A method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes
By adopting a packet loss modeling method based on Markov arrival process, the problem that existing technologies cannot accurately characterize the packet loss features of low-Earth orbit satellite networks is solved, and more accurate packet loss modeling and delay analysis are achieved, providing support for the simulation and coding design of low-Earth orbit satellite networks.
Patent Information
- Application Number
- CN202411968067.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2025-10-31
- Estimated Expiration
- 2044-12-30
AI Technical Summary
Existing packet loss models cannot accurately characterize the packet loss features of low-Earth orbit satellite networks, which limits the effectiveness of latency analysis and algorithm design.
A packet loss modeling method based on the Markov arrival process is adopted. By establishing a low-orbit satellite communication data measurement system, packet loss data is obtained, Markov arrival process model parameters are generated, and new packet loss data is constructed.
It effectively characterizes the frequent packet loss caused by high mobility in low-Earth orbit satellite networks, provides a more realistic packet loss modeling effect, and improves the accuracy of delay analysis and erasure coding design.
Smart Images
Figure CN119743402B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wireless communication technology, and in particular to a method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes. Background Technology
[0002] In recent years, Low-Earth-orbit Satellite Networks (LSNs) have developed rapidly. LSNs will play a core role in next-generation mobile communication systems (6G) and are an indispensable component of the vision of an integrated air-space-ground-sea network. However, compared with traditional terrestrial cellular networks, LSNs face severe challenges in terms of network reliability and Quality of Service (QoS). These problems stem from the dynamic nature of LSNs, including frequent satellite handovers, complex channel conditions, and high mobility.
[0003] In LSNs, due to the limited coverage area of each satellite and their high-speed movement, frequent switching occurs when satellites leave the coverage area of ground antennas. To maintain a continuous connection with users, the system needs to periodically switch to a new satellite and reconfigure the network. These frequent switchings cause dynamic changes in network paths and fluctuations in transmission latency, which can disrupt routing, congestion control, and packet loss recovery mechanisms, leading to severe packet loss. This packet loss is a critical factor affecting the end-to-end performance of LSN communication, resulting in increased end-to-end communication latency, decreased throughput, and consequently impacting the user experience and overall quality of service for real-time applications. Modeling these packet loss behaviors helps optimize network design and improve the performance of LSN communication systems.
[0004] In terrestrial cellular network scenarios, commonly used packet loss models include the Bernoulli model, the Gilbert-Elliott (GE) model, the 4-State Markov (4SM) model, the Loss Run-Length (LRL) model, and the Receive-Loss Run-Length (RLRL) model. Among these, the GE and 4SM models are frequently used by network simulators such as Netem to simulate packet loss in terrestrial cellular networks. However, because the packet loss characteristics of LSNs (Latent-Side Networks) differ significantly from those of terrestrial cellular networks, these commonly used packet loss models are insufficient to effectively characterize the packet loss features in LSN communication. Using models such as GE and 4SM to simulate LSN packet loss characteristics is inaccurate in some LSN-related algorithm design and system development. If a packet loss model cannot adequately characterize the packet loss characteristics of a network, its guiding effect on latency analysis, algorithm design, and other issues is very limited.
[0005] In LSNs scenarios, existing packet loss models face packet loss characteristics that are very different from those of terrestrial cellular networks. The high dynamic characteristics of LSNs result in frequent packet loss and large-scale fluctuations in latency in end-to-end connections. The randomness and suddenness of this frequent packet loss are beyond the capabilities of existing packet loss models. Summary of the Invention
[0006] To overcome the shortcomings of existing packet loss models, this invention provides a packet loss modeling method for low-Earth orbit (LEO) satellite networks based on Markov arrival processes. This method overcomes the problem that existing packet loss models cannot accurately characterize the packet loss features of LEO satellite networks, and provides assistance for LEO satellite network simulation, delay analysis, and erasure coding design.
[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:
[0008] A method for modeling packet loss in low-Earth orbit (LEO) satellite networks based on Markov arrival processes includes the following steps: S1, establishing a LEO satellite communication data measurement system and acquiring data packet information; S2, processing the measured data from the system and extracting packet loss data from the acquired data packet information; S3, obtaining the packet loss arrival time interval sequence. S4. Use the packet loss arrival time interval sequence S5. Generate the corresponding Markov arrival process model parameters; S6. Generate new packet loss data using the Markov arrival process model parameters.
[0009] As a further preferred technical solution of the present invention, in S1, the low-Earth orbit satellite communication data measurement system adopts a user-router-Starlink satellite terminal-Starlink satellite-ground base station-gateway-server communication data measurement system, with a data packet transmission time interval of 10ms; by running iRTT simultaneously on the user end and the server, the accurate transmission and reception time of each data packet is obtained; through the continuous communication of the measurement system and the continuous operation of iRTT, real low-Earth orbit satellite communication data based on the Starlink satellite communication system is obtained.
[0010] Furthermore, as a preferred embodiment of the present invention, in S2, packet loss data is obtained by processing real low-orbit satellite communication data. ,in It is the total length of the data. Indicates packet loss. This indicates that the data packet was received correctly. In the actual test system, when iRTT is running simultaneously at both the sending and receiving ends, if only the sending time is obtained at the sending end and the receiving time at the receiving end is empty, it means that the data packet was lost; if the receiving time at the receiving end is not empty, it means that the data packet was received correctly.
[0011] Furthermore, as a preferred embodiment of the present invention, in S3, the sequence of packet loss arrival time intervals is denoted as... ,in It is lost data. The total number of packets lost in the data. It is the first The first packet loss and the first The time interval between packet loss; sequence There are two methods for obtaining the time interval data: The first method obtains the actual time interval data by recording the packet transmission time obtained by the sender running iRTT each time a packet is lost; the actual time interval data is obtained by subtracting the former from the latter in sequence. The second method of generating time interval sequences is based on the premise of a constant data packet transmission rate, and includes lost data. The time interval between all elements and their adjacent elements is considered to be 1 second; a set of virtual whole-second time interval data is obtained by recording the number of 0s between two 1s. ; where 1 indicates packet loss and 0 indicates packet reception.
[0012] Furthermore, as a preferred embodiment of the present invention, in S4, the Markov arrival process... The matrix is fitted using either expectation maximization or moment matching; for a given matrix... The Markov arrival process of the order, its and All A matrix of order n; where, The infinitesimal generator that did not appear was not reached. It is the transfer rate matrix at the time of arrival; It has non-negative off-diagonal elements and negative diagonal elements. All matrix elements are non-negative; It is an infinitesimal generator of the underlying Markov chain; when using the expectation-maximization algorithm, it will... Defined as:
[0013]
[0014] in, ;set up and These are observable data vectors and unobservable data vectors, defining an arrival time series. , ;matrix The parameters in the algorithm are updated and calculated using the following formula:
[0015]
[0016] in, It is a temporary parameter vector; It is a state In time interval Total stay time; In the time interval from arrive The number of phase transitions; It is an indicator random variable, representing an instantaneous... From arrive Phase transition.
[0017] Furthermore, as a preferred embodiment of the present invention, in S4, parameters are used. Generate new time interval data The process involves calculating the dwell time in each state and constructing a probability matrix to guide state transitions. For each sample to be generated, state transitions are simulated, and the elapsed time is accumulated until the preset sample size requirement is met. During this process, whenever a new state is reached, the elapsed time interval and corresponding arrival type are recorded, and the current state is updated for the next iteration. All generated data is then organized into new time interval data. ; For the generated time interval data Data loss during the S3 process Data to time interval The inverse transformation of this process yields the packet loss data generated by the Markov arrival process model.
[0018] Furthermore, as a preferred embodiment of the present invention, it further includes selecting a new set of measured data to compare the effect of Markov arrival process fitting low-Earth orbit satellite network packet loss; to verify whether the packet loss data generated in step S5 has packet loss characteristics similar to the real data, the following definition is used:
[0019] Gap G is defined as a series of consecutive zeros between two 1s in a lost packet sequence, the length of which is equal to the number of consecutive zeros; error burst EB is defined as a sequence of zeros with greater than or equal to 1 at both ends. Long gap error clusters or error clusters and the intervals between them less than A sequence consisting of long gaps; an error-free burst (EFB) is defined as a sequence with a length greater than or equal to [a certain value]. The gap, in which It is a predefined positive integer;
[0020] The probability distribution function defined above is used to compare the effectiveness of the generated lost packet data in characterizing the lost and received features in the real data; if the data complexity is too high and the model fails to reasonably characterize the lost packet features of the real data, then... The matrix order is multiplied by 2, and new packet loss data is generated again. The fitting effect of the packet loss features is compared with the real data. For most packet loss time interval sequences, an 8th-order matrix is used. A matrix is sufficient to characterize its packet loss features.
[0021] The low-Earth orbit satellite network packet loss modeling method based on Markov arrival process described in this invention has the following technical advantages compared with existing technologies:
[0022] (1) This invention adopts the Markov arrival process as the packet loss model and proposes a new packet loss modeling scheme, providing a new solution for simulating packet loss in low-Earth orbit satellite networks. It effectively portrays the frequent packet loss caused by high mobility in low-Earth orbit satellite networks, achieving a modeling effect that is closer to real network packet loss, which is very beneficial for delay analysis and erasure coding design.
[0023] (2) This invention describes packet loss as a random arrival event and simulates packet loss in low-Earth orbit satellite networks through Markov arrival processes, which solves the problem that existing packet loss models cannot reasonably characterize the packet loss features of low-Earth orbit satellite networks and provides a new reference for the simulation of low-Earth orbit satellite networks. Attached Figure Description
[0024] Figure 1 This is a schematic diagram of a Starlink-based low-Earth orbit satellite communication data measurement system according to an embodiment of the present invention.
[0025] Figure 2 This is a schematic diagram of the error burst distribution (error burst probability distribution function) according to an embodiment of the present invention;
[0026] Figure 3 This is a schematic diagram of the error-free burst distribution (error-free burst probability distribution function) according to an embodiment of the present invention. Detailed Implementation
[0027] The present invention will be further explained in detail below with reference to the accompanying drawings, so that those skilled in the art can better understand and implement the present invention. However, the following examples are only used to explain the present invention and are not intended to limit the present invention.
[0028] A method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes includes the following steps:
[0029] S1. Establish a low-Earth orbit satellite communication data measurement system and acquire communication data packet information by collecting data. The system adopts a user-router-Starlink satellite terminal-Starlink satellite-ground base station-gateway-server communication data measurement system, such as... Figure 1As shown, the data packet transmission interval is 10ms. By running iRTT simultaneously on both the user end and the server, the accurate transmission and reception times of each data packet can be obtained. Through continuous communication of the actual test system and continuous operation of iRTT, real-world data on low-Earth orbit satellite communication based on the Starlink satellite communication system was obtained.
[0030] S2. Process the system's measured data to extract packet loss data from the acquired data packet information. Packet loss data can be obtained by processing the actual data. ,in It is the total length of the data. Indicates packet loss. This indicates that the data packet was received correctly. In the actual test system, when iRTT is running simultaneously at both the sending and receiving ends, but only the sending time is obtained at the sending end while the receiving time at the receiving end is empty, this means that the data packet was lost. Similarly, if the receiving time at the receiving end is not empty, it means that the data packet was received correctly.
[0031] S3. Obtain the packet loss arrival time interval sequence. Denote the packet loss arrival time interval sequence as... ,in It is lost data. The total number of packets lost in the data. It is the first The first packet loss and the first The time interval between packet losses. Sequence There are two ways to obtain the time interval data: The first method is to obtain the actual time interval data, which is to record the data packet transmission time obtained by the sender running iRTT each time a packet is lost; the actual time interval data is obtained by subtracting the former from the latter in turn. The second method of generating time interval sequences assumes a constant data packet transmission rate and includes lost data. The time interval between all elements and their adjacent elements is considered to be 1 second; by recording the number of 0s (received) between two 1s (packet loss), a set of virtual whole-second time interval data can be quickly obtained. .
[0032] S4. Use a sequence of packet loss arrival time intervals. Generate the corresponding Markov arrival process parameters. Specifically, the Markov arrival process... The matrix can be fitted using either expectation maximization or moment matching methods. For a given matrix... The Markov arrival process of the order, its and All A matrix of order n. Where, The infinitesimal generator that did not appear was not reached. It is the transfer rate matrix at the time of arrival; It has non-negative off-diagonal elements and negative diagonal elements. All matrix elements are non-negative; It is an infinitesimal generator of the underlying Markov chain. When using the expectation-maximization algorithm, it can be... Defined as:
[0033]
[0034] in, .set up and These are observable data vectors and unobservable data vectors, defining an arrival time series. , .matrix The parameters in the algorithm can be updated and calculated using the following formula:
[0035]
[0036] in, It is a temporary parameter vector; It is a state In time interval Total stay time; In the time interval from arrive The number of phase transitions; It is an indicator random variable, representing an instantaneous... From arrive Phase transition.
[0037] S5. Generate new packet loss data using the Markov arrival process model parameters. (Use the parameters...) New time interval data can be generated. The process calculates the dwell time in each state and constructs a probability matrix to guide state transitions. For each sample to be generated, state transitions are simulated, and the elapsed time is accumulated until a preset sample size requirement is met. During this process, whenever a new state is reached, the elapsed time interval and corresponding arrival type are recorded, and the current state is updated for the next iteration. All generated data is then organized into new time interval data. For the generated time interval data Data loss during the S3 process Data to time interval The inverse transformation of this process yields the packet loss data generated by the Markov arrival process model.
[0038] A new set of measured data is selected to compare the effectiveness of the Markov arrival process in fitting packet loss data of low-Earth orbit satellite networks. To verify whether the packet loss data generated in step S5 possesses packet loss characteristics similar to the real data, the following definition is used here:
[0039] The gap (G) is defined as the consecutive zeros between two 1s in the lost packet sequence, and its length is equal to the number of consecutive zeros.
[0040] An error burst (EB) is defined as an error that occurs at both ends with a value greater than or equal to 1. Long gap error clusters or error clusters and the intervals between them less than A sequence consisting of long gaps.
[0041] Error-free burst (EFB) is defined as having a length greater than or equal to The gap, in which It is a predefined positive integer.
[0042] The probability distribution function defined above is used to compare the effectiveness of the generated lost packet data in characterizing the lost and received features in the real data. If the data complexity is too high and the model fails to reasonably characterize the lost packet features of the real data, then... The matrix order is multiplied by 2, and new packet loss data is generated again. The fitting effect of the packet loss features is compared with the real data. For most packet loss time interval sequences, an 8th-order matrix is used. A matrix is sufficient to characterize its packet loss features.
[0043] This invention relates to the acquisition and processing of low-Earth orbit satellite communication data. Using real data from the Starlink satellite communication system as raw data, and by simultaneously running iRTT on the user end and the remote server, the acquired sender time and receiver time can be processed into 0 / 1 packet loss and packet loss arrival time interval data.
[0044] This invention uses a Markov arrival process as a packet loss model. Addressing the high mobility and frequent random packet loss caused by satellite handover in low-Earth orbit satellite communications, packet loss is treated as a random arrival event, and a Markov arrival process is used to characterize the packet loss features.
[0045] This invention underwent extensive testing in MATLAB and was compared with existing packet loss models such as Bernoulli, GE, 4SM, LRL, and RLR. Given that downlink communication packet loss is significantly higher than uplink communication loss, downlink communication data from the same time period over two consecutive days were selected for model parameter generation and evaluation metric comparison. The results of the three metrics are as follows: Figures 2-3As shown. The GE and 4SM models in the existing models are the most commonly used packet loss models in current network simulators. Figures 2-3 It can be seen that these methods are not suitable for fitting the packet loss characteristics of low-Earth orbit (LEO) satellite networks. The results show that the packet loss model described in this invention exhibits significant advantages and is more suitable for fitting the packet loss characteristics of LEO satellite networks than existing packet loss models.
[0046] The specific implementation schemes described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific implementation schemes of the present invention and are not intended to limit the scope of the present invention. Any equivalent changes and modifications made by those skilled in the art without departing from the concept and principles of the present invention should fall within the scope of protection of the present invention.
Claims
1. A method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes, characterized in that, Specifically, the following steps are included: S1. Establish a low-orbit satellite communication data measurement system and obtain communication data packet information by collecting data. S2. Process the system's measured data and extract lost packet data from the acquired data packet information; S3. Obtain the packet loss arrival time interval sequence. ; In S3, the sequence of packet loss arrival time intervals is denoted as... ,in It is lost data. The total number of packets lost in the data. It is the first The first packet loss and the first The time interval between packet loss; sequence There are two methods for obtaining the time interval data: The first method obtains the actual time interval data by recording the packet transmission time obtained by the sender running iRTT each time a packet is lost; the actual time interval data is obtained by subtracting the former from the latter in sequence. The second method of generating time interval sequences is based on the premise of a constant data packet transmission rate, and includes lost data. The time interval between all elements and their adjacent elements is considered to be 1 second; a set of virtual whole-second time interval data is obtained by recording the number of 0s between two 1s. Where 1 indicates packet loss and 0 indicates packet reception; S4. Use packet loss arrival time interval sequence Generate the corresponding Markov arrival process model parameters; In S4, the Markov arrival process The matrix is fitted using either expectation maximization or moment matching; for a given matrix... The Markov arrival process of the order, its and All A matrix of order n; where, The infinitesimal generator that did not appear was not reached. It is the transfer rate matrix at the time of arrival; It has non-negative off-diagonal elements and negative diagonal elements. All matrix elements are non-negative; It is an infinitesimal generator of the underlying Markov chain; when using the expectation-maximization algorithm, it will... Defined as: ； in, ;set up and These are observable data vectors and unobservable data vectors, defining an arrival time series. , ;matrix The parameters in the algorithm are updated and calculated using the following formula: ； in, It is a temporary parameter vector; It is a state In the time interval Total stay time; In the time interval from arrive The number of phase transitions; It is an indicator random variable, representing an instantaneous... From arrive Phase transition; In S4, parameters are used. Generate new time interval data The process involves calculating the dwell time in each state and constructing a probability matrix to guide state transitions. For each sample to be generated, state transitions are simulated, and the elapsed time is accumulated until the preset sample size requirement is met. During this process, whenever a new state is reached, the elapsed time interval and corresponding arrival type are recorded, and the current state is updated for the next iteration. All generated data is then organized into new time interval data. ; For the generated time interval data Data loss during the S3 process Data to time interval The inverse transformation of this process yields the packet loss data generated by the Markov arrival process model; S5. Generate new packet loss data using the parameters of the Markov arrival process model.
2. The method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes according to claim 1, characterized in that, In S1, the low-Earth orbit satellite communication data test system adopts a user-router-Starlink satellite terminal-Starlink satellite-ground base station-gateway-server communication data test system with a data packet transmission interval of 10ms. By running iRTT simultaneously on the user end and the server, the accurate transmission and reception time of each data packet is obtained. Through continuous communication of the test system and continuous operation of iRTT, real low-Earth orbit satellite communication data based on the Starlink satellite communication system is obtained.
3. The method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes according to claim 2, characterized in that, In S2, packet loss data is obtained by processing real low-Earth orbit satellite communication data. ,in It is the total length of the data. Indicates packet loss. This indicates that the data packet was received correctly. In the actual test system, when iRTT is running simultaneously at both the sending and receiving ends, if only the sending time is obtained at the sending end and the receiving time at the receiving end is empty, it means that the data packet was lost; if the receiving time at the receiving end is not empty, it means that the data packet was received correctly.
4. The method for modeling packet loss in low-Earth orbit satellite networks based on Markov arrival processes according to claim 3, characterized in that, This also includes selecting a new set of measured data to compare the effectiveness of Markov arrival process fitting for packet loss in low-Earth orbit satellite networks; using the following definitions: Gap G is defined as a series of consecutive zeros between two 1s in a lost packet sequence, the length of which is equal to the number of consecutive zeros; error burst EB is defined as a sequence of zeros with greater than or equal to 1s at both ends. Long gap error clusters or error clusters and the intervals between them less than A sequence consisting of long gaps; an error-free burst (EFB) is defined as a sequence with a length greater than or equal to [a certain value]. The gap, in which It is a predefined positive integer; The above definitions are used to compare the effectiveness of the generated lost packet data in characterizing the lost and received features in the real data; if the data complexity is too high and the model fails to reasonably characterize the lost packet features of the real data, then... The matrix order is multiplied by 2, and new packet loss data is generated again. The fitting effect of the packet loss features is compared with the real data.