specifically:

[0082] After completing the optimized compression of the transmitted data, a series of data blocks are obtained, each of which represents a data packet to be transmitted. A unique serial number is assigned to each data packet. For example, if the current packet is the fifth, the serial number should be set to 5. The checksum is generated using the AES-GCM encryption method of the network communication protocol QUIC. When the network communication protocol QUIC supports multi-path data transmission, a path identifier needs to be specified for each data packet to indicate which path the data packet should be transmitted through. For example, if the optimal path identifier selected is 2, PathID = 2 is added to the data packet header. The initial window size is set to 65535 bytes according to the requirements of the network communication protocol QUIC, and this value is filled in the corresponding field in the header.

[0083] After encapsulation is completed, a comprehensive check is performed to ensure that all sequence numbers, checksums, path identifiers, and window sizes have been added correctly, and that the packet size complies with the limits specified by the network communication protocol QUIC, that is, the maximum size of a single packet usually does not exceed 1200 bytes.

[0084] According to the predetermined optimal path, the function of the network communication protocol QUIC is used to arrange the order and timing of data packet sending. When the network conditions are good, the data packets are sent directly in sequence. If slight congestion is detected, the slow start mechanism is used to gradually increase the sending rate.

[0085] This transmission solution takes into account different network conditions and provides flexible adjustment mechanisms, making data transmission more adaptable to various complex network environments. Whether in situations with high latency, high packet loss rates, or limited bandwidth, intelligent scheduling and dynamic adjustment strategies ensure smooth and efficient data transmission.

[0086] S6. Use the multi-dimensional action recommendations of the deep reinforcement learning model DDPG to adjust the transmission strategy and optimize the link. Specifically:

[0087] A deep reinforcement learning (DDPG) model is constructed with an actor network and a critic network. The actor network determines the actions to be taken (i.e., the transmission strategy) based on the packet loss rate, round-trip time (RTT), and bandwidth utilization during data transmission. The critic network provides feedback based on the actions taken and the results. The packet loss rate, round-trip time (RTT), and bandwidth utilization are used as the training data set and fed into the deep reinforcement learning model for a certain number of training rounds (e.g., 50 rounds).

[0088] During the data transmission process, the proportion of lost data packets to the total number of sent data packets (packet loss rate) and the time required to transmit data from the source to the destination (round-trip time RTT) are continuously collected, as well as the ratio of the actually used bandwidth to the available bandwidth (bandwidth utilization). When high latency and high packet loss rate occur during data transmission, the packet loss rate, round-trip time RTT and bandwidth utilization are input into the pre-trained deep reinforcement learning model DDPG. The deep reinforcement learning model DDPG predicts the transmission strategy (retransmission threshold and transmission rate) for the current data transmission stage. In addition, a trigger condition can be set. For example, when the network delay exceeds 100ms and the packet loss rate is higher than a certain percentage (for example, 1%), the multi-dimensional action recommendation mechanism of the deep reinforcement learning model DDPG is triggered to adjust the retransmission threshold and transmission rate.

[0089] When the packet loss rate is greater than 1%, the deep reinforcement learning model DDPG recommends reducing the transmission rate to cope with the high packet loss rate. The transmission rate parameter in the network communication protocol QUIC is adjusted, which is expressed as:

[0090] R = V × (1-α·P);

[0091] Where R represents the dynamically adjusted transmission rate parameter, V represents the current transmission rate, α represents the adjustment coefficient of the current packet loss rate, and P represents the current packet loss rate.

[0092] When the network delay during data transmission exceeds, for example, 100ms, the deep reinforcement learning model DDPG recommends increasing the retransmission threshold to reduce the number of retransmissions, expressed as:

[0093] M = K + β·RTT;

[0094] Where M represents the dynamically adjusted retransmission threshold, K represents the current retransmission threshold, β represents the RTT adjustment coefficient, and RTT represents the current round-trip time.

[0095] The retransmission threshold is typically set between 1 and 5. Lower retransmission thresholds (such as 1 and 2) are suitable for latency-sensitive applications. In such cases, high latency can lead to data transmission failures, and quickly trying other paths is more important than waiting for retransmissions. Higher retransmission thresholds (such as 4 and 5) are more suitable for scenarios with high data integrity requirements and complex network conditions.

[0096] Compare the network status information before and after the adjustment to evaluate the effectiveness of the adjustment. For example, check whether the packet loss rate has decreased and the transmission rate has become more stable.

[0097] This method of adjusting transmission strategies emphasizes a continuous monitoring and optimization mechanism that is not limited to one-time adjustments, but also involves regular evaluation and adjustment of network performance, ensuring long-term stable low-latency data transmission capabilities.

[0098] After adjusting the transmission strategy, continue to use the Prometheus tool to count the total amount of data successfully transmitted from the source to the destination, ensuring that all successfully transmitted data is included and ignoring any retransmitted data. For each data item, record the send time and reception confirmation time. Calculate the actual transmission delay for each data item by subtracting the send time from the time the data was confirmed received. Format the collected successfully transmitted data volume and actual transmission delay for easy understanding and analysis. For example, the delay time of each data packet can be listed in chronological order and the transmitted data volume can be summarized. The total transmitted data volume, average delay time, maximum and minimum delay times are fed back to the deep reinforcement learning model (DDPG) to evaluate the effectiveness of the adjustment. Based on the received feedback report, the deep reinforcement learning model (DDPG) first evaluates the effectiveness of the existing link status and decision logic, and checks whether the low latency and high throughput meet the expected performance goals.

[0099] Key indicators for evaluating link quality are set. For example, if the average delay of a link exceeds 84ms and the packet loss rate is higher than 0.9%, it is considered that there is a problem with the link. When it is found that some links perform significantly below expectations and the average delay is higher than other links, these links are marked as objects that need to be adjusted. All available backup paths are obtained through the routing protocol, each available backup path is checked, and paths with low latency and high stability are selected as potential replacement options. In addition, the delay can be minimized by increasing bandwidth allocation, that is, contacting the network service provider and the internal IT department to request an increase in the bandwidth of the specified link. Combining the above methods, the present invention can significantly reduce the delay of data transmission while ensuring data security, and improve transmission efficiency and service response speed. As the network environment and user needs change, the system can continue to evolve and continue to provide efficient and secure services, thereby improving the user experience.

[0100] This embodiment further provides a low-latency data transmission system, including:

[0101] The initial configuration module uses the network communication protocol QUIC and performs initial configuration;

[0102] The path module uses the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path;

[0103] The processing module selects the best edge computing node to process the user request based on the data transmission location of the user request and the nearest neighbor principle;

[0104] The compression module compresses the transmitted data using adaptive Huffman coding technology based on the characteristics of the optimal edge node;

[0105] The adjustment module sends compressed transmission data from the source to the destination according to the network communication protocol QUIC and the optimal path. When high latency and packet loss rate occur during data transmission, the multi-dimensional action suggestion mechanism of the deep reinforcement learning model DDPG is used to adjust the transmission strategy, and the link effectiveness is evaluated based on the deep reinforcement learning model DDPG feedback mechanism.

[0106] This embodiment also provides a computer device suitable for implementing a low-latency data transmission method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the low-latency data transmission method proposed in the above embodiment.

[0107] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0108] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the low-latency data transmission method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0109] In summary, the present invention uses the reinforcement learning model PPO to enable the data transmission path to automatically learn and optimize according to the real-time network conditions, and uses adaptive Huffman coding technology to compress the transmitted data, which not only improves the security and efficiency of data transmission, but also can dynamically adjust the transmission strategy according to the real-time network conditions to ensure that data is always transmitted through the most appropriate path. It can greatly reduce the delay of data transmission while ensuring data security, improve transmission efficiency and service response speed, and continuously evolve with changes in network environment and user needs, continuously providing efficient and secure services, thereby improving user experience.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for achieving low-latency data transmission, characterized in that: include, Use the network communication protocol QUIC and perform initial configuration; Use the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path; According to the data transmission location of the user request, the best edge computing node is selected based on the nearest neighbor principle to process the user request; Adaptive Huffman coding technology is used to compress the transmitted data based on the characteristics of the optimal edge node; Compressed transmission data is sent from the source to the destination according to the network communication protocol QUIC and the optimal path. When high latency and packet loss rate occur during data transmission, the multi-dimensional action suggestion mechanism of the deep reinforcement learning model DDPG is used to adjust the transmission strategy, and the link effectiveness is evaluated based on the deep reinforcement learning model DDPG feedback mechanism.

2. The method for achieving low-latency data transmission according to claim 1, wherein: The initial configuration is as follows: Use the network communication protocol QUIC, select the encryption algorithm AES-128-GCM and the key exchange mechanism X25519 to configure the encryption parameters in the QUIC protocol; Adjust the QUIC handshake timeout based on the target network environment and use a congestion control algorithm. Enable the connection multiplexing function of the network communication protocol QUIC, so that multiple HTTP requests share the same network communication protocol QUIC connection.

3. The method for achieving low-latency data transmission according to claim 2, wherein: The reinforcement learning model PPO is used to analyze the network status of the network communication protocol QUIC and select the optimal path, specifically: Download Prometheus and Grafana to collect network information, and add logging capabilities to the receive buffer and send queue of the QUIC stack, a network communication protocol. The bandwidth utilization is calculated based on the actual amount of data transmitted, the total available bandwidth, and the time interval. The packet loss rate is calculated by dividing the number of lost packets by the total number of sent packets. The comprehensive score S of each potential path is calculated based on the weight of the bandwidth utilization and the weight of the packet loss rate. Create a PPO model instance and specify the multilayer perceptron (MLP) as the PPO model architecture. Input the comprehensive score S as the reward value into the PPO model. The PPO model outputs the expected reward value for each path and selects the path with the highest reward value as the optimal path.

4. The method for achieving low-latency data transmission according to claim 3, wherein: According to the data transmission location of the user request, the best edge computing node is selected according to the nearest neighbor principle to process the user request, specifically: Establish a database containing all edge computing nodes and their geographical locations, and update the database regularly; When a user initiates a data transmission request, the user's geographic location is recorded, and a ping request is sent from the user device to all edge computing nodes to calculate the average round-trip time (RTT). The nearest neighbor principle is used to select the edge node with the minimum distance value as the best edge node based on the distance between each edge computing node and the data transmission location.

5. The method for achieving low-latency data transmission according to claim 4, wherein: The adaptive Huffman coding technology is used to compress the transmitted data based on the characteristics of the optimal edge node, specifically: Use APIs to collect characteristics of optimal edge computing nodes; Initialize an adaptive Huffman encoder based on the characteristics of the optimal edge node and set the maximum resource consumption limit when the encoder is running; Based on previous data transmission records, the frequency of characters in the transmitted data is collected and an initial character frequency table is created to construct a Huffman tree for data compression.

6. The method for achieving low-latency data transmission according to claim 5, wherein: The compressed transmission data is sent from the source end to the destination end according to the network communication protocol QUIC and the optimal path, specifically: Assigning sequence numbers, checksums, marking path identifiers, and setting window sizes to the compressed transmission data to form data packets; According to the network communication protocol QUIC and the optimal transmission path, the encapsulated data packet is sent from the source end to the destination end using the network communication protocol QUIC.

7. The method for achieving low-latency data transmission according to claim 6, wherein: The transmission strategy is adjusted as follows: A multi-dimensional action suggestion mechanism based on the DRL model, setting the triggering conditions of the multi-dimensional action suggestion mechanism; During data transmission, network delay and packet loss rate are collected in real time and input into the DDPG model. After triggering the DDPG multi-dimensional action suggestion mechanism of the deep reinforcement learning model, the transmission rate and retransmission threshold during data transmission are adjusted. After adjusting the transmission rate and retransmission threshold, the link is evaluated and optimized based on the feedback of the adjustment results provided by the deep reinforcement learning model DDPG.

8. A low-latency data transmission system, based on the low-latency data transmission method according to any one of claims 1 to 7, characterized in that: include, The initial configuration module uses the network communication protocol QUIC and performs initial configuration; The path module uses the reinforcement learning model PPO to analyze the network status of the network communication protocol QUIC and select the optimal path; The processing module selects the best edge computing node to process the user request based on the data transmission location of the user request and the nearest neighbor principle; The compression module compresses the transmitted data using adaptive Huffman coding technology based on the characteristics of the optimal edge node; The adjustment module sends compressed transmission data from the source to the destination according to the network communication protocol QUIC and the optimal path. When high latency and packet loss rate occur during data transmission, the multi-dimensional action suggestion mechanism of the deep reinforcement learning model DDPG is used to adjust the transmission strategy, and the link effectiveness is evaluated based on the deep reinforcement learning model DDPG feedback mechanism.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of the method for realizing low-latency data transmission according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for realizing low-latency data transmission according to any one of claims 1 to 7 are implemented.

Images