Congestion detection method and device, equipment and storage medium

Abstract

The invention provides a congestion detection method, apparatus and device, and a storage medium. The method comprises the steps of obtaining TS samples; the tth sample comprises congestion states of P paths in the network link system; constructing a routing matrix according to the relationship between the P paths and the L links, and determining a constraint condition based on the congestion state of the paths depending on the congestion state of the links and the routing matrix; determining a congestion probability # imgabs0 # of a pth path in a tth sample according to congestion states corresponding to a plurality of time slots, determining an initial congestion probability # imgabs2 # of an lth link in the tth sample according to the congestion probability # imgabs1 # and constraint conditions, and constructing a historical congestion information set according to the initial congestion probability # imgabs3 #, inputting the historical congestion information set into a long-short term memory network model to obtain a predicted congestion state of the lth link; the method and the device are used for realizing low-redundancy and high-sensitivity congestion sensing and improving the industrial production efficiency and the system stability.

Description

Technical Field

[0001] The present application relates to the field of communication status detection, and in particular to a congestion detection method, apparatus, device and storage medium. Background Art

[0002] With the rapid development of fifth-generation (5G) networks, the vertical application of the Internet of Things (IoT) in the industrial sector, namely the Industrial IoT (IIoT), has gradually become one of the key technologies to meet the efficiency requirements of production automation. In IIoT systems, a large number of industrial devices generate massive amounts of data in real time. This data must be processed under strict time constraints to avoid significant safety issues and economic losses caused by unexpected delays or task interruptions. Therefore, facing the massive amount of data generated by large-scale distributed IIoT devices, congestion detection in the end-to-network collaboration process has become particularly important. Especially in the unique distributed structure of IIoT, congestion caused by excessive burst data or heavy load on a single node can cause the entire network to crash. Therefore, to effectively monitor and mitigate this potential congestion risk, existing research mainly focuses on two methods: network parameter monitoring and time series modeling. However, comprehensive collection of information from all nodes in the entire network leads to high redundancy and heavy processing burden, which affects detection efficiency and decision-making timeliness, thereby reducing production performance and even causing accidents. Summary of the Invention

[0003] This application provides a congestion detection method, apparatus, device, and storage medium. CS reduces monitoring costs and LSTMN enhances time dimension discrimination capabilities, achieving low-redundancy, high-sensitivity congestion perception and assisting in link selection and task scheduling, thereby improving industrial production efficiency and system stability. It is suitable for efficient congestion monitoring needs in complex scenarios.

[0004] In a first aspect, the present application provides a congestion detection method, the method comprising:

[0005] Get T S samples; wherein the t-th sample includes the congestion status of P paths in the network link system, where t=1,…,T S , the T S represents the number of samples, the network link system includes L links; the congestion state of the p-th path is determined according to the congestion states corresponding to the multiple time slots included in the t-th sample;

[0006] Constructing a routing matrix based on the relationship between the P paths and the L links in the network link system, and determining a constraint condition based on the fact that the congestion state of the path depends on the congestion state of the link and the routing matrix;

[0007] Determine the congestion probability of the pth path at the tth sample based on the congestion status corresponding to the multiple time slots

[0008] According to the congestion probability The initial congestion probability of the lth link at the tth sample is determined by the constraint condition

[0009] According to the initial congestion probability A historical congestion information set is constructed, and the historical congestion information set is input into a pre-constructed long short-term memory network (LSTMN) model to obtain a predicted congestion state of the lth link.

[0010] In one or more possible embodiments, the congestion state of the p-th path is determined according to the congestion states corresponding to the multiple time slots included in the t-th sample, including:

[0011] Get the pth path at the tth sample T M T corresponding to the time slot M Packet loss rate;

[0012] For a packet loss rate corresponding to any time slot, if the packet loss rate is greater than or equal to a path packet loss rate threshold of the p-th path, then the congestion state of the p-th path in the any time slot is determined to be congested; otherwise, the congestion state of the p-th path in the any time slot is determined to be non-congested;

[0013] According to the T M T corresponding to the time slot M congestion status of the p-th path, and determining the congestion status of the p-th path.

[0014] In one or more possible embodiments, the congestion probability of the p-th path at the t-th sample is determined based on the congestion states corresponding to the multiple time slots. include:

[0015] According to the T M T corresponding to the time slot M congestion state, determine the pth path in T M The congestion state in a time slot is the number of congestions;

[0016] According to the p-th path in T M The congestion state in a time slot is the number of congested times and the T M The ratio of the pth path to the tth sample determines the congestion probability of the pth path in the tth sample.

[0017] In one or more possible embodiments, the congestion probability The initial congestion probability of the lth link at the tth sample is determined by the constraint condition include:

[0018] Using the compressed sensing algorithm, the congestion probability and initial congestion probability Perform logarithmic transformation;

[0019] According to the congestion probability after logarithmic transformation and initial congestion probability Determine the initial congestion probability of the lth link that meets the constraint condition at the tth sample

[0020] In one or more possible embodiments, according to the initial congestion probability Construct a historical congestion information set, including:

[0021] According to the initial congestion probability Determining an initial congestion state of the lth link within a time window;

[0022] A historical congestion information set of the lth link is determined according to the initial congestion state of the lth link and the duration of the sliding window.

[0023] In one or more possible embodiments, inputting the historical congestion information set into a pre-built long short-term memory network (LSTMN) model to obtain a predicted congestion state of the lth link includes:

[0024] Expanding the historical congestion information set of the lth link according to the duration of the sliding window, wherein the initial congestion state in the expanded historical congestion information set is divided into a historical initial congestion state and a predicted congestion state;

[0025] The historical initial congestion state is input into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion state of the lth link.

[0026] In one or more possible embodiments, the pre-built LSTMN model includes: an input gate, an output gate, and a forget gate;

[0027] Inputting the historical initial congestion state into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion state of the lth link includes:

[0028] The historical initial congestion state is concatenated with the implicit state output by the output gate at the previous moment to form a joint input vector;

[0029] Determining an output of the forget gate according to the joint input vector, the weight matrix of the forget gate, and the bias vector of the forget gate;

[0030] Determining an output of the input gate according to the output of the forget gate, the joint input vector, a weight matrix of the input gate, and a bias vector of the input gate;

[0031] determining an output of the output gate according to the output of the input gate, the joint input vector, a weight matrix of the output gate, and a bias vector of the output gate;

[0032] According to a preset weight vector, the output of the output gate is mapped to the predicted congestion state of the lth link.

[0033] In a second aspect, the present application provides a congestion detection device, the device comprising:

[0034] Sample acquisition module, used to obtain T S samples; wherein the t-th sample includes the congestion status of P paths in the network link system, where t=1,…,T S , the T S represents the number of samples, the network link system includes L links; the congestion state of the p-th path is determined according to the congestion states corresponding to the multiple time slots included in the t-th sample;

[0035] a condition determination module, configured to construct a routing matrix based on the relationship between the P paths and the L links in the network link system, and determine a constraint condition based on the fact that the congestion state of a path depends on the congestion state of the link and the routing matrix;

[0036] A path congestion probability determination module is used to determine the congestion probability of the p-th path at the t-th sample based on the congestion states corresponding to the multiple time slots.

[0037] An initial congestion probability determination module is configured to determine the initial congestion probability based on the congestion probability. The initial congestion probability of the lth link at the tth sample is determined by the constraint condition

[0038] The predicted congestion state determination module is used to determine the initial congestion probability A historical congestion information set is constructed, and the historical congestion information set is input into a pre-constructed long short-term memory network (LSTMN) model to obtain a predicted congestion state of the lth link.

[0039] In a third aspect, the present application provides an electronic device, comprising:

[0040] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any one of the methods in the first aspect.

[0041] In a fourth aspect, the present application further provides a computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is used to enable a computer to execute any one of the methods in the first aspect.

[0042] According to the congestion detection method, device, equipment and storage medium provided in this application, CS is used to reduce monitoring costs, LSTMN is used to enhance the time dimension discrimination capability, achieve low-redundancy and high-sensitivity congestion perception, assist in link selection and task scheduling, thereby improving industrial production efficiency and system stability, and is suitable for efficient congestion monitoring needs in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0044] Figure 1 A flowchart of a congestion detection method provided according to an embodiment;

[0045] Figure 2 A schematic diagram of a link scenario provided according to an embodiment;

[0046] Figure 3 A schematic diagram of the specific structure of an LSTMN model provided according to an embodiment;

[0047] Figure 4 A schematic diagram of a congestion detection device provided according to an embodiment;

[0048] Figure 5 A schematic diagram of an electronic device provided according to an embodiment;

[0049] Figure 6 The present invention is a schematic diagram of a computer storage medium provided according to an embodiment. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It is apparent that the embodiments described are only some, not all, of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0051] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure as detailed in the appended claims.

[0052] Moreover, in the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0053] For ease of understanding, the terms involved in the embodiments of the present invention are explained below:

[0054] Compressed Sensing (CS) is an emerging signal processing theory that involves the technology of recovering the original signal from a small number of measurements under certain conditions. The core idea of ​​this technology is that if a signal is sparse or can be represented as a sparse vector (that is, the signal has only a few non-zero elements in a certain basis or dictionary), then this signal can be accurately reconstructed using a sampling rate far lower than that required by the traditional Nyquist sampling theorem.

[0055] LSTM (Long Short-Term Memory, LSTM) is a time recursive neural network that is suitable for processing and predicting important events in time series with relatively long intervals and delays. LSTM is a special recurrent neural network proposed to solve the gradient vanishing problem in the recurrent neural network QRNN structure.

[0056] With the rapid development of fifth-generation (5G) networks, vertical applications of the Internet of Things (IoT) are attracting increasing attention in research and industry. Specifically, the Industrial Internet of Things (IIoT) is considered a promising candidate technology for meeting the efficiency requirements of production automation. In IIoT systems, large-scale industrial equipment generates large amounts of real-time data that must be processed under strict constraints. Unexpected delays or task interruptions can lead to significant safety issues and significant production losses. Traditionally, industrial data is transmitted to a centralized cloud server (CCS) for processing. However, given the distributed topology of large-scale IIoT devices, CCS-based processing struggles to meet the latency requirements of time-sensitive services. To address this issue, mobile edge computing (MEC) reduces data transmission latency by allowing IIoT devices to offload processing tasks to nearby edge servers (ES). While ES has limited computing power compared to CCS, it can shorten task response times, thereby ensuring the security and effectiveness of real-time IIoT systems.

[0057] At the same time, given the massive amounts of data generated by large-scale industrial equipment, congestion detection in the end-to-end collaboration process has attracted widespread attention. It is worth noting that, given the unique distributed structure of IIoT end-to-end collaboration, congestion at a single node due to excessive burst data or data overload can lead to the collapse of the entire network. Current research can be divided into two major categories: On the one hand, congestion detection is usually completed by detecting network parameters of congestion signals. Specifically, whether the network is congested is determined based on parameters such as instantaneous queue size, average queue size, average sending rate, arrival rate, link utilization, and cache occupancy. On the other hand, based on the transmission properties of data packets in TCP / IP networks, time series methods are used to analyze network traffic, establish relevant models, and use recursive least squares methods to estimate relevant parameters and determine the corresponding congestion detection algorithm.

[0058] However, the two traditional congestion detection methods mentioned above do not take into account the sparsity of the congestion state of industrial IIoT. That is, in normal industrial production, only a few nodes experience congestion at irregular intervals with a certain probability, and the status of other nodes has a relatively low correlation with the congestion state. Therefore, the process of collecting information from all nodes in the entire network may generate a large amount of redundancy, which is not conducive to efficient congestion detection. This may lead to untimely detection and judgment decisions, affecting the production performance of the IIoT industrial network, and further leading to reduced industrial production capacity or even production accidents.

[0059] Therefore, this application proposes a congestion detection method, device, equipment and storage medium, which combines compressed sensing (CS) theory and estimates the congestion probability of each link based on the sparsity of the congestion state, reduces the number of sensor deployments, and improves the flexibility of the solution; and based on the preliminary results of CS, introduces a long short-term memory network (LSTMN) to extract the time correlation of the link congestion state, further improving the accuracy of congestion detection; in this application, CS is first used to obtain a preliminary estimate of the link congestion, and then the discrimination ability in the time dimension is enhanced with the help of LSTMN to obtain the final prediction result of the link congestion, which can reduce the monitoring cost and improve the accuracy of congested link detection; in addition, this solution supports centralized deployment on a cloud server (CCS) to assist in network-wide scheduling, and can also be distributedly deployed on end nodes (EN) to optimize local transmission path selection, which is suitable for efficient congestion monitoring needs in various IIoT scenarios.

[0060] Example 1

[0061] This application provides a congestion detection method, such as Figure 1 Shown, including:

[0062] Step 101, obtain T S samples; where the tth sample includes the congestion status of P paths in the network link system, and the above t=1,…,T S , the above T S represents the number of samples, the network link system includes L links; the congestion state of the p-th path is determined based on the congestion states corresponding to the multiple time slots included in the t-th sample;

[0063] Step 102: construct a routing matrix based on the relationship between the P paths and the L links in the network link system, and determine the constraint conditions based on the fact that the congestion status of the paths depends on the congestion status of the links and the routing matrix.

[0064] In one or more possible embodiments, specifically Figure 2 As shown, this is a schematic diagram of a link scenario provided by this application, including T S samples, each sample contains T M measurement time slots, where the congestion state of each sample is T M The above sampling method is used to reduce the impact of random fluctuations on link congestion detection; Figure 2 As shown, for example, the network link system includes N nodes, which can be user equipment, switches, routers, etc., and also includes L links and P paths, where P represents the number of paths from the input end to the output end of the measurement, and L represents the number of links from node to node; and Figure 2It can be seen that a path contains at least one link, and it can be concluded that P < L; for the t-th sample, a routing matrix is constructed denotes a matrix with elements R pl , with a size of P × L; the above routing matrix is a binary routing matrix, which is obtained according to the network routing topology relationship of the above network link system. If it means that the p-th path passes through the l-th link. If the p-th path does not pass through the l-th link, then Since the path from the input end to the output end is composed of different links, the congestion state of the path also reflects the congestion state of the link; for the t-th sample (i.e., the t-th sampling time), use to represent the congestion state of link l, and use​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​packet loss rate; for the packet loss rate corresponding to any time slot, if the packet loss rate is greater than or equal to the path packet loss rate threshold of the p-th path, the congestion state of the p-th path in the above-mentioned any time slot is determined to be congested; otherwise, the congestion state of the p-th path in the above-mentioned any time slot is determined to be non-congested; according to the above T M T corresponding to the time slot M congestion status, and determine the congestion status of the p-th path.

[0068] In one or more possible embodiments, for the t-th sample, as Figure 2 As shown, the congestion state of the pth path can be further divided into two types based on T M The congestion state of the time slot is expressed as Indicates that the p-th path is in the T-th M The congestion state of the time slot; the above Determined in the following way:

[0069]

[0070] Among them, λ P,p is the path packet loss rate threshold of the p-th path, is the sample t at the Tth M The packet loss rate of the pth path in the time slot is the real data collected; where the above λ P,p It is based on the link packet loss rate threshold λ L The number of links contained in the path p is also called the depth of the path p, which is represented by d. The specific calculation formula is: P,p =1-(λ L ) d ; Determine T M After determining the congestion status of the p-th path, the congestion status of the p-th path in sample t can be determined according to the preset determination conditions. For example, if T M If more than half of the congestion states of the p-th path are congested, the congestion state of the p-th path in sample t can be determined, or if T M If m consecutive congestion states in the congestion states of the p-th path are congested, the congestion state of the p-th path in sample t can be determined; other methods can also be used, which will not be specifically described.

[0071] Step 103: Determine the congestion probability of the p-th path in the t-th sample according to the congestion status corresponding to the multiple time slots.

[0072] In one or more possible embodiments, according to the above TM T corresponding to the time slot M congestion state, determine the pth path above in T M The congestion state in the time slot is the number of congestions; according to the above p-th path in T M The congestion state in a time slot is the number of congested times and the above T M The ratio of t to p determines the congestion probability of the p-th path in the t-th sample above. The specific method is as follows: According to T M The p-th path above is in the T M Congestion status of time slots Calculate the congestion probability of path p at the tth sample The specific calculation method is the following formula (2):

[0073]

[0074] Step 104: Based on the above congestion probability And the above constraints determine the initial congestion probability of the lth link in the above tth sample

[0075] In one or more possible embodiments, according to the above formula (1), after determining the congestion state of path p and the above routing matrix After that, the congestion status of link l can be Calculation is performed, but due to the R (t) is not a full rank matrix, so It is unsolvable, so this application introduces the CS algorithm to deal with the above problem; specifically: using the compressed sensing algorithm, the congestion probability and initial congestion probability Perform logarithmic transformation; according to the congestion probability after logarithmic transformation and initial congestion probability Determine the initial congestion probability of the lth link that meets the above constraints in the tth sample above

[0076] In one or more possible embodiments, the path-based congestion status depends on the congestion status of the link, and the congestion probability of path p at the tth sample is calculated based on the congestion probability of path p at the tth sample. as well as Determine the initial congestion probability of the lth link at the tth sample above The calculation method is the following formula (3):

[0077]

[0078] In order to solve the The unsolvable problem is solved by using the compressed sensing algorithm CS, which transforms the problem into the problem of reconstructing the congestion probability of the link based on the congestion probability of the path. Then, the unsolvable problem caused by insufficient matrix rank is solved as follows:

[0079] First, let Transform the above problem into a solution:

[0080]

[0081] Then construct an optimization problem to solve the congestion probability of the link That is, according to the principle of formula (3), the initial congestion probability of the lth link in the above tth sample is solved: (i.e., the estimated initial congestion probability ), solve formula (4) as the constraint condition of the following formula (5), the specific formula is as follows:

[0082]

[0083] In the above formula (5) Indicates non-zero Minimize the number of It means to find the minimum number of congested link combinations, and make the congestion probability of these links fully explain the observed path congestion. Finally, according to Reversely infer the initial congestion probability of the lth link in the tth sample (i.e., the estimated initial congestion probability ).

[0084] Step 105: Based on the initial congestion probability A historical congestion information set is constructed and input into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion status of the lth link.

[0085] In one or more possible embodiments, traditional LSTMN is widely used in the field of natural language processing (NLP) to analyze the correlation between consecutive words / sentences. It is worth noting that the correlation characteristics exhibited by the continuous trajectory formed by the movement of the terminal in space and the correlation characteristics exhibited by the change in traffic volume are similar to the correlation characteristics exhibited by the continuity of words / sentences in NLP. Therefore, this application applies LSTMN to the congestion status prediction solution. The specific process is as follows:

[0086] Get the estimated initial congestion probability Then, based on the above estimated initial congestion probability (Right now ), determine the initial congestion state of the lth link within the time window;

[0087] Determine a historical congestion information set of the lth link according to the initial congestion state of the lth link and the duration of the sliding window;

[0088] According to the duration of the sliding window, the historical congestion information set of the lth link is expanded, and the initial congestion state in the expanded historical congestion information set is divided into a historical initial congestion state and a predicted congestion state;

[0089] The above historical initial congestion state is input into the pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion state of the lth link.

[0090] In one or more possible embodiments, according to the above initial congestion probability The following formula (6) is used to determine the initial congestion state of the lth link in the time window:

[0091]

[0092] Among them, t in formula (6) LSTMN With the above initial congestion probability The t in the example is the same. To indicate that the purpose of this step is to make predictions based on the long short-term memory network LSTMN model, t LSTMN =1,…,T LSTMN , the above T LSTMN Represents the number of training data, and T S Same; in formula (6) Indicates that the CS algorithm is used in the process of prediction in the LSTMN model according to the above initial congestion probability The lth link determined above is in time window t LSTMN Initial congestion state The initial congestion state mentioned above It is 1 or 0. "1" indicates congestion and "0" indicates non-congestion.

[0093] According to the initial congestion state obtained in the above formula (6) and the duration of the sliding window, determine the historical congestion information set of the lth link, as shown in the following formula (7):

[0094]

[0095] N in the above formula (7) W is the duration of the sliding time window used to extract relevant congestion information, []T represents the transposition; According to the above formula (7), it can be seen that is a (N W -1)×1 vector; in order to predict the congestion state of the lth link at the next moment, the elements in formula (7) are expanded, that is, Adding ℓ as the last element to the above formula (7) yields the expanded historical congestion information set of the lth link; at this time, Become an N W ×1 vector, the first N W -1 element is used as historical information, and the last element is used as current information. It is worth noting that during the LSTMN model training process, the above historical information and the above current information are both known information that can be calculated. The LSTMN model is trained using the historical information as input and the current information as output. After the LSTMN model training is completed, when the LSTMN model is used to predict the congestion status of the link at the next moment, the last element is unknown information and needs to be predicted based on the historical information.

[0096] In one or more possible embodiments, the pre-built LSTMN model includes: an input gate, an output gate, and a forget gate. The specific structural diagram of the LSTMN model is as follows: Figure 3 As shown, for the convenience of expression, (t LSTMN +N W -1) is replaced by μ; the LSTMN model can be regarded as a time-related neural cell structure, each cell contains a forget gate, an input gate and an output gate, and each gate contains N G hidden units and use and Define the tth LSTMN The output of the forget gate, input gate, and output gate in each cell;

[0097] Specifically, such as Figure 3 As shown, the input gate contains two gate structures, that is, the weight matrix is ​​W I , the bias vector is b I The gate based on the sigmoid function, and the weight matrix is ​​W C , the bias vector is b C The gate based on the tanh function, in addition, the weight matrix of the forget gate and the output gate is defined as W F and W O , and define its bias vector as b F and b O , the size of each gate’s weight matrix and bias vector is N G ×(N W -1+N G ) and N G ×1, define wS The size is 1×N G The preset weight vector, at the same time, before the prediction starts W F , W I , W C , W O and w S Gaussian distribution N(0,0.01) is used for initialization b F , b I , b C and b O All are initialized to zero vector 0;

[0098] The specific process is as follows Figure 3 As shown, including the forget gate, the input and output flow of the input gate and the output gate, the connection structure between multiple gates, and the specific operations in the process, the above historical initial congestion state is spliced ​​with the implicit state of the output gate at the previous moment to form a joint input vector

[0099]

[0100] The [·,·] in formula (9) represents the vector concatenation operation, which can be obtained with a size of (N W -1+N G )×1 vector;

[0101] Determine the output of the forget gate according to the joint input vector, the weight matrix of the forget gate, and the bias vector of the forget gate;

[0102]

[0103] Determine the output of the input gate according to the output of the forget gate, the joint input vector, the weight matrix of the input gate, and the bias vector of the input gate;

[0104]

[0105] Among them, ★ represents the multiplication between two vector elements of the same size: [a i ] l×1 ★[b i ] l×1 =[a1b1,…,a l b l ] T ;

[0106] Determining an output of the output gate according to the output of the input gate, the joint input vector, the weight matrix of the output gate, and the bias vector of the output gate;

[0107]

[0108] Among them, the output result of the output gate, that is, It can be calculated by iterating equations (9) to (12), where the initialization process is:

[0109] According to the preset weight vector, the output of the above output gate is mapped to the predicted congestion status of the lth link;

[0110] final The prediction result can be calculated as:

[0111]

[0112] Among them, l in formula (9)-formula (13) is Figure 3 It is represented by n, which is the and Compared with the formula (9)-formula (13) and Those with the same meaning are not listed here one by one;

[0113] Similarly, the congestion status of L links in the network link system is obtained using the above method to determine the final output Is 1 or 0? If it is 1, it means the predicted congestion status of the l-th link is congested. If it is 0, it means the congestion status of the l-th link is non-congested.

[0114] According to a congestion detection method provided in this application, the problem of detection lag caused by the traditional method under the IIoT distributed structure not considering the sparsity and time correlation of the congestion state is solved; this application first uses CS theory to make a preliminary estimate of the link congestion probability based on the sparsity of congestion, effectively reducing the number of sensor deployments and improving flexibility. Combined with the long short-term memory network (LSTMN), it further extracts the time-related features of the link congestion state and improves detection accuracy; CS reduces monitoring costs, LSTMN enhances the time dimension discrimination capability, realizes low-redundancy, high-sensitivity congestion perception, assists link selection and task scheduling, thereby improving industrial production efficiency and system stability, and is suitable for efficient congestion monitoring needs in complex scenarios.

[0115] Example 2

[0116] Corresponding to the above congestion detection method, the present invention also provides a congestion detection device. Since the device embodiment of the present invention corresponds to the above method embodiment, details not disclosed in the device embodiment can be referred to the above method embodiment and will not be repeated in the present invention.

[0117] This application provides a congestion detection device, such as Figure 4 As shown, the above device includes:

[0118] Sample acquisition module 401, used to obtain T S samples; where the tth sample includes the congestion status of P paths in the network link system, and the above t=1,…,T S , the above T S represents the number of samples, the network link system includes L links; the congestion state of the p-th path is determined based on the congestion states corresponding to the multiple time slots included in the t-th sample;

[0119] Condition determination module 402 is configured to construct a routing matrix based on the relationship between the P paths and the L links in the network link system, and determine a constraint condition based on the fact that the congestion status of a path depends on the congestion status of a link and the routing matrix;

[0120] The path congestion probability determination module 403 is used to determine the congestion probability of the p-th path in the t-th sample according to the congestion status corresponding to the multiple time slots.

[0121] The initial congestion probability determination module 404 is configured to determine the initial congestion probability based on the above congestion probability. And the above constraints determine the initial congestion probability of the lth link in the above tth sample

[0122] The predicted congestion state determination module 405 is used to determine the initial congestion probability according to the above A historical congestion information set is constructed and input into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion status of the lth link.

[0123] Example 3

[0124] The present application also provides an electronic device comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so as to enable the at least one processor to perform the above-mentioned congestion detection method.

[0125] like Figure 5 As shown, the device includes a processor 501 , a memory 502 , a communication interface 503 and a bus 504 . The processor 501 , the memory 502 and the communication interface 503 are interconnected via the bus 504 .

[0126] The processor 501 is configured to read and execute instructions in the memory 502 , so that at least one processor can execute the congestion detection method provided in the above embodiment.

[0127] The memory 502 is used to store various instructions and programs of the congestion detection method provided in the above embodiment.

[0128] The bus 504 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0129] The processor 501 may be a central processing unit (CPU), a network processor (NP), a graphic processing unit (GPU), or any combination of a CPU, NP, and GPU. It may also be a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0130] Example 4

[0131] In addition, the present application also provides a computer-readable storage medium, such as Figure 6 As shown, the computer storage medium stores a computer program, and the computer program is used to enable a computer to execute any one of the methods in the above embodiments.

[0132] The memory may include readable media in the form of volatile memory, such as random access memory (RAM) 601 and / or cache memory 602 , and may further include read-only memory (ROM) 603 .

[0133] The memory may also include a program / utility 605 having a set (at least one) of program modules 604, such program modules 604 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0134] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0135] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0136] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0137] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0138] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A congestion detection method, characterized in that: The method comprises: Get T S samples; wherein the tth sample includes the congestion status of P paths in the network link system, where t=1,…,T S , the T S represents the number of samples, the network link system includes L links; the congestion state of the p-th path is determined according to the congestion states corresponding to the multiple time slots included in the t-th sample; Constructing a routing matrix based on the relationship between the P paths and the L links in the network link system, and determining a constraint condition based on the fact that the congestion state of the path depends on the congestion state of the link and the routing matrix; Determine the congestion probability of the pth path at the tth sample based on the congestion status corresponding to the multiple time slots According to the congestion probability The initial congestion probability of the lth link at the tth sample is determined by the constraint condition According to the initial congestion probability A historical congestion information set is constructed, and the historical congestion information set is input into a pre-constructed long short-term memory network (LSTMN) model to obtain a predicted congestion state of the lth link.

2. The method according to claim 1, characterized in that The congestion state of the p-th path is determined according to the congestion states corresponding to the multiple time slots included in the t-th sample, including: Get the pth path at the tth sample T M T corresponding to the time slot M Packet loss rate; For a packet loss rate corresponding to any time slot, if the packet loss rate is greater than or equal to a path packet loss rate threshold of the p-th path, then the congestion state of the p-th path in the any time slot is determined to be congested; otherwise, the congestion state of the p-th path in the any time slot is determined to be non-congested; According to the T M T corresponding to the time slot M congestion status of the p-th path, and determining the congestion status of the p-th path.

3. The method according to claim 2, characterized in that The congestion probability of the p-th path in the t-th sample is determined according to the congestion states corresponding to the multiple time slots. include: According to the T M T corresponding to the time slot M congestion state, determine the pth path in T M The congestion state in a time slot is the number of congestions; According to the p-th path in T M The congestion state in a time slot is the number of congested times and the T M The ratio of the pth path to the tth sample determines the congestion probability of the pth path in the tth sample.

4. The method according to claim 1, wherein The congestion probability The initial congestion probability of the lth link at the tth sample is determined by the constraint condition include: Using the compressed sensing algorithm, the congestion probability and initial congestion probability Perform logarithmic transformation; According to the congestion probability after logarithmic transformation and initial congestion probability Determine the initial congestion probability of the lth link that meets the constraint condition at the tth sample 5. The method according to any one of claims 1 to 4, characterized in that: According to the initial congestion probability Construct a historical congestion information set, including: According to the initial congestion probability Determining an initial congestion state of the lth link within a time window; A historical congestion information set of the lth link is determined according to the initial congestion state of the lth link and the duration of the sliding window.

6. The method according to claim 5, characterized in that Inputting the historical congestion information set into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion state of the lth link includes: Expanding the historical congestion information set of the lth link according to the duration of the sliding window, wherein the initial congestion state in the expanded historical congestion information set is divided into a historical initial congestion state and a predicted congestion state; The historical initial congestion state is input into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion state of the lth link.

7. The method according to claim 1, characterized in that The pre-built LSTMN model includes: an input gate, an output gate, and a forget gate; Inputting the historical initial congestion state into a pre-built long short-term memory network (LSTMN) model to obtain the predicted congestion state of the lth link includes: The historical initial congestion state is concatenated with the implicit state output by the output gate at the previous moment to form a joint input vector; Determining an output of the forget gate according to the joint input vector, the weight matrix of the forget gate, and the bias vector of the forget gate; Determining an output of the input gate according to the output of the forget gate, the joint input vector, a weight matrix of the input gate, and a bias vector of the input gate; determining an output of the output gate according to the output of the input gate, the joint input vector, a weight matrix of the output gate, and a bias vector of the output gate; According to a preset weight vector, the output of the output gate is mapped to the predicted congestion state of the lth link.

8. A congestion detection device, characterized in that: The device comprises: Sample acquisition module, used to obtain T S samples; wherein the tth sample includes the congestion status of P paths in the network link system, where t=1,…,T S , the T S represents the number of samples, the network link system includes L links; the congestion state of the p-th path is determined according to the congestion states corresponding to the multiple time slots included in the t-th sample; a condition determination module, configured to construct a routing matrix based on the relationship between the P paths and the L links in the network link system, and determine a constraint condition based on the fact that the congestion state of a path depends on the congestion state of the link and the routing matrix; A path congestion probability determination module is used to determine the congestion probability of the p-th path at the t-th sample based on the congestion states corresponding to the multiple time slots. An initial congestion probability determination module is configured to determine the initial congestion probability based on the congestion probability. The initial congestion probability of the lth link at the tth sample is determined by the constraint condition The predicted congestion state determination module is used to determine the initial congestion probability A historical congestion information set is constructed, and the historical congestion information set is input into a pre-constructed long short-term memory network (LSTMN) model to obtain a predicted congestion state of the lth link.

9. An electronic device, characterized in that: The electronic device comprises: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any one of the methods of claims 1-7.

10. A computer storage medium, characterized in that The computer storage medium stores a computer program, and the computer program is used to make a computer execute the method according to any one of claims 1 to 7.

Images