Multi-autonomous robot collaborative operation control method and system based on Internet of Things

Abstract

The invention relates to the technical field of the Internet of Things, in particular to a multi-autonomous-robot collaborative operation control method and system based on the Internet of Things. The method comprises the following steps: acquiring multi-point leaf surface micro-wind speed detection data of plants in a greenhouse environment; analyzing plant stomata opening and closing state change characteristics according to blade surface micro wind speed detection data, and determining plant health state grades; detecting a plant stomata activity time distribution rule according to plant health state grading; determining detection time window data of the autonomous robot according to the plant stomata activity time distribution rule; and determining a time sequence collaborative operation scheme based on the plant stomatal activity time distribution rule and the detection time window data. According to the invention, airflow disturbance generated by movement of the robot is converted into a reference signal for environmental perception, and intelligent compensation of environmental disturbance is realized through comparative analysis of expected disturbance and an actual detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a multi-autonomous robot collaborative operation control method and system based on the Internet of Things. Background Art

[0002] The IoT-based multi-autonomous robot collaborative operation control method connects multiple autonomous robots through IoT technology to enable information sharing, task allocation, path planning, and collaborative operations. The IoT platform, serving as the core link, leverages sensor networks and wireless communication technologies to collect real-time robot and environmental data and transmit this data to a control center for processing and analysis. The control center uses optimization algorithms to rationally allocate tasks based on task requirements and robot status. Through a distributed control architecture, each robot independently determines its path based on its own sensory information and data obtained from the platform, while also coordinating with other robots. Regarding communication mechanisms, the system leverages low-latency, high-bandwidth wireless communication technologies (such as 5G) to ensure real-time information exchange between robots, optimizing path planning and effectively avoiding obstacles. Furthermore, the system incorporates fault detection and fault tolerance mechanisms, enabling real-time monitoring of robot status and, in the event of a fault, rapid switching to a backup robot to redistribute tasks and ensure operational continuity. However, in the agricultural sector, most existing agricultural robots operate in a stand-alone mode, which cannot meet the complex management requirements of large-scale greenhouses. Although multi-robot collaborative operations can improve work efficiency, traditional collaborative control methods often ignore the physiological rhythms of plants themselves, resulting in a mismatch between robot operations and the optimal growth requirements of plants, affecting the accuracy and effectiveness of agricultural production. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a multi-autonomous robot collaborative operation control method and system based on the Internet of Things to solve at least one of the above technical problems.

[0004] To achieve the above objectives, a multi-autonomous robot collaborative operation control method based on the Internet of Things includes the following steps: Step S1: Acquire micro-wind speed detection data on multiple leaf surfaces of plants in a greenhouse environment; analyze the characteristics of plant stomatal opening and closing state changes based on the micro-wind speed detection data on the leaf surfaces to determine the plant health status classification; Step S2: Detecting the time distribution pattern of plant stomatal activity according to plant health status classification; determining detection time window data of the autonomous robot based on the time distribution pattern of plant stomatal activity; and determining a time sequence collaborative operation plan based on the time distribution pattern of plant stomatal activity and the detection time window data; Step S3: Based on the sequential collaborative operation scheme, the movement paths of multiple robots are planned, and the airflow disturbance parameters generated by the robot movements are determined according to the plant health status classification. The task information transmission between robots is controlled according to the task allocation requirements in the sequential collaborative operation scheme to obtain the collaborative operation execution status. Step S4: Based on the collaborative work execution status, the greenhouse space airflow change detection during the robot movement is performed, and the environmental interference compensation is performed in combination with the airflow disturbance parameters to obtain the greenhouse wind field distribution map; the robot position is corrected and the task is dynamically allocated according to the greenhouse wind field distribution map to obtain the corrected robot collaborative work configuration.

[0005] The present invention also provides a multi-autonomous robot collaborative operation control system based on the Internet of Things, which is used to execute the multi-autonomous robot collaborative operation control method based on the Internet of Things. The multi-autonomous robot collaborative operation control system based on the Internet of Things includes: The plant health monitoring module is used to obtain micro-wind speed detection data from multiple leaf surfaces of plants in a greenhouse environment; based on the micro-wind speed detection data from the leaf surfaces, the characteristics of changes in the opening and closing states of the plant stomata are analyzed to determine the plant health status classification; The collaborative task planning module is used to detect the temporal distribution of plant stomatal activity according to plant health status; determine the detection time window data of the autonomous robot based on the temporal distribution of plant stomatal activity; and determine the time-series collaborative operation plan based on the temporal distribution of plant stomatal activity and the detection time window data; The multi-robot control module is used to plan the movement paths of multiple robots based on the time-sequential collaborative operation scheme and determine the airflow disturbance parameters generated by the robot movements according to the plant health status classification. It also controls the task information transmission between robots based on the task allocation requirements in the time-sequential collaborative operation scheme to obtain the collaborative operation execution status. The adaptive adjustment module is used to detect changes in greenhouse airflow during robot movement based on the collaborative work execution status, and to compensate for environmental interference in combination with airflow disturbance parameters to obtain a greenhouse wind field distribution map. The robot position is corrected and tasks are dynamically allocated based on the greenhouse wind field distribution map to obtain the corrected robot collaborative work configuration.

[0006] The present invention's Internet of Things-based multi-autonomous robot collaborative operation control method significantly improves the accuracy and efficiency of multi-robot collaborative operations in greenhouse agriculture by deeply integrating environmental perception and intelligent decision-making technologies. First, the method uses multiple micro-wind speed sensors distributed on the surface of plant leaves to collect micro-wind speed data in real time. Combined with analysis of the changing characteristics of plant stomata opening and closing, it achieves dynamic grading of plant health status. This health grading method based on plant physiological characteristics breaks through the limitations of traditional reliance on visual or manual observation. It can more meticulously and objectively reflect plant physiological changes, help accurately understand plant growth conditions, and promptly identify abnormalities, thereby providing a scientific decision-making basis for subsequent robot operations. Second, the method analyzes the time distribution of plant stomata activity and scientifically divides the robot's detection time window, avoiding blind and disorderly operation scheduling. Based on the combination of time windows and plant health levels, the method can rationally allocate robot operation tasks, achieve a dynamic balance between detection density and time, and effectively alleviate task congestion and resource waste. Furthermore, the calculation of the number and spatial distribution of plants within the inspection time window, combined with parameters such as robot movement speed and single-plant inspection time, accurately assesses the inspection workload, ensures the rational allocation of robot resources, and improves overall operational efficiency. Rational planning of the number of robots and operating times reduces energy consumption and equipment wear, extends robot life, and saves operation and maintenance costs. Regarding path planning, this method, based on a time-series collaborative operation scheme, generates a basic movement path for each robot, consisting of a starting point, an end point, and multiple intermediate nodes. A detailed movement path planning scheme is then developed, combining speed parameters and turning and deceleration rules. Path planning fully considers operational efficiency and movement safety, reducing the mechanical burden caused by ineffective movement and sharp turns. Airflow disturbance intensity parameters, based on plant health levels, are used to scientifically estimate the airflow disturbances generated by robot movement. This allows for the rational prediction and control of the impact of robot operations on the plant's physiological environment, preventing wind disturbances caused by robot activity from misjudging or causing abnormal changes in plant stomata, thereby improving environmental adaptability. In addition, this method establishes a standardized task information sharing mechanism between robots, clarifies the communication protocol and information transmission frequency, ensures that robots exchange detection target allocation, time window arrangement, path planning and status information in real time, and improves the synchronization and coordination of collaborative execution. By continuously monitoring task progress and communication status, anomalies are promptly discovered and marked, and it is possible to quickly respond to abnormal situations in task execution, ensure the stability and reliability of collaborative operations, and avoid overall operation interruption or efficiency reduction due to the failure of a single robot. In terms of environmental interference compensation, this method uses an airflow detection sensor network to collect wind speed and direction data around the robot in real time. Combined with the robot's movement path and airflow disturbance parameters, it accurately identifies the airflow disturbance components caused by the robot's movement. Based on the event of airflow disturbance exceeding the standard, the robot's movement parameters are dynamically adjusted, and environmental interference compensation instructions are generated to ensure the stability of the working environment.Combined with spatial interpolation, continuous wind field distribution data, wind speed contours, and wind direction vectors within the greenhouse are generated. This provides detailed information on greenhouse airflow distribution characteristics, supporting robot position correction and dynamic task allocation, improving robot positioning accuracy and operational flexibility. By assessing wind field intensity and actual position deviations, immediate position correction and task reallocation are achieved when thresholds are exceeded, avoiding operational blind spots and environmental degradation, and improving overall operational adaptability and environmental friendliness. Overall, this approach implements intelligent collaborative control across the entire chain, from environmental perception, health status determination, time window division, path planning, collaborative communication, to environmental interference compensation, enhancing the autonomous decision-making capabilities and collaborative efficiency of the robotic system. In practical application scenarios, this approach significantly improves the real-time monitoring and precise operation capabilities of multiple robots for plant health in greenhouse environments, ensuring a stable plant growth environment, reducing human intervention, and enhancing the automation and intelligence of operations. By scientifically allocating inspection tasks and optimizing movement paths, conflicts and resource waste between robots are reduced, improving overall robot swarm utilization and operational continuity. Efficient communication and anomaly monitoring mechanisms among multiple robots ensure reliable system operation, reducing maintenance costs and operational risks. Environmental disturbance compensation technology effectively mitigates the negative impact of robotic operations on the greenhouse microenvironment and promotes harmonious coexistence between robots and the plant physiological environment. This approach not only enhances the intelligence and precision of multi-robot collaborative operations but also helps greenhouse agriculture achieve green and efficient production. Through data-driven health status grading and time management, the goals of precision agriculture are achieved, effectively improving the scientific nature and sensitivity of plant growth monitoring. The dynamic optimization capabilities of the robotic collaborative operation solution support rapid response and adjustment to emergencies, enhancing the system's robustness and adaptability. The overall system, through cloud-based data integration and IoT technology support, achieves seamless integration of data and equipment, promoting intelligent upgrades in greenhouse management. This approach has broad application value in the future, not only in greenhouse environments but also in other fields of precision agriculture and intelligent manufacturing, laying a solid foundation for the intelligent development of multi-robot collaborative operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1 This is a schematic diagram of the steps of a multi-autonomous robot collaborative operation control method based on the Internet of Things of the present invention; Figure 2 for Figure 1 Detailed step flow diagram of step S1; Figure 3 for Figure 1 Detailed step flow chart of step S3 in FIG. DETAILED DESCRIPTION

[0008] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0009] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0010] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0011] To achieve this, please refer to Figures 1 to 3 The present invention provides a multi-autonomous robot collaborative operation control method based on the Internet of Things, the method comprising the following steps: Step S1: Acquire micro-wind speed detection data on multiple leaf surfaces of plants in a greenhouse environment; analyze the characteristics of plant stomatal opening and closing state changes based on the micro-wind speed detection data on the leaf surfaces to determine the plant health status classification; In a greenhouse tomato cultivation environment, an embodiment of the present invention deploys three micro-wind speed sensors near the leaf surface of each plant, located 20 mm from the top, middle, and root leaves. Each sensor has a sampling frequency of 5 Hz and an output voltage range of 0.1 V to 3.3 V. The system uploads real-time voltage signals to an edge computing module via an IoT gateway node. First, a one-to-one mapping of sensor IDs with their installation coordinates is performed to generate a micro-wind sensing voltage matrix. The matrix data is then sorted by acquisition time to form a time-continuous micro-wind speed data sequence. The voltage is converted to wind speed (in meters per second) using a factory calibration curve and interpolated based on the spatial position of the leaves to generate continuous micro-wind speed field data covering the entire greenhouse plant area. In wind speed field analysis, a local weighted regression method is used to extract background wind speed trends. After subtracting this background wind speed, the net wind speed change caused by plant transpiration is obtained. Subsequently, the wind speed change rate over time, the amplitude of the sudden change point, and the amplitude of the periodic fluctuation at each sensing point are extracted to determine whether the stomata are actively transpiring, temporarily closed, or permanently closed. Each monitoring point is assigned a stomatal status marker based on wind speed change thresholds for open and closed states defined in an existing plant physiology experiment library (e.g., 1.2 meters per second or higher is considered open, while a speed below 0.3 meters per second with no fluctuation for more than 20 minutes is considered closed). Data from multiple monitoring points within the same area are integrated and statistically analyzed. Combined with statistical indicators such as stomatal opening frequency, duration of a single opening, and average daily opening ratio, plant health levels are classified into three categories: high activity (healthy), moderate activity (mildly stressed), and low activity (severely stressed). This generates a plant health status grading table, which serves as a physiological reference for subsequent robot scheduling.

[0012] Step S2: Detecting the time distribution pattern of plant stomatal activity according to plant health status classification; determining detection time window data of the autonomous robot based on the time distribution pattern of plant stomatal activity; and determining a time sequence collaborative operation plan based on the time distribution pattern of plant stomatal activity and the detection time window data; In the plant health status classification data table obtained in this embodiment of the present invention, each record includes the plant number, the location coordinates of the region to which it belongs, and the health level (e.g., healthy, sub-healthy, or unhealthy). The stomatal status changes of these plants are continuously monitored on a daily basis. For plants with higher health levels, the time points of stomatal opening and closing are collected over a 24-hour period, and the duration of the opening is counted. An opening duration exceeding 15 minutes is considered an "effective opening event." The frequency of these events is counted in four daily time periods (0-6:00, 6-12:00, 12-18:00, and 18-24:00) to obtain a frequency distribution of active time periods for plants of each health level. In a particular tomato planting area, it was found that the stomatal activity of healthy plants was concentrated in the 6-12 period, with an activity intensity coefficient (i.e., the number of effective opening events in this period divided by the total number for the entire day) exceeding 60%. Given the system's set activity threshold of 40%, this period was marked as the active time period for plants of that level. Based on the overlapping analysis of the active times of all plant health levels, a time period overlap relationship diagram is formed. For example, 6:30-9:30 is the active time interval for both healthy and sub-healthy plants. Based on this, the detection load within each time period is calculated. For example, if 60 plants need to be inspected in a certain time period, the robot takes 3 minutes to inspect each plant, and the maximum continuous operation time of the robot is 60 minutes, combined with the average robot movement speed of 1.2 meters per second, it is finally concluded that 4 robots need to be allocated for inspection in this high-density section. This forms a detection time window data table including parameters such as the detection time window number, corresponding health level, number of plants to be inspected, start and end times, and number of robots, providing a foundation for the next step of building a collaborative operation timing plan.

[0013] Step S3: Based on the sequential collaborative operation scheme, the movement paths of multiple robots are planned, and the airflow disturbance parameters generated by the robot movements are determined according to the plant health status classification. The task information transmission between robots is controlled according to the task allocation requirements in the sequential collaborative operation scheme to obtain the collaborative operation execution status. Based on the aforementioned detection time window data and plant health status grading results, the present embodiment prioritizes the time windows, for example, prioritizing time windows with high health levels and a large number of detected plants. Robot resources are pre-allocated. For example, if there are currently six robots, and the pre-allocation reveals that the 8:00-9:00 time window requires four robots and the 9:00-10:00 time window requires three robots, there is a resource conflict for one robot. The system prioritizes the requirements of the first time window and defers the execution of the second time window by 15 minutes or replaces it with another robot. After completing robot resource allocation, a path is planned for each robot based on the spatial coordinates of the plants, prioritizing connecting adjacent plants using the shortest path principle. If a path contains corners greater than 90 degrees, the system automatically adds a 2-second deceleration wait time at that point to ensure path safety. Furthermore, the system calculates "airflow disturbance intensity" for the robots based on their health status. For example, if the maximum allowable disturbance in an unhealthy area is 1.5 meters per second, the system reduces the movement speed of all robots passing through that area, creating a disturbance-controlled path. Finally, an information exchange node is set up to synchronize the robot's position, task completion, and status code every 15 minutes to ensure inter-robot coordination and consistency. Task information transmission uses a protocol format such as robot ID, current position coordinates, task number, completion percentage, and device status (0 - normal, 1 - delayed, 2 - abnormal). This process generates the robot collaborative task execution status, including key data such as task assignment completion, path trajectory arrangement, and communication status.

[0014] Step S4: Based on the collaborative work execution status, the greenhouse space airflow change detection during the robot movement is performed, and the environmental interference compensation is performed in combination with the airflow disturbance parameters to obtain the greenhouse wind field distribution map; the robot position is corrected and the task is dynamically allocated according to the greenhouse wind field distribution map to obtain the corrected robot collaborative work configuration.

[0015] When a robot begins a task, an embodiment of the present invention activates an array of wind speed and direction sensors installed in a greenhouse grid, sampling at a spacing of one node per square meter. The collected wind speed data is compared with the overlapping areas of the robot's current path to identify areas exceeding the acceptable disturbance threshold for plants. For example, if the system defines a wind speed disturbance threshold of 2.0 meters per second, a wind speed exceeding the threshold of 2.5 meters per second in a particular area triggers an airflow disturbance exceeding threshold event. The system automatically issues movement deceleration commands or path fine-tuning commands to the robot to compensate. After compensation, wind speed data is collected again, and an inverse distance weighted interpolation method is used to generate wind speed contour maps and wind direction vector maps covering the entire greenhouse, forming a greenhouse wind field map distributed on a 1 meter x 1 meter grid. On this greenhouse wind field map, the system compares the actual direction of the wind speed vector with the planned path. If the deviation exceeds 0.5 meters, a position correction command is issued. Furthermore, in areas with excessively strong wind speeds, which are deemed unsuitable for the current robot to continue operating, the system automatically reassigns these tasks to nearby idle robots or defers them to the next time window. Finally, the system updates the robot's target list, path planning, and detection tasks to form a corrected robot collaborative operation configuration, ensuring that the robot's detection activities are successfully completed in a safe and stable microenvironment and with good coordination of plant physiological rhythms.

[0016] Preferably, step S1 includes the following steps: Step S11: collecting data on real-time voltage signals from a plurality of micro-wind speed sensors distributed near the surface of plant leaves in the greenhouse, and spatially marking the voltage signals according to the installation position coordinates of the micro-wind speed sensors to obtain micro-wind sensing voltage matrix data; In an embodiment of the present invention, three micro-wind speed sensors were deployed near the main stem of each tomato plant in a smart greenhouse covering an area of ​​approximately 200 square meters. These sensors were located approximately 20 mm from the leaf surface, around the top, middle, and ground-level leaves. These sensors were ultra-low-speed hot-film anemometers with a sensitivity range of 0.05 to 2.5 meters per second and a voltage output range of 0.1 to 3.3 volts. The data acquisition system collected data from each sensor once per second, recording its output voltage in real time and transmitting the data to an edge gateway device via an RS485 bus. During data recording, each sensor's unique number was associated with its actual deployment coordinates (a Cartesian coordinate system established with the greenhouse's lower left corner as the origin). For example, sensor FS1021 was located at coordinates (12.3, 8.7, 0.6) meters. In this way, all collected voltage values ​​are attached with spatial coordinate information to form a set of multidimensional perception matrix data consisting of the four-tuple "sensor number-time stamp-voltage value-spatial coordinate", which is called breeze perception voltage matrix data.

[0017] Step S12: sorting the breeze sensing voltage matrix data by acquisition timestamp to obtain multi-point blade surface breeze speed detection data; The embodiment of the present invention imports the breeze sensing voltage matrix data collected in step S11 into the preprocessing module, first classifies and organizes them according to the sensor number, and then sorts the voltage series of each sensor according to the timestamp field to ensure the formation of strict time continuity data. The sorting adopts an ascending rule, arranging them one by one from the earliest time, and the sorting time granularity is 1 second. After sorting, the voltage data of each sensor at each time point are combined into a time series matrix. The rows of the matrix represent the time points, and the columns represent different sensors. The value of each cell is the voltage value at a certain position at a certain time point. Since the data is accompanied by coordinate information, the result obtained by the sorting process not only has time series properties, but also has spatial distribution characteristics, thereby forming a set of multi-point blade surface breeze speed detection data that can be used for wind speed modeling, providing stable input for subsequent wind speed conversion and interpolation analysis.

[0018] Step S13: performing sensor calibration conversion processing on the micro-wind speed detection data of multiple blade surfaces, and performing spatial interpolation calculation based on the spatial distribution of the blades to obtain micro-wind speed field data with continuous distribution in the greenhouse space; The present embodiment, based on sorted micro-wind speed data, first requires calibrating the raw voltage values ​​for wind speed. A pre-set nonlinear calibration curve is used to convert the sensor output voltage to actual wind speed. For example, for a hot-film sensor, calibration data indicates that 0.5 volts corresponds to approximately 0.1 meters per second, and 2.5 volts corresponds to approximately 2.0 meters per second. The calibration results are mapped from voltage to wind speed using a lookup table, while invalid values ​​are discarded (e.g., values ​​below 0.05 volts are considered disconnected, and values ​​above 3.3 volts are considered abnormal). After completing the voltage-to-wind speed conversion, wind speed values ​​at each spatial point at the same time are interpolated and reconstructed using a three-dimensional inverse distance weighted interpolation method. At each unknown grid point, the wind speed value is determined by the weighted average of the values ​​of several surrounding known points, with the weight being inversely proportional to the distance. Because the greenhouse space is a regular square, the system divides the entire space into voxel units measuring 1 meter by 1 meter by 0.5 meters, constructing approximately 2,400 interpolation points. This interpolation yields continuous wind speed field data covering the entire plant leaf layer. The wind speed field reflects the changes in local micro-airflow caused by plant transpiration, water vapor release, etc. at different locations, and is an important basis for the subsequent identification of stomatal activity.

[0019] Step S14: performing spatial variation characteristic analysis on the micro-wind speed field data, identifying background wind speed components, and performing component separation on the micro-wind speed field data based on the background wind speed components to obtain plant transpiration net wind speed change data; Based on the constructed wind speed field data, the embodiment of the present invention first performs a spatial variation feature analysis. The specific operation is to calculate the gradient of the wind speed value of each voxel unit in the spatial dimension, and estimate the "background wind speed component" by identifying the area where the wind speed changes slowly. That is, the continuous airflow background caused by non-plant factors caused by environmental sources such as the air conditioning circulation system and ventilation fans. In actual application, it was found that the vents at the top of the greenhouse can cause a wind speed gradient of about 0.3 meters per second per hour, which is stably distributed in the vertical direction. The system models this distribution as a background field. The original wind speed field is then subjected to "background stripping" processing, that is, the wind speed value of each point is subtracted from its corresponding background value to obtain the net wind speed change value under the dominant influence of plants, that is, the net wind speed data of plant transpiration. This component separation operation ensures that the subsequent judgment of the opening and closing status of the stomatal pores is not interfered by irrelevant airflow, so that the system only focuses on the real breeze disturbance formed between the plant leaves and the environment, thereby improving the reliability of health status analysis.

[0020] Step S15: Analyze the characteristics of plant stomatal opening and closing state changes based on the plant transpiration net wind speed change data to determine the plant health status classification.

[0021] The present embodiment utilizes the aforementioned plant transpiration net wind speed change data. First, a time series analysis is performed on the net wind speed data at each monitoring point. Parameters such as the rising slope, falling slope, extreme value frequency, and magnitude of the sudden change point within every 30 minutes are extracted to construct a wind speed temporal variation feature set. In conjunction with the plant physiology knowledge base, the system sets a threshold for rapid wind speed increases corresponding to the open state at 0.5 meters per second per 10 minutes and a threshold for wind speed decreases corresponding to the closed state at 0.3 meters per second per 10 minutes. A high-frequency opening threshold and an extremely low fluctuation range threshold are also set to identify abnormal states. The system compares the characteristic data of each monitoring point with the thresholds and records the stomatal opening and closing state at each moment as "open," "closed," or "uncertain." The system then calculates the total number of openings, average duration, and percentage of open time for each monitoring point per day. For example, a tomato plant opens 12 times in 24 hours, with an average duration of 18 minutes per opening. This percentage represents over 60% of the total open time. Based on the health threshold in the plant physiology database, the plant is labeled healthy. Another plant, on the other hand, opened its leaves only four times, for a total duration of just 12 minutes, and was assessed as being in a state of "severe stress." Ultimately, all plants were classified into three levels: "healthy," "sub-healthy," and "unhealthy," based on the intensity and response patterns of their stomatal activity. This structured plant health status grading data table was constructed, which served as input for scheduling robot inspection tasks.

[0022] Preferably, step S15 includes the following steps: Step S151: Calculate the time change rate and change amplitude of the wind speed at each monitoring point for the plant transpiration net wind speed change data, and calculate the periodic fluctuation pattern and mutation point to form the wind speed time change characteristics; After acquiring data on the net transpiration wind speed changes around each plant in the greenhouse, the system first imports and processes the wind speed time series for each monitoring point, with a data granularity of one record per minute, for a total of 1440 data points per day. A programmed algorithm then calculates the time rate of change of this series: the difference between the current wind speed and the previous wind speed divided by the interval time, yielding the wind speed change rate per minute. The difference between the maximum and minimum wind speeds in each time period is then calculated as an indicator of the magnitude of change. A sliding window method is then used to analyze the data for periodic fluctuations. For example, using a 120-minute window, the system determines whether the wind speed exhibits a regular increase-decrease cycle, thereby identifying the plant's transpiration rhythmicity. A sudden inflection point detection algorithm (such as the double mean shift algorithm) is then used to identify sudden inflection points in the wind speed curve. For example, a rise in wind speed from 0.1 to 0.9 meters per second within one minute is recorded as a possible sudden stomatal opening event. The final output of each monitoring point includes the wind speed time variation characteristics of "time change rate series", "wind speed amplitude index", "periodic change feature identifier" and "mutation point timestamp", providing quantifiable input for subsequent analysis of stomatal opening and closing status.

[0023] Step S152: determining the wind speed increment threshold for the stomatal open state and the wind speed decrement threshold for the stomatal closed state based on pre-acquired basic knowledge of plant physiology and experimental data, and establishing a corresponding relationship table between the wind speed variation amplitude and the stomatal opening and closing degree; The system in this embodiment incorporates a plant physiology knowledge base, constructed by a team of agricultural experts based on the transpiration characteristics of common greenhouse crops such as tomatoes, cucumbers, and peppers. This knowledge base includes multiple experimental data models. For tomatoes, laboratory measurements show that when the net wind speed on the leaf surface increases by more than 0.4 meters per second within 10 minutes, the majority of samples indicate fully open stomata. A decrease of more than 0.3 meters per second within 10 minutes signals stomatal closure. Therefore, the system sets a "wind speed increment threshold" of 0.4 meters per second and a "wind speed decrement threshold" of 0.3 meters per second, with a ±0.1 meter per second buffer zone to account for minor fluctuations. Based on this experimental data, a table is established that maps wind speed variation to stomatal opening and closing levels—a switch correspondence table. This table maps wind speed increment, decrement, and cycle frequency to open and closed states, categorized into three levels: "fully open," "partially open," and "stable closed." This table is fed into the robot control system in tabular form and serves as a reference for inferring the stomatal state at each location.

[0024] Step S153: comparing the wind speed time variation characteristics with the thresholds in the switch correspondence table one by one, and recording the stomata opening and closing status of each monitoring point at different times; This embodiment of the present invention compares the wind speed time-varying characteristics generated in the previous step with the wind speed increase and decrease thresholds defined in the switch correspondence table, point by point. Specifically, the system iterates over the wind speed change rate and amplitude values ​​at each monitoring point within each 10-minute window to determine whether they meet the "open threshold" or "close threshold." If so, the stomatal status at the current time point is recorded as "open" or "closed." If it falls within the buffer zone, it is marked as "uncertain." For example, at a monitoring point on a tomato plant, the wind speed increased from 0.25 to 0.71 meters per second from 10:10 to 10:20, an increase of 0.46, meeting the open criteria and the status being recorded as "open." Subsequently, from 10:30 to 10:40, the wind speed decreased from 0.73 to 0.38, a decrease of 0.35 meters per second, and the status was updated to "closed." Through this time-period comparison, each monitoring point generates a complete, time-series stomatal status record, with a data structure of "monitoring point number - timestamp - stomatal status," providing the multi-robot system with a key basis for determining the detection time window.

[0025] Step S154: Counting the stomatal opening and closing state change frequency and stomatal opening and closing duration of each monitoring point, and calculating the proportion of the opening time to the total observation time to obtain stomatal opening and closing state change data; Based on the sequence of stomatal opening and closing state records generated in the previous step, the system in this embodiment counts the total number of times each monitoring point appears "open" throughout the day, recording this as the opening frequency. The system also calculates the duration of each opening, taking into account the average and standard deviation. Furthermore, the total "open" state duration for the entire day is calculated and divided by the total observation time (24 hours) to determine the open time percentage. For example, monitoring point A appears open 12 times throughout the day, each lasting an average of approximately 22 minutes, for a total open time of 264 minutes, representing 18.3% of the total 1440 minutes of the day. The system also records the longest single opening time and the interval between openings, which are used to subsequently assess the stability of the plant's response to conditions such as light, temperature, and humidity. All of these statistical data constitute "stomatal opening and closing state change data," which is output as a structured table. Each row records indicators such as the monitoring point number, opening frequency, average opening duration, open time percentage, and maximum closed time, providing a quantifiable basis for subsequent plant health grading.

[0026] Step S155: Calculate the overall activity and response speed of plant stomatal activity based on the stomatal opening and closing state change data, and match them with the preset health standards to obtain the plant health status classification.

[0027] This embodiment of the present invention imports the stomatal activity data from each monitoring point obtained in step S154 into the health assessment module, where it is matched against built-in health standards for each crop type. Taking tomatoes as an example, the health status criteria are set as follows: daily opening frequency of at least 8 times, average opening duration of at least 15 minutes, opening time ratio of at least 20%, and maximum closing interval of no more than 3 hours. If all indicators are met, the plant is classified as "healthy." If some of the indicators are critical, the plant is classified as "sub-healthy." If multiple indicators deviate significantly, the plant is classified as "unhealthy." The system compares these criteria for each monitoring point and calculates the health level of each plant. It also uses the plant's location information in the greenhouse to perform spatial distribution annotation. For example, the tomato plant numbered PLT1021 is rated "healthy" due to its location (8.2 meters, 4.7 meters). The cucumber plant numbered PLT1083 is rated "sub-healthy" because it is located near a ventilation vent and is susceptible to cold drafts. Finally, a complete plant health status classification data table is generated, which will be used in robot detection time planning and path avoidance design to ensure that unhealthy plants are inspected and environmentally regulated first.

[0028] Preferably, in step S2, detecting the temporal distribution of plant stomatal activity according to the plant health status in a graded manner includes: Establishing a plant health status data table including plant number, health grade and location coordinates according to plant health status classification; Based on the health level information in the plant health status data table, the stomatal opening and closing activities of plants at each health level are continuously monitored within 24 hours, and the stomatal opening and closing time data of each monitoring point are recorded; The stomatal opening duration of each monitoring point is calculated based on the stomatal opening and closing time data. When the duration exceeds 15 minutes, it is marked as a valid opening event, and a valid opening event time series is generated. Based on the time series of effective opening events, the event frequency distribution data of stomatal activity of plants of each health level in four time periods of 0-6 o'clock, 6-12 o'clock, 12-18 o'clock, and 18-24 o'clock were statistically analyzed; The activity intensity of plants at each health level in each time period is calculated based on the event frequency distribution data. When the activity intensity coefficient is greater than the preset threshold, it is determined as the active time period data of the level; The temporal distribution pattern of plant stomatal activity is generated based on the active time period data of each health level.

[0029] After identifying the stomatal status and health grading of greenhouse plants, the system in this embodiment compiles basic information about each plant into a standardized data table. Fields include plant number (e.g., PLT_2023_0051), health level (categorized as "healthy," "sub-healthy," and "unhealthy"), and two-dimensional spatial coordinates (e.g., x = 6.5 meters, y = 3.8 meters). This data, automatically collected and aggregated by sensor networks and IoT edge computing devices deployed in the greenhouse environment, is consolidated into a structured data file called "Plant Health Status Data Table" in CSV or database table format. This data table serves as an important foundation for subsequent analysis of the distribution of plant stomatal activity. It allows the system to retrieve the distribution range and location of plants at a specific health level over different time periods, providing data support for subsequent time window allocation and path avoidance. Using the "Plant Health Status Data Table" as input, the system selects plant numbers and corresponding monitoring points for each health level, then continuously monitors changes in stomatal status for these plants over a 24-hour period. The system reads stomatal opening and closing status data once a minute and, for each monitoring point, records the timestamp of the transition from "closed" to "open" as the "opening time" and the timestamp of the transition from "open" to "closed" as the "closing time." For example, if PLT_2023_0051 is recorded as "open" at 8:05 and "closed" at 8:33, this period forms a pair of opening and closing times. All time-pair data is structured, keyed by time and indexed by the monitoring point number, forming a sequence of stomatal opening and closing time records for each monitoring point, providing a basis for subsequent screening of valid opening events. The system iterates over each pair of stomatal opening and closing time recorded in the previous step and calculates the duration between the opening and closing times. For example, the stomatal opening event from 8:05 to 8:33 lasted 28 minutes. The system sets a threshold of 15 minutes to determine the "validity" of stomatal activity. That is, if an opening event lasts for 15 minutes or longer, it is recorded as a "valid opening event"; otherwise, it is ignored. Taking tomato crops as an example, this threshold is determined based on their typical transpiration response time. After screening, the start and end times of valid events are recorded in the "Valid Opening Event Time Series" format: monitoring point number, opening timestamp, duration, and health level. This time series serves as the core data source for subsequent time period distribution statistics and eliminates invalid stomatal change events caused by transient disturbances or system misjudgments. Based on the valid opening event time series, the system assigns each event to one of four time periods: "0-6:00," "6-12:00," "12-18:00," or "18-24:00," and categorizes and counts them by health level. For example, PLT_2023_0051, with a "Sub-Health" level, had five valid opening events within 24 hours: one occurring between 6-12:00, two between 12-18:00, and two between 18-24:00. This results in a frequency distribution of 1:2:2.This statistical process uses a grouping and aggregation algorithm to generate a structured table structured as "health level - time period - event frequency." The event frequencies for multiple plants of the same level are combined and averaged to form a standardized basis for comparing activity intensity across different time periods, paving the way for determining active time periods. Based on the event frequency distribution data obtained in step 4, the "activity intensity coefficient" for each health level in each time period is further calculated. This coefficient is defined as the ratio of the frequency of valid events within that time period to the total number of plants of that level. For example, if 20 plants in the "health" level had 45 valid events recorded between 6:00 and 12:00, the activity intensity coefficient for that time period would be 2.25. A preset threshold of 1.5 is set; if the activity intensity coefficient is greater than 1.5, the time period is considered an "active time period" for that health level. Plants of different levels may have different active periods. Based on this determination, the system generates a corresponding active time period list for each health level, along with the activity intensity value and time period index, for use by the robot inspection scheduling module. The active time period data for plants of each health level, output from step 5, is combined to create a "table of plant stomatal activity temporal distribution patterns." This table, with health level as the primary key, records all time periods determined to be active for that level. For example, plants with a "sub-healthy" health level are active between 6:00 and 12:00 and 12:00 and 18:00, with activity intensity coefficients of 1.9 and 2.1, respectively. This distribution pattern table serves as a key input for robot inspection planning, enabling robots to perform operations when plant stomata are most active and gas exchange is most pronounced, improving inspection accuracy and task efficiency. Furthermore, the distribution pattern is used as a time weighting factor in subsequent time window partitioning, ensuring that robot operations are prioritized during periods of high activity, reducing the probability of omissions.

[0030] Preferably, determining the detection time window data of the autonomous robot according to the temporal distribution law of plant stomatal activity in step S2 includes: According to the time distribution of plant stomatal activity, the start and end time of each active time period are extracted, and an active time period list table is established; sorting the active time periods of each health level in chronological order based on the time information in the active time period list table, identifying the active time periods that overlap between plants of different health levels, and generating time period overlap relationship data; Based on the overlapping relationship data of time periods, the number of active plant health levels in the same time period is counted. When the number of active levels exceeds 2 at the same time, it is marked as a high-density detection period, and a detection density distribution table is formed; Based on the high-density detection period information in the detection density distribution table, each high-density period is divided into 30-minute sub-periods. For non-high-density periods, the original time period length is maintained to generate an initial time window division scheme; According to the duration of each time window in the initial time window division scheme and the plant distribution position in the plant health status data table, the number of plants that need to be inspected in each time window and the spatial distribution range are calculated to obtain the plant inspection load distribution data; Based on the plant inspection load distribution data, the inspection task workload in each time window is calculated according to the parameters of a robot movement speed of 1.2m / s and a single plant inspection time of 3 minutes; Based on the workload of the inspection task and the maximum working time of a single robot of 60 minutes, the number of robots required for each time window is determined, and a robot demand allocation table is generated; Autonomous robot inspection time window data including window number, start time, end time, target plant health level, number of inspected plants, and number of assigned robots are generated based on the robot demand allocation table.

[0031] After obtaining a table of plant stomatal activity temporal distribution patterns, the system of this embodiment extracts the start and end times of active periods for each plant health level. For example, for the "Healthy" level, the active periods are 06:30 to 09:00 and 14:00 to 16:30; for the "Sub-Healthy" level, the active periods are 08:00 to 10:00 and 15:00 to 17:00; and for the "Unhealthy" level, the active periods are 07:00 to 11:00 and 13:30 to 15:30. The system records this information as a triplet of "level - start time - end time" and stores it in an "Active Time Period List Table" in a unified CSV or SQL database format, with each row corresponding to one active period. This list facilitates subsequent comparison and overlap analysis of time periods across levels using time as the primary axis, providing a direct basis for preliminary time window demarcation. The system sorts all time periods in the "Active Time Period List Table" in ascending order by start time and then uses an interval overlap determination algorithm to compare the temporal overlap of plants at each health level. The system defines an overlap as long as two time periods overlap for more than 15 minutes. For example, if the "Healthy" level (06:30-09:00) overlaps with the "Unhealthy" level (07:00-11:00) for a period of one hour (07:00 to 09:00), this is considered an overlap. The system labels each pair of overlapping time periods and records the overlapping start and end times, as well as the names of the involved levels. This generates a "Time Period Overlap Relationship Data Table" with fields such as "Overlap Start Time," "Overlap End Time," and "Involved Level Set," providing basic data for subsequent identification of high-density detection periods. After obtaining the time period overlap data, the system counts the number of plant health levels involved in each overlapping interval. If a time period has two or more active plants of different health levels (e.g., 08:00-09:00 includes plants of different health levels, including Healthy, Sub-Healthy, and Unhealthy), the system marks that time period as a "High-Density Detection Period." All high-density detection periods are summarized in a "Detection Density Distribution Table," which includes fields such as the time period number, start and end time, participating level values, a list of level types, and a label (High-Density or Non-High-Density). This table is a reference framework for detection resource demand data that takes into account multiple levels of active behaviors. It helps to reasonably divide the time window granularity and guide task priority sorting in the future. The high-density detection period is divided into time sub-intervals based on the detection density distribution table. Each high-density period is divided into several consecutive sub-time periods of 30 minutes. For example, a 2-hour high-density period of 07:00-09:00 will be divided into four time windows: 07:00-07:30, 07:30-08:00, 08:00-08:30, and 08:30-09:00. For low-density or non-overlapping time periods, such as healthy plants that are active only from 14:00 to 16:30, no sub-division is performed and the original time span is maintained.Each time period records its window number, start and end times, and whether it is a high-density marker, forming an "initial time window partitioning scheme." This data is structured as a table or JSON, providing the smallest granularity for subsequent detection load calculations. Based on the time span information for each time window in the "initial time window partitioning scheme," combined with the health level and spatial coordinates of each plant in the previously established "Plant Health Status Data Table," the system selects the set of plants to be inspected within the current window. For example, within window W07 (07:30-08:00), there are 27 active plants, both healthy and unhealthy, covering the greenhouse area (312 meters by 26 meters). The system records the number of inspected plants and their XY coordinate boundaries for each time window, outputting the "Plant Detection Load Distribution Data Table." Fields include the time window number, the total number of inspected plants, the maximum and minimum horizontal and vertical coordinate values ​​(used for spatial density estimation), and level distribution statistics. After determining the number and distribution of plants to be inspected in each time window, the system estimates the theoretical workload for each window based on two parameters: the robot's movement speed (set at 1.2 m / s) and the per-plant inspection time (fixed at 3 minutes). The system first calculates the average distance between plants and divides the total distance by the movement speed to obtain the total movement time. This is then multiplied by the number of plants to obtain the total static inspection time. The sum of these two factors represents the total workload time. For example, if 20 plants are to be inspected within a window, the total movement path length is estimated to be 36 meters (accounting for loops), the movement time is approximately 30 seconds, and the inspection time is 60 minutes, resulting in a total workload of 60.5 minutes. The workload for each window is recorded in the "Time Window Task Workload Table," which includes fields such as the window number, estimated time (in minutes), number of plants to be inspected, path length, and area density, providing a reference for subsequent robot allocation decisions. Based on these workload estimates, the system determines the minimum number of robots required for each time window, subject to the constraint that each robot must work continuously for no more than 60 minutes. For example, if a task window lasts 95 minutes, at least two robots will be deployed and the task will be evenly divided. If it only takes 45 minutes, a single robot can complete the task. The system uses rounding up to ensure that the load does not overflow and that idle time is avoided. A "Robot Demand Allocation Table" is ultimately generated. Fields include the time window number, the number of robots required, the workload distribution ratio (e.g., the time each robot shares), and the task concentration level. This is used by the scheduling system to schedule the robots' scheduled departures. Combining the "Robot Demand Allocation Table" with the "Initial Time Window Division Scheme," the system integrates all information to generate the final "Autonomous Robot Inspection Time Window Data Table." This table records fields such as the window number, start time, end time, target plant health level (which can be a set, such as healthy + sub-healthy), the number of target plants for inspection, and the number of robots required.For example, the record for window W07 is: "W07, 07:30, 08:00, healthy + unhealthy, 27 strains, 2 robots." This table serves as the core input structure for robot task scheduling and collaborative path planning. It will be distributed to the robot control terminal via the IoT platform to achieve precise scheduling of time-series collaborative operations.

[0032] Preferably, the timing collaborative operation scheme determined based on the temporal distribution pattern of plant stomatal activity and the detection time window data in step S2 includes: According to the window number and time parameters in the detection time window data and the activity intensity information of each health level in the plant stomatal activity time distribution law, the time window is assigned a detection priority weight to obtain a time window priority allocation table; Pre-allocate robot resources for each time window based on the time window priority allocation table and the robot demand allocation table to form a robot resource pre-allocation plan; According to the robot allocation status of each time window in the robot resource pre-allocation plan, the robot resource conflicts between adjacent time windows are identified. When the same robot is allocated to consecutive windows with a time interval of less than 10 minutes, it is marked as a resource conflict and a resource conflict identification table is generated; The robots are reallocated for the conflicting time windows in the resource conflict identification table, the robot requirements of the high-priority windows are maintained, and the conflicting robots are allocated to the secondary priority windows or the execution time is delayed to obtain the robot allocation plan. According to the robot allocation plan and the position coordinate information in the plant health status data table, a specific inspection path sequence is planned for each robot. The plant inspection sequence within the same time window is arranged based on the proximity principle, and the robot inspection path planning data is generated. Based on the robot detection path planning data, the movement time and standby time of each robot between different time windows are calculated. When the movement time exceeds 5 minutes, the movement buffer time is added to the task sequence to form a detailed task sequence table; Set the information exchange nodes between robots according to the detailed task schedule, set a status synchronization time point every 15 minutes, and obtain the robot collaborative communication node data; Based on the robot collaborative communication node data, a time-series collaborative operation plan is generated, which includes robot numbering, execution time sequence, detection target allocation, movement path arrangement and coordinated communication plan.

[0033] This embodiment of the present invention assigns a detection priority weight to each window based on its window number, start and end times, and target plant health level in the "detection time window data," combined with the activity intensity values ​​for different health levels in each time period in the "plant stomatal activity temporal distribution pattern." For example, if the stomatal opening rate for the health level during the 7:30-08:00 period is 85%, while that for the sub-health level is only 50%, the system assigns a higher priority score to that time window, such as 0.9. This assignment can utilize a weighted model, comprehensively considering factors such as activity intensity, plant population, and level weight. This ultimately creates a "time window priority allocation table" containing fields such as window number, target health level, average activity intensity, and priority score, which guides subsequent robot scheduling priorities. The system combines this "time window priority allocation table" with the "robot demand allocation table" to perform a preliminary robot number matching for each window. This matching strategy uses a top-down allocation strategy based on priority sorting, prioritizing robot demand for high-priority windows. For example, if window W08 has a priority of 0.92 and requires two robots, the system prioritizes assigning robots R1 and R2 to it. It then assigns the remaining available robots to window W09, which has a priority of 0.86. This step does not consider conflicts; it only generates an initial "Robot Resource Pre-allocation Plan," which includes the time window number, a list of assigned robot numbers, priority values, and window period information. The allocation information in the "Robot Resource Pre-allocation Plan" is analyzed to check whether the same robot is assigned to consecutive windows with too short a time interval. For example, if robot R3 is assigned to W11 (08:00-08:30) and W12 (08:35-09:00) with a time interval of only 5 minutes, this is considered a resource conflict. Using a minimum safety buffer of 10 minutes, the system iterates through all robot numbers and their task sequences to generate a "Resource Conflict Identification Table." This table records fields such as the conflict type, involved robot numbers, conflict window numbers, time overlap range, and conflict severity level, providing a basis for resource reallocation. Based on the conflicting windows identified in the resource conflict identification table, the system adopts a "keep high, abandon low" strategy. This strategy maintains the robot assignment to the higher-priority window, removes the conflicting robot from the lower-priority window, and attempts to assign it to an idle robot. If no replacement robot is available, the task execution is deferred for one window period. For example, if W14 and W15 conflict, and W14 has a priority of 0.91 and W15 has a priority of 0.77, the system retains W14's assignment and adjusts W15 to W17 (the next available empty window). This strategy generates a final "robot allocation plan," which includes fields such as the window number, robot number, whether to adjust, the reason for the adjustment, and the final execution time, ensuring that task conflicts are resolved and resources are properly allocated.Based on the robot allocation plan and the actual coordinate positions of the plants in the greenhouse as listed in the plant health status data table, the system generates a detection path sequence for each robot within its assigned window. The path planning principle is "nearest detection first," meaning that within a given time window, the robot prioritizes undetected plants closest to its current location. The system employs a greedy strategy based on shortest path-first search, incorporating hotspot priority and obstacle avoidance logic. It outputs "detection path planning data" for each robot in each time window. Fields include the robot number, path point sequence, plant number, corresponding coordinates, task duration, and estimated travel distance. Based on the task connection between each robot across multiple time windows in the detection path planning data, the system calculates the time required for the robot to move between any two consecutive time windows. For example, if the distance between the last plant in window W20 and the first plant in window W21 for robot R4 is 240 meters, and the robot moves at a speed of 1.2 m / s, the time required is 200 seconds, or approximately 3.3 minutes. If this value exceeds 5 minutes, the system adds a "movement buffer" to the time sequence to ensure sufficient time for the robot to complete the scheduling switch. The resulting "detailed task sequence table" records the complete timeline for each robot from the start to the end of a task, including detection periods, buffer periods, and idle periods. Fields include time nodes, task types, target plants, and task status tags. To achieve state synchronization and avoid task conflicts during multi-robot collaborative operations, the system establishes fixed "information exchange nodes," establishing communication synchronization points every 15 minutes. At these points, all robots must upload key operational data such as their position, task progress, remaining energy status, and path deviation via their IoT communication modules, and receive status update commands from the scheduling platform. The communication node data table includes the node number, timestamp, robot number involved, synchronization field type (position information, energy consumption status, path deviation, etc.), and communication method (5G, LoRa, WiFi, etc.). By integrating all of this data, including the robot detection time windows, final assignment plan, detection path plan, task sequence table, and communication node data, the system automatically generates a complete "time-sequence collaborative operation plan." The plan includes each robot's ID, the time sequence of tasks to be performed (including detection, movement, buffering, and standby blocks), the ID of the corresponding target plant, a sequence of pathpoint coordinates, a list of communication synchronization node times, and inter-robot collaborative tasks (such as intersection avoidance). The entire plan is output as a structured JSON file that can be directly parsed and executed by the lower-level robot control module, ensuring conflict-free operation, state synchronization, and load balancing between robots.

[0034] Preferably, step S3 includes the following steps: Step S31: extracting the plant position coordinate sequence that each robot needs to visit based on the execution time sequence of each robot and the detection target allocation information in the time-series collaborative operation plan, and generating basic movement path data for each robot including the starting point, end point and intermediate nodes; This embodiment of the present invention extracts the coordinate information of all target plant points that each robot needs to visit based on the order in which each robot executes its tasks within different time windows in the "Time-Sequential Collaborative Operation Solution." For example, if robot R1 needs to sequentially inspect plants numbered P102, P107, and P111 within the 08:00-08:30 time window, and their corresponding coordinates are (5.2, 3.8), (6.5, 5.1), and (8.0, 4.4) meters, respectively, the system uses R1's current initial position (e.g., (3.0, 2.0)) as the path starting point, the coordinates of P102, P107, and P111 as intermediate nodes, and the coordinates of the last plant as the endpoint of the path segment. This path is recorded as "basic movement path data," which includes fields such as the robot number, starting point coordinates, sequence of intermediate nodes, endpoint coordinates, and task window number. This path data only reflects the spatial order and does not include dynamic factors such as speed, turning, and obstacle avoidance.

[0035] Step S32: Calculate the estimated movement time of each path segment based on the basic movement path data and the robot's speed parameters. When there is a node with a turning angle greater than 90 degrees in the path, add a deceleration time of 2 seconds to generate a detailed movement path plan. After acquiring basic path data, the system in this embodiment of the present invention further calculates the estimated travel time for each path segment, combining the robot's operating speed parameter (set to 1.2 m / s). The Euclidean distance between each two consecutive nodes is calculated by the coordinate difference, which is then divided by the movement speed to obtain the normal travel time. For example, if the distance between P102 and P107 is 1.9 meters, the normal travel time is approximately 1.58 seconds. The system also determines whether the path contains any significant corners, namely, whether the angle between three consecutive points is greater than 90 degrees. If so, such as if the angle between path points P102, P107, and P111 is 135 degrees, the system adds a 2-second "deceleration compensation time" to this corner. Ultimately, the system combines the estimated travel time, path nodes, and deceleration compensation information for each path segment into a "detailed path plan." This plan includes fields such as the start and end nodes of each segment, segment length, travel time, whether corner compensation is included, and total time, providing a time basis for airflow disturbance control.

[0036] Step S33: setting an airflow disturbance intensity parameter generated by the robot movement based on the health level according to the movement trajectory of each robot in the detailed movement path planning scheme and the plant health status classification information to form an airflow disturbance parameter; After obtaining the robot's detailed movement path, the system in this embodiment combines plant health classification information to set corresponding airflow disturbance intensity parameters for each plant location along the path. Specifically, the upper limit for wind speed disturbances tolerated by "healthy" plants is 0.5 m / s, the upper limit for "sub-healthy" plants is 0.3 m / s, and the upper limit for "unhealthy" plants is 0.2 m / s. The system calculates the potential airflow disturbance caused by the robot in each path segment based on the robot's speed and the closest distance between the path and the plant (for example, a high-impact segment is less than 0.5 m). The system then compares the disturbance with the health level tolerance. If the disturbance approaches or exceeds the tolerance threshold, the system marks the segment as a "high-risk disturbance segment" and adds a disturbance intensity label (e.g., 0.45 m / s). Output fields include the path segment number, robot number, adjacent plant number, plant level, estimated disturbance value, and disturbance risk level. The system ultimately generates an "airflow disturbance parameter table" for environmental compensation and path optimization.

[0037] Step S34: Control the task information transmission between robots according to the task allocation requirements in the time-series collaborative operation plan to obtain the collaborative operation execution status.

[0038] Based on a "timed collaborative operation scheme" and an "airflow disturbance parameter table," this embodiment of the present invention ensures information coordination and task status sharing among multiple robots within adjacent time windows. Therefore, the system leverages IoT communication protocols (such as MQTT or 5G URLLC) to establish a task information exchange mechanism between robots. Thirty seconds before the start of each time window, the scheduling system sends the execution sequence, path point data, and airflow risk segment warning indicators for each task to all robots about to initiate a task. Upon receiving these instructions, the robots establish a task execution queue using their local cache and return the initialization status to the master control platform. During operation, after completing each plant point inspection or path segment, the robot uploads its current status (e.g., "P102 inspection completed," "R1 passed high-disturbance segment 3") via a communication node. The master control platform records this real-time status data and synchronously updates the current collaborative task stage for all relevant robots. The resulting "collaborative operation execution status" table, which records the node number, execution timestamp, and information such as delays, disturbance anomalies, and task rescheduling for each robot's completed task within each time window, serves as a key intermediate result for the closed-loop control of the greenhouse multi-robot operation process.

[0039] It is particularly important that step S33 includes the following steps: Based on the movement trajectory coordinate data of each robot in the detailed movement path planning scheme and the plant position information in the plant health status data table, the plant area range passed by each robot's movement path is identified, and a robot trajectory area data table is generated; Set airflow disturbance intensity parameters for plant areas based on plant health status classification to form a health level disturbance parameter comparison table; According to the robot trajectory passing area data table and the health level disturbance parameter comparison table, the movement trajectory of each robot is associated and matched with the disturbance parameter value of the corresponding passing area to form the airflow disturbance parameter.

[0040] The autonomous robots deployed in greenhouse environments according to embodiments of the present invention are equipped with centimeter-level positioning systems, such as UWB and inertial navigation modules. The path planning module generates a complete trajectory for each robot within each mission time window, formatted as a timestamp and a two-dimensional or three-dimensional coordinate sequence (e.g., R1's trajectory in the 08:00-08:30 window contains a coordinate point every two seconds, resulting in a path sequence of [(5.0, 3.1), (5.4, 3.5), ..., (8.1, 4.5)]). The system also knows the fixed position coordinates of each plant in the greenhouse, which are maintained in a plant health status table by number and health status (e.g., P102's coordinates are (5.3, 3.3) and its health level is "medium"). The system uses a buffer zone determination method, defining a 0.8-meter impact buffer radius for each trajectory. Specifically, if the Euclidean distance between a trajectory point and a plant's coordinates is less than or equal to 0.8 meters, the robot is considered to have "passed through the plant area." The system traverses the spatial distance relationship between each robot trajectory and all plant coordinates, identifies the plant numbers within the influence area of ​​each robot path, and generates a "Robot Path Traversal Area Data Table." This table contains fields such as robot number, time window number, trajectory segment number, plant number within the influence area, plant coordinates, and plant health level. Based on the varying sensitivity of plants to airflow disturbances, system engineers have preset a correspondence between health status levels and maximum acceptable disturbance wind speeds. For example, plants with a "Good" health level typically have strong stomatal regulation and a high tolerance for airflow disturbances, with an upper limit of 0.5 m / s. Plants with a "Fair" health level, characterized by fragile leaves or moderate stomatal activity, have a lower upper limit of 0.3 m / s. Plants in "Poor" or "Critical" health states have stricter requirements, with a maximum disturbance limit of 0.2 m / s. Based on this, the system creates a "Health Level Disturbance Parameter Comparison Table," which records the disturbance threshold parameters corresponding to each health level. This table is not only used for the calculation of disturbance parameters in this step, but also serves as core reference data in subsequent greenhouse disturbance compensation, risk warning and other modules. The system receives the data table of the area through which the robot trajectory passes generated in the previous step, and uses each "plant number + health level" in the table as an index to retrieve the maximum acceptable disturbance wind speed corresponding to the plant from the "health level disturbance parameter comparison table". Subsequently, the system matches the robot path segment speed parameters in the trajectory information with the plant position, and calculates the estimated disturbance intensity generated by the robot on the adjacent plants when it runs on this path segment. The estimation model is based on empirical formulas, such as the robot's running speed, the shortest distance between the plant and the path, the greenhouse structure (whether there is an obstruction or ventilation device) and other parameters as input, combined with the wind tunnel experiment correction coefficient, to output an estimated disturbance value.For example, in one section, the robot was traveling at 1.2 m / s and only 0.4 m from a plant. The estimated path disturbance intensity for this section was approximately 0.35 m / s. If the upper limit allowed by the corresponding plant health level is 0.3 m / s, this would be recorded as a "high disturbance risk section." Ultimately, the system generates an "airflow disturbance parameter data table" with fields such as robot number, path segment number, plant number, health level, estimated disturbance value, allowable threshold, and disturbance level (normal / critical / exceeding limit). This provides a basis for subsequent airflow compensation and path readjustment.

[0041] Preferably, step S34 includes the following steps: Step S341: extracting task information that needs to be shared between robots based on the task allocation information of each robot in the time-series collaborative operation plan, and generating a robot task information requirement table, wherein the robot task information requirement table includes detection target allocation, time window arrangement, and path planning information; In the completed sequential collaborative operation scheme of the embodiment of the present invention, the system assigns each robot the detection target plant number, detection area, task start and end time, and detailed path planning data within a specific time window. In order to achieve collaborative operation control among multiple robots, the system first extracts task information that has spatial intersection or temporal overlap with other robots. For example, two robots are responsible for detection tasks in adjacent areas between 08:00 and 08:30, and there are intersections in the paths or shared areas of the detection objects. On this basis, the system generates a "robot task information requirement table" with fields including: robot number, time window number, target plant number, path coordinate sequence, description of the overlapping area with other robot tasks, whether it is a high-density area, etc. This information requirement table serves as the basis for the subsequent information transmission protocol, determining which information needs to be actively shared, who the sharing object is, and whether collaborative obstacle avoidance or rhythm adjustment is required.

[0042] Step S342: Establishing a communication protocol between robots based on the robot task information requirement table, setting the information transmission frequency to once every 5 minutes, and defining the transmission content format to include the robot number, current position, task progress percentage, and device status code, forming a task information transmission standard; To ensure the efficient and safe operation of multiple robots in a shared area, this embodiment of the present invention establishes a lightweight communication protocol based on the "shared task identifier" in the robot task information requirement table. This protocol utilizes low-power Wi-Fi communication and defines a fixed information structure template within the IoT communication management module. Each transmission includes four key fields: robot number (e.g., R3), current location (e.g., coordinates (6.2, 4.1)), task progress percentage (e.g., 72% completed), and device status code (e.g., 100 for normal operation, 201 for task failure, and 400 for communication loss). Each robot is set to transmit status data every five minutes, a frequency that balances communication bandwidth with the timeliness of status updates. The task information transmission standard also includes a list of target numbers that the receiving robot recognizes. This means that the receiving robot only receives information from robots that share tasks with its own, thus avoiding irrelevant broadcasts.

[0043] Step S343: Activate the information sending and receiving functions of each robot according to the task information transmission standard, control the robot to send status information to the system center at a set frequency, and generate real-time task information transmission records; In this embodiment of the present invention, each robot uniformly loads the task information transmission standard and initializes its communication module before the system issues a task. The system scheduler controls each robot to proactively transmit status information to the IoT system center at every 5-minute trigger (e.g., 08:00, 08:05, 08:10, etc.). This information is generated in real time by the robot based on its current navigation status and task execution status. For example, the current position is read from the navigation and positioning system (with an accuracy of less than 20 cm), the task progress is automatically calculated based on the percentage of completed plants, and a status code is assigned based on the self-test results. Upon receiving each robot's report, the system center automatically records the information in a "Real-time Task Information Transmission Log Table," indexed by the robot number and timestamp. This table allows the system to readily access the task execution status and spatial position status of all robots at any point in time.

[0044] Step S344: Monitor the task execution progress of each robot based on the real-time task information transmission record. When it is detected that the robot task completion rate is less than 80% of the expected level or the communication is interrupted for more than 20 minutes, it is marked as an abnormal state and a task execution abnormality event record is generated. In this embodiment of the present invention, a real-time monitoring rules engine is deployed in the dispatch center backend. This engine continuously reads the status data of all robots in various time windows from the "Real-time Task Information Transmission Log Table." The system compares each robot's task completion percentage every 10 minutes with the standard expected progress curve. For example, if all tasks are expected to be completed between 08:00 and 08:30, the robot will complete approximately 33% of its tasks every 10 minutes on average. If the robot's progress is only 50% at 08:20, falling below the expected lower limit of 80%, the system immediately marks it as a "progress anomaly." The system also monitors the status reporting interval. If no reports are received from a robot for 20 consecutive minutes (e.g., if the last recorded information is at 08:10 and there is still no update by 08:30), it is considered a "communication interruption anomaly." The system records these events in the "Task Execution Abnormal Event Record Table," which includes fields such as robot number, exception type, trigger time, and a list of uncompleted tasks, providing a basis for abnormal recovery and rescheduling.

[0045] Step S345: Calculate the working status, task completion status and communication connection status of each robot based on the task execution abnormal event record and the real-time task information transmission record to form the collaborative task execution status.

[0046] This embodiment of the present invention comprehensively analyzes the "Task Execution Abnormal Event Record Table" and the "Real-time Task Information Transmission Log Table" to compile statistics on each robot's current task execution time, the number of completed tasks, the number of failed tasks, the frequency of communication interactions with other robots, and the time of the most recent status update. Based on this data, the system assigns each robot a current "Collaborative Task Execution Status," which consists of three main dimensions: operating status (e.g., "Normal Operation," "Progress Delayed," "Task Failed"), communication status (e.g., "Normal Connection," "Delayed Update," "Communication Interrupted"), and collaboration status (e.g., "Collaboration Completing," "Awaiting Assistance," "Disconnected from Collaboration"). All execution status data is recorded in the "Collaborative Task Execution Status Table," providing comprehensive data support for subsequent steps such as path replanning, task reallocation, and energy consumption assessment.

[0047] Preferably, step S4 includes the following steps: Step S41: activating a pre-set airflow detection sensor network in the greenhouse according to the collaborative task execution state, collecting wind speed and direction data around the robot, and generating airflow change detection data for the robot movement; When the system identifies a collaborative task as "operating" or "recovering from an abnormal state," this embodiment of the present invention automatically activates a network of airflow detection sensors deployed within the greenhouse. This network consists of equally spaced ultrasonic wind speed and direction sensors (with a density of one per 10 square meters), with a measurement frequency of 1 Hz and an accuracy error of ±0.2 m / s. As the robot performs its task, each sensor node senses changes in the surrounding air flow in real time and transmits the collected wind speed (in meters per second) and wind direction (in polar coordinates) data to the greenhouse's central gateway via ZigBee or LoRa wireless protocols. The gateway then packages and uploads this data to the IoT platform's data center. The platform generates a "Robot Movement Airflow Change Detection Data Table" indexed by a timestamp based on each sensor's fixed coordinate location. The data fields include sensor ID, measurement time, wind speed, wind direction, and corresponding robot number, providing the raw data for subsequent analysis of robot-generated disturbances.

[0048] Step S42: identifying the airflow disturbance components generated by the robot movement based on the airflow change detection data and the airflow disturbance parameters, and generating an airflow disturbance exceeding standard event record; This embodiment of the present invention compares the "robot movement airflow change detection data" acquired in the previous step with the "airflow disturbance parameter table" previously determined by the path planning system. The disturbance parameter table records the maximum permissible disturbance wind speed values ​​for each robot within plant health zones of different levels (e.g., the upper limit for disturbance in a health level 1 zone is 0.8 m / s, while for a health level 2 zone, it is 1.0 m / s). During this comparison, the system determines whether the measurement point falls within the plant health zone corresponding to the robot's current path based on the spatial coordinates of the wind speed change. If the detected wind speed exceeds the upper limit, it is determined to have "exceeded disturbance." The system aggregates all exceeded-limit records into an "airflow disturbance exceeded-limit event record table." Each record includes the robot number, detection point coordinates, exceeded wind speed, corresponding plant health level, and trigger time. If the disturbance exceeds the limit for three consecutive periods (i.e., three consecutive seconds), the event is marked as "continuous disturbance," and subsequent movement strategy adjustments are made.

[0049] Step S43: adjusting the robot movement parameters according to the airflow disturbance exceeding standard event record, and generating an environmental interference compensation adjustment instruction; Upon receiving a disturbance exceeding the limit event, the system dispatch center in this embodiment of the present invention immediately reads the corresponding robot's current motion parameters (including speed, path turning angle, cruise mode, etc.) and makes compensatory adjustments based on the degree of disturbance. For example, if the current robot speed is 1.2 m / s, the system will adjust it down to 0.8 m / s in the event of sustained disturbance. If the path contains frequent sharp turns (e.g., continuous turns greater than 60 degrees), the system will adjust to a smoother path mode to reduce the magnitude of transient disturbances. The system generates an "environmental disturbance compensation adjustment command" containing the target robot number, the adjusted speed value, and the path correction method (e.g., angle smoothing coefficient). This command is then sent to the robot control terminal in real time via the IoT command channel, where the local controller executes the parameter update. This command triggers a status refresh of the robot's motion module and replans the current local path and speed in the next cycle.

[0050] Step S44: Acquire airflow detection data after the execution of the environmental interference compensation adjustment instruction, and perform spatial interpolation calculations based on the greenhouse spatial coordinate information to form continuous wind field distribution data covering the entire greenhouse; generate a greenhouse wind field distribution map containing wind speed contours and wind direction vectors based on the continuous wind field distribution data, and perform spatial discretization processing according to a 1 meter × 1 meter grid; After the adjustment instructions in this embodiment of the present invention are executed, the sensor network continues operating, and the newly collected airflow data is archived as "compensated airflow detection data." The system uses an inverse distance weighted (IDW) interpolation method to interpolate wind speed and direction for each 1-meter x 1-meter grid point in the greenhouse space based on the actual sensor coordinates and their measured wind speed values. This method weights the data based on sensor distance, ensuring that neighboring values ​​have a greater influence on the results, thereby generating a "continuous wind field distribution dataset." The system then converts this wind field data into a two-dimensional graphical wind field map, where wind speed is represented by color contours (e.g., 0-1 m / s green, 1-1.5 m / s yellow), and wind direction is displayed as a vector arrow, with azimuth angles measured clockwise from true north as 0 degrees. This "greenhouse wind field distribution map" is not only used to determine disturbance distribution but also provides a visual interface for greenhouse managers to intervene or further correct the coordinated path.

[0051] Step S45: Perform robot position correction and task dynamic allocation according to the greenhouse wind field distribution map to obtain a corrected robot collaborative operation configuration.

[0052] The embodiment of the present invention superimposes and analyzes the greenhouse wind field distribution map and the robot's current moving path to identify whether there are unfavorable conditions such as the robot's current position falling into an overlapping area of ​​wind speed disturbances, or the wind direction vector being opposite to the path direction. If the system detects that the robot has entered an area with strong disturbances, it immediately performs position offset correction based on the spatial grid position, instructing the robot to offset the current path by 0.5 to 1.0 meters to avoid sensitive areas; at the same time, it reads the task plan table. If the original task path is severely affected by the wind field and is within the adjustable range, the system re-evaluates the task priority and the robot resources that can be deployed, transfers the task to be executed to an idle robot in a more stable wind field area, and dynamically updates the "Robot Collaborative Operation Configuration Table". This table includes the updated robot number, task plant number, adjusted path coordinates, expected task time, etc., to ensure that the operation accuracy and detection coverage are maintained under controllable wind field disturbance conditions.

[0053] It is particularly important that step S45 includes the following steps: Step S451: Based on the wind field intensity information in the greenhouse wind field distribution map and the current position of the robot in the collaborative operation execution state, the deviation between the actual position of the robot and the planned path is identified, and a position correction instruction is generated when the deviation exceeds 0.5 meters; In an embodiment of the present invention, a real-time positioning module (such as a LiDAR SLAM or an ultra-wideband positioning system (UWB)) continuously acquires the current position coordinates of each robot, compares them with the target coordinates at the corresponding time point in the predetermined path planning, and calculates the position deviation distance. Combined with the greenhouse wind field distribution map, the system focuses on areas with significant wind field disturbances, as position deviations can easily exacerbate task errors. A deviation threshold is set at 0.5 meters. When the deviation between the robot's actual position and the corresponding point on the predetermined path exceeds this threshold, the system automatically generates a "position correction instruction," which includes the target correction coordinates, a correction time window, and a speed adjustment recommendation (such as slowing down to 0.6 m / s for precise adjustment). The instruction is sent to the corresponding robot in real time via IoT communication. The robot's local controller initiates a path correction program to adjust the current trajectory, reduce position errors, and ensure detection accuracy.

[0054] Step S452: Based on the airflow disturbance intensity distribution in the greenhouse wind field distribution map, the detection suitability of each area is evaluated. When the wind field intensity in a region exceeds the safety threshold, the task of the region is reallocated to other robots, forming a dynamic task reallocation plan; This embodiment of the present invention extracts wind speed intensity data from a wind field distribution map, divides the greenhouse space into a 1-meter-by-1-meter grid, compares the wind speed at each grid point with a set safety threshold (e.g., 1.2 meters per second), and marks areas where wind speed exceeds the standard. For planned robot inspection tasks, the system checks whether the location of each target plant falls within the exceeding-standard area. If so, the inspection suitability of the task area is considered reduced. Based on each robot's current load and position, the system invokes a task scheduling algorithm (based on a greedy algorithm or reinforcement learning) to reallocate tasks in areas with lower wind field intensity to nearby idle or lightly loaded robots, forming a "dynamic task reallocation scheme." This scheme specifies the new responsible robot number and task time window for each affected area, ensuring that other robots avoid high-disturbance areas and complete inspections, reducing the risk of environmental disturbances and improving overall inspection effectiveness.

[0055] Step S453: Based on the position correction instruction and the task dynamic reallocation plan, the target position, movement path and detection task list of each robot are updated to form a corrected robot collaborative operation configuration.

[0056] The embodiment of the present invention integrates all received position correction instructions and task dynamic reallocation schemes, combines the real-time status data of the robot, and uses the path replanning module to generate a new movement path and detection plan. This module calculates the optimized path based on the latest target position and task allocation to avoid duplicate coverage and path conflicts. At the same time, it takes into account the robot's kinematic limitations and the impact of the environmental wind field to generate a smooth path that meets the needs of the dynamic environment. The system updates the detection task list of each robot, adjusts the task execution order and time nodes, and forms a "corrected robot collaborative work configuration table", which includes the robot ID, the updated path coordinate sequence, the task plant list, the estimated completion time, etc. The updated configuration is synchronized to each robot in real time through the Internet of Things platform to ensure that it can continuously and efficiently complete the multi-point plant detection tasks in the greenhouse in the corrected state.

[0057] The present invention also provides a multi-autonomous robot collaborative operation control system based on the Internet of Things, which is used to execute the multi-autonomous robot collaborative operation control method based on the Internet of Things. The multi-autonomous robot collaborative operation control system based on the Internet of Things includes: The plant health monitoring module is used to obtain micro-wind speed detection data from multiple leaf surfaces of plants in a greenhouse environment; based on the micro-wind speed detection data from the leaf surfaces, the characteristics of changes in the opening and closing states of the plant stomata are analyzed to determine the plant health status classification; The collaborative task planning module is used to detect the temporal distribution of plant stomatal activity according to plant health status; determine the detection time window data of the autonomous robot based on the temporal distribution of plant stomatal activity; and determine the time-series collaborative operation plan based on the temporal distribution of plant stomatal activity and the detection time window data; The multi-robot control module is used to plan the movement paths of multiple robots based on the time-sequential collaborative operation scheme and determine the airflow disturbance parameters generated by the robot movements according to the plant health status classification. It also controls the task information transmission between robots based on the task allocation requirements in the time-sequential collaborative operation scheme to obtain the collaborative operation execution status. The adaptive adjustment module is used to detect changes in greenhouse airflow during robot movement based on the collaborative work execution status, and to compensate for environmental interference in combination with airflow disturbance parameters to obtain a greenhouse wind field distribution map. The robot position is corrected and tasks are dynamically allocated based on the greenhouse wind field distribution map to obtain the corrected robot collaborative work configuration.

[0058] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0059] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A multi-autonomous robot collaborative operation control method based on the Internet of Things, characterized in that: The following steps are involved: Step S1: Acquire micro-wind speed detection data on multiple leaf surfaces of plants in a greenhouse environment; analyze the characteristics of plant stomatal opening and closing state changes based on the micro-wind speed detection data on the leaf surfaces to determine the plant health status classification; Step S2: Detecting the time distribution pattern of plant stomatal activity according to plant health status classification; determining detection time window data of the autonomous robot based on the time distribution pattern of plant stomatal activity; and determining a time sequence collaborative operation plan based on the time distribution pattern of plant stomatal activity and the detection time window data; Step S3: Based on the sequential collaborative operation scheme, the movement paths of multiple robots are planned, and the airflow disturbance parameters generated by the robot movements are determined according to the plant health status classification. The task information transmission between robots is controlled according to the task allocation requirements in the sequential collaborative operation scheme to obtain the collaborative operation execution status. Step S4: Based on the collaborative work execution status, the greenhouse space airflow change detection during the robot movement is performed, and the environmental interference compensation is performed in combination with the airflow disturbance parameters to obtain the greenhouse wind field distribution map; the robot position is corrected and the task is dynamically allocated according to the greenhouse wind field distribution map to obtain the corrected robot collaborative work configuration.

2. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting data on real-time voltage signals from a plurality of micro-wind speed sensors distributed near the surface of plant leaves in the greenhouse, and spatially marking the voltage signals according to the installation position coordinates of the micro-wind speed sensors to obtain micro-wind sensing voltage matrix data; Step S12: sorting the breeze sensing voltage matrix data by acquisition timestamp to obtain multi-point blade surface breeze speed detection data; Step S13: performing sensor calibration conversion processing on the micro-wind speed detection data of multiple blade surfaces, and performing spatial interpolation calculation based on the spatial distribution of the blades to obtain micro-wind speed field data with continuous distribution in the greenhouse space; Step S14: performing spatial variation characteristic analysis on the micro-wind speed field data, identifying background wind speed components, and performing component separation on the micro-wind speed field data based on the background wind speed components to obtain plant transpiration net wind speed change data; Step S15: Analyze the characteristics of plant stomatal opening and closing state changes based on the plant transpiration net wind speed change data to determine the plant health status classification.

3. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 2 is characterized in that: Step S15 includes the following steps: Step S151: Calculate the time change rate and change amplitude of the wind speed at each monitoring point for the plant transpiration net wind speed change data, and calculate the periodic fluctuation pattern and mutation point to form the wind speed time change characteristics; Step S152: determining the wind speed increment threshold for the stomatal open state and the wind speed decrement threshold for the stomatal closed state based on pre-acquired basic knowledge of plant physiology and experimental data, and establishing a corresponding relationship table between the wind speed variation amplitude and the stomatal opening and closing degree; Step S153: comparing the wind speed time variation characteristics with the thresholds in the switch correspondence table one by one, and recording the stomata opening and closing status of each monitoring point at different times; Step S154: Counting the stomatal opening and closing state change frequency and stomatal opening and closing duration of each monitoring point, and calculating the proportion of the opening time to the total observation time to obtain stomatal opening and closing state change data; Step S155: Calculate the overall activity and response speed of plant stomatal activity based on the stomatal opening and closing state change data, and match them with the preset health standards to obtain the plant health status classification.

4. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 3 is characterized in that: In step S2, the stomatal activity time distribution pattern of the plant is detected according to the plant health status classification, including: Establishing a plant health status data table including plant number, health grade and location coordinates according to plant health status classification; Based on the health level information in the plant health status data table, the stomatal opening and closing activities of plants at each health level are continuously monitored within 24 hours, and the stomatal opening and closing time data of each monitoring point are recorded; The stomatal opening duration of each monitoring point is calculated based on the stomatal opening and closing time data. When the duration exceeds 15 minutes, it is marked as a valid opening event, and a valid opening event time series is generated. Based on the time series of effective opening events, the event frequency distribution data of stomatal activity of plants of each health level in four time periods of 0-6 o'clock, 6-12 o'clock, 12-18 o'clock, and 18-24 o'clock were statistically analyzed; The activity intensity of plants at each health level in each time period is calculated based on the event frequency distribution data. When the activity intensity coefficient is greater than the preset threshold, it is determined as the active time period data of the level; The temporal distribution pattern of plant stomatal activity is generated based on the active time period data of each health level.

5. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 4 is characterized in that: In step S2, the detection time window data of the autonomous robot is determined according to the temporal distribution law of plant stomatal activity, including: According to the time distribution of plant stomatal activity, the start and end time of each active time period are extracted, and an active time period list table is established; sorting the active time periods of each health level in chronological order based on the time information in the active time period list table, identifying the active time periods that overlap between plants of different health levels, and generating time period overlap relationship data; Based on the overlapping relationship data of time periods, the number of active plant health levels in the same time period is counted. When the number of active levels exceeds 2 at the same time, it is marked as a high-density detection period, and a detection density distribution table is formed; Based on the high-density detection period information in the detection density distribution table, each high-density period is divided into 30-minute sub-periods. For non-high-density periods, the original time period length is maintained to generate an initial time window division scheme; According to the duration of each time window in the initial time window division scheme and the plant distribution position in the plant health status data table, the number of plants that need to be inspected in each time window and the spatial distribution range are calculated to obtain the plant inspection load distribution data; Based on the plant inspection load distribution data, the inspection task workload in each time window is calculated according to the parameters of a robot movement speed of 1.2m / s and a single plant inspection time of 3 minutes; Based on the workload of the inspection task and the maximum working time of a single robot of 60 minutes, the number of robots required for each time window is determined, and a robot demand allocation table is generated; Autonomous robot inspection time window data including window number, start time, end time, target plant health level, number of inspected plants, and number of assigned robots are generated based on the robot demand allocation table.

6. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 5 is characterized in that: The timing collaborative operation scheme determined in step S2 based on the temporal distribution pattern of plant stomatal activity and the detection time window data includes: According to the window number and time parameters in the detection time window data and the activity intensity information of each health level in the plant stomatal activity time distribution law, the time window is assigned a detection priority weight to obtain a time window priority allocation table; Pre-allocate robot resources for each time window based on the time window priority allocation table and the robot demand allocation table to form a robot resource pre-allocation plan; According to the robot allocation status of each time window in the robot resource pre-allocation plan, the robot resource conflicts between adjacent time windows are identified. When the same robot is allocated to consecutive windows with a time interval of less than 10 minutes, it is marked as a resource conflict and a resource conflict identification table is generated; The robots are reallocated for the conflicting time windows in the resource conflict identification table, the robot requirements of the high-priority windows are maintained, and the conflicting robots are allocated to the secondary priority windows or the execution time is delayed to obtain the robot allocation plan. According to the robot allocation plan and the position coordinate information in the plant health status data table, a specific inspection path sequence is planned for each robot. The plant inspection sequence within the same time window is arranged based on the proximity principle, and the robot inspection path planning data is generated. Based on the robot detection path planning data, the movement time and standby time of each robot between different time windows are calculated. When the movement time exceeds 5 minutes, the movement buffer time is added to the task sequence to form a detailed task sequence table; Set the information exchange nodes between robots according to the detailed task schedule, set a status synchronization time point every 15 minutes, and obtain the robot collaborative communication node data; Based on the robot collaborative communication node data, a time-series collaborative operation plan is generated, which includes robot numbering, execution time sequence, detection target allocation, movement path arrangement and coordinated communication plan.

7. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 6 is characterized in that: Step S3 includes the following steps: Step S31: extracting the plant position coordinate sequence that each robot needs to visit based on the execution time sequence of each robot and the detection target allocation information in the time-series collaborative operation plan, and generating basic movement path data for each robot including the starting point, end point and intermediate nodes; Step S32: Calculate the estimated movement time of each path segment based on the basic movement path data and the robot's speed parameters. When there is a node with a turning angle greater than 90 degrees in the path, add a deceleration time of 2 seconds to generate a detailed movement path plan. Step S33: setting an airflow disturbance intensity parameter generated by the robot movement based on the health level according to the movement trajectory of each robot in the detailed movement path planning scheme and the plant health status classification information to form an airflow disturbance parameter; Step S34: Control the task information transmission between robots according to the task allocation requirements in the time-series collaborative operation plan to obtain the collaborative operation execution status.

8. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 7 is characterized in that: Step S34 includes the following steps: Step S341: extracting task information that needs to be shared between robots based on the task allocation information of each robot in the time-series collaborative operation plan, and generating a robot task information requirement table, wherein the robot task information requirement table includes detection target allocation, time window arrangement, and path planning information; Step S342: Establishing a communication protocol between robots based on the robot task information requirement table, setting the information transmission frequency to once every 5 minutes, and defining the transmission content format to include the robot number, current position, task progress percentage, and device status code, forming a task information transmission standard; Step S343: Activate the information sending and receiving functions of each robot according to the task information transmission standard, control the robot to send status information to the system center at a set frequency, and generate real-time task information transmission records; Step S344: Monitor the task execution progress of each robot based on the real-time task information transmission record. When it is detected that the robot task completion rate is less than 80% of the expected level or the communication is interrupted for more than 20 minutes, it is marked as an abnormal state and a task execution abnormality event record is generated. Step S345: Calculate the working status, task completion status and communication connection status of each robot based on the task execution abnormal event record and the real-time task information transmission record to form the collaborative task execution status.

9. The multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 8 is characterized in that: Step S4 includes the following steps: Step S41: activating a pre-set airflow detection sensor network in the greenhouse according to the collaborative task execution state, collecting wind speed and direction data around the robot, and generating airflow change detection data for the robot movement; Step S42: identifying the airflow disturbance components generated by the robot movement based on the airflow change detection data and the airflow disturbance parameters, and generating an airflow disturbance exceeding standard event record; Step S43: adjusting the robot movement parameters according to the airflow disturbance exceeding standard event record, and generating an environmental interference compensation adjustment instruction; Step S44: Acquire airflow detection data after the execution of the environmental interference compensation adjustment instruction, and perform spatial interpolation calculations based on the greenhouse spatial coordinate information to form continuous wind field distribution data covering the entire greenhouse; generate a greenhouse wind field distribution map containing wind speed contours and wind direction vectors based on the continuous wind field distribution data, and perform spatial discretization processing according to a 1 meter × 1 meter grid; Step S45: Perform robot position correction and task dynamic allocation according to the greenhouse wind field distribution map to obtain a corrected robot collaborative operation configuration.

10. A multi-autonomous robot collaborative operation control system based on the Internet of Things, characterized in that: Used to execute the multi-autonomous robot collaborative operation control method based on the Internet of Things according to claim 1, the multi-autonomous robot collaborative operation control system based on the Internet of Things includes: The plant health monitoring module is used to obtain micro-wind speed detection data from multiple leaf surfaces of plants in a greenhouse environment; based on the micro-wind speed detection data from the leaf surfaces, the characteristics of changes in the opening and closing states of the plant stomata are analyzed to determine the plant health status classification; The collaborative task planning module is used to detect the temporal distribution of plant stomatal activity according to plant health status; determine the detection time window data of the autonomous robot based on the temporal distribution of plant stomatal activity; and determine the time-series collaborative operation plan based on the temporal distribution of plant stomatal activity and the detection time window data; The multi-robot control module is used to plan the movement paths of multiple robots based on the time-sequential collaborative operation scheme and determine the airflow disturbance parameters generated by the robot movements according to the plant health status classification. It also controls the task information transmission between robots based on the task allocation requirements in the time-sequential collaborative operation scheme to obtain the collaborative operation execution status. The adaptive adjustment module is used to detect changes in greenhouse airflow during robot movement based on the collaborative work execution status, and to compensate for environmental interference in combination with airflow disturbance parameters to obtain a greenhouse wind field distribution map. The robot position is corrected and tasks are dynamically allocated based on the greenhouse wind field distribution map to obtain the corrected robot collaborative work configuration.

Images