Data communication system for multi-level linkage intelligent agent deduction

Abstract

The present invention relates to the field of data communication technology, and in particular to a data communication system for multi-level linkage intelligent agent deduction. At a data acquisition end, basic data of the intelligent agent is collected and verified, and the data is processed at multiple nodes to obtain deduced data; at a data transmission end, the deduced data is encrypted to obtain encrypted data, and the encrypted data is transmitted using blockchain technology using multiple industrial control communication protocols, and the encrypted data is verified and monitored during the transmission process; at a data receiving end, the encrypted data is decrypted, and error correction is performed on the decoded data based on verification and monitoring results at the data acquisition end and the data transmission end; distributed caching technology is used to synchronize the verified data to the corresponding intelligent agent for data update, and a deep learning algorithm is used to deduce and analyze the synchronized updated data, and the deduction and analysis results are defuzzified using a fuzzy mathematics algorithm to obtain a deduced decision of the intelligent agent.

Description

Technical Field

[0001] The present invention belongs to the field of data communication technology, and in particular relates to a data communication system for multi-level linkage intelligent agent deduction. Background Art

[0002] Agent-based simulation is a technical means of simulating action and decision-making processes. With the increasing complexity of modern agent control, traditional single-level agent-based simulation systems are no longer able to meet the needs of multi-level, multi-departmental collaboration. Therefore, multi-level, linked agent-based simulation systems have emerged, aiming to achieve seamless collaboration between agents at all levels through efficient data communication and collaboration mechanisms.

[0003] Existing technologies often use distributed architectures to achieve data synchronization in multi-level linked intelligent game simulation systems. For example, the paper "Design and Implementation of a Distributed Wargame Simulation System," published in the Journal of Military Simulation, proposes a data synchronization mechanism based on message queues, which enables data transfer and synchronization between nodes at all levels. This solution can effectively reduce data transmission latency, but in scenarios with massive amounts of data, message queue performance can become a bottleneck. Furthermore, existing technologies often use time synchronization protocols to achieve data consistency. For example, the paper "Research on Time Synchronization Technology for Multi-level Linked Wargame Simulation Systems," published in Computer Engineering and Applications, proposes a time synchronization mechanism based on NTP (Network Time Protocol) to ensure time consistency across all nodes. This solution effectively addresses data inconsistencies caused by time asynchrony, but time synchronization accuracy can be affected in complex network environments. In the data processing and analysis process of the multi-level linkage intelligent agent deduction system, the existing technology uses a big data processing framework based on Hadoop and Spark to efficiently process and analyze massive deduction data, or uses data mining and prediction methods based on deep learning and fuzzy mathematics to analyze the potential correlation between deduction data and predict the other party's actions. However, in scenarios with high real-time requirements, the existing methods have high demand for computing resources and will have certain delays. In the network security and data protection technology process of the multi-level linkage intelligent agent deduction system, the existing technology uses a multi-level security protection architecture based on firewalls, intrusion detection systems and encryption technology to ensure the security of data transmission and storage, or uses a data protection mechanism based on blockchain to ensure the integrity and security of data through the immutability and distributed storage characteristics of blockchain. The above methods can effectively prevent data leakage and tampering, but in complex network environments, the real-time performance of security protection may be affected, and there are certain limitations in real-time data updating and synchronization. Summary of the Invention

[0004] In view of this, the present invention aims to provide a data communication system for multi-level linkage intelligent agent deduction, which solves the shortcomings of existing technologies in accuracy, efficiency, security and real-time performance by integrating data verification and error correction, optimizing data processing algorithms, multi-level security protection architecture, improving data synchronization and sharing mechanisms, improving data correlation and mining capabilities, and enhancing dynamic data update capabilities.

[0005] To achieve the above object, the technical solution created by the present invention is implemented as follows:

[0006] A data communication system for multi-level linkage intelligent agent deduction includes a data acquisition terminal, a data transmission terminal, and a data receiving terminal; wherein:

[0007] In the data collection end, the basic data of the intelligent agent is collected and verified, and the collected and verified basic data is processed by multiple nodes to obtain deduction data;

[0008] At the data transmission end, the deduced data is verified and monitored, and the verified data is encrypted to obtain encrypted data; blockchain technology is used to transmit the encrypted data using multiple industrial control communication protocols, and the encrypted data is verified and monitored during the transmission process;

[0009] At the data receiving end, the encrypted data is decrypted and the decoded data is corrected based on the verification and monitoring results at the data acquisition and data transmission ends. Distributed caching technology is used to synchronize the verified data to the corresponding intelligent agent for data update. A deep learning algorithm is used to deduce and analyze the synchronized updated data, and the deduction and analysis results are defuzzified using a fuzzy mathematical algorithm to obtain the deduction decision of the intelligent agent.

[0010] An API development interface is set in the data collection end, the data transmission end, and the data receiving end. The API development interface is used to make customized settings in the data collection end, the data transmission end, and the data receiving end.

[0011] Furthermore, cyclic redundancy check technology is used to verify and monitor data at the data acquisition end and the data transmission end; when errors or damage are detected in the data acquisition end, and / or data loss or damage is detected in the data transmission end, ECC data is used to perform data error correction at the data receiving end.

[0012] Furthermore, in the process of multi-node processing in the data collection end to obtain inferred data, the basic data is distributed to multiple processing nodes through the Apache Kafka framework; in each processing node, the basic data is processed by Apache Flink to obtain inferred data.

[0013] Furthermore, in each processing node, when an anomaly in the deduction data is detected, an event-driven architecture is used to trigger the corresponding event processing process, generate early warning information or adjust the deduction strategy to adjust the deduction data.

[0014] Furthermore, at the data transmission end, the deduction data is encrypted using the AES encryption algorithm, and the encrypted key is encrypted using the RSA encryption algorithm.

[0015] Furthermore, in the process of transmitting encrypted data using blockchain technology, the encrypted data is stored in a dispersed manner in multiple blockchain nodes, and a hash value is added before the encrypted data in each blockchain node.

[0016] Furthermore, at the data receiving end, Redis is used to perform distributed caching of the verified data, and then the verified data is published to the corresponding intelligent agent through the Apache Kafka framework for data update, and the updated data is obtained in the corresponding intelligent agent in combination with the event-driven architecture.

[0017] Furthermore, after the corresponding agent in the data receiving end updates the data, it also includes: when the corresponding agent detects that the updated data is abnormal, the corresponding event is immediately triggered through the Kafka trigger to process the updated data.

[0018] Furthermore, the process of using a deep learning algorithm to deduce and analyze the synchronized update data at the data receiving end includes:

[0019] Use convolutional neural networks to extract features from structured data in the update data to obtain the agent's intention information and deduced situation information;

[0020] Use recurrent neural networks to analyze the time series data in the update data to predict the movement trajectory and state changes of the intelligent agent;

[0021] After the intention information and deduced situation information, as well as the motion trajectory and situation changes are integrated and feature extracted, they are input into the decision-making model to generate deduction and analysis results.

[0022] Furthermore, the data receiving end uses fuzzy mathematical algorithms to defuzzify the deduction and analysis results to obtain the deduction and decision-making process of the intelligent agent:

[0023] Utilize fuzzy logic algorithms to obtain decision support information for deductive decision-making from intention information and deductive situation information, including: using membership functions to map intention information and deductive situation information to obtain fuzzy values; establishing fuzzy rules based on intention information and deductive situation information; using fuzzy reasoning mechanisms combined with fuzzy rules to reason on fuzzy values ​​to obtain fuzzy sets; and using the centroid method to convert fuzzy sets into decision support information.

[0024] The fuzzy C-means algorithm is used to cluster the motion trajectories and obtain different types of action patterns in the deduction decision;

[0025] A fuzzy comprehensive evaluation is performed on the deduced situation information and situation changes to obtain a comprehensive evaluation result in the deduction decision, including: defining a fuzzy set of the deduced situation information and situation changes; constructing a fuzzy matrix, wherein the elements in the fuzzy matrix represent the membership of the evaluation level of the deduced situation information or situation changes; determining a weight vector of the deduced situation information and situation changes; multiplying the weight vector with the fuzzy matrix to obtain a comprehensive evaluation vector; and using the center of gravity method or the maximum membership method to convert the comprehensive evaluation vector into a comprehensive evaluation result.

[0026] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0027] The present invention creates the data communication system for multi-level linkage intelligent agent deduction: first, by integrating data verification and error correction mechanisms, the accuracy and reliability of the deduction results are ensured, providing solid support for decision-making; second, by optimizing the data processing algorithm and architecture, the data processing efficiency is significantly improved, the system's requirements for real-time performance are met, and the timeliness of the deduction results is ensured; third, a multi-level security protection architecture is introduced to effectively prevent data leakage and tampering, greatly enhancing the security of the system; fourth, by optimizing the data synchronization algorithm and architecture, the data synchronization efficiency between multi-level nodes is improved, delays and inconsistencies are reduced, and efficient data sharing is achieved; fifth, by adopting advanced data mining algorithms, the potential correlations between data are deeply analyzed, data correlation and mining capabilities are significantly improved, and more comprehensive support is provided for decision-making; finally, through the real-time data update mechanism, it is ensured that the deduction results can reflect the latest deduction situation, significantly enhance the dynamic data update capability, and improve the accuracy of the deduction results. In summary, the present invention aims to address the technical deficiencies of the prior art in multi-level linkage wargaming simulation systems. By integrating data verification and error correction mechanisms, optimizing data processing algorithms and architectures, introducing a multi-level security protection architecture, improving data synchronization and sharing mechanisms, enhancing data relevance and mining capabilities, and strengthening dynamic data updating capabilities, the accuracy, efficiency, security, and real-time performance of the system are comprehensively improved, providing more reliable support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0029] Figure 1 A structural diagram of the data communication system for multi-level linkage intelligent agent deduction described in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation of the present invention.

[0031] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0032] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second" and the like are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, features defined as "first", "second" and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0033] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art can understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0034] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0035] like Figure 1As shown, the data communication system for multi-level linkage intelligent agent deduction described in the embodiment of the present invention includes a data acquisition terminal, a data transmission terminal, and a data receiving terminal. Among them, in the data acquisition terminal, the basic data of the intelligent agent is collected and verified, and the collected and verified basic data is subjected to multi-node processing to obtain deduction data; in the data transmission terminal, the deduction data is verified and monitored, and the verified data is encrypted to obtain encrypted data; blockchain technology is used to transmit the encrypted data using multiple industrial control communication protocols; in the data receiving terminal, the encrypted data is decrypted, and the decoded data is corrected according to the verification and monitoring results in the data acquisition terminal and the data transmission terminal; distributed caching technology is used to synchronize the verified data to the corresponding intelligent agent for data update, and a deep learning algorithm is used to deduce and analyze the synchronized updated data, and the deduction and analysis results are defuzzified using a fuzzy mathematical algorithm to obtain the intelligent agent's deduction decision. In addition, the present invention provides an API development interface in the data acquisition terminal, the data transmission terminal, and the data receiving terminal. The API development interface is used to customize the settings in the data acquisition terminal, the data transmission terminal, and the data receiving terminal.

[0036] In embodiments of the present invention, the data transmission terminal supports multiple industrial control communication protocols (such as Modbus and OPC UA), enhancing compatibility. In other embodiments, a protocol conversion gateway can be used to enable data interoperability between different protocols, or the Common Industrial Protocol (CIP) can be used instead of Modbus and OPC UA to improve protocol compatibility. Furthermore, in some other embodiments, scripting languages ​​(such as Python or Lua) can be used instead of API development interfaces to provide more flexible development support. Visual programming tools (such as Node-RED) can also be used to lower the user development threshold. Open source frameworks (such as Spring Boot or Django) can also be used to provide a richer set of development resources. Furthermore, in embodiments of the present invention, a microservices architecture is used instead of a modular design to achieve greater flexibility and scalability. In some other embodiments, a plugin architecture can be used to allow users to dynamically load and unload functional modules. Containerization technologies (such as Docker) and orchestration tools (such as Kubernetes) can also be used to enable rapid system deployment and expansion.

[0037] Because errors or deviations may occur during data collection and transmission as the agent automatically detects and corrects data, to ensure data accuracy, in some embodiments, cyclic redundancy check (CRC) technology is used to verify and monitor data at the data collection and transmission ends. When errors or corruption are detected at the data collection end and / or data loss or corruption is detected at the data transmission end, ECC data is used to perform data error correction at the data receiving end. In this embodiment of the present invention, the verification and error correction process specifically includes: First, the agent uses CRC technology to verify and monitor the collected and transmitted data in real time. By generating a checksum and comparing it with the original data, CRC can quickly identify whether errors or corruption have occurred during data transmission. During data transmission, if data packets are lost or corrupted due to network fluctuations or hardware failures, CRC can immediately detect the anomaly and trigger an error correction process. After detecting data errors, the agent automatically corrects the data using ECC technology. ECC not only detects errors but also repairs erroneous data through built-in redundant information, ensuring data integrity and accuracy. During the deduction data collection process, if the sensor produces deviated data due to environmental interference, ECC can automatically correct these deviations to ensure the reliability of the input data.

[0038] Specifically, the mathematical expression of CRC is:

[0039] R(x)=(M(x)×x l )mod(G(x));

[0040] Among them, M(x) represents the original data polynomial, G(x) represents the generating polynomial, and x l It means to shift the data left by 1 bit, R(x) is the calculated remainder, that is, the CRC check code, and mod means to take the remainder.

[0041] ECC codes have strong error correction capabilities, among which Reed-Solomon codes (RS codes) and low-density parity-check codes (LDPC codes) are two codes with very strong error correction capabilities. In the embodiment of the present invention, RS codes and LDPC codes are preferably used for cascade coding to complete error correction. That is, the data is first encoded into RS codes and then further encoded using LDPC codes. In this way, the RS code can handle burst errors, while the LDPC code can handle random errors. The cascade method of the two is used to further improve the error correction capability.

[0042] Specifically, Reed-Solomon codes are block codes based on finite fields and are particularly suitable for correcting burst errors. RS codes are usually expressed as RS(n,k), where n represents the codeword length, k represents the message length, and nk represents the number of check symbols. The generated codeword c(x) in the RS code is obtained by multiplying the message polynomial m(x) by the generator polynomial g(x), that is:

[0043] c(x)=m(x)·g(x).

[0044] Among them, the message polynomial m(x) is expressed as:

[0045] m(x)=m0+m1x+m2x 2 +...+m k-1 x k-1 ;

[0046] The generating polynomial g(x) is a 2q-degree polynomial, where q represents the number of correctable error symbols, namely:

[0047] g(x)=(x-α 1 )(x-α 2 )...(x-α 2q );

[0048] Where α represents the finite field GF(2 m ) is an original data.

[0049] The error correction capability of RS code can be expressed as follows:

[0050] ;

[0051] That can be corrected A symbol error.

[0052] LDPC code is a linear block code based on a sparse matrix with very strong error correction capability, especially suitable for high-noise environments. LDPC code is defined by the check matrix H, which is a (nk)×n sparse matrix containing a large number of zeros. The codeword C in the LDPC code satisfies HC T = 0. The generator matrix G of the LDPC code can be obtained from the check matrix H through GH T = 0. LDPC codes typically use iterative decoding algorithms, such as the belief propagation (BP) algorithm or the message passing (MP) algorithm. In each iteration, the probability information of the nodes is updated based on the sparse structure of the check matrix H, gradually approaching the correct codeword.

[0053] The present invention adopts a multi-level verification mechanism, that is, verification and correction are performed at the data acquisition end, the data transmission end, and the data receiving end to ensure the accuracy of the data in each link. At the data acquisition end, the system will perform a preliminary verification of the original data; at the data transmission end, real-time monitoring is performed through CRC technology; at the data receiving end, the ECC technology is used to make the final correction to ensure the reliability of the data throughout the entire process. In addition, in order to improve the fault tolerance of the system, the embodiment of the present invention combines the redundant design of the 3-copy strategy, that is, in the process of transmitting key data at the data transmission end, multiple identical copies are directly generated based on the original data, and the received data is compared and verified with the multiple copies at the data receiving end to ensure that even if part of the data is lost or damaged, the complete information can still be restored through redundant data.

[0054] The present invention can effectively avoid distortion of deduction results caused by data errors or deviations through the above-mentioned data verification and error correction mechanism, and provide highly accurate and reliable data support for the multi-level linkage war game deduction system, thereby laying a solid foundation for decision-making.

[0055] In some other embodiments, Hamming Code may be used instead of CRC and ECC technology to implement data verification and error correction; Parity Check may be used in combination with a retransmission mechanism to ensure data accuracy; Forward Error Correction Code (FEC) technology, such as Reed-Solomon code, may be used to improve data error correction capabilities.

[0056] The present invention utilizes streaming data processing technology at the data acquisition end to process massive amounts of data in real time, significantly reducing data processing delays and ensuring the efficient operation of the multi-level linkage intelligent agent deduction system in complex deduction environments. This technology not only improves the real-time performance of the system but also provides more timely and accurate support for decision-making. Specifically, in some embodiments, during the multi-node processing at the data acquisition end to obtain deduction data, the basic data is distributed to multiple processing nodes via the Apache Kafka framework; within each processing node, the basic data is processed via Apache Flink to obtain deduction data.

[0057] In this embodiment of the present invention, Apache Kafka is used as a distributed message queue in the data collection terminal to achieve real-time data collection and transmission. Apache Kafka can efficiently process high-throughput data streams and ensure reliable data transmission between multiple nodes. In a simulation environment, massive amounts of data generated by sensors, radars, and other equipment can be collected and transmitted to data processing nodes in real time using Apache Kafka, avoiding data backlogs or loss. Apache Kafka's distributed architecture also ensures high data availability and reliability.

[0058] The real-time data collection process of the Apache Kafka framework itself does not directly involve complex mathematical expressions. It is mainly a data transmission and processing mechanism. The following mathematical symbols and functions are used to abstractly represent the real-time data flow and processing process of the Apache Kafka framework:

[0059] In the Apache Kafka framework, the data stream is represented as D(t), where t represents the time variable, i.e., the change of the data stream over time; the process of sending data can be represented as the producer P i Send data to topic T j The function f p (D(t))=(T j ,d n ), d n Represents a specific data item sent at time t; topic T j It can be expressed as a set, namely T j ={p1,p2,...,p k}, p k Represents a topic partition; the process of receiving data can be represented as the cth consumer C c From the jth topic T j Function f of consuming data c (T j ) = D'(t), where D'(t) represents the data stream received by the consumer at time t. The data processing process of the Apache Kafka framework includes:

[0060] In the process of data transmission, the transmission of data from the producer to multiple nodes can be expressed as a mapping process φ: D(t)→T j , ensuring that data is correctly routed to the topic; during data storage and persistence, the storage of data in the Kafka cluster can be expressed as: , representing the accumulation of data from the initial time t0 to the current time t. During real-time data processing, the consumer's real-time processing can be represented by a processing function f(D'(t)) = ProcssedData(t), which converts the raw data stream into processed data. ProcssedData represents the data delta function, which is essentially a dynamic operation on the raw data in stream processing. Specific implementations include aggregation (such as SUM, COUNT, AVG), state calculation (such as cumulative values, sliding windows), transformation (such as format standardization, field mapping), pattern recognition (such as anomaly detection, complex event processing), and machine learning reasoning (such as real-time classification and prediction). The choice of function depends on the business goal, such as monitoring, analysis, alerting, or real-time response. In code, these functions are typically implemented using operators (such as map, filter, aggregate, and window) in stream processing frameworks (such as Kafka Streams and Apache Flink).

[0061] In an embodiment of the present invention, the Apache Flink framework is used in the data collection end to perform real-time processing of streaming data. Apache Flink is a high-performance stream processing engine that supports low-latency, high-throughput data processing. The data collection end can use Apache Flink to perform real-time analysis of the collected deduction data, including the opponent's action trajectory, deduction situation changes, etc., and quickly generate deduction results to provide real-time support for decision-making. Apache Flink supports a fault recovery mechanism that can automatically restore data processing tasks when a node fails, ensuring the continuity and stability of the system. The real-time data flow and processing process of the Apache Flink framework are abstractly represented by the following mathematical symbols and functions:

[0062] In the Apache Flink framework, the data flow is represented as S(t)={s1,s2,...,s s}, s s Represents the sth data item or event at time t; an operator is represented as a function that maps the input data stream to the output data stream. Apache Flink uses operators to transform and process data streams. The data stream processing process of the Apache Flink framework can be abstracted as a series of function combinations, represented as: S output (t)=f n (f n-1(...f2(f1(S(t)))...)), where each f represents a specific operation, including map operation, filter operation, reduce operation and window operation. The map operation maps each element in the input data stream to another element, expressed as S'(t)=f map (S(t)); The filtering operation is used to filter out elements that do not meet the conditions from the data stream, expressed as S'(t)={s s |condition(s s ),s s ∈S(t)}, |condition means filtering out elements that do not meet the conditions. In the embodiment of the present invention, the condition function is a Boolean judgment on the data item, for example:

[0063] ;

[0064] Among them, s.userID represents the user identification field in the data item, Indicates the target user set, used for filtering conditions.

[0065] Aggregation operations are used to aggregate data streams (such as summation, averaging, etc.), expressed as S'(t)=f reduce (S(t)); The block operation is used to process the data stream in blocks, which is often used for batch computing and is expressed as W(t)={S(t1),S(t2),...,S(t w )}.

[0066] In some other embodiments, Apache Storm can be used instead of Apache Kafka and Apache Flink to achieve real-time data processing; Google Cloud Dataflow or AWS Kinesis can be used as a streaming data processing platform; and a custom distributed message queue system, such as RabbitMQ or ZeroMQ, can be used to achieve data stream processing.

[0067] In some embodiments, at each processing node, when an anomaly in deduction data is detected, an event-driven architecture is used to trigger a corresponding event processing process, generate warning information, or adjust the deduction strategy to adjust the deduction data. The present invention combines streaming data processing technology with an event-driven architecture to ensure that the system can respond to events in the data stream in real time. When an abnormal action by the other party or a sudden change in the deduction situation is detected, the system can immediately trigger the corresponding event processing process, generate warning information, or adjust the deduction strategy to ensure the timeliness of the deduction results.

[0068] This invention combines blockchain technology and encryption algorithms at the data transmission end to ensure data security during transmission and storage in a multi-level linked intelligent agent deduction system, preventing data leakage and tampering. Specifically, in some embodiments, the data transmission end uses the AES (Advanced Encryption Standard) encryption algorithm to encrypt deduction data, and the encryption key is encrypted using RSA (Reliable Asymmetric Encryption Algorithm) to generate encrypted data, thereby preventing data leakage and tampering. During the encrypted data transmission process using blockchain technology at the data transmission end, the encrypted data is distributed and stored across multiple blockchain nodes, and a hash value is added to the encrypted data in each blockchain node.

[0069] Blockchain technology, through its decentralized storage, immutability, and traceability, provides strong protection for data security. Specifically, blockchain technology's decentralized storage means that data is no longer stored centrally on a single node, but rather distributed across multiple nodes in the blockchain network, reducing the risk of single points of failure and data leaks. Simulation data can be distributed across multiple blockchain nodes, ensuring complete data recovery even if some nodes are attacked. Blockchain technology's immutability is demonstrated by its hash chain structure, where each data block contains the hash value of the previous block. Once data is written to the blockchain, it cannot be tampered with. Once a record of an adversary's actions is written to the blockchain, any attempt to modify it will be detected and rejected by the system. Blockchain technology's traceability is demonstrated by its complete record of data flow, enabling the tracing of data sources and changes. The entire process of generating simulation results is fully documented, ensuring data transparency and credibility.

[0070] In an embodiment of the present invention, before the deduction data is written to the blockchain, it is first encrypted using the AES algorithm and the RSA algorithm at the data transmission end to ensure that even if the blockchain node is attacked, the data cannot be cracked; the deduction data is written to the blockchain after encryption, and even if the attacker obtains the data block, the content cannot be decrypted; combining the hash verification of the blockchain and the digital signature of the encryption algorithm, dual verification of the data is achieved. During the transmission process, the data not only needs to be verified for integrity through the hash value, but also needs to be verified for source authenticity through the digital signature. In an embodiment of the present invention, a self-executing smart contract deployed in the blockchain is also used to realize the automated management of data access rights. Only authorized nodes can decrypt and access specific data, preventing unauthorized access. By combining blockchain technology with encryption algorithms, the present invention provides multi-level security protection for the multi-level linkage intelligent agent deduction system, ensuring the security of data during transmission and storage, preventing data leakage and tampering, and providing more reliable support for decision-making.

[0071] Specifically, in the embodiment of the present invention, the encryption process of the deduction data in the data transmission end includes:

[0072] Use the AES algorithm to encrypt the deduction data P and obtain the ciphertext CIP, that is, CIP=AES K (P), K represents the encryption key of AES algorithm;

[0073] Use the RSA algorithm to encrypt the key K according to the public key Kpub of the data receiving end to obtain the key E K , that is: E K =RSA Kpub (K);

[0074] Use the RSA algorithm to sign the hash value H(P) of the blockchain according to the private key Kpriv of the data collection end to obtain the digital signature S, that is: S=RSA Kpriv (H(P));

[0075] At this time, the encrypted data includes the ciphertext CIP, the encrypted AES key E K , and digital signature S. Correspondingly, the process of decrypting the encrypted data includes:

[0076] Use the RSA algorithm to encrypt the encrypted AES key E according to the private key Kpriv' of the data receiver. K Decrypt and get the key K, that is, K=RSA -1 Kpriv’ (E K ), where RSA -1 Indicates the decryption algorithm corresponding to the RSA algorithm;

[0077] Use the AES algorithm to decrypt the ciphertext CIP according to the key K, that is, P=AES -1 K (CIP), AES -1 Indicates the decryption algorithm of the AES algorithm;

[0078] The data receiving end uses the RSA algorithm to verify the digital signature S according to the public key Kpub' of the data collection end, that is, Verify=(H(P)=RSA v Kpub’ (S)), where RSA v Indicates the RSA algorithm used to verify the digital signature S.

[0079] During the distributed transmission of the blockchain, the hash value of each data block is obtained by the following formula:

[0080] ;

[0081] Among them, data d Indicates the data in the current data block, Represents the hash value of the previous data block. The current hash value is linked to the hash value in the previous data block to form a hash chain.

[0082] The data processing process of distributed blockchain transmission in the embodiment of the present invention includes:

[0083] The encrypted data is broadcast to the entire blockchain network. After receiving the encrypted data, each node in the blockchain network verifies whether the digital signature and content of the encrypted data are legal; consensus is reached through the consensus mechanism of Byzantine Fault Tolerance (PBFT) or Proof of Stake (PoS) to confirm the validity of the encrypted data.

[0084] In some other embodiments, elliptic curve cryptography (ECC) can be used instead of the RSA encryption algorithm to improve encryption efficiency; national secret algorithms (such as SM2 and SM4) can be used instead of the AES encryption algorithm to meet the security requirements of specific countries or regions; and zero-knowledge proof (Zero-Knowledge Proof) technology can be used to enhance data privacy protection.

[0085] This invention utilizes distributed caching technology and an event-driven architecture to improve data synchronization efficiency across multiple nodes, reduce latency and inconsistencies, and ensure the real-time and accuracy of deduction results. Specifically, in some embodiments, the data receiving end uses Redis for distributed caching of error-corrected data. This data is then published to the corresponding agent via the Apache Kafka framework for data update, and the updated data is then obtained by the corresponding agent in conjunction with the event-driven architecture.

[0086] In the present invention, Redis is used as a distributed cache system at the data receiving end. With its high performance and low latency, Redis enables rapid data storage and retrieval. Decrypted and corrected deduction data can be stored in the Redis cache, allowing each node to quickly access the data when needed, reducing the pressure on direct data access. Furthermore, the present invention leverages the distributed nature of Redis and the Apache Kafka framework to synchronize data across multiple nodes in real time, ensuring data consistency. Specifically, when a node updates deduction data, Redis automatically synchronizes the update to other nodes, avoiding data inconsistencies. Furthermore, the present invention incorporates an event-driven architecture at the data receiving end, enabling efficient processing of data change events. Specifically, when the Apache Kafka framework receives a data update event, the system immediately retrieves the latest data from the Redis cache and updates the local state, ensuring real-time and consistent data. The event-driven architecture responds in real time to data changes or specific conditions. Specifically, when an abnormal action or sudden change in the deduction situation is detected, the system immediately triggers a corresponding event, notifying the relevant nodes for processing. The event-driven architecture supports asynchronous processing, ensuring efficient system operation even under high load. Data update tasks can be distributed asynchronously through Apache Kafka to avoid blocking the main thread and improve the overall performance of the system.

[0087] In the embodiment of the present invention, Redis nodes and Kafka partitions can also be used to process larger-scale data and more concurrent events to meet the needs of a multi-level linkage deduction system.

[0088] In some other embodiments, Memcached can be used instead of Redis to implement distributed caching functions; ActiveMQ or RabbitMQ can be used instead of Kafka to implement an event-driven architecture; and real-time communication protocols based on HTTP / 2 or WebSocket can be used to implement data synchronization and sharing.

[0089] The present invention uses deep learning to simulate the structure and function of the human brain's neural network, automatically extracting features from massive amounts of data and learning complex nonlinear relationships between data. Specifically, in some embodiments, the process of using a deep learning algorithm at the data receiving end to perform deductive analysis on synchronized update data includes: using a convolutional neural network (CNN) to extract features from the structured data in the update data to obtain the agent's intention information and deduced situation information; using a recurrent neural network (RNN) to analyze the time series data in the update data to predict the agent's movement trajectory and state changes; integrating the intention information and deduced situation information, as well as the movement trajectory and state changes, with feature extraction, and then inputting them into a decision model to generate a deduction analysis result containing deduction recommendations.

[0090] In an embodiment of the present invention, a CNN is used to extract features and recognize patterns in structured data (such as images and maps). In a simulation environment, the CNN can analyze satellite imagery or radar data, identify the agent's intent and simulate situational information, and provide data support for simulation. Specifically, the CNN process of extracting features and recognizing patterns in structured data includes:

[0091] Perform normalization and denoising preprocessing on input structured data (such as images, maps) and time series data (such as movement trajectories, situation changes) to improve the training effect of the model. Use the convolution layer to extract local features of the preprocessed input data, namely:

[0092] F p =ReLU(W×X+b);

[0093] Among them, F p Represents the extracted local features, ReLU represents the ReLU activation function, X represents the preprocessed input data, W represents the weight of the convolutional layer, and b represents the bias of the convolutional layer;

[0094] For the extracted local features F p Perform pooling dimensionality reduction, namely:

[0095] F l =pool(F);

[0096] Among them, F l Represents the features after pooling, and pool represents the pooling operation;

[0097] The fully connected layer and softmax function are used to integrate the pooled features, and the agent's intention information and deduction situation information are extracted from the integrated features, namely:

[0098] F cnn =softmax(W fc ×Flatten(F l )+b fc );

[0099] Among them, F cnn It represents the feature information including the agent’s intention information and deduction situation information, Flatten represents the fully connected layer, and W fc and b fc denote the weights and biases of the fully connected layer respectively.

[0100] In an embodiment of the present invention, an RNN network is used to analyze the time series data in the update data to predict the motion trajectory and state changes of the intelligent agent. Specifically, the RNN network processing process includes:

[0101] The RNN unit in the RNN network is used to recursively process the input time series through time steps, that is, h t =σ(W h ·h t-1 +W x ·x in-t +b t ), h t represents the output characteristics of the RNN unit at time t in the RNN network, σ represents the activation function, and W h represents the weight of the hidden layer in the RNN network, b t Represents the bias of the hidden layer in the RNN network, h t-1 The input value of the hidden node, W x is the weight of the input node, x in-t represents the time series of input;

[0102] Using the softmax function, from the output feature h t The motion trajectory and state changes of the predicted intelligent agent, namely F rnn =softmax(W out ·h t +b out ), F rnn Represents the features including motion trajectory and state change, where W out and b out Represent the weight and bias of the output layer of the RNN network respectively.

[0103] The feature F containing intention information and deduced situation information cnn , and the features F including motion trajectory and situation change rnn Integration and feature extraction are performed to automatically extract the action mode and situation changes of the intelligent agent, namely F auto =Combine(F rnn ,F cnn ), F auto Indicates the features including the action mode and situation change information of the intelligent agent. In the embodiment of the present invention, the feature F auto It includes the agent's action mode (such as the agent's behavior type, resource allocation ratio, historical behavior pattern, etc.), situation change information (including the agent's environmental status (terrain, weather), agent dynamics, real-time event triggering, etc.), feature F auto It can be expressed mathematically as:

[0104] Fauto={action,state,history,event};

[0105] Among them, action represents the action mode of the agent, state represents the environmental situation feature vector of the agent, histroy represents the historical interaction data of the agent, and event represents the real-time event triggering of the agent (such as an emergency). The historical interaction data histroy can be adaptively adjusted according to the actual situation to determine whether it is used as the feature F. auto One of the items.

[0106] Combine represents integration and feature extraction. In the embodiment of the present invention, the Combine operation specifically includes mapping two features to the same space and concatenating them, thereby achieving feature enhancement and interaction:

[0107] ;

[0108] in, Indicates the features obtained by the Combine operation, and Represent the weight and bias of the RNN network respectively, and Represent the weights and biases of the CNN network respectively.

[0109] The feature F containing the agent's action mode and situation change information auto Input into the decision model to generate deduction analysis results containing deduction suggestions. The above process can be expressed as follows:

[0110] D output =Decision(F auto );

[0111] Among them, D output Indicates the deduction and analysis results, Decision indicates the decision model. In the embodiment of the present invention, the decision model Decision includes a first convolutional layer, a ReLU activation layer, a second activation layer, and a softmax layer. The feature F auto The input is fed into the first convolutional layer for feature extraction, and then the extracted features are activated by the ReLU activation layer and the convolution operation of the second activation layer in turn. Finally, the processed features are fed into the softmax layer to generate a probability distribution, and the deduction analysis result D is obtained. output , thereby completing the complex action pattern classification or high-dimensional situation deduction of the intelligent agent. The above process can be expressed as follows:

[0112] D output =softmax(W2⋅ReLU(W1⋅Fauto+b1)+b2);

[0113] Where W1 and b1 represent the weight and bias of the first convolutional layer, respectively; W2 and b2 represent the weight and bias of the second convolutional layer, respectively.

[0114] In some other embodiments, support vector machines (SVM) or random forests may be used instead of convolutional neural networks to implement data correlation analysis; reinforcement learning technology may also be used to optimize data analysis models.

[0115] In some embodiments, in the process of defuzzifying the deduction and analysis results using a fuzzy mathematical algorithm at the data receiving end to obtain the deduction and decision of the intelligent agent:

[0116] First, a fuzzy logic algorithm is used to obtain decision support information for deductive decision-making from intention information and deductive situation information. This includes: using a membership function to map intention information and deductive situation information to obtain fuzzy values; establishing fuzzy rules based on intention information and deductive situation information; using a fuzzy inference mechanism combined with fuzzy rules to infer fuzzy values ​​to obtain fuzzy sets; and using the center of gravity method to convert fuzzy sets into decision support information. In this embodiment of the present invention, the following membership function μ(x) is used to map the input intention information and deductive situation information x to obtain fuzzy values:

[0117] ;

[0118] Among them, a and b represent the endpoints of the interval, and the endpoints a and b of the interval are adaptively adjusted according to the actual situation;

[0119] Combine intention information and deduced situation information to establish fuzzy rules, such as "if the situation is tense and the agent's intention is unclear, the risk is high";

[0120] Use fuzzy reasoning mechanism (such as Mamdani or Sugeno method) combined with fuzzy rules to reason about fuzzy values ​​and obtain fuzzy set R;

[0121] The fuzzy set is converted into decision support information using the following centroid method:

[0122] .

[0123] Where y represents decision support information.

[0124] Then, the fuzzy C-means algorithm is used to cluster the motion trajectories to obtain different types of action modes in the deduction decision. In the embodiment of the present invention, the cluster number Ω and the initial cluster center v are selected. γ ; Calculate the membership μ by the following formula γη :

[0125] ;

[0126] Among them, V represents the fuzzy factor, ω represents the number of clusters, γ represents the behavior cluster, that is, the γth behavior cluster, η represents the behavior, that is, the ηth behavior, Represents the data point, that is, the motion trajectory of the current agent.

[0127] Finally, a fuzzy comprehensive evaluation is performed on the deduced situation information and situation changes to obtain a comprehensive evaluation result in the deduction decision, including: defining a fuzzy set of deduced situation information and situation changes; constructing a fuzzy matrix, in which the elements in the fuzzy matrix represent the membership of the evaluation level of the deduced situation information or situation changes; determining the weight vector of the deduced situation information and situation changes; multiplying the weight vector with the fuzzy matrix to obtain a comprehensive evaluation vector; and using the center of gravity method or the maximum membership method to convert the comprehensive evaluation vector into a comprehensive evaluation result.

[0128] By combining deep learning with fuzzy mathematical methods, this invention can deeply explore potential connections between data, revealing complex nonlinear relationships. This provides more comprehensive and accurate support for decision-making in multi-level intelligent agent deduction systems, significantly enhancing the system's intelligence and decision-making capabilities. In other embodiments, Bayesian networks can be used in place of fuzzy mathematical methods for reasoning under uncertainty.

[0129] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

[0130] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A data communication system for multi-level linkage intelligent agent deduction, characterized in that: It includes data acquisition end, data transmission end and data receiving end; among which: In the data collection terminal, basic data of the intelligent agent is collected and verified, and the collected and verified basic data is processed by multiple nodes to obtain deduction data; At the data transmission end, the deduced data is verified and monitored, and the verified data is encrypted to obtain encrypted data; the encrypted data is transmitted using blockchain technology using multiple industrial control communication protocols, and the encrypted data is verified and monitored during the transmission process; At the data receiving end, the encrypted data is decrypted, and the decoded data is corrected based on the verification and monitoring results at the data acquisition end and the data transmission end; the verified data is synchronized to the corresponding intelligent agent using distributed cache technology for data update, and the synchronized updated data is deduced and analyzed using a deep learning algorithm, and the deduction and analysis results are defuzzified using a fuzzy mathematical algorithm to obtain the deduction decision of the intelligent agent; The process of using a deep learning algorithm in the data receiving end to deduce and analyze the synchronized update data includes: using a convolutional neural network to extract features from the structured data in the update data to obtain the intention information and deduced situation information of the intelligent agent; using a recurrent neural network to analyze the time series data in the update data to predict the movement trajectory and state changes of the intelligent agent; integrating the intention information and the deduced situation information, as well as the movement trajectory and state changes, with feature extraction, and then inputting them into the decision model to generate a deduction and analysis result; in the process of defuzzifying the deduction and analysis result using a fuzzy mathematical algorithm in the data receiving end to obtain the deduction decision of the intelligent agent: using a fuzzy logic algorithm to obtain the decision support information in the deduction decision from the intention information and the deduced situation information, including: using a membership function to map the intention information and the deduced situation information to obtain a fuzzy value; establishing a fuzzy rule in combination with the intention information and the deduced situation information; using a fuzzy inference mechanism in combination with the fuzzy rule to infer the fuzzy value to obtain a fuzzy set; using the center of gravity method to convert the fuzzy set into the decision support information; Clustering the motion trajectories using a fuzzy C-means algorithm to obtain different categories of action modes in the deduction decision; performing a fuzzy comprehensive evaluation on the deduction situation information and the situation change to obtain a comprehensive evaluation result in the deduction decision, including: defining a fuzzy set of the deduction situation information and the situation change; constructing a fuzzy matrix, wherein elements in the fuzzy matrix represent the membership of the evaluation level of the deduction situation information or the situation change; determining a weight vector for the deduction situation information and the situation change; multiplying the weight vector by the fuzzy matrix to obtain a comprehensive evaluation vector; and converting the comprehensive evaluation vector into the comprehensive evaluation result using a center of gravity method or a maximum membership method. An API development interface is set in the data acquisition end, the data transmission end, and the data receiving end, and the API development interface is used to perform customized settings in the data acquisition end, the data transmission end, and the data receiving end.

2. The data communication system for multi-level linkage intelligent agent deduction according to claim 1 is characterized in that: Cyclic redundancy check technology is used to check and monitor data at the data acquisition end and the data transmission end; when errors or damage are detected in the data acquisition end, and / or data loss or damage is detected in the data transmission end, ECC data is used to perform data error correction at the data receiving end.

3. The data communication system for multi-level linkage intelligent agent deduction according to claim 1 is characterized in that: In the process of obtaining the deduced data through multi-node processing in the data acquisition terminal, the basic data is distributed to multiple processing nodes through the Apache Kafka framework; in each processing node, the basic data is processed by Apache Flink to obtain the deduced data.

4. The data communication system for multi-level linkage intelligent agent deduction according to claim 3 is characterized in that: In each processing node, when an anomaly in the deduction data is detected, an event-driven architecture is used to trigger a corresponding event processing process, generate warning information or adjust the deduction strategy to adjust the deduction data.

5. The data communication system for multi-level linkage intelligent agent deduction according to claim 1 is characterized in that: In the data transmission end, the deduction data is encrypted using the AES encryption algorithm, and the encrypted key is encrypted using the RSA encryption algorithm.

6. The data communication system for multi-level linkage intelligent agent deduction according to claim 1 is characterized in that: In the process of transmitting the encrypted data using blockchain technology, the encrypted data is dispersedly stored in multiple blockchain nodes, and a hash value is added before the encrypted data in each blockchain node.

7. The data communication system for multi-level linkage intelligent agent deduction according to claim 1 is characterized in that: In the data receiving end, Redis is used to perform distributed caching of the verified data, and then the verified data is published to the corresponding intelligent agent through the Apache Kafka framework for data update, and the updated data is obtained in the corresponding intelligent agent in combination with the event-driven architecture.

8. The data communication system for multi-level linkage intelligent agent deduction according to claim 7 is characterized in that: After the corresponding agent in the data receiving end updates the data, it also includes: when the corresponding agent detects that the updated data is abnormal, the corresponding event is immediately triggered through the Kafka trigger to process the updated data.

Images