RedCap terminal delay jitter compensation method, device and equipment and storage medium

Abstract

The invention discloses a RedCap terminal delay jitter compensation method, device and equipment and a storage medium, and belongs to the technical field of terminal access, and the method comprises the following steps: selecting a target LSTM model matched with a memory resource demand according to the memory capacity of a RedCap terminal; training a target LSTM model based on TCP related data obtained in advance; constructing a delay jitter compensation system based on the trained LSTM model; deploying a delay jitter compensation system to the RedCap terminal so as to send state data in real time according to a TCP, and feeding back a real-time data sending interval after optimization of delay jitter to the RedCap terminal; and enabling the RedCap terminal to adjust the data transmission interval of the TCP data packet according to the real-time data transmission interval after the delay jitter optimization, and completing the delay jitter compensation of the RedCap terminal. According to the invention, the problem of insufficient accuracy and adaptability of business flow delay jitter performance optimization is solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of terminal access, and particularly relates to a RedCap terminal time delay jitter compensation method, device, equipment and storage medium. BACKGROUND

[0002] RedCap (Reduced Capability) is a lightweight 5G technology. Compared with traditional 5G terminals, RedCap terminals have the advantages of halving the cost and power consumption, while still inheriting the slicing, low latency, high reliability and other characteristics of 5G networks, which is more conducive to the large-scale deployment of 5G applications. For concentrators and energy controllers that need to be deployed on a large scale, RedCap is a cost-effective choice. The typical application scenarios of RedCap are industrial wireless sensors, video surveillance and wearable devices, and a variety of services supported in these three application scenarios have high requirements for time delay jitter. In addition, RedCap can fully utilize the characteristics of 5G networks to meet the business needs of the power industry, thereby further reducing the cost of terminals. Smart grid services are mainly divided into control and collection, and control services have high requirements for network latency and reliability, with a service demand of 20ms-50ms. Therefore, certain time delay jitter control and optimization are also needed.

[0003] Current schemes for optimizing time delay jitter performance include: (1) introducing edge computing at the network layer to reduce the long-distance transmission delay of data to and from the cloud. However, the computing resources of edge nodes are relatively limited, and when faced with large-scale complex data processing tasks, they may cause some tasks to queue and wait, thereby causing new time delay jitter. Moreover, due to the dispersion of edge nodes, the cost of operation and maintenance is high, and it is difficult to ensure security. (2) Traffic shaping technology allocates and controls network traffic according to the priority of different services. However, traffic shaping requires accurate understanding of the traffic characteristics of various services. Once the business model changes or new services appear, the previous settings may become invalid, and it takes time and effort to reconfigure. During this period, traffic control vulnerabilities may occur, causing time delay fluctuations. (3) Buffering mechanisms are also commonly used to smooth jitter in data transmission processes. However, a large buffer can introduce additional delay, which seriously affects real-time interactive experience. If the buffer is too small, it is difficult to effectively respond to sudden time delay jitter, and there is still a risk of data flow interruption. (4) From the perspective of optimizing transmission protocols, new protocols with low latency, such as QUIC protocol, can replace traditional TCP protocols to speed up connection establishment, optimize congestion control, and improve transmission efficiency. However, the popularization of new protocols requires the comprehensive upgrade and adaptation of existing network devices, servers and client software, which is costly. On the other hand, in complex network environments, compatibility is poor, and when working with some old systems, it may even cause abnormalities, leading to increased time delay. SUMMARY

[0004] In order to solve the above problems in the prior art, the application provides a RedCap terminal time delay jitter compensation method, device, equipment and storage medium, which intelligently control the sending rate, and solves the problems of insufficient accuracy and adaptability of service flow time delay jitter performance optimization.

[0005] In order to achieve the above application purposes, the application adopts the technical solutions of: In a first aspect, the application provides a RedCap terminal time delay jitter compensation method, which includes the following steps: S1, selecting a target LSTM model matched with the memory resource requirement according to the memory capacity of the RedCap terminal; S2, training the target LSTM model based on the pre-acquired TCP related data; S3, constructing a time delay jitter compensation system based on the trained LSTM model; S4, deploying the time delay jitter compensation system to the RedCap terminal, so as to feed back the real-time data sending interval after optimizing the time delay jitter to the RedCap terminal according to the TCP sending state data of the RedCap terminal; S5, adjusting the data sending interval of the TCP data packet according to the real-time data sending interval after optimizing the time delay jitter, and completing the RedCap terminal time delay jitter compensation.

[0006] Further, the S1 includes the following steps: S11, acquiring the memory capacity of the GPU hardware device when the RedCap terminal runs, as the memory capacity of the RedCap terminal; S12, taking the GPU memory required by the LSTM model when running as the memory resource requirement of the LSTM model; S13, in the case that the memory resource requirement of the LSTM does not exceed the memory capacity of the RedCap terminal, selecting the LSTM model with the largest LSTM memory resource requirement as the target LSTM model.

[0007] Further, the S2 includes the following steps: S21, pre-collecting the TCP sending state data of the RedCap terminal based on the SOCKET connection by using the first state collection module, and calculating the pre-collection data round-trip delay and congestion window size when the collected data amount meets the LSTM input sequence length; S22, normalize and pre-process the pre-collected round-trip delay and congestion window size by using the first pre-processing module to obtain pre-collected TCP related data, wherein the pre-collected TCP related data includes normalized pre-collected congestion window size, normalized pre-collected round-trip delay size, and pre-collected data sending interval after optimizing delay jitter; S23, repeat S21-S22 for a preset collection number of times, take the pre-collected TCP related data obtained multiple times as training sample data respectively, and form an LSTM training set; S24, according to the LSTM training set and a target loss function, based on a target LSTM model using an encoding-decoding model, perform semantic encoding on the normalized pre-collected congestion window size and the normalized pre-collected round-trip delay size in each training sample data, and take the pre-collected data sending interval after optimizing delay jitter as a decoding result for training, adjust the network parameters in the target LSTM model, and obtain a trained LSTM model.

[0008] Further, the calculation expression of the pre-collected data round-trip delay and congestion window size in S21 is as follows: , , , wherein, represents the minimum round-trip delay of the i-th pre-collected data packet, represents the receiving rate of TCP when the i-th pre-collected data packet is sent, represents the sending rate of TCP when the i-th pre-collected data packet is sent, represents the congestion window size when the pre-collected data is collected, represents the total number of pre-collected data packets, represents the congestion window size when the i-th pre-collected data packet is sent, represents the TCP sending state data of the i-th pre-collected data packet. Further, the calculation expression of the pre-collected TCP related data in S22 is as follows: , ,

[0009] , , , , wherein, ​​​denotes a normalized pre-collected congestion window size, denotes a maximum value of a pre-collected congestion window size, denotes a normalized round-trip time size of a pre-collected data packet, denotes a minimum round-trip time of a pre-collected data packet, denotes a maximum round-trip time of a pre-collected data packet, denotes a round-trip time size of a pre-collected data packet, denotes a time jitter between a pre-collected data packet and a pre-collected data packet, denotes a pre-collected data sending interval after optimizing time jitter, denotes a data sending interval.

[0010] Further, the time jitter compensation system in S3 comprises: a second state collection module, configured to collect TCP sending state data of the RedCap terminal in real time based on a SOCKET connection, until the amount of collected data meets a length of an input sequence of an LSTM, to calculate a real-time data round-trip time and a congestion window size, and to transmit the real-time data round-trip time and the congestion window size to a second preprocessing module; the second preprocessing module, configured to perform normalization processing on the real-time data round-trip time and the congestion window, to obtain a normalized real-time congestion window size and a normalized real-time round-trip time size, and to transmit the normalized real-time congestion window size and the normalized real-time round-trip time size to an LSTM module; the LSTM module, configured to use a trained LSTM model to predict a real-time data sending interval based on the normalized real-time congestion window size and the normalized real-time round-trip time size, and to transmit the real-time data sending interval to a rate control module; the rate control module, configured to optimize and adjust the data sending interval based on a detection and adjustment model according to the real-time data sending interval, to obtain a real-time data sending interval after optimizing time jitter, and to transmit the real-time data sending interval after optimizing time jitter to a parameter configuration module; the parameter configuration module, configured to send the real-time data sending interval after optimizing time jitter to the RedCap terminal based on the SOCKET connection.

[0011] Further, a calculation expression of the detection and adjustment model is as follows: , , , wherein, denotes a real-time data sending interval after optimizing time jitter between a real-time data packet and a real-time data packet outside a real-time congestion window, represents the first real-time data packet, represents a real-time congestion window size, represents an assignment, represents a cumulative data sending interval of the real-time data packet in the real-time congestion window, represents a round-trip delay size of the first real-time data packet, represents the first real-time data packet, represents a round-trip delay size of the first real-time data packet, represents the first real-time data packet, represents the first real-time data packet, represents a data sending interval between the first real-time data packet and the second real-time data packet. In a second aspect, the present application provides a RedCap terminal latency jitter compensation device, comprising: An LSTM matching module is configured to select a target LSTM model matched in memory resource demand according to the memory capacity of the RedCap terminal;

[0012] An LSTM training module is configured to train the target LSTM model based on pre-acquired TCP related data; A compensation system construction module is configured to construct a latency jitter compensation system based on the trained LSTM model; A latency jitter optimization module is configured to deploy the latency jitter compensation system to the RedCap terminal, so as to feed back the real-time data sending interval after optimizing the latency jitter to the RedCap terminal in real time according to the TCP sending state data of the RedCap terminal; A latency jitter compensation module is configured to enable the RedCap terminal to adjust the data sending interval of the TCP data packet according to the real-time data sending interval after optimizing the latency jitter, so as to complete the latency jitter compensation of the RedCap terminal. In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to execute the RedCap terminal latency jitter compensation method.

[0013] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the RedCap terminal latency jitter compensation method.

[0014] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the RedCap terminal latency jitter compensation method.

[0015] ​The application has the advantages that the RedCap terminal time delay jitter compensation method, device, equipment and storage medium provided by the application configure LSTM models with different memory resource requirements for the memory resource differences of the RedCap terminal, preferentially use the model with the optimal performance under the memory constraint condition of the RedCap terminal, and effectively avoid the risk that the model is too large to run on some small RedCap; the application uses the PACING sending mechanism in the TCP congestion control, the time delay jitter compensation system constructed based on the trained LSTM model, controls the data packet sending interval, can cope with complex changing RedCap terminal network scenes, optimizes the time delay jitter performance of different services of the RedCap 5G terminal system, checks and corrects the time delay jitter compensation result of the LSTM model through the rate control module, does not reduce the sending rate of the service under the premise of ensuring that the end system time delay jitter performance is improved, guarantees the stability of the end-to-end data transmission throughput, and realizes the steady improvement of the network performance; the method can adapt to different network and terminal scenes, the feedback loop of the whole algorithm can still work normally even when the RedCap terminal network scene changes, effectively adjusts the service data packet sending interval, and quickly adapts to the new network scene.

[0016] Other advantages of the application will be analyzed in more detail in the subsequent embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 The step flow chart of the RedCap terminal time delay jitter compensation method in the embodiment of the application.

[0019] Figure 2 The block diagram of the time delay jitter compensation system in the embodiment of the application.

[0020] Figure 3 The block diagram of the RedCap terminal time delay jitter compensation device in the embodiment of the application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.

[0022] The RedCap terminal communicates end-to-end through an operator network, wherein one RedCap terminal acts as a sender and another RedCap terminal acts as a receiving receiver. The present application is a scheme for compensating for the delay jitter of the RedCap terminal as a sender. Therefore, a delay jitter compensation system is deployed to the RedCap terminal as a sender to provide the RedCap terminal with a real-time data sending interval after optimizing the delay jitter, so as to achieve RedCap terminal delay jitter compensation.

[0023] As shown in the figure, in one embodiment of the present application, the present application provides a RedCap terminal delay jitter compensation method, comprising the following steps: Figure 1 S1, selecting a target LSTM model matched in memory resource demand according to the memory capacity of the RedCap terminal; In the present application, the Long Short-Term Memory (LSTM) is a kind of recurrent neural network (RNN), which can effectively avoid the gradient disappearance or gradient explosion phenomenon existing in the traditional RNN when processing long sequence data, so as to better capture the long-term dependence relationship in the sequence. The core of LSTM lies in its unique cell state and a series of "gate structures" for controlling information flow, mainly including input gate, forget gate and output gate. These gate structures output values between 0~1 through sigmoid activation function to determine the degree of information transmission and retention. At each time step t , the input data and the hidden state at the last time step enter the forget gate, the input gate and the output gate in turn; the forget gate decides how much information to discard according to the hidden state at the last time step; the input gate decides how much new information to retain and adds the retained new information to the cell state; the output gate decides what kind of hidden state to output according to the current cell state and passes it to the next time step.

[0024] The S1 comprises the following steps: ​S11, acquire the memory capacity of the GPU hardware device running at the RedCap terminal as the memory capacity of the RedCap terminal; S12, take the GPU memory required by the LSTM model running as the memory resource requirement of the LSTM model; The conventional LSTM model matching operation is that a single model learns based on different network scenes, and the final model structure is consistent, and different groups of LSTM model parameters are obtained, that is, the parameter value size in the LSTM neuron is different. In the present scheme, the functions of the LSTM models are consistent with the network scenes, and are used to control the data packet sending interval. The difference lies in that the LSTM model structure, input parameter dimension, input parameter data volume, neuron number, output parameter dimension, output parameter number, and model parameter compression package file size are different.

[0025] S13, in the case where the memory resource requirement of the LSTM does not exceed the memory capacity of the RedCap terminal, select the LSTM model with the largest LSTM memory resource requirement as the target LSTM model.

[0026] In the present embodiment, the LSTM model requires a certain GPU memory capacity when running, and the occupied GPU memory capacity is related to the number of neurons, the number of parameters and the forward propagation memory inside the LSTM model. As an example, the running parameters and memory resource requirement table of the LSTM model is shown in Table 1: Table 1: Running parameters and memory resource requirement table of the LSTM model As an example, if the memory capacity of the RedCap terminal is only 10MB, then model 2 is selected for deployment to the RedCap terminal, and models 3 and 4 cannot be effectively run on the RedCap terminal. The present scheme sets LSTM models of different sizes according to the memory capacity resource differences of the RedCap terminal, and by preferentially using the model with the best performance under the memory constraint condition of the RedCap terminal, both the effect and efficiency of the delay jitter compensation of the RedCap terminal are guaranteed, and the situation that the LSTM model is too large to be effectively run on the RedCap terminal is avoided.

[0027] S2, train the target LSTM model based on the pre-acquired TCP related data; The S2 includes the following steps: S21, pre-collect the TCP sending state data of the RedCap terminal based on the SOCKET connection by using the first state collection module, and when the collected data volume meets the LSTM input sequence length, calculate the pre-collected data round-trip delay and congestion window size; Based on ACK confirmation data packet, the round-trip delay size of data packet and the statistical congestion window size can be calculated.

[0028] The calculation expression of the pre-collected data round-trip delay and congestion window size in S21 is as follows: , , , Among them, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet.

[0029] S22, using the first preprocessing module to normalize and preprocess the pre-collected data round-trip delay and congestion window size, to obtain the pre-collected TCP related data, wherein the pre-collected TCP related data includes normalized pre-collected congestion window size, normalized pre-collected round-trip delay size, and pre-collected data sending interval after optimizing delay jitter; The calculation expression of the pre-collected TCP related data in S22 is as follows: , , , , Among them, represents the normalized pre-collected congestion window size, represents the maximum value of the pre-collected congestion window size, represents the normalized round-trip delay size of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the minimum round-trip delay of the first pre-collected data packet, represents the maximum round-trip delay of the first pre-collected data packet, represents the first pre-collected data packet and the first pre-collected data packet, represents the first pre-collected data packet and the first pre-collected data packet, The delay jitter of the pre-collected data packets, Indicates the The round trip delay of the pre-collected data packets, Indicates the interval for sending pre-collected data after optimizing delay jitter. Indicates the data sending interval.

[0030] In this embodiment, Determined by the window size field in the TCP header, there are 16 bits in total, so it is 0xffff.

[0031] S23, repeat S21-S22 for the preset number of collections, and use the TCP related data obtained in advance as training sample data to form an LSTM training set; in this embodiment, when the preset number of collections is set, it can meet the requirement of 10 training sample data. 4 ~10 5 In actual implementation, it can be adjusted according to the actual situation.

[0032] S24. According to the LSTM training set and the target loss function, based on the target LSTM model using the encoding-decoding model, the normalized pre-collected congestion window size and the normalized pre-collected round-trip delay size in each training sample data are semantically encoded, and the pre-collected data sending interval after optimizing the delay jitter is used as the decoding result for training, and the network parameters in the target LSTM model are adjusted to obtain a trained LSTM model.

[0033] In this embodiment, when training the target LSTM model according to the LSTM training set, the L2 loss function is used as the target loss function: the LSTM model is trained based on each training sample until the difference in L2 loss values ​​between consecutive adjacent training batches exceeding a preset number of stable times is less than a loss fluctuation threshold, thereby obtaining a trained LSTM model; in this embodiment, the preset number of stable times ranges from 5 to 7 times, and the loss fluctuation threshold ranges from 1e-3 to 1e-5.

[0034] S3. Build a delay jitter compensation system based on the trained LSTM model; like Figure 2 As shown, the delay jitter compensation system in S3 includes: The second state collection module is used to collect the TCP sending state data of the RedCap terminal in real time based on the SOCKET connection. When the amount of collected data meets the LSTM input sequence length, the real-time data round-trip delay and congestion window size are calculated and transmitted to the second preprocessing module; In this embodiment, when the second status collection module calculates the real-time data round-trip delay and congestion window size, the calculation method adopted is the same as the method used by the first status collection module to calculate the pre-collected data round-trip delay and congestion window size.

[0035] The second preprocessing module is used to normalize the real-time data round-trip delay and congestion window, obtain the normalized real-time congestion window size and real-time round-trip delay size, and transmit them to the LSTM module; In this embodiment, the second preprocessing module only needs to perform normalization processing on the round-trip delay of real-time data and the congestion window, and the normalization processing calculation method used is the same as the normalization calculation method in the first preprocessing module.

[0036] The LSTM module uses the trained LSTM model to predict the real-time data transmission interval based on the normalized real-time congestion window size and real-time round-trip delay, and transmits it to the rate control module. In this embodiment, after the normalized real-time congestion window size and the real-time round-trip delay size are input into the trained LSTM module, the data sending interval can be predicted, providing a basis for controlling and adjusting the data sending interval.

[0037] The rate control module is used to optimize and adjust the data transmission interval based on the detection and adjustment model according to the real-time data transmission interval, obtain the real-time data transmission interval after optimizing the delay jitter, and transmit it to the parameter configuration module; The calculation expressions of the detection and adjustment model are as follows: , , , in, Indicates the number of packets outside the real-time congestion window. The real-time data packet and The real-time data transmission interval after optimizing the delay jitter between real-time data packets, Indicates the Real-time data packets, Indicates the real-time congestion window size, Indicates assignment, Indicates the cumulative data sending interval of real-time data packets within the real-time congestion window. Indicates the The round trip delay of a real-time data packet, Indicates the Real-time data packets, Indicates sending The real-time data packet and The data sending interval between real-time data packets.

[0038] In the scheme, in the congestion control, the sending time of the data packet needs to be strictly controlled to match the network bandwidth or delay requirement, when , it indicates that the sending is too fast, which may cause queue accumulation or packet loss, by detecting and adjusting the model, reducing , the sending rate of the service can be ensured not to be reduced, the end-to-end data transmission throughput is ensured to be stable, and the overall performance is steadily improved.

[0039] The parameter configuration module is configured to send the optimized real-time data sending interval after delay jitter to the RedCap terminal based on the SOCKET connection.

[0040] S4, deploy the delay jitter compensation system to the RedCap terminal to feed back the optimized real-time data sending interval after delay jitter to the RedCap terminal in real time according to the TCP sending state data of the RedCap terminal. S5, instruct the RedCap terminal to adjust the data sending interval of the TCP data packet according to the optimized real-time data sending interval after delay jitter, and complete the delay jitter compensation of the RedCap terminal.

[0041] As shown in Figure 3 , another embodiment of the application provides a RedCap terminal delay jitter compensation device, comprising: The LSTM matching module is configured to select a target LSTM model matched in memory resource demand according to the memory capacity of the RedCap terminal. The LSTM training module is configured to train the target LSTM model based on the pre-acquired TCP related data. The compensation system construction module is configured to construct a delay jitter compensation system based on the trained LSTM model. The delay jitter optimization module is configured to deploy the delay jitter compensation system to the RedCap terminal to feed back the optimized real-time data sending interval after delay jitter to the RedCap terminal in real time according to the TCP sending state data of the RedCap terminal. The delay jitter compensation module is configured to instruct the RedCap terminal to adjust the data sending interval of the TCP data packet according to the optimized real-time data sending interval after delay jitter, and complete the delay jitter compensation of the RedCap terminal.

[0042] The RedCap terminal delay jitter compensation device is configured to execute the RedCap terminal delay jitter compensation method.

[0043] In another embodiment of the present application, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to execute the RedCap terminal delay jitter compensation method.

[0044] The computer device comprises a memory, a processor and one or more computer programs, wherein the one or more computer programs can be stored in the memory and configured to be executed by the one or more processors, and the one or more application programs are configured to execute the RedCap terminal delay jitter compensation method.

[0045] The processor can comprise one or more processing cores. The processor connects various parts in the entire computer device through various interfaces and lines, and executes various functions of the computer device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Optionally, the processor can be implemented in at least one of the forms of digital signal processing, field programmable gate array and programmable logic array. The processor can be integrated with one or a combination of the central processing unit, the data reporting verifier and the modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing display content; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor, but can be realized by a separate communication chip.

[0046] The memory can comprise a random access memory and a read-only memory. The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory can comprise a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for implementing at least one function, instructions for implementing the above-mentioned various method embodiments, etc. The data storage area can also store data created by the terminal device in use, etc.

[0047] In another embodiment of the present application, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the RedCap terminal delay jitter compensation method.

[0048] The computer readable storage medium can be an electronic storage, such as a flash memory, an electrically erasable programmable read only memory (EEPROM), an erasable programmable read only memory (EPROM), a hard disk, or a ROM. Alternatively, the computer readable storage medium includes a non-transitory computer readable medium. The computer readable storage medium has a storage space for storing program codes for executing any of the method steps described above. The program codes can be read from or written to one or more computer program products in an appropriate form.

[0049] The above description is merely a specific implementation of the present application. However, the scope of protection of the present application is not limited in this way. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, and all such changes or replacements should be covered within the scope of protection of the present application.

Claims

1. A RedCap terminal delay jitter compensation method, characterized in that: The steps include: S1. Select the target LSTM model that matches the memory resource requirements based on the memory capacity of the RedCap terminal. S2. Train the target LSTM model based on the pre-acquired TCP related data; S3. Build a delay jitter compensation system based on the trained LSTM model; S4. Deploy the delay jitter compensation system to the RedCap terminal to send status data of the TCP of the RedCap terminal in real time and provide feedback to the RedCap terminal on the real-time data sending interval after optimizing the delay jitter; S5. Instruct the RedCap terminal to adjust the data transmission interval of the TCP data packet according to the real-time data transmission interval after the delay jitter is optimized, thereby completing the delay jitter compensation of the RedCap terminal.

2. The RedCap terminal delay jitter compensation method according to claim 1, characterized in that: The S1 comprises the following steps: S11. Obtain the memory capacity of the GPU hardware device when the RedCap terminal is running, as the memory capacity of the RedCap terminal; S12. The GPU memory required for the LSTM model to run is used as the memory resource requirement of the LSTM model. S13. When the memory resource requirement of the LSTM does not exceed the memory capacity of the RedCap terminal, the LSTM model with the largest LSTM memory resource requirement is selected as the target LSTM model.

3. The RedCap terminal delay jitter compensation method according to claim 1, characterized in that: The S2 comprises the following steps: S21. Using the first state collection module based on the SOCKET connection, pre-collect the TCP sending state data of the RedCap terminal until the amount of collected data meets the LSTM input sequence length, and calculate the round-trip delay and congestion window size of the pre-collected data; S22. Normalize and preprocess the round-trip delay and congestion window size of the pre-collected data using the first pre-processing module to obtain pre-collected TCP-related data, wherein the pre-collected TCP-related data includes a normalized pre-collected congestion window size, a normalized pre-collected round-trip delay, and a pre-collected data sending interval after delay jitter optimization. S23, repeat S21-S22 for the preset number of collections, and use the TCP-related data obtained multiple times as training sample data to form an LSTM training set; S24. According to the LSTM training set and the target loss function, based on the target LSTM model using the encoding-decoding model, the normalized pre-collected congestion window size and the normalized pre-collected round-trip delay size in each training sample data are semantically encoded, and the pre-collected data sending interval after optimizing the delay jitter is used as the decoding result for training, and the network parameters in the target LSTM model are adjusted to obtain a trained LSTM model.

4. The RedCap terminal delay jitter compensation method according to claim 3, characterized in that: The calculation expressions for the round trip delay of the pre-collected data and the congestion window size in S21 are as follows: ， ， ， in, Indicates the The minimum round trip delay of pre-collected data packets, Indicates the The TCP receive rate when the pre-collected packets are sent, Indicates the The TCP sending rate when the pre-collected data packets are sent, Indicates the congestion window size during pre-collection. Indicates the total number of pre-collected data packets, Indicates the The congestion window size when sending pre-collected data packets, Indicates the TCP send status data for a pre-collected data packet.

5. The RedCap terminal delay jitter compensation method according to claim 3, characterized in that: The calculation expression of the TCP related data pre-acquired in S22 is as follows: ， ， ， ， in, Indicates the normalized pre-collected congestion window size, Indicates the maximum value of the pre-collected congestion window size. Indicates the The normalized round-trip delay of the pre-collected data packets, Indicates the minimum round-trip delay of pre-collected data packets, Indicates the maximum round trip delay of pre-collected data packets, Indicates the The first pre-collected data package and the The delay jitter of the pre-collected data packets, Indicates the The round trip delay of the pre-collected data packets, Indicates the interval for sending pre-collected data after optimizing delay jitter. Indicates the data sending interval.

6. The RedCap terminal delay jitter compensation method according to claim 3, characterized in that: The delay jitter compensation system in S3 includes: The second state collection module is used to collect the TCP sending state data of the RedCap terminal in real time based on the SOCKET connection. When the amount of collected data meets the LSTM input sequence length, the real-time data round-trip delay and congestion window size are calculated and transmitted to the second preprocessing module; The second preprocessing module is used to normalize the real-time data round-trip delay and congestion window, obtain the normalized real-time congestion window size and real-time round-trip delay size, and transmit them to the LSTM module; The LSTM module uses the trained LSTM model to predict the real-time data transmission interval based on the normalized real-time congestion window size and real-time round-trip delay, and transmits it to the rate control module. The rate control module is used to optimize and adjust the data transmission interval based on the detection and adjustment model according to the real-time data transmission interval, obtain the real-time data transmission interval after optimizing the delay jitter, and transmit it to the parameter configuration module; The parameter configuration module is used to send the real-time data transmission interval after delay jitter optimization to the RedCap terminal based on the SOCKET connection.

7. The RedCap terminal delay jitter compensation method according to claim 6, characterized in that: The calculation expressions of the detection and adjustment model are as follows: ， ， ， in, Indicates the number of packets outside the real-time congestion window. The real-time data packet and The real-time data transmission interval after optimizing the delay jitter between real-time data packets, Indicates the Real-time data packets, Indicates the real-time congestion window size, Indicates assignment, Indicates the cumulative data sending interval of real-time data packets within the real-time congestion window. Indicates the The round trip delay of a real-time data packet, Indicates the Real-time data packets, Indicates sending The real-time data packet and The data sending interval between real-time data packets.

8. A RedCap terminal delay jitter compensation device, characterized in that: include: The LSTM matching module is used to select the target LSTM model that matches the memory resource requirements based on the memory capacity of the RedCap terminal; LSTM training module, used to train the target LSTM model based on pre-acquired TCP-related data; Compensation system building module, used to build a delay jitter compensation system based on the trained LSTM model; The delay jitter optimization module is used to deploy the delay jitter compensation system to the RedCap terminal to send status data of the TCP of the RedCap terminal in real time and feedback the real-time data sending interval after optimizing the delay jitter to the RedCap terminal; The delay jitter compensation module is used to enable the RedCap terminal to adjust the data transmission interval of the TCP data packet according to the real-time data transmission interval after optimizing the delay jitter, so as to complete the delay jitter compensation of the RedCap terminal.

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, it performs the RedCap terminal delay jitter compensation method 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 RedCap terminal delay jitter compensation method according to any one of claims 1 to 7 are implemented.

Images