Industrial device agent-based encapsulation and communication method based on a2a protocol

Abstract

This invention relates to a method for intelligent agent encapsulation and communication of industrial equipment based on the A2A protocol, belonging to the interdisciplinary field of Industrial Internet of Things (IIoT) and Artificial Intelligence (AI). Based on the A2A protocol, this method models various types of equipment in the industrial field through an intelligent agent encapsulation mechanism. Without altering the original physical structure and control logic of the equipment, it constructs a unified intelligent agent model for each type of industrial equipment. It also constructs an A2A semantic communication extension model oriented towards industrial scenarios, including content unit design, Message objects, and Artifact object extensions. Furthermore, it builds an A2A interaction mechanism driven by industrial tasks, and then realizes communication between intelligent agents of industrial equipment based on this interaction mechanism. This invention can solve the problems of lack of unified intelligent agent abstraction, inconsistent semantics of interaction between devices, rigid communication modes, and low collaborative efficiency in industrial equipment under heterogeneous computing power conditions.

Description

Technical Field

[0001] This invention belongs to the field of interdisciplinary technology of industrial Internet of Things and artificial intelligence, and relates to an intelligent encapsulation and communication method for industrial equipment based on the A2A protocol. Background Technology

[0002] With the rapid evolution of next-generation artificial intelligence, the Industrial Internet, and Multi-Agent Systems (MAS) technologies, industrial sites are undergoing a profound transformation from "automatic control" to "intelligent collaboration." Under this trend, industrial field equipment is gradually evolving from traditional "passive execution and data acquisition units" to intelligent units with environmental perception, status analysis, and a certain degree of autonomous decision-making capabilities. The communication needs between devices are no longer limited to simple data acquisition, status reporting, and control command issuance, but are leaping towards higher-level interactive needs oriented towards task objectives, collaborative relationships, and operational semantics. This means that devices need to be able to communicate effectively at the intent level, addressing questions such as "What is the current state? Who is performing the task? How can we dynamically collaborate?"

[0003] However, the deep integration of heterogeneous devices and AI agents faces a serious "semantic gap" problem. Existing industrial communication standards and protocols, such as OPC UA, Modbus, CAN, and PROFIBUS, have played an important role in traditional automation systems, but they are essentially communication mechanisms "data access-oriented" or "process control-oriented." For example, Modbus and CAN mainly provide low-level byte stream or register read / write operations, and while OPC UA has some information modeling capabilities, its communication semantics are still limited to static descriptions of device attributes and controlled objects. These mechanisms lack the ability to uniformly model and express high-level semantics at the "agent level," such as the dynamic capabilities of devices, operational intentions, task states, and collaborative behaviors. When industrial equipment needs to act as an intelligent agent for autonomous collaborative interaction, existing mechanisms cannot output semantic messages that AI agents can directly understand. They often have to rely on writing a large amount of complex and customized protocol parsing and conversion logic at the application layer, resulting in extremely high system coupling, poor flexibility, and high expansion costs.

[0004] Meanwhile, industrial sites face a structural contradiction between "limited edge computing resources" and "high-level intelligent communication needs." In actual industrial production environments, a large number of resource-constrained devices, such as sensors and actuators, are widely distributed. These devices typically use low-power MCUs with extremely limited computing and storage resources, and their communication interfaces are complex, making it impossible for them to natively support complex intelligent algorithms or run high-level semantic communication protocols. In the wave of intelligent upgrades to industrial systems, how to seamlessly integrate these resource-constrained devices into new intelligent collaborative systems without large-scale replacement of existing outdated equipment has become a core bottleneck restricting the implementation of large-scale industrial models and multi-agent technologies in industrial settings. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide an intelligent agent encapsulation and communication method for industrial equipment based on the A2A protocol, to solve the problems of industrial equipment lacking a unified intelligent agent abstraction, inconsistent semantics of interaction between devices, rigid communication modes, and low collaborative efficiency under heterogeneous computing power conditions. Furthermore, addressing the current situation where existing industrial equipment mainly relies on traditional industrial protocols or proprietary interfaces for communication, has difficulty directly accessing native intelligent agent collaborative networks, and cannot perform standardized task interaction and semantic collaboration with native intelligent agents, this invention, through intelligent agent encapsulation, enables industrial equipment to possess a unified identity, capability description, and standard communication interface, fundamentally breaking down the interaction barriers between industrial equipment and native intelligent agents.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for intelligent packaging and communication of industrial equipment based on the A2A protocol, the method comprising: S1. Based on the A2A protocol, various types of equipment in the industrial field are modeled through an intelligent agent encapsulation mechanism. Without changing the original physical structure and control logic of the equipment, a unified intelligent agent model of industrial equipment is built for each type of industrial equipment. S2. Construct an A2A semantic communication extension model for industrial scenarios, design the content unit Part in the A2A protocol, and extend the Message object and Artifact object for industrial applications. S3. Construct an A2A interaction mechanism based on industrial task-driven operation, define the tasks and task states of industrial equipment agents, and define a matching mechanism for selecting interaction modes based on task characteristics; realize communication between industrial equipment agents based on the interaction mechanism.

[0007] Furthermore, in step S1, various types of equipment in the industrial site are classified into resource-constrained equipment and resource-abundant equipment, wherein... For resource-constrained devices, an external agent adaptation approach is adopted: the resource-constrained device is connected to an independent external intelligent agent packaging device, where an agent-type industrial device intelligent agent running on the device represents the physical device in A2A communication; the agent-type industrial device intelligent agent maintains the AgentCard, receives and parses A2A Messages, creates and updates A2A Tasks, and maps underlying industrial protocol data, register data, sensor sampling values, or actuator control interfaces to Part, Message, Artifact, and Task.status in A2A standard objects; the mapping between the resource-constrained physical device and the agent-type industrial device intelligent agent in the external intelligent agent packaging device is one-to-one or one-to-many. For resource-rich equipment, an industrial equipment intelligent agent encapsulation module is directly deployed in the equipment body or its built-in control unit, enabling it to issue AgentCards to the outside world as an independent industrial equipment intelligent agent and participate in task collaboration and interaction directly through the A2A protocol.

[0008] Furthermore, in step S1, a unified model of industrial equipment intelligent agents is constructed based on the A2A standard AgentCard, enabling industrial equipment to declare its access points, protocol interaction capabilities, input / output modes, and executable industrial task capabilities to external systems in the form of a standard intelligent agent; the industrial equipment intelligent agent AgentCard includes at least the following fields: capabilities represent the interactive capabilities that an intelligent agent supports at the A2A protocol level, and are expressed as:

[0009] in, This indicates the protocol capabilities of intelligent agents in industrial equipment; Indicates whether the agent supports streaming responses; Indicates whether the agent supports sending push notifications for asynchronous task updates; Indicates whether the agent supports a history of task state transitions; skills represent the industrial task capabilities that the intelligent agent of the industrial equipment can invoke; let the skill set of the intelligent agent be... for:

[0010] Each task capability Defined as:

[0011] in, Indicates a unique identifier for a skill; Indicates the skill name; Indicates skill description; Indicate skill tags; This indicates the types of input content that the skill supports; Indicates the types of output content supported by this skill; Provide sample hints or scenarios that can be handled.

[0012] Furthermore, in step S2, the content unit Part is the basic unit in the A2A protocol used to carry specific content; a Message or Artifact consists of one or more Parts. Through Parts, intelligent agents in industrial equipment can simultaneously transmit text descriptions, structured parameters, and file content, that is:

[0013] Among them, TextPart is used to carry text content; DataPart is used to carry structured data; and FilePart is used to carry file content.

[0014] Furthermore, in step S2, Message is the standard message object for interaction between intelligent agents in the A2A protocol; after receiving the Message, the industrial equipment intelligent agent creates or updates a Task based on its content; Message is used to carry task input, supplementary information, status descriptions, or other interactive content, that is:

[0015] in, Indicates the message role; This represents a collection of message content units, containing the task body; This represents message-level industrial metadata, used to express message-level industrial scheduling constraints and processing auxiliary information. It is defined as follows:

[0016] in, Indicates the message semantic type; Used to constrain the expected completion time of message-triggered tasks; Used to determine whether the current message request or task wait has timed out.

[0017] Furthermore, in step S2, Artifact is the task result object attached to Task in the A2A protocol; during or after task execution, the industrial equipment intelligent agent encapsulates the result content into Artifact and returns or pushes it to the task initiator through Task, which is represented as follows:

[0018] in, Indicates the name of the result; Indicates the results; This represents a collection of result content units used to carry the main text of the result. This represents result-level industrial metadata, used to describe the time, unit, quality, and order attributes of the result data itself, and is represented as follows:

[0019] in, Indicates the time when the result was generated; Indicates the result sequence number, used for sorting results in subscription-based or phased tasks; Indicates the unit of data, used to indicate the reliability of the result; To represent data quality, let the set of result quality be:

[0020] in, This indicates that the result is valid; This indicates an abnormal result; This indicates that the result is unavailable.

[0021] Furthermore, in step S3, the industrial task is a task instance triggered by one or more A2A Messages and executed by the target industrial equipment agent; the j-th task of the i-th industrial equipment agent is defined. for:

[0022] in, This represents a unique identifier for the task; Represents a session identifier, used to associate multiple tasks or multiple rounds of interaction within the same context; Indicates the current state of the task; This indicates the message history related to the task; This represents the collection of Artifacts generated during or after the task execution process; This represents task-level industrial metadata; Used to describe task scheduling and execution constraints, defined as follows:

[0023]

[0024]

[0025] in, This indicates the task priority; the higher the value, the more urgent the task. These indicate the service quality level, corresponding to real-time priority processing, transactional reliable transmission, and normal best-effort transmission, respectively. Indicates the execution interval of a periodic task; Based on AgentCard.skills, determine whether the agent has the ability to process the task input corresponding to the Message. When the ability matching result is true, the target agent creates a Task and enters the task execution process; when the ability matching result is false, the task enters a failure state. An object used to describe the current state of a task, through Track the execution status of tasks:

[0026] in, This represents an A2A message related to the current state. Indicates the state generation time; The task status is represented and defined as follows:

[0027] in, This indicates that the task has been submitted. Indicates that the task is in progress. This indicates that the task is complete. This indicates that the task failed.

[0028] Furthermore, in step S3, the mode selection is completed based on the expected execution time of the task, the task feedback cycle, the expected number of results, and the single response adaptability derived from the first three factors. The interaction feature vector of industrial tasks is defined as follows:

[0029] in, Indicates the execution time stress factor; Indicates the periodic feedback demand factor; Indicates a factor that returns multiple results; Represents the single-response adaptation factor; defines the normalization function:

[0030] Execution duration stress factor This is used to measure whether the estimated execution time of a task is close to or exceeds the acceptable waiting time for the task initiator, and it is expressed as:

[0031] Periodic feedback demand factor This is used to measure whether a task requires periodic status feedback, and it is expressed as:

[0032] in, Indicates the feedback cycle required for the task; This indicates the system's preset high-frequency feedback reference period.

[0033] Multi-result return factor This is used to determine whether a task will produce multiple stage results or multiple artifacts, and it is represented as:

[0034] in, This indicates the number of results or artifacts that the current task is expected to produce.

[0035] Single Response Adaptation Factor This indicates whether the current task is suitable for a single request and response via tasks / send, and is represented as:

[0036] The interaction mode selection function is:

[0037]

[0038] in,

[0039]

[0040] The final matching rules are: .

[0041] Furthermore, in step S3, the task-driven A2A collaborative interaction process is as follows: (a) The system receives the industrial task message sent to it; (b) Read the AgentCard corresponding to the target industrial equipment to be docked; (c) Obtain the AgentCard profile data and real-time operating status of the target intelligent agent; (d) Determine whether the task Message.parts of the industrial task message matches the capability set AgentCard.skills in the agent file. If the match fails, jump directly to the end of the process and terminate the current interaction process; if the match succeeds, enter the interaction mode determination calculation stage (e). (e) Solving for the interaction mode determination factor ; (f) Calculation of model score , ,like If the condition is met, a subscription-based interaction will be executed; otherwise, a request-response interaction will be executed.

[0042] Furthermore, in step S3, under the request-response interaction mode, we have: The task-initiating agent sends a Message to the target industrial equipment agent via tasks / send. The target agent parses the Part in the Message, reads the constraint information in Message.metadata, and determines whether it has the ability to execute based on AgentCard.skills; After a successful match, the target agent creates a Task and sets Task.status to submitted, then enters the working state; After the task is completed, the target agent generates an Artifact and returns the Task containing Task.status and Artifact to the task initiator. In a subscription-based interaction model, there are: The task-initiating agent submits a message to the target industrial equipment agent via tasks / sendSubscribe and subscribes to subsequent state changes and result updates of the task. After receiving the Message, the target industrial equipment intelligent agent parses Message.parts and Message.metadata, performs capability matching based on AgentCard.skills, and creates a Task after a successful match. After the target agent creates a Task, it updates Task.status multiple times during task execution and generates staged Artifacts based on task progress. When a task is finally completed or fails, the target industrial equipment agent identifies the end of the task through a final task status update event and terminates the corresponding subscription stream.

[0043] The beneficial effects of this invention are as follows: This invention addresses the problems of industrial equipment lacking a unified intelligent agent abstraction, inconsistent semantics in inter-device interactions, rigid communication modes, and low collaborative efficiency under heterogeneous computing power conditions. This invention constructs a unified intelligent agent encapsulation mechanism that automatically extracts, maps, and converts multi-source heterogeneous industrial protocol data from the underlying layer into semantic expressions understandable by an AI Agent.

[0044] This invention addresses the current situation where industrial equipment primarily relies on traditional industrial protocols or proprietary interfaces for communication, making it difficult to directly connect to native intelligent agent collaborative networks and hindering standardized task interaction and semantic collaboration with native intelligent agents. By encapsulating intelligent agents, this invention endows industrial equipment with unified identity identification, capability descriptions, and standard communication interfaces, fundamentally breaking down the interaction barriers between industrial equipment and native intelligent agents. This invention introduces an Agent-to-Agent (A2A) communication mechanism, abstracting and encapsulating the functions, states, and communication capabilities of industrial equipment into standard intelligent agent interfaces. This shields the heterogeneity of underlying hardware, interfaces, and protocols, enabling industrial equipment to participate in system collaboration as a unified intelligent agent. This achieves task-oriented and semantic-based inter-device interaction, thus providing fundamental support for building open, flexible, and scalable industrial intelligent collaborative systems.

[0045] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0046] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the overall process of the intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the A2A communication network after the intelligent packaging of industrial equipment according to an embodiment of the present invention; Figure 3 This is a flowchart of a task-driven A2A collaborative interaction in an embodiment of the present invention; Figure 4 This is a sequence diagram of the request-response pattern interaction under an embodiment of the present invention; Figure 5 This is a timing diagram of the subscription-based interaction in an embodiment of the present invention. Detailed Implementation

[0047] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0048] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0049] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0050] Please see Figures 1-5 This is a method for intelligent encapsulation and communication of industrial equipment based on the A2A protocol.

[0051] Example 1 This embodiment first provides a method for intelligent encapsulation and communication of industrial equipment based on the A2A protocol, such as... Figure 1 As shown, the whole process includes the following steps: S1. Based on the A2A protocol, various types of equipment in the industrial field are modeled through an intelligent agent encapsulation mechanism. Without changing the original physical structure and control logic of the equipment, a unified intelligent agent model of industrial equipment is built for each type of industrial equipment. S2. Construct an A2A semantic communication extension model for industrial scenarios, design the content units in the A2A protocol, and extend the Message object and Artifact object for industrial applications. S3. Construct an A2A interaction mechanism based on industrial task-driven operation, define the tasks and task states of industrial equipment agents, and define a matching mechanism for selecting interaction modes based on task characteristics; realize communication between industrial equipment agents based on the interaction mechanism.

[0052] like Figure 2 As shown, the encapsulated device agent can access the inter-agent communication network. This invention follows the object structure and interaction semantics of the A2A standard protocol, without altering the basic field definitions of standard objects such as AgentCard, Message, Part, Task, Task.status, and Artifact. Extended information required in industrial scenarios, such as device scheduling constraints, task priorities, service levels, feedback cycles, result quality, and time constraints, is expressed through the metadata field in the A2A standard object or the structured data in the Part. Specifically, Message.parts carries the task input text, Artifact.parts carries the task output results, and metadata is only used to express industrial extended information related to task scheduling, execution constraints, and result attributes; it does not replace A2A standard fields, nor does it change the basic mechanisms of task creation, task status updates, and task result returns in the A2A protocol. In this way, the encapsulated industrial device agent can maintain interoperability with the native A2A agent while meeting the extended requirements of industrial sites for task semantics, execution constraints, and result traceability.

[0053] The definitions of terms used in this invention are shown in Table 1: Table 1

[0054] In step S1 of this embodiment, the industrial equipment intelligent agent encapsulation modeling specifically includes: In this invention, various types of equipment in the industrial field can be divided into resource-constrained equipment and resource-abundant equipment, all of which are mapped into independent industrial device agents (IDAs) through an intelligent agent encapsulation mechanism. Based on the A2A protocol, through industrial equipment intelligent agent encapsulation modeling, a unified industrial device agent model is constructed for each type of industrial equipment without changing the original physical structure and control logic of the equipment, abstracting the physical capabilities, data interfaces, and control behaviors of the equipment into standardized intelligent agent capability units.

[0055] First, industrial field equipment is classified, including but not limited to: (1) resource-constrained equipment; (2) resource-abundant equipment.

[0056] For resource-constrained devices, due to limitations in computing, storage, and communication capabilities, they typically cannot independently run a complete A2A protocol stack, nor can they directly maintain A2A standard objects such as AgentCard, Task, Message, and Artifact. Therefore, this invention adopts an external agent adaptation approach: the resource-constrained device is connected to an independent external intelligent agent encapsulation device, where a proxy-type industrial equipment intelligent agent running on this device represents the physical device in A2A communication.

[0057] In this approach, the intelligent agents in the A2A communication network are agent-type industrial equipment intelligent agents on external intelligent agent packaging devices, rather than the resource-constrained physical devices themselves. The agent-type industrial equipment intelligent agents are responsible for maintaining the AgentCard, receiving and parsing A2A Messages, creating and updating A2A Tasks, and mapping underlying industrial protocol data, register data, sensor sampling values, or actuator control interfaces to Part, Message, Artifact, and Task.status objects in the A2A standard objects. The mapping between resource-constrained physical devices and agent-type industrial equipment intelligent agents in external intelligent agent packaging devices can be one-to-one or one-to-many.

[0058] For resource-rich equipment, since it has the ability to run intelligent agent encapsulation logic, maintain A2A standard objects, and conduct network communication, industrial equipment intelligent agent encapsulation modules can be directly deployed in the equipment body or its built-in control unit, so that it can issue AgentCards to the outside world as an independent industrial equipment intelligent agent and directly participate in task collaboration and interaction through the A2A protocol.

[0059] Next, a unified model of intelligent agents for industrial equipment is constructed. In the A2A protocol, the AgentCard is a standard descriptive object for an intelligent agent to publish its basic information, access methods, protocol capabilities, and callable skills. This invention constructs a unified model of intelligent agents for industrial equipment based on the A2A standard AgentCard, enabling industrial equipment to declare its access points, protocol interaction capabilities, input / output modalities, and executable industrial tasks to external systems in the form of a standard intelligent agent.

[0060] The AgentCard for industrial equipment intelligent agents does not directly expose the register addresses, private protocol messages, internal control logic, or hardware driver details of the underlying devices. This underlying information is maintained internally by the agent's encapsulation layer, and the mapping between the A2A standard object and the underlying industrial protocol is completed during task execution. The AgentCard primarily declares to the outside world the standardized capabilities of the industrial equipment intelligent agent to be discovered, understood, and invoked by other intelligent agents. This invention constructs a unified model for industrial equipment intelligent agents based on the standard A2A AgentCard. The meanings of the main fields in the Industrial Equipment Intelligent AgentCard are shown in Table 2 below: Table 2

[0061] The capabilities in AgentCard represent the interactive capabilities that an agent supports at the A2A protocol level, rather than the physical capabilities of the underlying industrial equipment. It can be represented as:

[0062] in, This indicates the protocol capabilities of intelligent agents in industrial equipment; Indicates whether the agent supports streaming responses; Indicates whether the agent supports sending push notifications for asynchronous task updates; This indicates whether the agent supports a history of task state transitions.

[0063] `skills` represents the task capabilities that an industrial equipment agent can perform, and is the field in AgentCard most directly related to the expression of industrial equipment capabilities. Let the agent's skill set be... for:

[0064] Each task capability Defined as:

[0065] in, Indicates a unique identifier for a skill; Indicates the skill name; Indicates skill description; Indicate skill tags; This indicates the types of input content that the skill supports; Indicates the types of output content supported by this skill; Provide sample hints or scenarios that can be handled.

[0066] By reading the AgentCard of the target industrial equipment, the task initiator or other intelligent agents can determine whether the target intelligent agent has the ability to handle specific industrial tasks and further select an appropriate A2A task interaction method.

[0067] In step S2 of this embodiment, communication between agents in the A2A protocol is mainly accomplished through standard objects such as Message, Part, Task, Task.status, and Artifact. For industrial equipment collaboration scenarios, it is necessary to ensure that task input, execution constraints, task status, interim results, and final results can be uniformly expressed and tracked without changing the structure of the A2A standard objects.

[0068] Therefore, this invention constructs an extended A2A semantic communication model for industrial scenarios, rather than redefining the A2A protocol standard. This model is based on standard A2A communication objects, using Message.parts to carry industrial task input, Artifact.parts to carry industrial task output, Task and Task.status to maintain the task lifecycle, and metadata fields to express extended attributes such as industrial task scheduling, time constraints, priority, service level, data quality, and result order.

[0069] Specifically, a Part is the basic unit in the A2A protocol used to carry specific content. A Message or Artifact can consist of one or more Parts. Through Parts, intelligent agents in industrial equipment can simultaneously transmit text descriptions, structured parameters, and file content, that is:

[0070] The TextPart is used to carry text content. Text content is suitable for expressing task descriptions, status explanations, causes of anomalies, handling suggestions, or human-readable conclusions. The DataPart is used to carry structured data. Control parameters, acquisition parameters, status data, and diagnostic indicators in industrial tasks should preferably be expressed through DataPart. DataPart has a clear data structure, facilitating parsing by the target intelligent agent and conversion into underlying device operations. The FilePart is used to carry file content. Industrial tasks may involve detection files, log files, configuration files, or report files; this type of content can be transmitted through FilePart.

[0071] Message is the standard message object used for interaction between intelligent agents in the A2A protocol. After receiving a Message, an industrial equipment intelligent agent can create or update a Task based on its content. The core function of a Message is to carry task input, supplementary information, status descriptions, or other interactive content, namely:

[0072] in, Indicates the message role; Represents a collection of message content units; This represents message-level industrial metadata.

[0073] The task body in a Message is carried by parts. For industrial control, data acquisition, query, or diagnostic tasks, specific parameters should be expressed through DataPart; human-readable descriptions can be expressed through TextPart; and file-type inputs can be expressed through FilePart.

[0074] Message It is not used to carry the task content, nor to replace the skill descriptions in AgentCard.skills. It is primarily used to express message-level industrial scheduling constraints and processing auxiliary information. Its definition is as follows:

[0075] in, Indicates the message semantic type; Used to constrain the expected completion time of message-triggered tasks; Used to determine whether the current message request or task wait has timed out.

[0076] To achieve rapid classification, routing, and processing of different industrial messages, this invention defines a set of industrial semantic message types:

[0077] The message type definitions are shown in Table 3: Table 3

[0078] In this way, Message maintains the standard A2A object structure while providing the necessary time constraints and category information for industrial task scheduling.

[0079] An Artifact is a task result object attached to a Task in the A2A protocol. During or after task execution, the industrial equipment agent encapsulates the result content into an Artifact and returns or pushes it to the task initiator via the Task. An Artifact in an industrial scenario can be defined as follows:

[0080] in, Indicates the name of the result; Indicates the results; This represents a set of result content units; This represents result-level industrial metadata.

[0081] In Artifact, the main body of the results is contained in parts. Structured content such as measured values, status values, control results, and diagnostic indicators should be placed in DataPart; explanatory text should be placed in TextPart; and file-type results should be placed in FilePart.

[0082] Artifact metadata is used to describe the time, unit, quality, and order attributes of the resulting data itself. This invention defines:

[0083] in, Indicates the time when the result was generated; Indicates the unit of data; Indicates data quality; This indicates the result sequence number, used for sorting results in subscription-based or phased tasks.

[0084] The field is used to represent the reliability of the results. Let the set of result quality be:

[0085] in, This indicates that the result is valid; This indicates an abnormal result; This indicates that the result is unavailable.

[0086] In step S3 of this embodiment, after completing the intelligent encapsulation of industrial equipment and the design of semantic communication objects, it is necessary to further define how industrial tasks are created, executed, updated, and return results between A2A agents. The Task in the A2A protocol is the core object of task collaboration, carrying the task lifecycle, task status, message history, and task results. This invention constructs a collaborative interaction mechanism for industrial tasks based on the A2A standard Task, Task.status, and the standard task methods tasks / send and tasks / sendSubscribe.

[0087] An industrial task is a task instance triggered by one or more A2A Messages and executed by a target industrial device agent. This task instance exists in the form of an A2A Task, containing a task identifier, session identifier, task state, message history, result object, and task metadata. Define the j-th task of the i-th industrial device agent. for:

[0088] in, This represents a unique identifier for the task; Represents a session identifier, used to associate multiple tasks or multiple rounds of interaction within the same context; Indicates the current state of the task; This indicates the message history related to the task; This represents the collection of Artifacts generated during or after the task execution process; This represents task-level industrial metadata.

[0089] Task It is not used to carry the task text, nor to express the target skill or operation type, but to describe task scheduling and execution constraints, as defined below:

[0090]

[0091]

[0092] in, This indicates the task priority; the higher the value, the more urgent the task. These indicate the service quality level, corresponding to real-time priority processing, transactional reliable transmission, and normal best-effort transmission, respectively. This indicates the execution interval for periodic tasks.

[0093] Before creating a Task, the target industrial equipment agent should determine whether it has the capability to process the task input corresponding to the Message based on AgentCard.skills. When the capability matching result is true, the target agent creates the Task and enters the task execution process; when the capability matching result is false, the task enters a failure state.

[0094] It is an object in A2A Task used to describe the current state of a task, through... It allows tracking of task execution status. Industrial tasks typically have a well-defined lifecycle, therefore requiring... It indicates the status changes of a task from submission and execution to completion or failure.

[0095]

[0096] in, Indicates the task status; This represents an A2A message related to the current state. Indicates the time when the state was generated.

[0097] This invention targets industrial execution scenarios and uses only four states: submitted, working, completed, and failed. The set of industrial task states is defined as follows:

[0098] The meanings of each state are shown in Table 4 below:

[0099] The A2A protocol offers two task interaction modes: tasks / send and tasks / sendSubscribe. Industrial tasks vary in duration, status feedback requirements, result generation methods, and response methods. Therefore, it is necessary to select the appropriate interaction mode based on the task characteristics to optimize network bandwidth and computing resource utilization.

[0100] To reduce the computational complexity of industrial equipment agents in the interaction pattern matching process, this invention employs a lightweight four-factor model. The target industrial equipment agent selects a pattern based solely on the expected task execution time, task feedback cycle, expected number of results, and the single-response fit derived from the first three factors.

[0101] The interaction feature vector of industrial tasks is defined as follows:

[0102] in, Indicates the execution time stress factor; Indicates the periodic feedback demand factor; Indicates a factor that returns multiple results; This represents the adaptation factor for a single response. Define the normalization function:

[0103] (1) Execution duration stress factor This is used to measure whether the estimated execution time of a task is close to or exceeds the acceptable waiting time for the task initiator, and it is expressed as:

[0104] (2) Periodic feedback demand factor This is used to measure whether a task requires periodic status feedback, and it is expressed as:

[0105] in, Indicates the feedback cycle required for the task; This indicates the system's preset high-frequency feedback reference period.

[0106] (3) Multi-outcome return factor This is used to determine whether a task will produce multiple stage results or multiple artifacts, and it is represented as:

[0107] in, This indicates the number of results or artifacts that the current task is expected to produce.

[0108] (4) Single response adaptation factor This indicates whether the current task is suitable for a single request and response via tasks / send, and is represented as:

[0109] The interaction mode selection function is:

[0110]

[0111] in,

[0112]

[0113] The final matching rules are:

[0114] Task-driven A2A collaborative interaction process, such as Figure 3 As shown, the interaction flow is as follows: (a) The system receives the industrial task message sent to it; (b) Read the AgentCard corresponding to the target industrial equipment to be docked; (c) Obtain the AgentCard profile data and real-time operating status of the target intelligent agent; (d) Determine whether the task Message.parts of the industrial task message matches the capability set AgentCard.skills in the agent file. If the match fails, jump directly to the end of the process and terminate the current interaction process; if the match succeeds, enter the interaction mode determination calculation stage (e). (e) Solving for the interaction mode determination factor ; (f) Calculation of model score , ,like If the condition is met, a subscription-based interaction will be executed; otherwise, a request-response interaction will be executed.

[0115] Specifically, in the request-response interaction mode, the task-initiating agent sends a Message to the target industrial equipment agent via tasks / send. The target agent parses the Part in the Message, reads the constraint information in Message.metadata, and determines whether it has execution capability based on AgentCard.skills. Upon successful matching, the target agent creates a Task, sets Task.status to submitted, and then enters the working state. After task execution, the target agent generates an Artifact and returns the Task containing Task.status and Artifact to the task initiator. The characteristics of tasks / send tasks are: short task duration, low frequency of status feedback, low degree of phased result generation, and high adaptability of a single response. Request-response interaction is as follows: Figure 4 As shown.

[0116] In the subscription-based interaction mode, the task-initiating agent submits a Message to the target industrial equipment agent via `tasks / sendSubscribe` and subscribes to subsequent state changes and result updates for the task. Upon receiving the Message, the target industrial equipment agent parses `Message.parts` and `Message.metadata`, performs capability matching based on `AgentCard.skills`, and creates a Task upon successful matching. After creating the Task, the target agent can update `Task.status` multiple times during task execution and generate staged artifacts based on task progress. When the task is finally completed or fails, the target industrial equipment agent identifies the task's end through a final task status update event and terminates the corresponding subscription stream. The characteristics of `tasks / sendSubscribe` tasks are: long task duration, high frequency of status feedback, high degree of staged result generation, and low adaptability of single responses. Subscription-based interaction is as follows: Figure 5 As shown.

[0117] Example 2 This embodiment verifies the feasibility and effectiveness of the proposed A2A protocol-based intelligent integrated packaging and communication method for industrial equipment. The verification environment is built based on a typical industrial scenario, including one Modbus TCP temperature and humidity sensor node, one PLC device with fan control function, and one Raspberry Pi as an external intelligent integrated packaging device. Focusing on the industrial control task of "starting the fan when the temperature exceeds the limit," the entire process of verifying the intelligent integrated packaging of industrial equipment, constructing industrial A2A semantic communication objects, and task-driven A2A collaborative interaction is completed sequentially.

[0118] In this embodiment, the temperature and humidity sensor and PLC fan control device are packaged into intelligent agents according to resource-constrained and resource-abundant devices, and their capabilities are released externally through the standard AgentCard. Based on the device's computing power, storage resources, and protocol support capabilities, the experimental equipment is divided into two categories and corresponding packaging strategies are adopted: Resource-constrained device: Modbus TCP temperature and humidity sensor. This device only supports Modbus TCP register read and write, has no operating system, and lacks sufficient computing and storage resources to independently run the A2A protocol stack, making it unable to natively generate semantic messages. An external intelligent agent encapsulation device proxy strategy is adopted: the sensor is connected to a Raspberry Pi as an external intelligent agent encapsulation device. The Raspberry Pi runs the complete A2A protocol stack, periodically reads sensor register data, completes the conversion from underlying industrial protocols to A2A semantics, and provides a standard intelligent agent interface to the outside world.

[0119] Resource-rich equipment: PLC fan control unit. This device has Ethernet communication capabilities and local computing resources, and can deploy a lightweight A2A protocol adaptation layer to directly parse semantic messages and execute local control logic. It adopts direct intelligent agent encapsulation: extending the A2A interface layer on the PLC's native control logic, enabling it to directly connect to the A2A collaborative network as an independent industrial intelligent agent.

[0120] The device intelligent agent model is built based on the A2A protocol standard AgentCard, strictly following the field specifications defined in Table 2 of the invention, and fully covering core elements such as identity identification, access entry, protocol capabilities, input and output modalities, and callable skills.

[0121] For the temperature and humidity intelligent agent, it is generated by the Raspberry Pi edge node and registered to the A2A collaborative communication environment.

[0122] Capabilities protocol capability calculation: This agent supports streaming responses, push notifications, and task state transition history, namely:

[0123] Skills set calculation: This agent contains 2 industrial skills, and the skill set is as follows:

[0124] in For temperature and humidity data acquisition skills, For the abnormal event reporting skill, each field strictly matches the seven-tuple definition.

[0125] After filling in all the fields of AgentCard, the complete AgentCard fields for the temperature and humidity intelligent agent are as follows: { "name": "Temperature and Humidity Intelligent Agent", "description": "Encapsulated by an edge node proxy, supporting real-time temperature and humidity data acquisition", "url": "coap: / / 192.168.1.50:5683 / ida / sensor_th_001", "provider": "External intelligent integrated packaging device", "version": "1.0.0", "documentationUrl": "null", "capabilities": { "streaming": true, "pushNotifications": true, "stateTransitionHistory": true }, "defaultInputModes": ["application / json"], "defaultOutputModes": ["application / json"], "skills": [ { "id": "skill_temp_humi_collect", "name": "Temperature and Humidity Data Acquisition", "description": "Collects real-time temperature and humidity values", "tags": ["data_acquisition", "environment_monitor"], "inputModes": [], "outModes": ["application / json"], Examples: ["Get current temperature and humidity data"] }, { "id": "skill_event_notify", "name": "Abnormal Event Reporting", "description": "When environmental parameters are abnormal, proactively report the abnormal event to the target intelligent agent". "tags": ["event_notification", "threshold_alarm"], "inputModes": ["application / json"], "outModes": ["application / json"], Examples: ["Temperature exceeding 35℃ threshold reported"] } ] } For PLC fan control equipment, it can be directly packaged into an intelligent agent for industrial equipment.

[0126] Capabilities protocol capability calculation: This agent also supports streaming responses, push notifications, and state history records, resulting in:

[0127] Skills set calculation: This agent has 1 fan control skill, and the skill set is:

[0128] in For fan operation control skills.

[0129] After filling in all the fields of AgentCard, the complete AgentCard fields for the PLC fan control agent are as follows: { "name": "PLC Fan Control Agent", "description": "Supports fan start / stop and speed adjustment, with a maximum supported speed of 2000rpm", "url": "coap: / / 192.168.1.60:5683 / ida / plc_fan_001", "provider": "Siemens", "version": "1.0.0", "documentationUrl": "null", "capabilities": { "streaming": true, "pushNotifications": true, "stateTransitionHistory": true }, "defaultInputModes": ["application / json"], "defaultOutputModes": ["application / json"], "skills": [ { "id": "skill_fan_motion_ctrl", "name": "Fan Operation Control", "description": "Controls fan start / stop and sets target speed, supporting speed range 0-2000rpm", "tags": ["motion_control", "actuator"], "inputModes": ["application / json"], "outModes": ["application / json"], "examples": ["Start the fan, target speed 1200rpm", "Stop the fan"] } ] } This embodiment, without changing the standard A2A object structure, uses Part to carry the industrial task text, Message to carry task input and message-level constraints, Artifact to carry the task results, and uses metadata fields to express extended attributes such as industrial scheduling, time, quality, and order.

[0130] Following the Part classification standard, this embodiment uses a combination of TextPart and DataPart to represent the "temperature exceeds the limit and the fan starts" task.

[0131] Request Part set (temperature and humidity agent → PLC fan control agent) [ { "type": "TextPart", "text": "Ambient temperature exceeds the 35℃ threshold, requesting fan activation and setting the target speed to 1200rpm." }, { "type": "DataPart", "mimeType": "application / json", "data": { "skill_id": "skill_fan_motion_ctrl", "action": "fan_on", "target_rpm": 1200 } } ] TextPart: A natural language description that carries the task intent, facilitating human operation and maintenance as well as semantic understanding by intelligent agents.

[0132] DataPart: Carries structured control parameters, including target skill ID, action instructions, and speed parameters, which are easily parsed by the PLC and mapped to underlying hardware operations.

[0133] Response Part Collection (PLC Fan Control Agent → Temperature and Humidity Agent) [ { "type": "TextPart", "text": "Fan start command executed successfully, current operating status is normal" }, { "type": "DataPart", "mimeType": "application / json", "data": { "result": "success", "current_rpm": 1200, "output_point": "Q0.1", "point_status": "SET" } } ] Following the Message structure specification, a message consists of three parts: the sending role, a set of content units, and industry metadata. The metadata extends the message's semantic type and timeliness constraints.

[0134] Fan start request Message (action_request type) Constructed by a temperature and humidity intelligent agent, after the selected mode is determined, the data is sent to the PLC fan control intelligent agent. The complete structure is as follows: { "role": "sensor_th_001", "parts": [ { "type": "TextPart", "text": "Ambient temperature exceeds the 35℃ threshold, requesting fan activation and setting the target speed to 1200rpm." }, { "type": "DataPart", "mimeType": "application / json", "data": { "skill_id": "skill_fan_motion_ctrl", "action": "fan_on", "target_rpm": 1200 } } ], "metadata": { "message_type": "action_request", "deadline": 1700000149500, "timeout": 500 } } Field descriptions: role: The identity identifier of the intelligent agent that sent the message.

[0135] parts: A collection of message content units, corresponding to the request Part mentioned above.

[0136] metadata: Strictly corresponds to message-level industrial metadata, including: message_type: The message semantic type is action_request (behavior request), corresponding to the industrial message type definition in Table 3, used for quick message classification, routing, and scheduling. deadline: The expected completion time stamp of the task (unit: ms), constraining the PLC to complete instruction execution before this time. timeout: The request timeout period (unit: ms), if no response is received after this time, it is considered a communication failure.

[0137] Following the Artifact structure specification, the task result object consists of a name, description, content unit, and result metadata. The metadata describes the time, unit, quality, and order attributes of the result.

[0138] After the PLC fan control agent executes the control command, it generates this artifact, attaches it to the Task, and returns it to the initiator. The complete structure is as follows: { "name": "Fan start execution result", "description": "Execution status and real-time operating parameters of fan start control commands", "parts": [ { "type": "TextPart", "text": "Fan start command executed successfully, current operating status is normal" }, { "type": "DataPart", "mimeType": "application / json", "data": { "result": "success", "current_rpm": 1200, "output_point": "Q0.1", "point_status": "SET" } } ], "metadata": { "timestamp": 1700000149400, "unit": { "current_rpm": "rpm", "result": "enum }, "quality": "valid", "sequence": 1 } } Field descriptions: name: The name of the response result.

[0139] description: A description of the response result.

[0140] parts: A collection of result content units, corresponding to the response Part mentioned above.

[0141] Metadata: Strictly corresponds to result-level industrial metadata, including: timestamp: Result generation timestamp (unit: ms), precisely marking the time the data was generated. unit: Dimensional description of the result data, specifying the physical unit. quality: Result quality is valid, corresponding to the result quality set definition, indicating data reliability. sequence: Result sequence number is 1, indicating that this is a single task producing only one result, used for sorting in subscription-based multi-result scenarios.

[0142] After completing the intelligent encapsulation of industrial equipment and the design of semantic communication objects, this embodiment further verifies how industrial tasks can be created, executed, updated in status, and returned in return using the A2A standard Task.

[0143] Following the Task structure specification, define a "start fan if temperature exceeds limit" task instance, which includes task identifier, session, status, message history, result set and task metadata.

[0144] After receiving the request message, the PLC fan control agent first performs a message timeliness check. After confirming that the message has not expired, it then performs the following capability matching and parameter verification: Read the task input parameters from DataPart in Message.parts; The controlled object is identified as fan, the control action is fan_on, and the target speed is 1200 rpm; Check your AgentCard.skills to confirm the existence of the skill_fan_motion_ctrl skill; Determine if skill_fan_motion_ctrl supports fan start and target speed setting; Determine if target_rpm=1200 is within the allowed range of 0-2000rpm; If the capability matching result is true, create a Task.

[0145] After successful capability matching, the PLC fan control agent creates an initial Task with a submitted status. { "id": "task_temp_fan_ctrl_001", "sessionId": "session_line01_env_monitor_20231114", "status": { "state": "submitted", "message": "msg_request_fan_on_001", "timestamp": 1700000149000 }, "history": ["msg_request_fan_on_001"], "artifacts": [], "metadata": { "priority": 4, "qosLevel": "transactional", "period": 0 } } Field descriptions: id: A globally unique identifier for the task.

[0146] sessionId: Session identifier.

[0147] status: The task status object, initially in the state of submitted.

[0148] history: A list of message history associated with the task, providing a complete record of the entire interaction chain.

[0149] artifacts: The collection of task results, initially empty.

[0150] Metadata: Task-level industrial metadata, including: Priority: Task priority is 4, corresponding to levels 0-5, with higher values ​​indicating greater urgency. qosLevel: Service quality level is transactional (reliable transactional transmission), corresponding to the QoS level definition, ensuring traceable and lossless instruction execution. Period: Task period is 0, indicating a single-execution task with no periodic scheduling.

[0151] The task lifecycle transition process is as follows: Submitted status: The PLC fan control agent receives the Message, completes the timeliness verification, capability matching and parameter verification, and then creates a Task. Working state: The PLC calls the underlying control logic to execute fan start and target speed setting; Completed status: Fan started successfully, PLC generated Artifact and attached it to Task; Failed status: If there is an illegal parameter, the target speed exceeds the limit, the PLC output point is abnormal, or the task times out, the task will enter the failed status.

[0152] Following a lightweight four-factor model, the system calculates the compatibility scores of the two interaction modes based on task characteristics and automatically selects the optimal communication mode.

[0153] (1) Setting of characteristic parameters For this "temperature exceeds limit and fan starts" task, the input characteristic parameters are as follows: Expected execution time =200ms (total time for PLC coil setting + status feedback); Allowable waiting time for the task. =1000ms; This is a single control task with no need for periodic status feedback; Expected number of results k=1 (only one execution result will be returned); Weighting coefficient: =0.4, =0.3, =0.3.

[0154] (2) Factor calculation Calculate each characteristic factor in turn: Execution duration stress factor :

[0155] Periodic feedback demand factor This is a one-time control task with no need for periodic feedback, therefore... =0.

[0156] Multi-result return factor :

[0157] Single Response Adaptation Factor :

[0158] Pattern Scoring and Decision: Calculating the fit score between two types of interaction patterns: Request-Response Pattern Rating: =0.8 Subscription model rating: =0.4×0.2+0.3×0+0.3×0=0.08 Matching rules: < Therefore, this task adopts a request-response (tasks / send) interaction mode, which is perfectly matched with the characteristics of short task duration, single result, and no periodic feedback.

[0159] The complete execution steps of the request-response sequence logic corresponding to the invention content are as follows: Task Initiation: The Raspberry Pi reads the temperature and humidity sensor value as 38.6℃ via Modbus TCP, which exceeds the 35℃ threshold; the temperature and humidity agent constructs the above action_request type Message and sends it to the PLC fan control agent through the tasks / send interface; Message parsing and verification: After receiving the Message, the PLC fan control agent performs three-layer verification: 1) Timeliness verification: The current timestamp is 1700000149100, which is earlier than the deadline=1700000149500 in the metadata, so the message has not timed out; 2) Ability matching: Query the skills field of AgentCard to confirm the existence of the skill_fan_motion_ctrl skill and that the input parameter format matches; 3) Constraint verification: target_rpm=1200rpm is within the 0-2000rpm range of the skill constraint, so the parameter is valid.

[0160] Task creation and execution: After verification, the PLC fan control agent creates the above Task instance and sets its status to submitted; then it calls the underlying native instruction PLC_COIL_Q0.1_SET to set the output point Q0.1, starts the fan, and the task status is synchronously switched to working.

[0161] Result generation and return: After the fan starts successfully, the PLC generates the above Artifact object, attaches it to the artifacts collection of the Task, and updates the task status to completed; finally, the complete Task object is returned to the initiating temperature and humidity intelligent agent.

[0162] The final returned Task instance in the completed state { "id": "task_temp_fan_ctrl_001", "sessionId": "session_line01_env_monitor_20231114", "status": { "state": "completed", "message": "msg_response_fan_on_001", "timestamp": 1700000149400 }, "history": ["msg_request_fan_on_001", "msg_response_fan_on_001"], "artifacts": [ { "name": "Fan start execution result", "description": "Execution status and real-time operating parameters of fan start control commands", "parts": [ { "type": "TextPart", "text": "Fan start command executed successfully, current operating status is normal" }, { "type": "DataPart", "mimeType": "application / json", "data": { "result": "success", "current_rpm": 1200, "output_point": "Q0.1", "point_status": "SET" } } ], "metadata": { "timestamp": 1700000149400, "unit": { "current_rpm": "rpm", "result": "enum }, "quality": "valid", "sequence": 1 } } ], "metadata": { "priority": 4, "qosLevel": "transactional", "period": 0 } } Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent encapsulation and communication of industrial equipment based on the A2A protocol, characterized in that: The method includes: S1. Based on the A2A protocol, various types of equipment in the industrial field are modeled through an intelligent agent encapsulation mechanism. Without changing the original physical structure and control logic of the equipment, a unified intelligent agent model of industrial equipment is built for each type of industrial equipment. S2. Construct an A2A semantic communication extension model for industrial scenarios, design the content unit Part in the A2A protocol, and extend the Message object and Artifact object for industrial applications. S3. Construct an A2A interaction mechanism based on industrial task-driven operation, define the tasks and task states of industrial equipment agents, and define a matching mechanism for selecting interaction modes based on task characteristics; realize communication between industrial equipment agents based on the interaction mechanism.

2. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 1, characterized in that: In step S1, various types of equipment in the industrial site are classified into resource-constrained equipment and resource-abundant equipment. For resource-constrained devices, an external agent adaptation approach is adopted: the resource-constrained device is connected to an independent external intelligent agent packaging device, and an agent-type industrial device intelligent agent running in this device represents the physical device in A2A communication; the agent-type industrial device intelligent agent maintains AgentCard, receives and parses A2A Message, creates and updates A2ATask, and maps underlying industrial protocol data, register data, sensor sampling values, or actuator control interfaces to Part, Message, Artifact, and Task.status in A2A standard objects; the mapping between the resource-constrained physical device and the agent-type industrial device intelligent agent in the external intelligent agent packaging device is one-to-one or one-to-many. For resource-rich equipment, an industrial equipment intelligent agent encapsulation module is directly deployed in the equipment body or its built-in control unit, enabling it to issue AgentCards to the outside world as an independent industrial equipment intelligent agent and participate in task collaboration and interaction directly through the A2A protocol.

3. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 2, characterized in that: In step S1, a unified model of industrial equipment intelligent agents is constructed based on the A2A standard AgentCard, enabling industrial equipment to declare its access points, protocol interaction capabilities, input / output modes, and executable industrial task capabilities to external systems in the form of a standard intelligent agent. The Industrial Equipment Intelligent Agent Card should include at least the following fields: capabilities represent the interactive capabilities that an intelligent agent supports at the A2A protocol level, and are expressed as: in, This indicates the protocol capabilities of intelligent agents in industrial equipment; Indicates whether the agent supports streaming responses; Indicates whether the agent supports sending push notifications for asynchronous task updates; Indicates whether the agent supports a history of task state transitions; skills represent the industrial task capabilities that the intelligent agent of the industrial equipment can invoke; let the skill set of the intelligent agent be... for: Each task capability Defined as: in, Indicates a unique identifier for a skill; Indicates the skill name; Indicates skill description; Indicate skill tags; This indicates the types of input content that the skill supports; Indicates the types of output content supported by this skill; Provide sample hints or scenarios that can be handled.

4. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 1, characterized in that: In step S2, the content unit Part is the basic unit in the A2A protocol used to carry specific content; a Message or Artifact consists of one or more Parts. Through Parts, intelligent agents in industrial equipment can simultaneously transmit text descriptions, structured parameters, and file content, that is: Among them, TextPart is used to carry text content; DataPart is used to carry structured data; and FilePart is used to carry file content.

5. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 1, characterized in that: In step S2, Message is the standard message object for interaction between intelligent agents in the A2A protocol; after receiving the Message, the industrial equipment intelligent agent creates or updates a Task based on its content; Message is used to carry task input, supplementary information, status descriptions, or other interactive content, that is: in, Indicates the message role; This represents a collection of message content units, containing the task body; This represents message-level industrial metadata, used to express message-level industrial scheduling constraints and processing auxiliary information. It is defined as follows: in, Indicates the message semantic type; Used to constrain the expected completion time of message-triggered tasks; Used to determine whether the current message request or task wait has timed out.

6. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 1, characterized in that: In step S2, Artifact is the task result object attached to Task in the A2A protocol; during or after task execution, the industrial equipment intelligent agent encapsulates the result content into Artifact and returns or pushes it to the task initiator through Task, which is represented as follows: in, Indicates the name of the result; Indicates the results; This represents a collection of result content units used to carry the main text of the result. This represents result-level industrial metadata, used to describe the time, unit, quality, and order attributes of the result data itself, and is represented as follows: in, Indicates the time when the result was generated; Indicates the result sequence number, used for sorting results in subscription-based or phased tasks; Indicates the unit of data, used to indicate the reliability of the result; To represent data quality, let the set of result quality be: in, This indicates that the result is valid; This indicates an abnormal result; This indicates that the result is unavailable.

7. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 1, characterized in that: In step S3, an industrial task is a task instance triggered by one or more A2A Messages and executed by the target industrial equipment agent; the j-th task of the i-th industrial equipment agent is defined. for: in, This represents a unique identifier for the task; Represents a session identifier, used to associate multiple tasks or multiple rounds of interaction within the same context; Indicates the current state of the task; This indicates the message history related to the task; This represents the collection of Artifacts generated during or after the task execution process; This represents task-level industrial metadata; Used to describe task scheduling and execution constraints, defined as follows: in, This indicates the task priority; the higher the value, the more urgent the task. These indicate the service quality level, corresponding to real-time priority processing, transactional reliable transmission, and normal best-effort transmission, respectively. Indicates the execution interval of a periodic task; Based on AgentCard.skills, determine whether it has the ability to process the task input corresponding to the Message. When the ability matching result is true, the target agent creates a Task and enters the task execution process; when the ability matching result is false, the task enters a failure state. An object used to describe the current state of a task, through Track the execution status of tasks: in, This represents an A2A message related to the current state. Indicates the state generation time; The task status is represented and defined as follows: in, This indicates that the task has been submitted. Indicates that the task is in progress. This indicates that the task is complete. This indicates that the task failed.

8. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 7, characterized in that: In step S3, the mode selection is completed based on the expected execution time of the task, the task feedback cycle, the expected number of results, and the single response adaptability derived from the first three factors. The interaction feature vector of industrial tasks is defined as follows: in, Indicates the execution time stress factor; Indicates the periodic feedback demand factor; Indicates a factor that returns multiple results; Represents the single-response adaptation factor; defines the normalization function: Execution duration stress factor This is used to measure whether the estimated execution time of a task is close to or exceeds the acceptable waiting time for the task initiator, and it is expressed as: Periodic feedback demand factor This is used to measure whether a task requires periodic status feedback, and it is expressed as: in, Indicates the feedback cycle required for the task; This indicates the system's preset high-frequency feedback reference period; Multi-result return factor This is used to determine whether a task will produce multiple stage results or multiple artifacts, and it is represented as: in, This indicates the number of results or artifacts that the current task is expected to produce; Single Response Adaptation Factor This indicates whether the current task is suitable for a single request and response via tasks / send, and is represented as: The interaction mode selection function is: in, The final matching rules are: 。 9. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 8, characterized in that: In step S3, the task-driven A2A collaborative interaction process is as follows: (a) The system receives the industrial task message sent to it; (b) Read the AgentCard corresponding to the target industrial equipment to be docked; (c) Obtain the AgentCard profile data and real-time operating status of the target intelligent agent; (d) Determine whether the task Message.parts of the industrial task message matches the capability set AgentCard.skills in the agent file. If the match fails, jump directly to the end of the process and terminate the current interaction process. If the match is successful, proceed to the interactive mode judgment and calculation stage (e). (e) Solving for the interaction mode determination factor ; (f) Calculation of model score , ,like If the condition is met, a subscription-based interaction will be executed; otherwise, a request-response interaction will be executed.

10. The intelligent encapsulation and communication method for industrial equipment based on the A2A protocol according to claim 9, characterized in that: In step S3, under the request-response interaction mode, the following applies: The task-initiating agent sends a Message to the target industrial equipment agent via tasks / send. The target agent parses the Part in the Message, reads the constraint information in Message.metadata, and determines whether it has the ability to execute based on AgentCard.skills; After a successful match, the target agent creates a Task and sets Task.status to submitted, then enters the working state; After the task is completed, the target agent generates an Artifact and returns the Task containing Task.status and Artifact to the task initiator. In a subscription-based interaction model, there are: The task-initiating agent submits a message to the target industrial equipment agent via tasks / sendSubscribe and subscribes to subsequent state changes and result updates of the task. After receiving the Message, the target industrial equipment intelligent agent parses Message.parts and Message.metadata, performs capability matching based on AgentCard.skills, and creates a Task after a successful match. After the target agent creates a Task, it updates Task.status multiple times during task execution and generates staged Artifacts based on task progress. When a task is finally completed or fails, the target industrial equipment agent identifies the end of the task through a final task status update event and terminates the corresponding subscription stream.

Images