Crop growth knowledge base-based agricultural intelligent agent decision control method and system

Abstract

The present application relates to the field of agricultural intelligent control technology, in particular to an agricultural intelligent agent decision control method and system based on a crop growth knowledge base. In view of the problem that external environment deviation in agricultural field regulation and control is difficult to reflect the hidden physiological limited state of crops, leading to the problem that intelligent agents directly generating control actions are prone to mismatch, a decision control scheme based on equivalent regulation and control carrier is proposed, the necessary regulation and control action set is determined by acquiring agricultural field state, equipment capacity state and control demand data, the equivalent regulation and control carrier template and the crop safety boundary are retrieved from the crop growth knowledge base, the target carrier action is executed and the crop response is collected, the hidden physiological channel state is obtained by response demodulation, and the subsequent control instruction of the agricultural intelligent agent is constrained and the template and demodulation rules are updated, which is suitable for crop regulation and control in the greenhouse and agricultural internet of things scene.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent control technology, specifically to an agricultural intelligent agent decision-making control method and system based on a crop growth knowledge base. Background Technology

[0002] Intelligent agricultural control methods are commonly used in greenhouses, plant factories, open-field facility agriculture, and irrigation and fertilization control. These methods generate control actions such as ventilation, supplemental lighting, irrigation, fertilization, heating, or cooling based on crop growth status, environmental conditions, equipment status, and production goals. Current methods typically collect data such as temperature, humidity, light intensity, carbon dioxide concentration, soil moisture content, nutrient solution parameters, and crop images using sensors. This data is then combined with threshold rules, empirical models, historical data, or intelligent algorithms to determine current control needs and convert the results into control commands for agricultural equipment to maintain the crop growth environment within preset ranges.

[0003] In real-world agricultural scenarios, changes in the external state of crops do not correspond simply to internal physiological constraints. Changes in the same temperature, humidity, light, or root zone data may be caused by different physiological factors. Furthermore, the amplitude of action, response delay, and direction of action of different agricultural equipment can lead to variations in crop responses. Existing methods often focus on directly judging environmental variables or apparent crop data, lacking effective utilization of the relationship between control actions and crop responses. This can easily result in mismatched control actions when the causes of crop responses are unclear, and insufficient correction of subsequent feedback parameters. Therefore, agricultural intelligent agent decision-making and control methods and systems based on crop growth knowledge bases are needed to address these issues. Summary of the Invention

[0004] To address the above problems, this invention provides the following technical solution: an agricultural intelligent agent decision-making and control method based on a crop growth knowledge base, comprising: S1. Acquire on-site status data, equipment capacity status, and control requirement data to determine the necessary set of control actions; S2. Retrieve equivalent control carrier templates and crop safety boundaries that match the set of necessary control actions from the crop growth knowledge base, and generate a set of candidate carrier actions. The equivalent control carrier template includes equivalent control amount constraints, carrier structure parameters, response sampling window and response demodulation rules. S3. Based on the equipment capability status, crop safety boundary and equivalent control quantity constraint, the candidate carrier action set is screened. If there are no remaining candidate carrier actions, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Otherwise, the target carrier action is selected and executed. S4. Collect crop response data within the response sampling window corresponding to the target carrier action, and generate a crop response demodulation object; S5. Demodulate the crop response demodulation object according to the response demodulation rule to determine the hidden physiological channel status; S6. Invoke the agricultural intelligent agent to generate subsequent agricultural equipment control commands based on the hidden physiological channel state, and suppress control actions that do not match the hidden physiological channel state; S7. When executing the subsequent agricultural equipment control command, update the equivalent modulation carrier template and response demodulation rule according to the crop response after execution.

[0005] Furthermore, the control demand data includes the variable to be controlled, the target control direction, and the control demand amount; the necessary control action set includes the applicable equipment, the basic action type, and the total control amount; determining the necessary control action set includes: Read the variable to be regulated, the target regulation direction, and the regulation requirement from the control demand data; Based on the variable to be controlled and the target control direction, retrieve the applicable devices and basic action types from the preset device function mapping table; Calculate the total control amount corresponding to each basic action type based on the control demand. Remove the operable devices whose device capability status does not meet the executable conditions, and associate the remaining operable devices, basic action types, and total control quantities to generate the necessary control action set.

[0006] Furthermore, the generation of the candidate carrier action set includes: Read the basic action type and total control amount from the set of necessary control actions; Retrieve an equivalent control carrier template from the crop growth knowledge base that matches the action type index with the basic action type; Under the condition of satisfying the equivalent control amount constraint, the carrier structure parameters are embedded into the action time sequence or action spatial distribution corresponding to the basic action type to generate candidate carrier actions; The candidate carrier action, the crop safety boundary, the response sampling window, and the response demodulation rule are associated to generate the candidate carrier action set.

[0007] Furthermore, the selection and execution of the target carrier action includes: Read the device's executable range, minimum action resolution, and action switching latency from the device capability status; Candidate carrier actions whose carrier structure parameters exceed the range that the device can execute, are below the minimum action resolution, or do not meet the action switching delay are eliminated. The reference response amplitude is read from the response demodulation rules, the expected response state of the candidate carrier action is calculated based on the equivalent control amount constraint and the reference response amplitude, and candidate carrier actions whose expected response state exceeds the crop safety boundary are eliminated. The carrier solvable scheduling of the remaining candidate carrier actions is calculated based on the carrier structure parameters, response sampling window, sensor sampling frequency in field status data, and reference response amplitude. The remaining candidate carrier action with the highest carrier solvability is selected as the target carrier action and executed.

[0008] Furthermore, the generated crop response demodulation object includes: Record the action time sequence and carrier structure parameters of the target carrier action, and calculate the equivalent control amount of the target carrier action according to the equivalent control amount constraint; Within the response sampling window, canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root region response sequence are collected; The action time series was time-aligned with the canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root zone response sequence. Calculate the response gain, response hysteresis, and phase offset between the target carrier action and each response sequence; The crop response demodulation object is generated by associating the action time sequence, carrier structure parameters, equivalent control amount, response gain, response hysteresis, and phase offset.

[0009] Furthermore, determining the state of the hidden physiological channel includes: Physiological channel reference features, demodulation thresholds, and discrimination thresholds are read from the response demodulation rules. The physiological channel reference features include stomatal regulation-limited reference features, root region absorption-limited reference features, photosynthetic utilization-limited reference features, and transpiration coupling-abnormal reference features. The response gain, response hysteresis, and phase offset in the crop response demodulation object are matched with the physiological channel reference features to generate channel matching values ​​for each physiological channel. When the highest channel matching value reaches the corresponding demodulation threshold, and the difference between the highest channel matching value and the second highest channel matching value reaches the distinction threshold, the physiological channel corresponding to the highest channel matching value is determined as the hidden physiological channel state. When two or more channel matching values ​​reach the corresponding demodulation threshold and the difference between the highest channel matching value and the second highest channel matching value does not reach the distinction threshold, a combined hidden physiological channel state consisting of two or more corresponding physiological channels is generated. When the conditions for generating a single hidden physiological channel state or a combination of hidden physiological channel states are not met, the hidden physiological channel state is marked as an unidentified state, and the generation of subsequent agricultural equipment control commands based on the hidden physiological channel state is prohibited.

[0010] Furthermore, the generation of subsequent agricultural equipment control commands and the suppression of control actions that do not match the state of the hidden physiological channel include: When the hidden physiological channel is in an unidentified state, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Retrieve from the crop growth knowledge base a whitelist of control actions and a control action inhibition table corresponding to the hidden physiological channel state; Input the hidden physiological channel status, equipment capability status, and control action whitelist into the agricultural intelligent agent to generate candidate subsequent control actions; Eliminate candidate subsequent control actions that match the control action inhibition table; Convert the candidate subsequent control actions that were not eliminated into subsequent agricultural equipment control commands; When all candidate subsequent control actions are eliminated, maintain the current control state and prohibit the generation of subsequent agricultural equipment control commands.

[0011] Furthermore, updating the equivalent modulated carrier template and response demodulation rules includes: When the subsequent agricultural equipment control command is generated and executed, crop response data after the execution of the subsequent agricultural equipment control command is collected, and a post-execution response object is generated. The expected response features are read from the crop growth knowledge base based on the hidden physiological channel state. The post-execution response object is compared with the expected response characteristics to generate a response deviation object, which includes response gain deviation, response hysteresis deviation and channel matching deviation. When the response hysteresis deviation exceeds a preset hysteresis deviation threshold, the response sampling window in the corresponding equivalent modulated carrier template is corrected according to the response hysteresis deviation. When the response gain deviation exceeds a preset gain deviation threshold, the carrier structure parameters in the corresponding equivalent modulated carrier template are corrected according to the response gain deviation. When the channel matching deviation exceeds the preset matching deviation threshold, the physiological channel reference features and demodulation threshold in the corresponding response demodulation rules are corrected according to the channel matching deviation.

[0012] This invention also provides an agricultural agent decision control system based on a crop growth knowledge base, used to implement the above-mentioned agricultural agent decision control method based on a crop growth knowledge base, including: The necessary control action determination module is used to acquire field status data, equipment capacity status and control requirement data, and determine the set of necessary control actions. The carrier action set generation module is used to retrieve equivalent control carrier templates and crop safety boundaries that match the necessary control action set from the crop growth knowledge base, and generate a candidate carrier action set. The equivalent control carrier template includes equivalent control amount constraints, carrier structure parameters, response sampling window and response demodulation rules. The target carrier action execution module is used to filter the candidate carrier action set based on the equipment capability status, crop safety boundary and equivalent control amount constraint. If there are no remaining candidate carrier actions, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Otherwise, the target carrier action is selected and executed. The response demodulation object generation module is used to collect crop response data within the response sampling window corresponding to the target carrier action and generate a crop response demodulation object. The hidden physiological channel demodulation module is used to demodulate the crop response demodulation object according to the response demodulation rules and determine the hidden physiological channel status. The intelligent agent control instruction generation module is used to call the agricultural intelligent agent to generate subsequent agricultural equipment control instructions based on the hidden physiological channel state, and to suppress control actions that do not match the hidden physiological channel state. The template rule update module is used to update the equivalent modulation carrier template and response demodulation rules according to the crop response after execution when executing the subsequent agricultural equipment control commands.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The agricultural intelligent agent decision-making and control method and system based on crop growth knowledge base retrieves equivalent control carrier templates and crop safety boundaries that match the set of necessary control actions from the crop growth knowledge base. Under the constraints of equipment capability status, crop safety boundaries, and equivalent control quantities, target carrier actions are selected and executed, enabling the crop to generate collectable and demodulated response data under controlled actions. By processing the crop response demodulation object according to response demodulation rules, the hidden physiological channel state is determined, allowing for the differentiation of changes in the same environmental variable or apparent state based on action-response relationships, reducing the likelihood of directly generating control actions based solely on environmental thresholds or apparent data. Subsequent agricultural equipment control commands are generated based on the hidden physiological channel state, and control actions that do not match the hidden physiological channel state are suppressed, excluding candidate control actions that do not conform to the current crop response cause from the control process. By updating the equivalent control carrier template and response demodulation rules based on the executed crop response, the response sampling window, carrier structure parameters, and channel identification criteria in subsequent control can be corrected with actual feedback, thereby reducing mismatched control command outputs when the cause of the crop response is unclear. Attached Figure Description

[0014] Figure 1 A schematic diagram of the overall agricultural intelligent agent decision control method based on crop growth knowledge base provided by the present invention; Figure 2 A flowchart illustrating the agricultural intelligent agent decision-making and control method based on a crop growth knowledge base provided by this invention; Figure 3 This is a schematic diagram of the agricultural intelligent agent decision control system based on crop growth knowledge base provided by the present invention. Detailed Implementation

[0015] 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.

[0016] 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.

[0017] Please see Figure 1 and Figure 2 , Figure 1 A schematic diagram of the overall agricultural intelligent agent decision control method based on crop growth knowledge base provided by the present invention; Figure 2This is a flowchart illustrating the agricultural agent decision-making and control method based on a crop growth knowledge base provided by the present invention. In one embodiment of the present invention, the agricultural agent decision-making and control method based on a crop growth knowledge base includes the following steps: S1. Acquire on-site status data, equipment capacity status, and control requirement data to determine the necessary set of control actions; Specifically, this step is performed by an edge gateway or agricultural controller deployed at the agricultural site. Agricultural site status data is uploaded by greenhouse environmental sensors, root zone sensors, crop image acquisition devices, and crop canopy temperature acquisition devices; equipment capability status is fed back by the controllers of each agricultural device, including at least device identification, online status, fault status, executable range, minimum action resolution, and action switching latency; control requirement data is written by the upper-level control task, manually input targets, or the target decomposition results of the agricultural agent. The edge gateway caches the above data according to the acquisition time and device identification, and, starting with the control requirement data, determines the basic control actions that must be performed by the agricultural equipment, thus forming a necessary set of control actions. This set of necessary control actions serves as the input for retrieving the equivalent control carrier template in subsequent steps.

[0018] Furthermore, in one embodiment provided by the present invention, the control demand data includes the variable to be controlled, the target control direction, and the control demand amount; the necessary control action set includes the applicable equipment, the basic action type, and the total control amount; determining the necessary control action set includes: Read the variable to be regulated, the target regulation direction, and the regulation requirement from the control demand data; Based on the variable to be controlled and the target control direction, retrieve the applicable devices and basic action types from the preset device function mapping table; Calculate the total control amount corresponding to each basic action type based on the control demand. Remove the operable devices whose device capability status does not meet the executable conditions, and associate the remaining operable devices, basic action types, and total control quantities to generate the necessary control action set.

[0019] Specifically, the variable to be regulated in the control demand data indicates the agricultural object that needs to be changed, such as canopy temperature, air humidity, carbon dioxide concentration, light intensity, or root zone water content; the target regulation direction indicates whether to increase, decrease, maintain, or suppress fluctuations; and the regulation demand quantity represents the difference between the variable to be regulated and the current or target state. When the edge gateway reads the control demand data, it writes the variable to be regulated into the variable field. Write the target control direction into the direction field. Write the regulated demand quantity into the demand quantity field. At the same time, the edge gateway initializes the necessary set of control actions, reserving for each candidate record the applicable device field, basic action type field, total control amount field, and executable flag field.

[0020] The preset equipment action mapping table is stored in the crop growth knowledge base or the edge gateway's local database. This table includes at least the following fields: variable field, direction field, equipment type field, basic action type field, unit action coefficient field, and execution condition field. During retrieval, the edge gateway uses the variable to be controlled... and target control direction As the matching key, filter the variable fields in the preset device function mapping table. Consistent and the direction field is the same as The system maintains consistent records and temporarily stores the retrieved device type and basic action type as candidate action records. If no matching record is found, the current control requirement cannot be directly responded to by the registered agricultural equipment, the edge gateway does not generate the necessary set of control actions, and outputs manual review information.

[0021] For each basic action type retrieved, the edge gateway reads the unit action coefficient corresponding to that basic action type. And calculate the total control amount based on the control demand:

[0022] in, Indicates the basic action type The corresponding total control amount, The amount of regulation demand comes from the control demand data. The unit action coefficient field is from the preset device action mapping table. (Calculated) The total control amount field written to the candidate action record is used to determine whether the candidate carrier action meets the equivalent control amount constraint.

[0023] When removing an applicable device, the edge gateway reads the device's online status, fault status, executable range, minimum action resolution, and action switching latency from the device capability status. If the device is offline, in a fault state, its executable range cannot cover the total control amount, its minimum action resolution is greater than the minimum action range required for the basic action type, or its action switching latency does not meet the execution requirements of the basic action type, then the corresponding applicable device is removed. After removal, the edge gateway uses the device identifier and basic action type as the association key to combine the remaining applicable devices, basic action types, and total control amount into a necessary control action set; each record in this set can be used by S2 to retrieve the equivalent control carrier template.

[0024] S2. Retrieve equivalent control carrier templates and crop safety boundaries that match the set of necessary control actions from the crop growth knowledge base, and generate a set of candidate carrier actions. The equivalent control carrier template includes equivalent control amount constraints, carrier structure parameters, response sampling window and response demodulation rules. Specifically, the crop growth knowledge base stores action type indexes, equivalent control carrier templates, crop safety boundaries, and response demodulation rules. After reading the necessary control action set, the edge gateway does not directly convert the necessary control actions into ordinary device control commands. Instead, it first retrieves equivalent control carrier templates that can embed a demodulated carrier structure while maintaining the overall control effect. The equivalent control amount constraints in the equivalent control carrier template limit the overall control amount of the carrier action, the carrier structure parameters limit the action amplitude, action interval, carrier period, or partition distribution, the response sampling window limits the time range for subsequent crop response acquisition, and the response demodulation rules are used to identify hidden physiological channel states in S5. The retrieved templates and crop safety boundaries jointly constrain the generation of the candidate carrier action set.

[0025] Furthermore, in one embodiment of the present invention, the generation of the candidate carrier action set includes: Read the basic action type and total control amount from the set of necessary control actions; Retrieve an equivalent control carrier template from the crop growth knowledge base that matches the action type index with the basic action type; Under the condition of satisfying the equivalent control amount constraint, the carrier structure parameters are embedded into the action time sequence or action spatial distribution corresponding to the basic action type to generate candidate carrier actions; The candidate carrier action, crop safety boundary, response sampling window, and response demodulation rule are associated to generate the candidate carrier action set.

[0026] Specifically, the edge gateway reads the basic action type from each record in the set of necessary control actions. and total regulation The basic action type is used to retrieve templates, and the total control amount is used to verify the equivalent control amount of the carrier action. The equivalent control carrier template table in the crop growth knowledge base includes at least an action type index, equivalent control amount constraints, carrier structure parameters, response sampling window identifier, response demodulation rule identifier, and crop safety boundary identifier. During retrieval, the basic action type is used as the retrieval method. The matching is performed using the action type index, and the matched equivalent control carrier template is written into the template candidate cache. If multiple templates match the same basic action type, multiple template records with different carrier structure parameters are retained so that S3 can continue filtering.

[0027] When generating candidate carrier actions, the edge gateway first reads the equivalent control quantity constraints in the template, which include the allowable deviation range. The minimum duration and maximum number of action segments are determined. For agricultural equipment operating in a time-continuous manner, the carrier structure parameters are embedded into the action time series to obtain multiple action segments. The equivalent control amount of this time series is determined as follows:

[0028] in, This represents the equivalent control amount for the candidate carrier action. Indicates the first The execution range of each action segment, Indicates the first The duration of an action segment Indicates the number of action segments. If If so, the action time series satisfies the equivalent control quantity constraint.

[0029] For devices with partitioned execution capabilities, the edge gateway reads the set of partitioned addresses from the device's capability status and embeds the carrier structure parameters into the action space distribution. The equivalent control quantity corresponding to the space distribution is determined as follows:

[0030] in, Indicates the first The execution scope of each partition, Indicates the first The effective area of ​​each partition This indicates the weight of the partition. This indicates the number of partitions. The time series or spatial distribution of actions that satisfy the equivalent control quantity constraints are written into the candidate carrier action field. Subsequently, the edge gateway associates the candidate carrier actions, crop safety boundaries, response sampling windows, and response demodulation rules with the template identifier as the association key to form a candidate carrier action set; this set continues to be screened in S3 based on equipment capabilities and crop safety constraints.

[0031] S3. Based on the equipment capability status, crop safety boundary and equivalent control quantity constraint, the candidate carrier action set is screened. If there are no remaining candidate carrier actions, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Otherwise, the target carrier action is selected and executed. Specifically, the edge gateway performs a two-level screening of the candidate carrier action set. The first level of screening determines whether the candidate carrier action can be actually executed by the field equipment, based on the equipment's capability status. The second level of screening determines whether the expected crop response of the candidate carrier action is still within a safe range, based on crop safety boundaries and equivalent control constraints. If no remaining candidate carrier actions exist after the two-level screening, the edge gateway will not release subsequent control tasks to the agricultural agent, but will maintain the current control state and record the reason why no carrier action can be executed. If remaining candidate carrier actions exist, the candidate carrier action with the highest demodulation capability will be selected as the target carrier action and executed.

[0032] Furthermore, in one embodiment of the present invention, selecting and executing the target carrier action that satisfies the equivalent control amount constraint includes: Read the device's executable range, minimum action resolution, and action switching latency from the device capability status; Candidate carrier actions whose carrier structure parameters exceed the range that the device can execute, are below the minimum action resolution, or do not meet the action switching delay are eliminated. The reference response amplitude is read from the response demodulation rules, the expected response state of the candidate carrier action is calculated based on the equivalent control amount constraint and the reference response amplitude, and candidate carrier actions whose expected response state exceeds the crop safety boundary are eliminated. The carrier solvable scheduling of the remaining candidate carrier actions is calculated based on the carrier structure parameters, response sampling window, sensor sampling frequency in agricultural field status data, and reference response amplitude. The remaining candidate carrier action with the highest carrier solvability is selected as the target carrier action and executed.

[0033] Specifically, when reading device capability status, the edge gateway uses the device identifier in the candidate carrier action as an index to read the executable amplitude of the corresponding device. Minimum motion resolution Action switching latency Among them, the executable range limits the maximum range of motion the device can perform, the minimum motion resolution limits the smallest adjustment unit the device can distinguish, and the motion switching delay limits the shortest allowed switching interval between adjacent motion segments. All of these fields are derived from the device controller's capability reporting data or the device parameter table.

[0034] When eliminating candidate carrier actions, the edge gateway reads the action amplitude, the difference between adjacent action amplitudes, and the interval between adjacent action segments for each candidate carrier action. If any action amplitude is greater than... If the candidate carrier action exceeds the device's executable range, then the action is determined to be within the range that the device can execute; if the difference in amplitude between adjacent actions is less than 1 / 3... If the candidate carrier action is below the minimum action resolution, then the action segment interval is determined to be less than the minimum action resolution. If the candidate carrier action fails to meet the action handover delay requirement, it is determined that the candidate carrier action does not meet the requirement. The eliminated candidate carrier actions are added to the non-executable list, and the remaining candidate carrier actions are entered into the crop safety boundary verification.

[0035] When calculating the expected response state, the edge gateway reads the reference response amplitude from the response demodulation rules. Read the current status from agricultural field status data And read the equivalent modulation amount of the candidate carrier action. In one embodiment, the expected response state is determined as follows:

[0036] in, This indicates the expected response status after the candidate carrier action is executed. This represents the corresponding variable value in the current agricultural field status data. This indicates the reference response amplitude in the demodulation rules. This represents the equivalent control amount for candidate carrier actions. The edge gateway will... Compared with the upper and lower limits in the crop safety boundary, if If the behavior exceeds the crop safety boundary, the candidate carrier action will be removed to avoid causing the crop state to go out of bounds in order to identify hidden physiological channels.

[0037] When calculating carrier unschedulable operations, the edge gateway reads the carrier structure parameters and response sampling window length of the remaining candidate carrier actions. Sensor sampling frequency in agricultural field status data and reference response amplitude In one embodiment, carrier deschedulability is determined as follows:

[0038] in, This indicates that the carrier can be descheduled. The carrier frequency is shown in the carrier structure parameters. This indicates the minimum number of sampling points required by the demodulation rule. This represents the minimum demodulated response amplitude. This indicates the weights configured by the response demodulation rules. Edge gateway selection. The largest remaining candidate carrier action is taken as the target carrier action, and the target carrier action is converted into a device action sequence and sent to the corresponding agricultural equipment controller for execution.

[0039] S4. Collect crop response data within the response sampling window corresponding to the target carrier action, and generate a crop response demodulation object; Specifically, after the target carrier action is executed, the edge gateway initiates data acquisition based on the response sampling window associated with the target carrier action. The response sampling window defines the start time, end time, and sampling interval of the acquisition, ensuring that the acquired crop response corresponds to the carrier structure of the target carrier action. The acquired crop response data is not directly used for control but is organized into a crop response demodulation object; this object is used in S5 to determine the hidden physiological channel state according to the response demodulation rules.

[0040] Furthermore, in one embodiment of the present invention, the generation of the crop response demodulation object includes: Record the action time sequence and carrier structure parameters of the target carrier action, and calculate the equivalent control amount of the target carrier action according to the equivalent control amount constraint; Within the response sampling window, canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root region response sequence are collected; The action time series was time-aligned with the canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root zone response sequence. Calculate the response gain, response hysteresis, and phase offset between the target carrier action and each response sequence; The crop response demodulation object is generated by associating the action time sequence, carrier structure parameters, equivalent control amount, response gain, response hysteresis, and phase offset.

[0041] Specifically, when recording target carrier actions, the edge gateway saves the start time, end time, execution amplitude, device identifier, and basic action type for each action segment, and also saves the carrier frequency, carrier period, amplitude combination, or spatial partition parameters from the carrier structure parameters. The equivalent control amount of the target carrier action is obtained according to the equivalent control amount calculation method in S2 and written into the equivalent control amount field of the target carrier action record for subsequent calculation of response gain.

[0042] Within the response sampling window, canopy temperature change sequences were acquired by infrared temperature sensors or thermal imaging devices; canopy image change sequences were acquired by image acquisition devices, and canopy coverage, leaf posture changes, and canopy color changes were obtained through canopy region segmentation; environmental response sequences were acquired by temperature, humidity, carbon dioxide, and light sensors; and root zone response sequences were acquired by matrix moisture content and matrix conductivity sensors. All of the above sequences were timestamped and written to the response sampling buffer.

[0043] During time alignment, the edge gateway uses the action time series of the target carrier as the reference time axis to resample the canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root region response sequence to a uniform sampling interval. If a sensor sampling time is inconsistent with the reference time axis, the response value at the corresponding time is obtained by linear interpolation of adjacent sampling points; if a sampling point is missing, the previous valid sampling value is used to maintain it if the missing length does not exceed the allowable missing length, and the response sequence is marked as unusable for demodulation if the missing length exceeds the allowable missing length.

[0044] When calculating the response gain, the edge gateway records the target carrier action sequence as follows: Let any response sequence be denoted as In one embodiment, the response gain is determined as follows:

[0045] in, Indicates response gain. This represents a constant to prevent the denominator from being zero. The response hysteresis is calculated by considering the sequence of actions. With response sequence The time offset corresponding to the maximum cross-correlation value is obtained; the phase offset is obtained by comparing the phase difference between the main frequency components of the action sequence and the main frequency components of the response sequence. After the calculation is completed, the edge gateway associates the action time sequence, carrier structure parameters, response gain, response hysteresis, and phase offset according to the device identifier and response sequence identifier to generate a crop response demodulation object.

[0046] S5. Demodulate the crop response demodulation object according to the response demodulation rule to determine the hidden physiological channel status; Specifically, the response demodulation rules are used to convert the externally observable response of the crop to the target carrier into an internal physiological channel state. The edge gateway reads the response gain, response hysteresis, and phase offset from the crop response demodulation object and compares them with the physiological channel reference features in the response demodulation rules. The hidden physiological channel state is used to indicate which type of physiological process the crop is currently primarily constrained by, such as stomatal regulation limitation, root zone absorption limitation, photosynthetic utilization limitation, or transpiration coupling anomaly. This state serves as an input constraint in S6 for the agricultural agent to generate subsequent control commands.

[0047] Furthermore, in one embodiment of the present invention, determining the state of the hidden physiological channel includes: Physiological channel reference features, demodulation thresholds, and discrimination thresholds are read from the response demodulation rules. The physiological channel reference features include stomatal regulation-limited reference features, root region absorption-limited reference features, photosynthetic utilization-limited reference features, and transpiration coupling-abnormal reference features. The response gain, response hysteresis, and phase offset in the crop response demodulation object are matched with the physiological channel reference features to generate channel matching values ​​for each physiological channel. When the highest channel matching value reaches the corresponding demodulation threshold, and the difference between the highest channel matching value and the second highest channel matching value reaches the distinction threshold, the physiological channel corresponding to the highest channel matching value is determined as the hidden physiological channel state. When two or more channel matching values ​​reach the corresponding demodulation threshold and the difference between the highest channel matching value and the second highest channel matching value does not reach the distinction threshold, a combined hidden physiological channel state consisting of two or more corresponding physiological channels is generated. When all channel matching values ​​fail to reach the corresponding demodulation threshold, the hidden physiological channel state is marked as unrecognized, and the generation of subsequent agricultural equipment control commands based on the hidden physiological channel state is prohibited.

[0048] Specifically, the response demodulation rules store a set of reference feature vectors for each physiological channel. These reference feature vectors include at least a response gain reference value, a response hysteresis reference value, and a phase offset reference value. The demodulation threshold is used to determine whether a channel has sufficient matching ability, and the discrimination threshold is used to determine whether the highest matching channel and the second highest matching channel can be distinguished. These thresholds can be obtained from historical crop response samples or read from empirical parameter tables corresponding to crop varieties, growth stages, and equipment types.

[0049] During matching, the edge gateway combines the response gain, response hysteresis, and phase offset from the crop response demodulation object into an observation feature vector. And the reference characteristics of each physiological channel are recorded as In one embodiment, the channel matching value is calculated as follows:

[0050] in, Indicates physiological channels The channel matching value, From crop response demodulation objects, From response demodulation rules, This represents a constant to prevent the denominator from being zero. The calculated value... Write the channel matching results into the table and sort them from highest to lowest value.

[0051] When the highest channel matching value Reaching its corresponding demodulation threshold ,and Matching value with the second highest channel The difference reaches the discrimination threshold At that time, the edge gateway will channel The state is determined to be a hidden physiological channel. This occurs when the matching values ​​of two or more channels all reach the corresponding demodulation threshold, and... At this time, the edge gateway does not forcibly select a single channel, but generates a combination of hidden physiological channel states and records the channel identifiers participating in the combination. When all channel matching values ​​fail to reach the corresponding demodulation threshold, the edge gateway writes the hidden physiological channel state as an unrecognized state; this state will trigger the prohibition control branch in S6 to prevent the agricultural agent from outputting mismatch control actions when the physiological channels are unclear.

[0052] S6. Invoke the agricultural intelligent agent to generate subsequent agricultural equipment control commands based on the hidden physiological channel state, and suppress control actions that do not match the hidden physiological channel state; Specifically, the agricultural agent does not directly generate control commands based solely on external environmental deviations. Instead, it uses hidden physiological channel states as pre-constraints for control. The edge gateway inputs the hidden physiological channel states, equipment capability states, and control action constraints from the knowledge base into the agricultural agent. The agent first generates candidate subsequent control actions, which are then filtered by the edge gateway using a control action inhibition table. Only the filtered candidate subsequent control actions are converted into agricultural equipment control commands, thus preventing control actions that are incompatible with the crop's internal physiological constraints from being executed.

[0053] Furthermore, in one embodiment of the present invention, generating subsequent agricultural equipment control commands and suppressing control actions that do not match the state of the hidden physiological channel includes: When the hidden physiological channel is in an unidentified state, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Retrieve from the crop growth knowledge base a whitelist of control actions and a control action inhibition table corresponding to the hidden physiological channel state; Input the hidden physiological channel status, equipment capability status, and control action whitelist into the agricultural intelligent agent to generate candidate subsequent control actions; Eliminate candidate subsequent control actions that match the control action inhibition table; Convert the candidate subsequent control actions that were not eliminated into subsequent agricultural equipment control commands; When all candidate subsequent control actions are eliminated, maintain the current control state and prohibit the generation of subsequent agricultural equipment control commands.

[0054] Specifically, when the hidden physiological channel state is unidentified, the edge gateway does not invoke the agricultural agent to generate subsequent control actions. Instead, it maintains the current control state and writes the reason for unidentification, the target carrier action identifier, and the crop response demodulation object into the review log. This process ensures that subsequent agricultural equipment control commands must be based on an identifiable physiological channel state.

[0055] When the hidden physiological channel state is not unrecognized, the edge gateway uses the hidden physiological channel state identifier as the matching key to retrieve the whitelist and suppression table of control actions from the crop growth knowledge base. The whitelist of control actions includes at least the allowed device type, allowed basic action type, allowed action amplitude range, and allowed duration range; the suppression table includes at least the hidden physiological channel state identifier, suppressed device type, suppressed action type, suppression trigger condition, and alternative action prompt. The retrieved whitelist is used to limit the range of actions that the agricultural agent can generate, and the suppression table is used for secondary filtering of candidate subsequent control actions after generation.

[0056] After receiving the hidden physiological channel status, equipment capability status, and a whitelist of control actions, the agricultural intelligent agent generates candidate subsequent control actions. Each candidate subsequent control action includes a device identifier, action type, action amplitude, duration, execution order, and corresponding hidden physiological channel status. During elimination, the edge gateway matches the device type, action type, and hidden physiological channel status of the candidate subsequent control actions against the control action suppression table one by one; if a candidate subsequent control action hits the suppression trigger condition, it is eliminated, and the reason for elimination is recorded.

[0057] Candidate subsequent control actions that are not eliminated are converted into subsequent agricultural equipment control commands. During the conversion, the edge gateway generates the target device address, action code, action amplitude field, duration field, and execution time field according to the device communication protocol, and sends them to the corresponding agricultural equipment controller. When all candidate subsequent control actions are eliminated, it indicates that the actions generated by the agricultural agent do not match the current hidden physiological channel state. The edge gateway maintains the current control state and prohibits the generation of subsequent agricultural equipment control commands.

[0058] S7. When executing the subsequent agricultural equipment control command, update the equivalent modulation carrier template and response demodulation rule according to the crop response after execution.

[0059] Specifically, after S6 generates and executes subsequent agricultural equipment control commands, the edge gateway continues to collect the crop response after execution and compares the executed response with the expected response characteristics in the crop growth knowledge base. The comparison result is used to correct the equivalent control carrier template and response demodulation rules, so that when generating subsequent target carrier actions, carrier structure parameters, response sampling windows, and demodulation thresholds that are more suitable for the current crop, equipment, and environmental conditions can be adopted. If S6 does not generate subsequent agricultural equipment control commands, the edge gateway does not perform template updates, but only saves the reason for not generating and the response demodulation record.

[0060] Furthermore, in one embodiment of the present invention, updating the equivalent modulated carrier template and the response demodulation rule includes: Collect crop response data after executing the subsequent agricultural equipment control commands, and generate a post-execution response object; The expected response features are read from the crop growth knowledge base based on the hidden physiological channel state. The post-execution response object is compared with the expected response characteristics to generate a response deviation object, which includes response gain deviation, response hysteresis deviation and channel matching deviation. When the response hysteresis deviation exceeds a preset hysteresis deviation threshold, the response sampling window in the corresponding equivalent modulated carrier template is corrected according to the response hysteresis deviation. When the response gain deviation exceeds a preset gain deviation threshold, the carrier structure parameters in the corresponding equivalent modulated carrier template are corrected according to the response gain deviation. When the channel matching deviation exceeds the preset matching deviation threshold, the physiological channel reference features and demodulation threshold in the corresponding response demodulation rules are corrected according to the channel matching deviation.

[0061] Specifically, the post-execution response object is generated by the edge gateway within the feedback window after the subsequent agricultural equipment control commands are executed. This object includes at least the post-execution response gain, post-execution response hysteresis, post-execution channel matching value, corresponding device identifier, corresponding action type, and corresponding hidden physiological channel state. The expected response features are retrieved from the crop growth knowledge base based on the hidden physiological channel state and action type, and include at least the expected response gain, expected response hysteresis, and expected channel matching value.

[0062] When generating the response bias object, the edge gateway calculates the response gain bias, response hysteresis bias, and channel matching bias respectively. The response hysteresis bias is determined as follows:

[0063] in, This indicates the response lag deviation. This indicates the actual response delay in the response object after execution. This represents the expected response hysteresis in the expected response characteristics. The response gain deviation is determined as follows:

[0064] in, Indicates the response gain deviation. Indicates the actual response gain after execution. This represents the expected response gain. Channel matching bias is determined as follows:

[0065] in, Indicates channel matching deviation. This indicates the actual channel matching value after execution. This represents the expected channel matching value. All deviations mentioned above are written to the response deviation object and used as the trigger for template and rule updates.

[0066] When the response hysteresis deviation exceeds the preset hysteresis deviation threshold, it indicates that the original response sampling window is too early or too late relative to the actual crop response. The edge gateway then... The response sampling window is modified. In one embodiment, it is modified as follows:

[0067] in, and These represent the start and end times of the corrected response sampling window, respectively. and These represent the start and end times of the response sampling window before correction. This represents the window correction factor.

[0068] When the response gain deviation exceeds the preset gain deviation threshold, it indicates that the actual response strength corresponding to the carrier structure parameters is inconsistent with the expectation. The edge gateway then... Modify the carrier structure parameters in the corresponding equivalent control carrier template. For example, when the actual response gain is lower than the expected response gain, increase the effective amplitude of the carrier action or extend the duration of the action segment without exceeding the equipment capacity and crop safety boundaries; when the actual response gain is higher than the expected response gain, decrease the effective amplitude of the carrier action or shorten the duration of the action segment.

[0069] When the channel matching deviation exceeds a preset matching deviation threshold, it indicates that there is a deviation between the physiological channel reference feature or demodulation threshold in the response demodulation rule and the current crop state. The edge gateway corrects the physiological channel reference feature based on the response object after execution. In one embodiment, it is updated as follows:

[0070] in, This indicates the updated physiological channel reference characteristics. This indicates the physiological channel reference characteristics before the update. This represents the actual response feature vector formed by the response object after execution. This represents the reference feature update coefficient. The updated equivalent control carrier template and response demodulation rules are written back to the crop growth knowledge base for use in subsequent control cycles.

[0071] Please see Figure 3 , Figure 3This is a schematic diagram of the system structure of the agricultural agent decision control method based on a crop growth knowledge base provided by the present invention; in one embodiment of the present invention, an agricultural agent decision control system based on a crop growth knowledge base is also provided to implement the above-mentioned agricultural agent decision control method based on a crop growth knowledge base, including: The necessary control action determination module is used to acquire agricultural field status data, equipment capacity status and control demand data, and determine the set of necessary control actions. The carrier action set generation module is used to retrieve equivalent control carrier templates and crop safety boundaries that match the necessary control action set from the crop growth knowledge base, and generate a candidate carrier action set. The equivalent control carrier template includes equivalent control amount constraints, carrier structure parameters, response sampling window and response demodulation rules. The target carrier action execution module is used to filter the candidate carrier action set based on the equipment capability status, crop safety boundary and equivalent control amount constraint. If there are no remaining candidate carrier actions, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Otherwise, the target carrier action is selected and executed. The response demodulation object generation module is used to collect crop response data within the response sampling window corresponding to the target carrier action and generate a crop response demodulation object. The hidden physiological channel demodulation module is used to demodulate the crop response demodulation object according to the response demodulation rules and determine the hidden physiological channel status. The intelligent agent control instruction generation module is used to call the agricultural intelligent agent to generate subsequent agricultural equipment control instructions based on the hidden physiological channel state, and to suppress control actions that do not match the hidden physiological channel state. The template rule update module is used to update the equivalent modulation carrier template and response demodulation rules according to the crop response after execution when executing the subsequent agricultural equipment control commands.

[0072] Specifically, the aforementioned system can be deployed in a greenhouse edge gateway, an agricultural control server, or a collaborative platform composed of an edge gateway and a cloud server. The necessary control action determination module receives sensor data, equipment controller feedback, and control requirement data, and outputs a set of necessary control actions; the carrier action set generation module calls the crop growth knowledge base and outputs a set of candidate carrier actions; the target carrier action execution module converts the selected target carrier actions into an agricultural equipment action sequence and issues it; the response demodulation object generation module receives data returned by the sensor within the response sampling window and generates a crop response demodulation object; the hidden physiological channel demodulation module outputs the hidden physiological channel status according to the response demodulation rules; the intelligent agent control instruction generation module generates subsequent agricultural equipment control instructions under the constraints of the control action whitelist and control action inhibition table; and the template rule update module updates the equivalent control carrier template and response demodulation rules in the knowledge base based on the executed crop response, thus completing this invention.

[0073] 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.

[0074] 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. An agricultural intelligent agent decision-making and control method based on a crop growth knowledge base, characterized in that, include: S1. Acquire on-site status data, equipment capability status and control requirement data, and determine the necessary set of control actions, which includes the applicable equipment, basic action types and total control amount; S2. Retrieve equivalent control carrier templates and crop safety boundaries from the crop growth knowledge base that match the basic action types in the necessary control action set. The equivalent control carrier template includes equivalent control quantity constraints, carrier structure parameters, response sampling windows, and response demodulation rules. Under the condition of satisfying the equivalent control quantity constraints, embed the carrier structure parameters into the action time series or action spatial distribution corresponding to the basic action type to generate candidate carrier actions. Associate the candidate carrier actions, the crop safety boundary, the response sampling window, and the response demodulation rules to generate the candidate carrier action set. The carrier structure parameters are used to limit the action amplitude, action interval, carrier period, or partition distribution. The response demodulation rules are used to identify hidden physiological channel states. The equivalent control quantity constraints include an allowable deviation range. When the absolute value of the difference between the equivalent control quantity of the candidate carrier action and the total control quantity of the corresponding basic action type does not exceed the allowable deviation range, the candidate carrier action satisfies the equivalent control quantity constraints. S3. Based on the equipment capability status, crop safety boundary and equivalent control quantity constraint, the candidate carrier action set is screened. If there are no remaining candidate carrier actions, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Otherwise, the target carrier action is selected and executed. S4. Collect crop response data within the response sampling window corresponding to the target carrier action, calculate the response gain, response hysteresis, and phase offset between the target carrier action and each response sequence in the crop response data, and associate the action time sequence, carrier structure parameters, equivalent control amount, response gain, response hysteresis, and phase offset of the target carrier action to generate a crop response demodulation object. S5. Read physiological channel reference features, demodulation thresholds, and discrimination thresholds from the response demodulation rules. The physiological channel reference features include stomatal regulation-limited reference features, root zone absorption-limited reference features, photosynthetic utilization-limited reference features, and transpiration coupling anomaly reference features. Match the response gain, response hysteresis, and phase shift in the crop response demodulation object with the physiological channel reference features to generate channel matching values ​​for each physiological channel. When the highest channel matching value reaches the corresponding demodulation threshold, and the difference between the highest and second-highest channel matching values ​​reaches the discrimination threshold, determine the physiological channel corresponding to the highest channel matching value as the hidden physiological channel state. When two or more channel matching values ​​reach the corresponding demodulation thresholds, and the difference between the highest and second-highest channel matching values ​​does not reach the discrimination threshold, generate a combined hidden physiological channel state consisting of two or more corresponding physiological channels. When the generation conditions for a single hidden physiological channel state or a combined hidden physiological channel state are not met, mark the hidden physiological channel state as an unidentified state. S6. When the hidden physiological channel is in an unidentified state, maintain the current control state and prohibit the generation of subsequent agricultural equipment control commands; when the hidden physiological channel is not in an unidentified state, call the agricultural agent to generate subsequent agricultural equipment control commands according to the hidden physiological channel state, and suppress control actions that do not match the hidden physiological channel state. S7. When the subsequent agricultural equipment control command is generated and executed, the equivalent control carrier template is updated according to the crop response after execution, wherein updating the equivalent control carrier template includes updating the response demodulation rules therein.

2. The agricultural intelligent agent decision-making and control method based on crop growth knowledge base according to claim 1, characterized in that, The control demand data includes the variable to be regulated, the target regulation direction, and the regulation demand quantity. Determining the set of necessary regulation actions includes: Read the variable to be regulated, the target regulation direction, and the regulation requirement from the control demand data; Based on the variable to be controlled and the target control direction, retrieve the applicable devices and basic action types from the preset device function mapping table; Calculate the total control amount corresponding to each basic action type based on the control demand. Remove the operable devices whose device capability status does not meet the executable conditions, and associate the remaining operable devices, basic action types, and total control quantities to generate the necessary control action set.

3. The agricultural intelligent agent decision-making and control method based on crop growth knowledge base according to claim 2, characterized in that, The generation of the candidate carrier action set includes: Read the basic action type and total control amount from the set of necessary control actions; Retrieve an equivalent control carrier template from the crop growth knowledge base that matches the action type index with the basic action type; Under the condition of satisfying the equivalent control amount constraint, the carrier structure parameters are embedded into the action time sequence or action spatial distribution corresponding to the basic action type to generate candidate carrier actions; The candidate carrier action, the crop safety boundary, the response sampling window, and the response demodulation rule are associated to generate the candidate carrier action set.

4. The agricultural intelligent agent decision-making and control method based on crop growth knowledge base according to claim 1, characterized in that, The selection and execution of the target carrier action includes: Read the device's executable range, minimum action resolution, and action switching latency from the device capability status; Candidate carrier actions whose carrier structure parameters exceed the range that the device can execute, are below the minimum action resolution, or do not meet the action switching delay are eliminated. The reference response amplitude is read from the response demodulation rules, the expected response state of the candidate carrier action is calculated based on the equivalent control amount constraint and the reference response amplitude, and candidate carrier actions whose expected response state exceeds the crop safety boundary are eliminated. Based on the carrier structure parameters, response sampling window, sensor sampling frequency in the field status data, and the reference response amplitude, the carrier solvable scheduling of the remaining candidate carrier actions is calculated according to the following formula: ; in, This indicates that the carrier can be descheduled. This refers to the carrier frequency in the carrier structure parameters. Indicates the sensor sampling frequency. Indicates the length of the response sampling window. This indicates the minimum number of sampling points required by the demodulation rule. Indicates the reference response range. This represents the minimum demodulated response amplitude. This indicates the weights configured by the response demodulation rules; The remaining candidate carrier action with the highest carrier solvability is selected as the target carrier action and executed.

5. The agricultural intelligent agent decision-making and control method based on crop growth knowledge base according to claim 1, characterized in that, The generated crop response demodulation object includes: Record the action time sequence and carrier structure parameters of the target carrier action, and calculate the equivalent control amount of the target carrier action according to the equivalent control amount constraint; Within the response sampling window, canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root region response sequence are collected; The action time series was time-aligned with the canopy temperature change sequence, canopy image change sequence, environmental response sequence, and root zone response sequence. Calculate the response gain, response hysteresis, and phase offset between the target carrier action and each response sequence; The crop response demodulation object is generated by associating the action time sequence, carrier structure parameters, equivalent control amount, response gain, response hysteresis, and phase offset.

6. The agricultural intelligent agent decision-making and control method based on crop growth knowledge base according to claim 1, characterized in that, The generation of subsequent agricultural equipment control commands and the suppression of control actions that do not match the state of the hidden physiological channel include: When the hidden physiological channel is in an unidentified state, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Retrieve from the crop growth knowledge base a whitelist of control actions and a control action inhibition table corresponding to the hidden physiological channel state; Input the hidden physiological channel status, equipment capability status, and control action whitelist into the agricultural intelligent agent to generate candidate subsequent control actions; Eliminate candidate subsequent control actions that match the control action inhibition table; Convert the candidate subsequent control actions that were not eliminated into subsequent agricultural equipment control commands; When all candidate subsequent control actions are eliminated, maintain the current control state and prohibit the generation of subsequent agricultural equipment control commands.

7. The agricultural intelligent agent decision-making and control method based on crop growth knowledge base according to claim 1, characterized in that, The update of the equivalent modulated carrier template includes: When the subsequent agricultural equipment control command is generated and executed, crop response data after the execution of the subsequent agricultural equipment control command is collected, and a post-execution response object is generated. The expected response features are read from the crop growth knowledge base based on the hidden physiological channel state. The post-execution response object is compared with the expected response characteristics to generate a response deviation object, which includes response gain deviation, response hysteresis deviation and channel matching deviation. When the response hysteresis deviation exceeds a preset hysteresis deviation threshold, the response sampling window in the corresponding equivalent modulated carrier template is corrected according to the response hysteresis deviation. When the response gain deviation exceeds a preset gain deviation threshold, the carrier structure parameters in the corresponding equivalent modulated carrier template are corrected according to the response gain deviation. When the channel matching deviation exceeds the preset matching deviation threshold, the physiological channel reference features and demodulation threshold in the corresponding response demodulation rules are corrected according to the channel matching deviation.

8. An agricultural intelligent agent decision control system based on a crop growth knowledge base, characterized in that, The method for implementing the agricultural agent decision-making and control method based on a crop growth knowledge base as described in any one of claims 1 to 7 includes: The necessary control action determination module is used to acquire on-site status data, equipment capacity status and control demand data, and determine the set of necessary control actions, which includes the applicable equipment, basic action types and total control amount. A carrier action set generation module is used to retrieve equivalent control carrier templates and crop safety boundaries from a crop growth knowledge base that match the basic action types in the necessary control action set. The equivalent control carrier template includes equivalent control quantity constraints, carrier structure parameters, a response sampling window, and a response demodulation rule. Under the condition of satisfying the equivalent control quantity constraints, the carrier structure parameters are embedded into the action time series or action spatial distribution corresponding to the basic action type to generate candidate carrier actions. The candidate carrier actions, the crop safety boundary, the response sampling window, and the response demodulation rule are associated to generate the candidate carrier action set. The carrier structure parameters are used to limit the action amplitude, action interval, carrier period, or partitioned distribution; the response demodulation rule is used to identify hidden physiological channel states. The equivalent control quantity constraints include an allowable deviation range. When the absolute value of the difference between the equivalent control quantity of the candidate carrier action and the total control quantity of the corresponding basic action type does not exceed the allowable deviation range, the candidate carrier action satisfies the equivalent control quantity constraints. The target carrier action execution module is used to filter the candidate carrier action set based on the equipment capability status, crop safety boundary and equivalent control amount constraint. If there are no remaining candidate carrier actions, the current control state is maintained and the generation of subsequent agricultural equipment control commands is prohibited. Otherwise, the target carrier action is selected and executed. The response demodulation object generation module is used to collect crop response data within the response sampling window corresponding to the target carrier action, calculate the response gain, response hysteresis and phase offset between the target carrier action and each response sequence in the crop response data, and associate the action time sequence, carrier structure parameters, equivalent control amount, response gain, response hysteresis and phase offset of the target carrier action to generate a crop response demodulation object. A hidden physiological channel demodulation module is used to read physiological channel reference features, demodulation thresholds, and discrimination thresholds from the response demodulation rules. The physiological channel reference features include stomatal regulation-limited reference features, root zone absorption-limited reference features, photosynthetic utilization-limited reference features, and transpiration coupling anomaly reference features. The module matches the response gain, response hysteresis, and phase shift in the crop response demodulation object with the physiological channel reference features to generate channel matching values ​​for each physiological channel. When the highest channel matching value reaches the corresponding demodulation threshold, and the difference between the highest and second-highest channel matching values ​​reaches the discrimination threshold, the physiological channel corresponding to the highest channel matching value is identified as the hidden physiological channel state. When two or more channel matching values ​​reach the corresponding demodulation thresholds, and the difference between the highest and second-highest channel matching values ​​does not reach the discrimination threshold, a combined hidden physiological channel state consisting of two or more corresponding physiological channels is generated. When the generation conditions for a single hidden physiological channel state or a combined hidden physiological channel state are not met, the hidden physiological channel state is marked as an unidentified state. The intelligent agent control command generation module is used to maintain the current control state and prohibit the generation of subsequent agricultural equipment control commands when the hidden physiological channel is in an unrecognized state; when the hidden physiological channel is not in an unrecognized state, it calls the agricultural intelligent agent to generate subsequent agricultural equipment control commands according to the hidden physiological channel state and suppresses control actions that do not match the hidden physiological channel state. The template rule update module is used to update the equivalent control carrier template according to the crop response after the subsequent agricultural equipment control command is generated and executed. The update of the equivalent control carrier template includes updating the response demodulation rules therein.

Images