An intelligent rateless coding system and method for industrial scenarios

Abstract

The present invention discloses an intelligent rateless coding system and method for industrial scenarios, including an industrial application management module for distinguishing between elastic services and inelastic services in industrial scenarios; an application data sending module containing data streams in different industrial application scenarios, including from industrial application 1 to industrial application n, for different business data streams, which are sent to an encoder module for encoding; an encoder module for encoding the original data stream according to a degree value provided by a rateless coding intelligent agent module; a channel including a binary deletion channel and a wireless channel for transmitting the data stream encoded by the encoder module; a decoder module for decoding the coding symbols transmitted via the channel to obtain the original data stream; a rateless coding intelligent agent module for providing a degree value to the encoder module through a deep reinforcement learning algorithm based on the data stream state decoded by the decoder module; and an application data receiving module for receiving the data stream output by the decoder module.

Description

Technical Field

[0001] The present invention relates to the field of trusted and secure coding communication technology, and in particular to an intelligent rateless coding system and method for industrial scenarios. Background Art

[0002] With the further development of communications and artificial intelligence technologies, the industrial sector is gradually entering an era of intelligent production. To achieve interconnection, interoperability, and shared collaboration between industrial resources and capabilities and the information sector, information technology and operational technology are gradually converging. Enterprise systems and control systems are transitioning from an internationally standardized five-tier architecture to an intelligent three-tier structure consisting of cloud computing platforms, edge computing platforms, and end devices. Within this "cloud-edge-end" architecture, industrial control algorithms, previously running on local computing devices, can be seamlessly migrated to cloud or edge computing platforms leveraging the computing and storage capabilities of cloud computing. This enables a variety of factory-, workshop-, and field-level applications and services under the umbrella of Industry 4.0. For example, industrial data can be stored in the cloud and intelligent algorithms such as AI can be used for large-scale data processing. However, industrial scenarios place stringent requirements on the real-time and reliability of point-to-point communication between the cloud, edge, and end. Furthermore, the packet sizes and communication requirements vary significantly across different scenarios.

[0003] First, the real-time performance of industrial scenarios focuses more on the determinism and boundedness of the end-to-end communication delay. That is, non-elastic services such as motion control and machine tools in industrial scenarios have a fixed upper limit on delay. Once the data transmission exceeds the upper limit on delay, it will be discarded. Elastic services such as workshop video surveillance and industrial robot visual information in industrial scenarios do not have fixed delay bounds. Even if the data transmission exceeds the upper limit on delay, it can still be received.

[0004] Another important indicator in communication transmission is reliability, which emphasizes that end-to-end communication should correctly transmit as much business data as possible while meeting certain delay constraints.

[0005] For the above-mentioned "cloud-edge-end" architecture, rateless coding is used as the application layer coding technology. The core component of rateless coding is the degree distribution (a probability distribution function formed by randomly sampling a degree value d for each coded symbol). For elastic services and inelastic services in industrial scenarios, the size of data packets, transmission delay, transmission reliability, etc. vary greatly. If a fixed degree distribution is used, it is difficult to meet the needs of various services in the industrial field. When the business scenario is switched, the flexibility is poor. In particular, for inelastic services under bounded delay constraints, it is necessary to meet the constraints of deterministic delay and ensure high-reliability transmission as much as possible. Summary of the Invention

[0006] In order to overcome the deficiencies in the prior art, the present invention provides an intelligent rateless coding system and method for industrial scenarios.

[0007] In order to achieve the above-mentioned purpose of the invention, the technical solutions adopted to solve the technical problems are as follows:

[0008] The present invention discloses an intelligent rateless coding system for industrial scenarios, comprising:

[0009] Industrial application management module, used to distinguish between elastic and non-elastic services in industrial scenarios;

[0010] An application data sending module, which includes motion control signal data in industrial scenarios, communication data between different controllers, and real-time data streams of video surveillance, including different business data streams from industrial application 1, industrial application 2 to industrial application n, used to distinguish different business data streams of the industrial application management module, and sends them to the encoder module for encoding;

[0011] An encoder module, configured to encode a data stream of a specific application according to a degree value provided by the rateless encoding intelligent agent module;

[0012] Channels, including binary erasure channels and wireless channels, for transmitting the data stream encoded by the encoder module;

[0013] The decoder module is used to decode the encoded symbols transmitted via the channel to obtain the original data stream of the application sending end;

[0014] A rateless coding intelligent agent module is used to return a reward for the decoded data stream based on the data stream state decoded by the decoder module. Through the deep reinforcement learning algorithm of the rateless coding intelligent agent module, the reward value is continuously increased, and the optimal degree value is provided to the encoder module, so that the entire rateless coding system meets the requirements of latency and reliability;

[0015] The application data receiving module is used to receive the data stream output by the decoder module.

[0016] Furthermore, the QoS parameters included in each industrial application include packet size, bounded latency, and reliability.

[0017] Furthermore, the decoder module includes a belief propagation algorithm and a Gaussian elimination algorithm. The combination of the two decoding algorithms is called deactivation decoding. The decoding status is divided into three cases:

[0018] Decoding failed: The generator matrix A corresponding to the original data to the decoded data is not full rank;

[0019] Loss of activity decoding successful: Decoding is successful using the loss of activity decoding algorithm;

[0020] Belief propagation decoding is successful: decoding is successful only using the BP algorithm.

[0021] The present invention further discloses an intelligent rateless encoding method for industrial scenarios. If the industrial application is a flexible application, there is no need to use the rateless encoding intelligent agent module to optimize and adjust the degree distribution in real time. Only conventional encoding and decoding are required. The method includes the following steps:

[0022] Step A1: Determine QoS parameters for elastic applications;

[0023] Step A2: transmitting the data stream of the application data sending module to the encoder module and encoding it according to the degree value provided by the general degree distribution;

[0024] Step A3: The data output by the encoder module is output via a channel and sent to the decoder module;

[0025] Step A4: The decoder module selects the BP decoding algorithm for decoding;

[0026] Step A5: The decoder module sends the decoded data to the application data receiving module.

[0027] The present invention also discloses an intelligent rateless coding method for industrial scenarios. If the industrial application is a non-elastic application, it is necessary to optimize and adjust the intelligence distribution in real time through a rateless coding intelligent agent module, including the following steps:

[0028] Step B1: Determine the QoS parameters of the non-elastic application, and determine the data packet size, bounded delay, and reliability of the data flow of the application data sending module for different industrial applications;

[0029] Step B2: transmitting the data stream of the application data sending module to the encoder module for encoding;

[0030] Step B3: The data output by the encoder module is output via a channel and sent to the decoder module;

[0031] Step B4: The decoder module selects an appropriate decoding algorithm based on the characteristics of the data stream. At the same time, the number of successful decodings using the BP algorithm is recorded.

[0032] At the same time, the reward value r is output to the rateless coding intelligent agent module according to the decoding status. The reward is calculated as follows:

[0033]

[0034] Where rank(A) is the rank of the generator matrix A, and p is the decoding success rate;

[0035] Step B5: The rateless coding intelligent agent module performs neural network training;

[0036] Step B6: Obtain the intelligent degree distribution function π and provide the degree value to the encoder module;

[0037] Step B7: The decoder module sends the decoded data to the application data receiving module.

[0038] Furthermore, step B5 includes the following steps:

[0039] Step B51: Initialize the parameters of the action network and the value network: the parameters τ of the online policy network π , the parameter τ′ of the target policy network π , parameter τ of the online V network v and the parameter τ′ of the target V network v ;

[0040] Step B52: Set the time when the data stream is generated by the application sending end to T=0;

[0041] Step B53: Set the derivative dτ′ of the network parameter π , dτ′ v :dτ′ π ←τ π , dτ′ v ←τ v ;

[0042] Step B54: Obtain the encoding status at different times s t(t=1, 2, 3...);

[0043] Step B55: The rateless coding intelligent proxy module generates the degree value d t , and record the probability distribution function of each degree value as π;

[0044] Step B56: Determine whether the encoding state exceeds the bounded delay Γ or has been decoded, and calculate the network reward R:

[0045]

[0046] Step B57: Gradient update:

[0047]

[0048]

[0049] Step B58: Parameter update: τ π ←τ′ π τ v ←τ′ v ;

[0050] Step B59: Repeat the network training until convergence.

[0051] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art:

[0052] 1. The present invention provides an industrial application management module in an intelligent rateless coding system for industrial scenarios, which can distinguish between elastic and inelastic services in industrial scenarios, maximizing the coding efficiency of both services. For elastic applications, normal transmission is generally performed directly through the encoder, channel, and decoder modules, and a general degree distribution (such as an ideal solitary wave distribution or a robust solitary wave distribution) is used to select the degree value. For inelastic applications, the degree value is selected using an intelligent degree distribution generated by the rateless coding intelligent agent module, so that the inelastic application meets bounded delay constraints and reliability requirements.

[0053] 2. The present invention provides a rateless coding intelligent agent module in an intelligent rateless coding system in an industrial scenario. By using a deep reinforcement learning algorithm, it can more flexibly handle various applications in industrial scenarios, so that the overall coding performance meets the bounded delay and reliability requirements.

[0054] 3. The present invention adds the decoding success rate p when setting the reward r. The characteristics of the present invention are that reliability is given a higher priority and the rank of the generator matrix and the number of times BP decoding is used are combined, so that both the coding efficiency and the decoding success rate can be reflected in the reward r.

[0055] 4. When calculating the network reward R, the present invention takes into account the two situations of exceeding the bounded delay and decoding completion, and sets the corresponding network reward R=0.

[0056] 5. When the network report R is updated, the present invention can adjust the parameter γ for different industrial applications to make the training process converge quickly and meet the communication needs of industrial applications. For example, if the data packet of the industrial application is large, it is difficult for the training process to converge quickly, and γ∈(0, 0.5] should be adjusted; if the data packet of the industrial application is small, γ∈(0.5, 1) should be adjusted. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0058] Figure 1 This is a diagram of the intelligent rateless coding system in the industrial scenario of the present invention;

[0059] Figure 2 is a deep reinforcement learning network diagram of the present invention;

[0060] Figure 3 This is a flow chart of rateless coding transmission in industrial scenarios according to the present invention. DETAILED DESCRIPTION

[0061] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0062] The present invention is aimed at the real-time and high reliability requirements for point-to-point communication between cloud, edge and end in industrial scenarios. In addition, the size of data packets sent and the communication requirements in different scenarios vary greatly. The real-time performance of industrial scenarios focuses more on the certainty and boundedness of the delay of end-to-end communication. The degree distribution has a significant impact on the overall efficiency of rateless codes. In many scenarios where the degree distribution is optimized, it is usually optimized based on complexity, overhead, translatable set size and other single performance factors. However, in industrial scenarios, the performance requirements for various communication indicators in different business fields are significantly different. Taking into account the constraints of certain delay and reliability, the present invention is based on a rateless intelligent agent module and uses deep reinforcement learning to obtain intelligent degree distribution, thereby improving the transmission performance of rateless coding in different application scenarios in the industrial field.

[0063] Example 1

[0064] like Figure 1 As shown, the present invention discloses an intelligent rateless coding system for industrial scenarios, comprising:

[0065] Industrial application management module, used to distinguish between elastic and non-elastic services in industrial scenarios;

[0066] The application data sending module includes motion control signal data in industrial scenarios, communication data between different controllers, and real-time data streams of video surveillance, including different business data streams from industrial application 1, industrial application 2 to industrial application n, which are used to distinguish different business data streams of the industrial application management module and send them to the encoder module for encoding; wherein, the QoS (Quality of Service) parameters contained in each industrial application include data packet size k, bounded delay Γ, and reliability p.

[0067] An encoder module, configured to encode a data stream of a specific application according to a degree value provided by the rateless encoding intelligent agent module;

[0068] Channels, including various transmission media such as binary erasure channels and wireless channels, for transmitting the data stream encoded by the encoder module;

[0069] The decoder module decodes the encoded symbols transmitted via the channel to obtain the original data stream for the application sender. This module uses the Belief Propagation (BP) algorithm and the Gaussian Elimination (GE) algorithm. The combination of these two decoding algorithms is called inactive decoding. The decoding status is divided into three cases:

[0070] Decoding failed: The generator matrix A corresponding to the original data to the decoded data is not full rank;

[0071] Loss of activity decoding successful: Decoding is successful using the loss of activity decoding algorithm;

[0072] Belief propagation decoding is successful: decoding is successful only using the BP algorithm.

[0073] The rateless coding intelligent agent module is used to return the reward of the decoded data stream according to the data stream state decoded by the decoder module. Through the deep reinforcement learning algorithm of the rateless coding intelligent agent module, the reward value is continuously increased and the optimal degree value is provided to the encoder module, so that the entire rateless coding system meets the requirements of latency and reliability. It contains an action network and a value network, both of which are optimized using Adam, such as Figure 2 shown.

[0074] The application data receiving module is used to receive the data stream output by the decoder module.

[0075] Example 2

[0076] like Figure 3 As shown, the present invention further discloses an intelligent rateless encoding method for industrial scenarios. If the industrial application is a flexible application, there is no need to use the rateless encoding intelligent agent module to optimize and adjust the degree distribution in real time. Only the usual encoding and decoding (using the general distribution) is required, which includes the following steps:

[0077] Step A1: Determine QoS parameters for elastic applications;

[0078] Step A2: transmitting the data stream of the application data sending module to the encoder module and encoding it according to the degree value provided by the general degree distribution;

[0079] Step A3: The data output by the encoder module is output via a channel and sent to the decoder module;

[0080] Step A4: The decoder module selects the BP decoding algorithm for decoding;

[0081] Step A5: The decoder module sends the decoded data to the application data receiving module.

[0082] Example 3

[0083] The present invention also discloses an intelligent rateless coding method for industrial scenarios. If the industrial application is a non-elastic application, it is necessary to optimize and adjust the intelligence distribution in real time through a rateless coding intelligent agent module, including the following steps:

[0084] Step B1: Determine the QoS parameters of the non-elastic application, and determine the QoS parameters such as the data packet size, bounded delay, and reliability of the data flow of the application data sending module for different industrial applications;

[0085] Step B2: transmitting the data stream of the application data sending module to the encoder module for encoding;

[0086] Step B3: The data output by the encoder module is output via a channel and sent to the decoder module;

[0087] Step B4: The decoder module selects a suitable decoding algorithm based on the characteristics of the data stream. At the same time, the number of successful decodings using the BP algorithm is recorded as n. bp ;

[0088] At the same time, the reward value r is output to the rateless coding intelligent agent module according to the decoding status. The reward is calculated as follows:

[0089]

[0090] Where rank(A) is the rank of the generator matrix A, and p is the decoding success rate;

[0091] Step B5: The rateless coding intelligent agent module performs neural network training;

[0092] Furthermore, step B5 includes the following steps:

[0093] Step B51: Initialize the parameters of the action network and the value network: the parameters τ of the online policy network π , the parameter τ′ of the target policy network π , parameter τ of the online V network v and the parameter τ′ of the target V network v ;

[0094] Step B52: Set the time when the data stream is generated by the application sending end to T=0;

[0095] Step B53: Set the derivative dτ′ of the network parameter π , dτ′ v :dτ′ π ←τ π , dτ′ v ←τ v ;

[0096] Step B54: Obtain the encoding state s at different times t (t=1, 2, 3...);

[0097] Step B55: The rateless coding intelligent proxy module generates the degree value d t , and record the probability distribution function of each degree value as π;

[0098] Step B56: Determine whether the encoding state exceeds the bounded delay Γ or has been decoded, and calculate the network reward R:

[0099]

[0100] Step B57: Gradient update:

[0101]

[0102]

[0103] Step B58: Parameter update: τ π ←τ′ π τ v ←τ′ v ;

[0104] Step B59: Repeat the network training until convergence.

[0105] Step B6: Obtaining an intelligent degree distribution function and providing a degree value to the encoder module;

[0106] Step B7: The decoder module sends the decoded data to the application data receiving module.

[0107] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An intelligent rateless coding system for industrial scenarios, characterized by: include: Industrial application management module, used to distinguish between elastic and non-elastic services in industrial scenarios; An application data transmission module, which includes motion control signal data in industrial scenarios, communication data between different controllers, and real-time data streams of video surveillance, including different business data streams from Industrial Application 1, Industrial Application 2, to Industrial Application N, used to distinguish different business data streams from the Industrial Application Management Module, and sends them to the encoder module for encoding. If the industrial application is a flexible application, there is no need for real-time optimization and degree distribution adjustment by the rateless encoding intelligent agent module, and only conventional encoding and decoding is required; An encoder module, configured to encode a data stream of an inelastic industrial application according to a degree value provided by the rateless encoding intelligent agent module; Channels, including binary erasure channels and wireless channels, for transmitting the data stream encoded by the encoder module; The decoder module is used to decode the coded symbols transmitted via the channel to obtain the original data stream of the application sender; Step B4: The decoder module selects a suitable decoding algorithm according to the characteristics of the data stream; and records the number of successful decodings using the BP algorithm as n bp ; At the same time, the reward value r is output to the rateless coding intelligent agent module according to the decoding status. The reward is calculated as follows: Where rank(A) is the rank of the generator matrix A, and p is the decoding success rate; A rateless coding intelligent agent module is configured to return a reward for decoding the data stream based on the data stream state decoded by the decoder module. The rateless coding intelligent agent module continuously increases the reward value through a deep reinforcement learning algorithm to obtain an intelligence distribution function π, and provides the optimal intelligence distribution function π to the encoder module, so that the entire rateless coding system meets the requirements of latency and reliability. The application data receiving module is used to receive the data stream output by the decoder module.

2. The intelligent rateless coding system for industrial scenarios according to claim 1, characterized in that: The QoS parameters included in each industrial application are packet size k, bounded delay Γ, and reliability p.

3. The intelligent rateless coding system for industrial scenarios according to claim 1 is characterized in that: The decoder module includes the belief propagation algorithm and the Gaussian elimination algorithm. The combination of the two decoding algorithms is called deactivation decoding. The decoding status is divided into three cases: Decoding failed: The generator matrix A corresponding to the original data to the decoded data is not full rank; Loss of activity decoding successful: Decoding is successful using the loss of activity decoding algorithm; Belief propagation decoding is successful: decoding is successful only using the BP algorithm.

4. An intelligent rateless coding method for industrial scenarios, characterized in that: The method is based on encoding and decoding in the intelligent rateless coding system in an industrial scenario according to any one of claims 1 to 3, and includes the following steps: Step A1: Determine QoS parameters for elastic applications; Step A2: transmitting the data stream of the application data sending module to the encoder module and encoding it according to the degree value provided by the general degree distribution; Step A3: The data output by the encoder module is output via a channel and sent to the decoder module; Step A4: The decoder module selects the BP decoding algorithm for decoding; Step A5: The decoder module sends the decoded data to the application data receiving module.

5. An intelligent rateless coding method for industrial scenarios, characterized in that: The method is based on the intelligent rateless coding system in the industrial scenario described in any one of claims 1-3, and performs real-time optimization and adjustment of intelligence distribution, comprising the following steps: Step B1: Determine the QoS parameters of the non-elastic application, and determine the data packet size k, bounded delay Γ, and reliability p of the data flow of the application data sending module for different industrial applications; Step B2: transmitting the data stream of the application data sending module to the encoder module for encoding; Step B3: The data output by the encoder module is output via a channel and sent to the decoder module; Step B4: The decoder module selects a suitable decoding algorithm according to the characteristics of the data stream; and records the number of successful decodings using the BP algorithm as n. bp ; At the same time, the reward value r is output to the rateless coding intelligent agent module according to the decoding status. The reward is calculated as follows: Where rank(A) is the rank of the generator matrix A, and p is the decoding success rate; Step B5: The rateless coding intelligent agent module performs neural network training; Step B6: Obtain the intelligent degree distribution function π and provide the degree value to the encoder module; Step B7: The decoder module sends the decoded data to the application data receiving module.

6. The intelligent rateless coding method in an industrial scenario according to claim 5, characterized in that: Step B5 includes the following steps: Step B51: Initialize the parameters of the action network and the value network: the parameters τ of the online policy network π , the parameter τ′ of the target policy network π , parameter τ of the online V network v and the parameter τ′ of the target V network υ ; Step B52: Set the time when the data stream is generated by the application sending end to T=0; Step B53: Set the derivative of the network parameter dτ′ π ,dτ′ υ :dτ′ π ←τ π ,dτ′ v ←τ v ; Step B54: Obtain the encoding state s at different times t (t=1,2,3…); Step B55: The rateless coding intelligent proxy module generates the degree value d t , and record the probability distribution function of each degree value as π; Step B56: Determine whether the encoding state exceeds the bounded delay Γ or has been decoded, and calculate the network reward R: Step B57: Gradient update: Step B58: Parameter update: τ π ←τ′ π τ v ←τ′ v ; Step B59: Repeat the network training until convergence.

Images