Greenhouse crop regulation method and system based on digital twinning and agent reasoning

Abstract

The present application relates to the technical field of intelligent regulation and control of greenhouse crops, in particular to a greenhouse crop regulation and control method and system based on digital twinning and intelligent agent reasoning. In view of the problem that in the regulation and control of multiple devices in a greenhouse, the execution capability of the devices, the influence range of the actions and the connection of the front and rear controls are difficult to confirm, regulation and control state data containing device operation state, controller receipt and sensor response are obtained, a joint state object of crops, environment and devices is generated, a device execution commitment object and an engineering scope segment are further generated, a device regulation and control relay diagram is constructed, and a device regulation and control relay chain is formed by intelligent agent reasoning, control instructions are gradually released after the response echo verifies the handover triggering condition, and when the verification is not satisfied, the release is suspended and the digital twinning and the device execution commitment object are updated, and the method is suitable for multi-device linkage regulation and control of greenhouse crops.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for greenhouse crops, specifically to a method and system for controlling greenhouse crops based on digital twins and agent reasoning. Background Technology

[0002] Greenhouse crop control typically utilizes environmental sensors, crop image acquisition equipment, root zone monitoring equipment, and actuator controllers to acquire data such as temperature, humidity, light intensity, carbon dioxide concentration, soil moisture content, crop growth, and equipment operating status. Based on threshold rules, expert experience, model predictions, or optimization algorithms, control commands for ventilation, shading, supplemental lighting, irrigation, fertilization, and temperature control are generated. With the development of digital twin and intelligent reasoning technologies, existing greenhouse control systems also utilize virtual models to simulate crop growth and changes in the greenhouse environment to aid in determining control objectives and equipment control strategies.

[0003] In actual greenhouse operation, the actions of different actuators involve response delays, operational margins, mutual exclusion conditions, and side effects, and there is often a time lag between controller feedback and sensor changes. Existing methods often focus on directly generating control quantities based on environmental deviations or crop needs, without adequately considering the actual executable state of the equipment, the scope of impact after the action, and the connection between preceding and subsequent control commands. This can easily lead to situations such as the premature release of subsequent control commands, mutual cancellation of actions between different devices, unconfirmed equipment responses, or external disturbances being mistaken for control results. Consequently, it is difficult to maintain continuous, verifiable, and correctable multi-device collaborative control during greenhouse regulation. Therefore, a greenhouse crop regulation method and system based on digital twins and agent reasoning is needed to solve these problems. Summary of the Invention

[0004] To address the above problems, this invention provides the following technical solution: a greenhouse crop regulation method based on digital twins and agent reasoning, comprising: S1. Obtain control status data including equipment operating status, controller feedback and sensor response, and generate a joint status object of crops, environment and equipment; S2. Generate a device execution commitment object based on the device operating status, controller feedback, and sensor response in the crop, environment, and device joint state object; S3. Input the crop, environment and equipment joint state object and the equipment execution commitment object into the digital twin to generate an engineering scope fragment; S4. Construct a device control relay diagram based on the engineering domain segment and the device execution commitment object. The device control relay diagram includes takeover, offsetting, and verification relationships between device actions. S5. The intelligent agent inference engine generates a device control relay chain that includes the device actions to be executed and the conditions for the interaction of the devices, based on the preset crop control target and the device control relay diagram. S6. Release the greenhouse execution equipment control command according to the equipment control relay chain, and verify the interaction triggering condition based on the response echo formed by the controller acknowledgment and sensor response; when the verification result is satisfied, release the control command for the next equipment action to be executed; when the verification result is not satisfied, pause the release and update the digital twin and the equipment execution commitment object.

[0005] Furthermore, the regulation status data includes crop phenotypic data, greenhouse environment data, root zone status data, actuator status data, controller feedback data, sensor response data, and future weather window data; the generation of a joint status object for crops, environment, and equipment includes: The crop images in the crop phenotypic data are segmented into canopy regions and leaf posture features and canopy cover features are extracted. The current crop status is calculated by combining the leaf temperature data in the crop phenotypic data. The greenhouse environment data is synchronized over time to generate the greenhouse status. Calculate the root region state based on the root region state data; The device operating mode, drive current, valve opening, fan speed, lamp power, and pump pressure in the execution device status data are analyzed to generate the device operating status; Parse the device identifier, receipt status, and receipt timestamp in the controller receipt data to generate a controller receipt; The sensor response data is parsed to extract the sensor identifier, response field, and response timestamp to generate a sensor response. Determine the sequence of future external disturbances based on the aforementioned future weather window data; The current crop status, greenhouse status, root zone status, equipment operating status, controller feedback, sensor response, and future external disturbance sequence are associated according to the acquisition time to generate the joint state object of crop, environment, and equipment.

[0006] Furthermore, the device execution commitment object includes an executable range, an executable time window, response latency, device margin, mutual exclusion conditions, observable response channels, and side effect pointers; the generation of the device execution commitment object includes: Extract the device type, current operating mode, protection status, device margin, and receipt timestamp from the device operating status and controller feedback; Based on the device type and current operating mode, the device capability table is retrieved to obtain the executable range, executable time window, and mutual exclusion conditions. The response delay is calculated based on the receipt timestamp and the corresponding sensor response timestamp; Match observable response channels and side effect pointers according to the device action mapping table; The executable range, executable time window, response latency, device margin, mutual exclusion conditions, observable response channels, and side effects are associated with the device execution commitment object.

[0007] Furthermore, the generation of the engineering scope fragment includes: The actions of the device to be simulated are generated based on the executable range and executable time window in the device execution commitment object; The actions of the device to be simulated are input into a digital twin along with the combined state objects of the crop, environment, and device to simulate the impact of the actions of the device to be simulated on the crop state, greenhouse state, and other device executable windows. Extract the affected status field, direction of action, magnitude of action, response delay, and affected device identifier; The actions of the device to be simulated, the affected state fields, the direction of action, the magnitude of action, the response delay, and the identifier of the affected device are encapsulated into an engineering scope fragment.

[0008] Furthermore, the step of constructing the equipment control relay diagram based on the engineering domain segment and the equipment execution commitment object includes: The engineering state after the equipment is activated is used as the graph node; The side effects of the previous device action are matched with the affected state field of the next device action, and a takeover relationship is generated when the effect of the previous device action falls within the executable range of the next device action. For the same affected status field, connect the graph nodes whose action direction of the previous device is opposite to that of the action direction of the next device to generate an offsetting relationship; Read the observable response channel and response delay of the previous device action from the device execution commitment object, match the response field in the observable response channel with the affected state field of the next device action, and write the matched response field, response direction and response delay between the corresponding graph nodes to generate a verification relationship; The equipment control relay diagram is composed of the graph nodes, takeover relationships, offset relationships, and verification relationships.

[0009] Furthermore, the generation device regulates the relay chain, including: The current engineering status is determined based on the combined status of crops, environment, and equipment, and the target engineering status is determined based on the preset crop control target in the equipment control relay diagram. Starting from the graph node corresponding to the current project state, retrieve the graph path pointing to the target project state, and use the graph nodes and connections in the graph path as candidate relay paths; Each subsequent graph node retains a candidate relay path that has a verification relationship triggered by the observable response channel of the previous graph node. The device actions, action sequences, and verification relationships in the retained candidate relay paths are converted into device actions to be executed, execution sequences, and cross-trigger conditions to generate a device control relay chain.

[0010] Furthermore, the response echo includes a controller acknowledgment item and a sensor response item; the step of releasing the greenhouse according to the device control relay chain to execute the device control command includes: Release the greenhouse execution device control command corresponding to the first pending device action in the device control relay chain, and mark the next pending device action as pending release; Collect the controller feedback and sensor response corresponding to the current action of the device to be executed, compare the controller feedback and sensor response with the interaction triggering conditions respectively, and generate response echo; When the response echo satisfies the interaction triggering condition, the greenhouse execution device control command corresponding to the next pending device action in the pending release state is released; When the response echo does not meet the interaction triggering condition, the pending release state of the next device action to be executed is maintained, the digital twin and the device execution commitment object are updated according to the response echo, and the engineering scope fragment, device control relay diagram and device control relay chain are regenerated.

[0011] This invention also provides a greenhouse crop control system based on digital twins and agent reasoning, for implementing the above method, comprising: The joint state generation module is used to acquire control state data including equipment operating status, controller feedback and sensor response, and generate a joint state object of crops, environment and equipment; The commitment object generation module is used to generate a device execution commitment object based on the device operating status, controller feedback and sensor response in the crop, environment and device joint state object; The scope fragment generation module is used to input the crop, environment and equipment joint state object and the equipment execution commitment object into the digital twin to generate engineering scope fragments; The relay diagram construction module is used to construct a device control relay diagram based on the engineering scope segment and the device execution commitment object. The device control relay diagram includes takeover, offsetting, and verification relationships between device actions. The relay chain generation module is used to generate a device control relay chain containing the actions of the devices to be executed and the conditions for the interaction of the devices, based on the preset crop control target and the device control relay diagram by the intelligent agent inference engine. The control release module is used to release the greenhouse execution equipment control command according to the equipment control relay chain, and to verify the exchange triggering condition based on the response echo formed by the controller acknowledgment and sensor response; when the verification result is satisfied, the control command for the next equipment action to be executed is released; when the verification result is not satisfied, the release is paused and the digital twin and the equipment execution commitment object are updated.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The greenhouse crop control method and system based on digital twins and agent reasoning organizes equipment operating status, controller feedback, and sensor responses into a joint state object of crops, environment, and equipment. It further generates equipment execution commitment objects, engineering domain fragments, and equipment control relay diagrams. This enables the agent reasoner to generate a control relay chain under constraints such as equipment executable windows, response delays, mutual exclusion conditions, action impact ranges, and the inter-control, cancellation, and verification relationships between equipment actions. Simultaneously, by verifying the trigger conditions of the interaction through response echoes before releasing the next pending equipment action, it improves upon existing methods that fail to adequately consider the actual executable state of the equipment, the impact range after the action, and the connection relationships between preceding and subsequent control commands. This reduces situations such as premature release of subsequent control commands, mutual cancellation of different equipment actions, continued control without confirmation of equipment responses, and external disturbances being mistaken for control results. Attached Figure Description

[0013] Figure 1 A schematic diagram of the greenhouse crop regulation method based on digital twin and intelligent agent reasoning provided by the present invention; Figure 2 A schematic diagram of the greenhouse crop regulation method based on digital twin and intelligent agent reasoning provided by the present invention; Figure 3 This is a schematic diagram of the structure of the greenhouse crop control system based on digital twin and intelligent agent reasoning provided by the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] Please see Figure 1 and Figure 2 , Figure 1 A schematic diagram of the greenhouse crop regulation method based on digital twin and intelligent agent reasoning provided by the present invention; Figure 2 This is a schematic diagram of the greenhouse crop regulation method based on digital twins and agent reasoning provided by the present invention; the present invention provides a greenhouse crop regulation method based on digital twins and agent reasoning, including: S1. Obtain control status data including equipment operating status, controller feedback and sensor response, and generate a joint status object of crops, environment and equipment; Specifically, in this embodiment, the greenhouse edge gateway follows a state cycle. Acquire regulation status data and write the data within the same status cycle into a crop, environment, and equipment joint status object. The crop, environment, and equipment joint status object is denoted as... ,in The current crop status, It is in a greenhouse state. This is the root region state. For equipment operating status, For controller feedback, For sensing response, This represents a sequence of future external disturbances.

[0017] Furthermore, in one embodiment provided in this application, the regulation state data includes crop phenotypic data, greenhouse environment data, root zone state data, actuator state data, controller feedback data, sensor response data, and future weather window data; the generation of a joint state object of crops, environment, and equipment includes: The crop images in the crop phenotypic data are segmented into canopy regions and leaf posture features and canopy cover features are extracted. The current crop status is calculated by combining the leaf temperature data in the crop phenotypic data. The greenhouse environment data is synchronized over time to generate the greenhouse status. Calculate the root region state based on the root region state data; The device operating mode, drive current, valve opening, fan speed, lamp power, and pump pressure in the execution device status data are analyzed to generate the device operating status; Parse the device identifier, receipt status, and receipt timestamp in the controller receipt data to generate a controller receipt; The sensor response data is parsed to extract the sensor identifier, response field, and response timestamp to generate a sensor response. Determine the sequence of future external disturbances based on the aforementioned future weather window data; The current crop status, greenhouse status, root zone status, equipment operating status, controller feedback, sensor response, and future external disturbance sequence are associated according to the acquisition time to generate the joint state object of crop, environment, and equipment.

[0018] Specifically, the edge gateway inputs crop images into the canopy segmentation model to obtain a set of canopy pixels. and the set of effective pixels of the image Canopy cover characteristics according to Calculation: The average tilt angle of the blade skeleton line within the canopy region is calculated to obtain the blade attitude characteristics. ; leaf temperature Air temperature in the same state cycle The difference is calculated to obtain the leaf temperature deviation value. Current crop status as follows: Generation. When synchronizing greenhouse environmental data over time, if the sensor data acquisition time is... The state period is Then select The two most recent sampled values ,according to When synchronized to the state cycle, air temperature, air humidity, light intensity, carbon dioxide concentration, and wind speed constitute the greenhouse state. Root region state according to Calculation, where This represents the root zone water content. The root region conductivity is... The root zone temperature, This refers to the return liquid volume. Equipment operating status. Generates data based on equipment identification, recording equipment operating mode, drive current, valve opening, fan speed, lamp power, and pump pressure. Controller receipt. It consists of device identifier, receipt status, and receipt timestamp; sensor response. It consists of the response sensor identifier, response field, response direction, response amplitude, and response timestamp. Future weather window data is arranged in future time slice order. Each This includes outdoor temperature, solar radiation, wind speed, and rainfall indicators.

[0019] S2. Generate a device execution commitment object based on the device operating status, controller feedback, and sensor response in the crop, environment, and device joint state object; Specifically, the commitment object generation module starts from... Read device operating status and controller receipt Generate a device commitment object using the device identifier as the primary key. For equipment The object of the equipment's commitment is represented as ,in The executable range, For the executable time window, In response to latency, This is for equipment margin. For mutually exclusive conditions, For observable response channels, This refers to a side effect. The object is limited to a specific device. The range of actions and feedback verifications that can be undertaken within the current state cycle.

[0020] Furthermore, in one embodiment provided in this application, the device execution commitment object includes an executable range, an executable time window, a response latency, device margin, mutual exclusion conditions, an observable response channel, and a side effect pointer; the generation of the device execution commitment object includes: Extract the device type, current operating mode, protection status, device margin, and receipt timestamp from the device operating status and controller feedback; Based on the device type and current operating mode, the device capability table is retrieved to obtain the executable range, executable time window, and mutual exclusion conditions. The response delay is calculated based on the receipt timestamp and the corresponding sensor response timestamp; Match observable response channels and side effect pointers according to the device action mapping table; The executable range, executable time window, response latency, device margin, mutual exclusion conditions, observable response channels, and side effects are associated with the device execution commitment object.

[0021] Specifically, the commitment object generation module first reads the device type. Current operating mode Protection status Current continuous runtime and receipt timestamp Equipment margin is based on Calculation, where The maximum continuous operating time recorded in the equipment capacity table. This refers to the equipment availability factor; when the protection status is protection shutdown. When the drive current, pump pressure, or lamp power is within the allowable range When it is within the range of reduction Take the reduction factor. Then use... To retrieve the device capability table using the index, the executable range is obtained. Executable time window and mutual exclusion conditions The commitment object generation module then responds to the sensor. Search and Device The response item whose device identifier, response field, and response direction match, has its response timestamp recorded as... Response latency The calculation involves using the device action mapping table, indexed by device type and action type, to output observable response channels. and side effects The above results, after being associated with the device identifier, form the device's commitment fulfillment object. .

[0022] S3. Input the crop, environment and equipment joint state object and the equipment execution commitment object into the digital twin to generate an engineering scope fragment; Specifically, the scope fragment generation module will As the current state of the digital twin, each device will fulfill its commitment. As the action boundary, the actions of the device to be simulated are generated. In this embodiment, the actions of the device to be simulated are represented as follows: ,in For equipment identification, For the range of motion, The moment the action begins. Duration of the action; amplitude of the action From the executable range The middle is selected according to the current state deviation direction, and satisfies the following conditions: The start time and duration of the action satisfy the following conditions: , .

[0023] Furthermore, in one embodiment provided in this application, the generation of the engineering scope fragment includes: The actions of the device to be simulated are generated based on the executable range and executable time window in the device execution commitment object; The actions of the device to be simulated are input into a digital twin along with the combined state objects of the crop, environment, and device to simulate the impact of the actions of the device to be simulated on the crop state, greenhouse state, and other device executable windows. Extract the affected status field, direction of action, magnitude of action, response delay, and affected device identifier; The actions of the device to be simulated, the affected state fields, the direction of action, the magnitude of action, the response delay, and the identifier of the affected device are encapsulated into an engineering scope fragment.

[0024] Specifically, the digital twin has a pre-set device action response matrix. Matrix elements Indicates equipment The unit action amplitude for the state field The influence coefficient of the device to be simulated. After input, the digital twin is pressed Calculate the status field The change in, of which For response delay decay factor, For future external disturbance sequences, pair the state field The amount of disturbance. If If the response value exceeds the field response threshold, then the field will be... Record as the affected status field; when The direction of action is recorded as rising, when The direction of action is denoted as decreasing, and the magnitude of action is denoted as decreasing. The response latency is directly taken from the device's execution commitment object. The affected device identifier is obtained by retrieving the device capability table from the affected status field. That is, when a change in a certain status field alters the executable time window or executable range of another device, the identifier of that other device is written into the affected device identifier. The scope fragment generation module ultimately... The affected status field, direction of action, magnitude of action, response delay, and affected device identifier are encapsulated into an engineering scope fragment. .

[0025] S4. Construct a device control relay diagram based on the engineering domain segment and the device execution commitment object. The device control relay diagram includes takeover, offsetting, and verification relationships between device actions. Specifically, the relay graph construction module uses project scope fragments. As the source of graph nodes, the engineering state of each device to be simulated after its action is registered as a graph node. Graph nodes Including equipment operation The affected status field includes the direction of action, the magnitude of action, the response delay, and the affected device identifier. The relay diagram construction module sequentially calculates the takeover relationship, offset relationship, and verification relationship between the nodes in the diagram to obtain the equipment control relay diagram. ,in For a set of graph nodes, For the set of takeover relationships, For the set of offsetting relationships, To verify the set of relations.

[0026] Furthermore, in one embodiment provided in this application, the step of constructing a device control relay diagram based on the engineering domain segment and the device execution commitment object includes: The engineering state after the equipment is activated is used as the graph node; The side effects of the previous device action are matched with the affected state field of the next device action, and a takeover relationship is generated when the effect of the previous device action falls within the executable range of the next device action. For the same affected status field, connect the graph nodes whose action direction of the previous device is opposite to that of the action direction of the next device to generate an offsetting relationship; Read the observable response channel and response delay of the previous device action from the device execution commitment object, match the response field in the observable response channel with the affected state field of the next device action, and write the matched response field, response direction and response delay between the corresponding graph nodes to generate a verification relationship; The equipment control relay diagram is composed of the graph nodes, takeover relationships, offset relationships, and verification relationships.

[0027] Specifically, the relay graph construction module compares the nodes of the previous graph one by one according to the order of the graph nodes. and the node in the next diagram When generating a takeover relationship, first read the side effects pointer of the previous device's action. And the affected status field of the subsequent device action ,when At that time, continue to judge the magnitude of the action of the previous device. Does it fall within the executable range corresponding to the action of the next device? , that is, judgment When satisfied and A takeover relationship is established between them. When an offsetting relationship is established, the data is read... and The affected state field and the direction of action are considered. When the two have the same affected state field and opposite directions of action, an offset relationship is generated.

[0028] When generating verification relationships, the relay graph construction module reads the nodes from the previous graph. Observable response channel corresponding to device action And response latency, and read the nodes of the next graph. The set of affected status fields for corresponding device actions Among them, the observable response channel Each item in is represented as , For response fields, In response to the direction, To account for latency, the relay graph construction module calculates the verifiable response matching result of the previous device's action to the next device's action using the following formula: .when At that time, a verification relationship is generated: and the verification relationship Write to the verification relationship set The verification relationship The response field, response direction, and response delay in the data serve as the basis for determining the triggering conditions when the subsequent intelligent agent inference generator controls the relay chain.

[0029] S5. The intelligent agent inference engine generates a device control relay chain that includes the device actions to be executed and the conditions for the interaction of the devices, based on the preset crop control target and the device control relay diagram. Specifically, the intelligent agent inference engine determines the current engineering state based on the joint state object of crops, environment, and equipment. And determine the target engineering state based on the preset crop regulation target. Current project status as follows: Generate; target project status according to Generate, where For the target crop status, For the target greenhouse condition, This represents the target root region state. The agent inference engine retrieves the state from the device control relay graph. point to The candidate relay paths are identified, and those with complete verification relationships are converted into equipment control relay chains.

[0030] Furthermore, in one embodiment provided in this application, the generation device regulates the relay chain, including: The current engineering status is determined based on the combined status of crops, environment, and equipment, and the target engineering status is determined based on the preset crop control target in the equipment control relay diagram. Starting from the graph node corresponding to the current project state, retrieve the graph path pointing to the target project state, and use the graph nodes and connections in the graph path as candidate relay paths; Each subsequent graph node retains a candidate relay path that has a verification relationship triggered by the observable response channel of the previous graph node. The device actions, action sequences, and verification relationships in the retained candidate relay paths are converted into device actions to be executed, execution sequences, and cross-trigger conditions to generate a device control relay chain.

[0031] Specifically, the agent inference engine first calculates the distance between the current project state and each graph node. The node with the smallest distance is used as the starting point for retrieval; then the distance between each node and the target project status is calculated. The graph node with the smallest distance is selected as the search endpoint. Then, a graph path is searched from the search starting point along the takeover and verification relationships, and paths containing offset relationships are assigned an offset cost. The total cost of candidate relay paths is calculated as follows: Calculation, where This is the sum of the response delays of the actions of each device in the path. This is for equipment margin. and The agent inference engine ranks candidate relay paths by total cost from smallest to largest. For each candidate relay path, the agent inference engine checks whether the next graph node has a verification relationship triggered by the observable response channel of the previous graph node; if any subsequent graph node lacks a verification relationship, the path is eliminated. The path with the lowest total cost among the remaining candidate relay paths is converted into a device control relay chain, where device actions in graph nodes are converted into device actions to be executed, graph node order is converted into execution order, and response fields, response directions, and response delays in verification relationships are converted into cross-trigger conditions.

[0032] S6. Release the greenhouse execution equipment control command according to the equipment control relay chain, and verify the interaction triggering condition based on the response echo formed by the controller acknowledgment and sensor response; when the verification result is satisfied, release the control command for the next equipment action to be executed; when the verification result is not satisfied, pause the release and update the digital twin and the equipment execution commitment object.

[0033] Specifically, the control release module reads the equipment control relay chain, converts the action of the first device to be executed into a control command for the greenhouse execution equipment, and sends it to the corresponding equipment controller. The greenhouse execution equipment control command includes the equipment identifier, command number, action type, action amplitude, start time, and duration. After the command is sent, the control release module collects the controller's acknowledgment and sensor response, generating a response echo. ,in For controller receipt items, This is the sensor response item. The response echo is used to determine whether the current device action has been acknowledged by the controller and whether a response consistent with the interaction triggering conditions has been generated on the observable response channel.

[0034] Furthermore, in one embodiment provided in this application, the response echo includes a controller acknowledgment item and a sensor response item; the step of releasing the greenhouse according to the device control relay chain to execute the device control command includes: Release the greenhouse execution device control command corresponding to the first pending device action in the device control relay chain, and mark the next pending device action as pending release; Collect the controller feedback and sensor response corresponding to the current action of the device to be executed, compare the controller feedback and sensor response with the interaction triggering conditions respectively, and generate response echo; When the response echo satisfies the interaction triggering condition, the greenhouse execution device control command corresponding to the next pending device action in the pending release state is released; When the response echo does not meet the interaction triggering condition, the pending release state of the next device action to be executed is maintained, the digital twin and the device execution commitment object are updated according to the response echo, and the engineering scope fragment, device control relay diagram and device control relay chain are regenerated.

[0035] Specifically, after the control release module releases the control command for the greenhouse equipment at the beginning of the sequence, it writes the next pending equipment action in the equipment control relay chain into the pending release state. (Controller receipt item) Includes device identifier, command number, receipt status, and receipt timestamp; sensor response items. This includes the response sensor identifier, response fields, response direction, response amplitude, and response timestamp. During verification, the control release module first determines... Check if the device identifier and instruction sequence number match the current action to be executed, then check if the receipt status is "execution completed"; then check... The system checks whether the response fields in the code belong to the response fields in the cross-triggered condition, determines whether the response direction is consistent with the response direction in the cross-triggered condition, and also checks the response timestamp. Does it meet the requirements? ,in For the timestamp of the receipt, For the response latency in the device's execution commitment object, To allow for time delay deviations, when all the above conditions are met, the response echo satisfies the cross-trigger condition, and the control release module releases the greenhouse execution equipment control command corresponding to the next pending equipment action that is in the pending release state; if any condition is not met, the response echo does not satisfy the cross-trigger condition, and the control release module maintains the pending release state of the next pending equipment action. During updates, the response delay is adjusted according to... Update, equipment reserve according to Updated, motion response coefficients in digital twins are calculated according to... Update, in which For the change observed in the sensor response, To predict changes in digital twins. To update the step size, To prevent constants with a denominator of zero, after the update, the system regenerates the project scope fragment, equipment control relay diagram, and equipment control relay chain; the current process ends when there is no next equipment action to be executed in the equipment control relay chain.

[0036] Please see Figure 3 The present invention also provides a greenhouse crop control system based on digital twins and agent reasoning, for implementing the above-mentioned greenhouse crop control method based on digital twins and agent reasoning, comprising: The joint state generation module is used to acquire control state data including equipment operating status, controller feedback and sensor response, and generate a joint state object of crops, environment and equipment; The commitment object generation module is used to generate a device execution commitment object based on the device operating status, controller feedback and sensor response in the crop, environment and device joint state object; The scope fragment generation module is used to input the crop, environment and equipment joint state object and the equipment execution commitment object into the digital twin to generate engineering scope fragments; The relay diagram construction module is used to construct a device control relay diagram based on the engineering scope segment and the device execution commitment object. The device control relay diagram includes takeover, offsetting, and verification relationships between device actions. The relay chain generation module is used to generate a device control relay chain containing the actions of the devices to be executed and the conditions for the interaction of the devices, based on the preset crop control target and the device control relay diagram by the intelligent agent inference engine. The control release module is used to release the greenhouse execution equipment control command according to the equipment control relay chain, and to verify the exchange triggering condition based on the response echo formed by the controller acknowledgment and sensor response; when the verification result is satisfied, the control command for the next equipment action to be executed is released; when the verification result is not satisfied, the release is paused and the digital twin and the equipment execution commitment object are updated.

[0037] Specifically, the aforementioned system is deployed at the greenhouse edge gateway. The joint state generation module connects crop image acquisition equipment, environmental sensors, root zone sensors, and execution device controllers. The commitment object generation module calls the device capability table and device action mapping table. The scope fragment generation module calls the device action response matrix in the digital twin. The relay graph construction module generates graph nodes, takeover relationships, offset relationships, and verification relationships. The relay chain generation module generates device control relay chains according to the candidate relay path costs and the integrity of the verification relationships. The control release module sends control commands to the greenhouse execution device controller and triggers updates to response delay, device margin, and action response coefficients when the response echo does not meet the crossover triggering conditions.

[0038] In one embodiment's logic verification scenario, greenhouse temperature control, humidity control, supplemental lighting control, and root zone irrigation control can be used as continuous control task inputs. When the response echo does not meet the cross-triggering condition, the control release module maintains the pending release state of the next device action to be executed and regenerates the engineering domain fragment, device control relay diagram, and device control relay chain; when the response echo meets the cross-triggering condition, the next control command is released. This scenario illustrates the correspondence between control release logic and technical effects.

[0039] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A greenhouse crop regulation method based on digital twin and agent reasoning, characterized in that, include: S1. Obtain control status data including equipment operating status, controller feedback and sensor response, and generate a joint status object of crops, environment and equipment; S2. Generate a device execution commitment object based on the device operating status, controller feedback, and sensor response in the crop, environment, and device joint state object; S3. Input the crop, environment and equipment joint state object and the equipment execution commitment object into the digital twin to generate an engineering scope fragment; S4. Construct a device control relay diagram based on the engineering domain segment and the device execution commitment object. The device control relay diagram includes takeover, offsetting, and verification relationships between device actions. S5. The intelligent agent inference engine generates a device control relay chain that includes the device actions to be executed and the conditions for the interaction of the devices, based on the preset crop control target and the device control relay diagram. S6. Release the greenhouse execution equipment control command according to the equipment control relay chain, and verify the interaction triggering condition based on the response echo formed by the controller acknowledgment and sensor response; when the verification result is satisfied, release the control command for the next equipment action to be executed; when the verification result is not satisfied, pause the release and update the digital twin and the equipment execution commitment object.

2. The greenhouse crop regulation method based on digital twin and agent reasoning according to claim 1, characterized in that, The regulation status data includes crop phenotypic data, greenhouse environment data, root zone status data, actuator status data, controller feedback data, sensor response data, and future weather window data. The generated crop, environment, and equipment joint state object includes: The crop images in the crop phenotypic data are segmented into canopy regions and leaf posture features and canopy cover features are extracted. The current crop status is calculated by combining the leaf temperature data in the crop phenotypic data. The greenhouse environment data is synchronized over time to generate the greenhouse status. Calculate the root region state based on the root region state data; The device operating mode, drive current, valve opening, fan speed, lamp power, and pump pressure in the execution device status data are analyzed to generate the device operating status; Parse the device identifier, receipt status, and receipt timestamp in the controller receipt data to generate a controller receipt; The sensor response data is parsed to extract the sensor identifier, response field, and response timestamp to generate a sensor response. Determine the sequence of future external disturbances based on the aforementioned future weather window data; The current crop status, greenhouse status, root zone status, equipment operating status, controller feedback, sensor response, and future external disturbance sequence are associated according to the acquisition time to generate the joint state object of crop, environment, and equipment.

3. The greenhouse crop regulation method based on digital twin and agent reasoning according to claim 1, characterized in that, The device's execution commitment includes the executable range, executable time window, response latency, device margin, mutual exclusion conditions, observable response channels, and side effect indications. The generating device executes the commitment object, including: Extract the device type, current operating mode, protection status, device capacity, and receipt timestamp from the device operating status and controller feedback; Based on the device type and current operating mode, the device capability table is retrieved to obtain the executable range, executable time window, and mutual exclusion conditions. The response delay is calculated based on the receipt timestamp and the corresponding sensor response timestamp. Match observable response channels and side effect pointers according to the device action mapping table; The executable range, executable time window, response latency, device margin, mutual exclusion conditions, observable response channels, and side effects are associated with the device execution commitment object.

4. The greenhouse crop regulation method based on digital twin and agent reasoning according to claim 1, characterized in that, The generated engineering scope fragment includes: The actions of the device to be simulated are generated based on the executable range and executable time window in the device execution commitment object; The actions of the device to be simulated are input into a digital twin along with the combined state objects of the crop, environment, and device to simulate the impact of the actions of the device to be simulated on the crop state, greenhouse state, and other device executable windows. Extract the affected status field, direction of action, magnitude of action, response delay, and affected device identifier; The actions of the device to be simulated, the affected state fields, the direction of action, the magnitude of action, the response delay, and the identifier of the affected device are encapsulated into an engineering scope fragment.

5. The greenhouse crop regulation method based on digital twin and agent reasoning according to claim 1, characterized in that, The step of constructing a device control relay diagram based on the engineering scope segment and the device execution commitment object includes: The engineering state after the equipment is activated is used as the graph node; The side effects of the previous device action are matched with the affected state field of the next device action, and a takeover relationship is generated when the effect of the previous device action falls within the executable range of the next device action. For the same affected status field, connect the graph nodes whose action direction of the previous device is opposite to that of the action direction of the next device to generate an offsetting relationship; Read the observable response channel and response delay of the previous device action from the device execution commitment object, match the response field in the observable response channel with the affected state field of the next device action, and write the matched response field, response direction and response delay between the corresponding graph nodes to generate a verification relationship; The equipment control relay diagram is composed of the graph nodes, takeover relationships, offset relationships, and verification relationships.

6. The greenhouse crop regulation method based on digital twin and agent reasoning according to claim 5, characterized in that, The generation device control relay chain includes: The current engineering status is determined based on the combined status of crops, environment, and equipment, and the target engineering status is determined based on the preset crop control target in the equipment control relay diagram. Starting from the graph node corresponding to the current project state, retrieve the graph path pointing to the target project state, and use the graph nodes and connections in the graph path as candidate relay paths; Each subsequent graph node retains a candidate relay path that has a verification relationship triggered by the observable response channel of the previous graph node. The device actions, action sequences, and verification relationships in the retained candidate relay paths are converted into device actions to be executed, execution sequences, and cross-trigger conditions to generate a device control relay chain.

7. The greenhouse crop regulation method based on digital twin and agent reasoning according to claim 6, characterized in that, The response echo includes a controller feedback item and a sensor response item; the step of releasing the greenhouse according to the equipment control relay chain to execute the equipment control command includes: Release the greenhouse execution device control command corresponding to the first pending device action in the device control relay chain, and mark the next pending device action as pending release; Collect the controller feedback and sensor response corresponding to the current action of the device to be executed, compare the controller feedback and sensor response with the interaction triggering conditions respectively, and generate response echo; When the response echo satisfies the interaction triggering condition, the greenhouse execution device control command corresponding to the next pending device action in the pending release state is released; When the response echo does not meet the interaction triggering condition, the pending release state of the next device action to be executed is maintained, the digital twin and the device execution commitment object are updated according to the response echo, and the engineering scope fragment, device control relay diagram and device control relay chain are regenerated.

8. A greenhouse crop control system based on digital twins and agent reasoning, used to implement the greenhouse crop control method based on digital twins and agent reasoning as described in any one of claims 1-7, characterized in that, include: The joint state generation module is used to acquire control state data including equipment operating status, controller feedback and sensor response, and generate a joint state object of crops, environment and equipment; The commitment object generation module is used to generate a device execution commitment object based on the device operating status, controller feedback and sensor response in the crop, environment and device joint state object; The scope fragment generation module is used to input the crop, environment and equipment joint state object and the equipment execution commitment object into the digital twin to generate engineering scope fragments; The relay diagram construction module is used to construct a device control relay diagram based on the engineering scope segment and the device execution commitment object. The device control relay diagram includes takeover, offsetting, and verification relationships between device actions. The relay chain generation module is used to generate a device control relay chain containing the actions of the devices to be executed and the conditions for the interaction of the devices, based on the preset crop control target and the device control relay diagram by the intelligent agent inference engine. The control release module is used to release the greenhouse execution equipment control command according to the equipment control relay chain, and to verify the exchange triggering condition based on the response echo formed by the controller acknowledgment and sensor response; when the verification result is satisfied, the control command for the next equipment action to be executed is released; when the verification result is not satisfied, the release is paused and the digital twin and the equipment execution commitment object are updated.

Images