An agricultural variable operation control system and method based on edge computing

Abstract

The application provides an agricultural variable operation control system and method based on edge computing. In a pre-seeding stage, an edge computing unit generates a pre-prescription model P_P based on sampled soil information S_D and a soil model M. In a seeding stage, a soil real-time information monitoring unit keeps sampling the soil of an operation plot in real time, and the edge computing unit forms an actual seeding prescription P based on the pre-prescription model P_P and the real-time sampled soil information. In a crop harvesting stage, the edge computing unit receives crop yield information and latitude and longitude information of the operation plot sent by a crop yield measuring device and a satellite positioning device to generate yield-position information H_W, and imports the yield-position information H_W into a model updating and management center to correct the soil model M of the operation plot. Based on the joint regulation idea of edge learning, the application realizes efficient interactive regulation, continuously collects and processes yield information to improve model accuracy, optimizes variable seeding operation efficiency, and improves crop yield.

Description

Technical Field

[0001] This invention belongs to the field of intelligent agricultural equipment, specifically relating to an agricultural variable operation control system and method based on edge computing. Background Technology

[0002] Variable-rate cropping is a planting technique that maximizes agricultural efficiency. It includes prescription-based variable-rate cropping and sensor-based variable-rate cropping. The difference lies in that the former generates the prescription in advance, while the latter generates it during operation using sensor technology. Typically, a variable-rate cropping system consists of sensors, a controller, and actuators. Different types of sensors collect different types of data. The collected information is aggregated via a bus to the controller, which parses the data, converts it into calibrated parameters, and sends them to the actuators for execution. Taking seeding control as an example, traditional seeding control systems individually detect soil pressure, soil moisture, or soil nutrients, sending a single parameter to the controller, which then makes a decision to control the downward pressure of the seedlings. This method is simple and easy to implement. However, to more efficiently control the specific adjustment parameters of variable-rate cropping, it is necessary to comprehensively consider multiple factors related to the interaction between the soil and the machinery. The microcontrollers commonly used as controllers do not have the ability to perform large-scale data reasoning and computation. Simply combining the outputs of multiple controllers not only involves complex hardware interfaces but also presents a significant challenge in ensuring that the final crop growth results meet expectations. Edge computing is an emerging control concept in the field of artificial intelligence. Since traditional controllers do not have high computing power, high-performance computing hardware platforms are placed at the "edge" of the control system to perform a large amount of computing without occupying the controller's computing power, and can improve the overall intelligence level of the control system. Summary of the Invention

[0003] To address the aforementioned problems, a variable-operation control system and method based on edge computing was developed. The system uses a microcontroller as the mechanism controller, supplemented by a dedicated edge computing unit for data inference and computation to achieve comprehensive data inference and computation based on time-series data. A cloud model data center and model calibration algorithm are also introduced. The system can infer and analyze the interaction relationship between soil, machinery, and plants based on multi-sensor data and actual sowing results, and output the optimal sowing control strategy to the mechanism controller.

[0004] On the one hand, this invention proposes a variable operation control system based on edge computing, including: a model update and management center, a real-time soil information monitoring unit, a satellite positioning device, a crop yield measurement device, an edge computing unit, an operation controller, a terminal control node, and a terminal execution mechanism;

[0005] The model update and management center transmits information with the real-time soil information monitoring unit, crop yield measurement equipment, and edge computing unit; it is used to send the soil information model of the corresponding plot to the edge computing unit according to the location of the work plot, and update the soil information model of the plot according to the yield information of the plot fed back by the crop yield measurement equipment.

[0006] The real-time soil information monitoring unit is used to measure the soil characteristics of the work site and send the measured real-time soil information to the edge computing unit.

[0007] The satellite positioning device is connected to the edge computing unit and is used to determine the location of the work site;

[0008] Crop yield measurement equipment is used to measure the crop yield of the work plot and send the measured crop yield information of the work plot to the edge computing unit;

[0009] The edge computing unit is used to process the information transmitted by each unit and transmit the processed information to the job controller.

[0010] The job controller connects the edge computing unit and each terminal control node, receives information from the edge computing unit, and generates control information to send to the corresponding terminal control node.

[0011] Multiple terminal control nodes, each corresponding to an independent terminal execution mechanism, wherein the terminal control node receives control information from the job controller and sends it to the terminal execution mechanism;

[0012] The terminal actuator performs operations based on the received execution information.

[0013] Preferably, the control system performs operation control according to the pre-sowing stage, the sowing stage, and the crop harvesting stage;

[0014] In the pre-sowing stage, the model update and management center sends the soil model M of the plot to be operated to the edge computing unit; the real-time soil information monitoring unit samples the soil of the plot to be operated and transmits the sampled soil information S_D to the edge computing unit, and the edge computing unit generates a pre-prescription model P_P based on the sampled soil information S_D and the soil model M.

[0015] During the sowing stage, the real-time soil information monitoring unit keeps sampling the soil of the work site in real time and sends the real-time sampled soil information to the edge computing unit. The edge computing unit forms the actual sowing prescription P based on the pre-prescription model P_P and the real-time sampled soil information, and sends it to the work controller to control the terminal execution mechanism to execute it.

[0016] During the crop harvesting stage, the edge computing unit receives crop yield information and latitude and longitude information of the work plot sent by the crop yield measurement equipment and satellite positioning equipment, binds them to generate yield-location information H_W, and imports the yield-location information H_W into the model update and management center for correction of the soil model M of the work plot.

[0017] Preferably, in the pre-sowing stage, the real-time soil information monitoring unit performs multiple random samplings of the plot to be operated on, and feeds back the collected soil information S_D of the plot to be operated on to the edge computing unit in vector form; the edge computing unit compares the difference between the collected soil information S_D and the received soil model M, generates a one-stage fitting function η(x) for the soil model M, and the edge computing unit combines the soil model M and the fitting function η(x) to calculate the pre-prescription model P_P of the plot to be operated on.

[0018] Preferably, the pre-prescription model P_P includes sowing decision schemes under different latitude and longitude coordinates within the plot to be operated.

[0019] Preferably, the one-stage fitting function η(x) for the soil model M is generated in a linear, piecewise, or polynomial manner.

[0020] Preferably, during the sowing stage, the edge computing unit uses the pre-prescription model P_P to calculate the pre-regulation amount C_P for each terminal execution structure. The real-time soil information monitoring unit acquires the soil information S_R at the work point in real time during sowing, and binds the soil information S_R and the current work point location information W to form S_R-W information. The S_R-W information is transmitted to the edge computing unit every sampling interval T. The edge computing unit determines whether a complete soil information sampling process has been completed based on whether the sampling interval has been reached. It performs a second-stage fitting based on the soil information S_R and the soil model after the first-stage fitting to obtain the soil model weight φ. The model weight φ and the pre-regulation amount C_P are used to comprehensively calculate the actual operation regulation amount C, and the actual regulation amount C is transmitted to the operation controller to control the operation of the terminal execution mechanism.

[0021] Preferably, if the edge computing unit determines that the sampling interval T has not been reached, it performs a second-stage fitting using the soil information S_R(T-1) at sampling time T-1 and the soil model after a first-stage fitting to obtain the soil model weight φ; if the edge computing unit determines that the sampling interval T has been reached, it performs a second-stage fitting using the soil information S_R(T) at time T and the soil model after a first-stage fitting to obtain the soil model weight φ.

[0022] Preferably, the weight φ is set according to the sampling frequency of the real-time soil information monitoring unit used. The higher the sampling frequency, the higher the weight φ is attached to the real-time sampled soil information S_R.

[0023] Preferably, during the crop harvesting stage, the edge computing unit generates yield-location information H_W based on the yield data H of the work plot measured in real time by the yield measuring equipment and the latitude and longitude data of the work plot obtained by the satellite equipment. After the crop harvest is completed, the edge computing unit sends the yield-location information H_W to the model update and management center. The model update and management center generates estimated yield information P_H based on the actual prescription P, compares the estimated yield information P_H with the yield-location information H_W, calculates the correction function τ(x) of the corrected soil model M, and sends the corrected soil model M_C before the next sowing operation begins.

[0024] On the other hand, this invention proposes a variable-operation control method based on edge computing, which is used to control the aforementioned variable-operation control system based on edge computing.

[0025] In the pre-sowing stage, the model update and management center sends the soil model M of the plot to be operated to the edge computing unit; the real-time soil information monitoring unit samples the soil of the plot to be operated and transmits the sampled soil information S_D to the edge computing unit, and the edge computing unit generates a pre-prescription model P_P based on the sampled soil information S_D and the soil model M.

[0026] During the sowing stage, the real-time soil information monitoring unit keeps sampling the soil of the work site in real time and sends the real-time sampled soil information to the edge computing unit. The edge computing unit forms the actual sowing prescription P based on the pre-prescription model P_P and the real-time sampled soil information, and sends it to the work controller to control the terminal execution mechanism to execute it.

[0027] During the crop harvesting stage, the edge computing unit receives crop yield information and latitude and longitude information of the work plot sent by the crop yield measurement equipment and satellite positioning equipment, binds them to generate yield-location information H_W, and imports the yield-location information H_W into the model update and management center for correction of the soil model M of the work plot.

[0028] The beneficial effects of this invention are as follows: Traditional variable seeding operations often adjust working components based on a single data parameter, neglecting the interaction between various adjustment parameters and rarely considering the temporal optimization of the model used. This invention employs a joint control approach based on edge learning. By introducing a variable seeding model set, a multi-layer control network, a computationally intensive edge computing unit, and a model update management center for storing soil models from different regions, it can achieve efficient interactive adjustment while continuously improving the model's accuracy through the collection and processing of yield information, thereby optimizing the efficiency of variable seeding operations and increasing crop yield. Attached Figure Description

[0029] Figure 1 This is an overall block diagram of a variable operation control system based on edge computing.

[0030] Figure 2 The overall flow of the variable operation control system based on edge computing;

[0031] Figure 3 This is a schematic diagram of the sampling process;

[0032] Figure 4 The flowchart shows the pre-cast correction algorithm for a variable operation control system based on edge computing.

[0033] Figure 5 The flowchart shows the seeding operation control method of the variable operation control system based on edge computing.

[0034] Figure 6 The flowchart shows the post-sowing yield correction algorithm for a variable operation control system based on edge computing.

[0035] Figure 7 This is a model set and iterative optimization method for variable operations. Detailed Implementation

[0036] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0037] like Figure 1 As shown, a variable operation control system based on edge computing includes a model update and management center, a real-time soil information monitoring unit, a satellite positioning device, a crop yield measurement device, an edge computing unit, an operation controller, a terminal control node, and a terminal execution mechanism.

[0038] The model update and management center transmits information with the real-time soil information monitoring unit, crop yield measurement equipment, and edge computing unit; it is used to send the soil information model of the corresponding work plot to the edge computing unit according to the location of the work plot, and update the soil information model of the work plot according to the yield information of the work plot fed back by the crop yield measurement equipment; in this embodiment of the invention, information transmission is completed through the network, and to ensure the speed and real-time transmission of information, a 4G network or above is preferred.

[0039] The real-time soil information monitoring unit is used to measure the soil characteristics of the work site and transmit the measured real-time soil information to the edge computing unit. The soil characteristics measured by the real-time soil information monitoring unit include soil nitrogen content, soil moisture content, organic matter content, organic carbon content, and pH value. The real-time soil information monitoring unit can be handheld or airborne, and both can independently measure the above-mentioned soil characteristics.

[0040] The satellite positioning device is connected to the edge computing unit to determine the location of the work site. The satellite positioning device includes a GPS receiver and a satellite antenna. In a preferred embodiment of the invention, the satellite positioning device is connected to the edge computing unit via a CAN bus.

[0041] Crop yield measurement equipment is used to measure the crop yield of the work plot and send the measured crop yield information to the edge computing unit.

[0042] The edge computing unit is used to process the information transmitted from each unit and then transmit the processed information to the job controller.

[0043] The job controller connects the edge computing unit and each terminal control node, receives information from the edge computing unit, and generates control information to send to the corresponding terminal control node. In an embodiment of the invention, the job controller is connected to the edge computing unit via a high-speed USB bus and to each terminal control node via a CAN bus.

[0044] Multiple terminal control nodes are used, each corresponding to an independent terminal actuator. Each terminal control node receives control information from the job controller and sends it to the terminal actuator. In this embodiment, the terminal control nodes and the terminal actuators are connected via a CAN bus.

[0045] The terminal actuator performs operations based on the received execution information. The terminal actuator includes a seed dispensing device, a fertilizer dispensing device, a depth limiting device, and a compaction device.

[0046] Figure 2The diagram shown is an overall flowchart of a variable operation control method for a variable operation control system based on edge computing according to an embodiment of the present invention. The entire process can be divided into a pre-sowing stage, a sowing stage, and a crop harvesting stage.

[0047] In the pre-planting stage, the model update and management center sends the soil model M of the plot to be planted to the edge computing unit; the real-time soil information monitoring unit samples the soil of the plot to be planted and transmits the sampled soil information S_D to the edge computing unit. The sampling process is as follows: Figure 3 As shown, the edge computing unit generates a pre-prescription model P_P based on the sampled soil information S_D and the soil model M.

[0048] Figure 4 The method for forming a pre-prescription model before sowing is specifically illustrated. The sowing machine arrives at the plot to be sown, and satellite positioning equipment acquires the plot's location signal and uploads this location information to the model update and management center. The model update and management center queries the local server for the soil model M for that plot and sends it to the edge computing unit. Simultaneously, test personnel, carrying a handheld real-time soil information monitoring unit, conduct five random samplings of the plot and feed back the collected soil information S_D to the edge computing unit in vector form. The edge computing unit compares the differences between the collected soil information S_D and the received model M, fitting the model using linear, piecewise, or polynomial methods, and calculates the first-stage fitting function η(x). The edge computing unit combines the model M and the fitting function η(x) to calculate the pre-prescription model P_P. The pre-prescription model includes sowing decision schemes calculated using the yield model at different latitude and longitude coordinates within the plot, including specific optimization decision schemes for each terminal control node under the influence of these sowing decision schemes.

[0049] During the sowing stage, the real-time soil information monitoring unit continuously samples the soil of the work area and sends the sampled soil information to the edge computing unit. The edge computing unit generates the actual sowing prescription P based on the pre-prescription model P_P and the sampled soil information, and sends it to the work controller to control the terminal execution mechanism to execute it. In this embodiment of the invention, the real-time soil information monitoring unit during the sowing stage is an onboard unit, installed at the front end of the seeder.

[0050] Figure 5The diagram illustrates the operation control method during the sowing stage. The edge computing unit uses the pre-prescription model P_P formed before sowing to calculate the pre-control quantities C_P for each terminal execution structure, including the sowing amount S_P, fertilizer application amount F_P, depth wheel position D_P, downforce N_P, and compaction pressure R_P. An onboard real-time soil information monitoring unit is placed in the soil in front of the sowing implement and is propelled forward by the implement in its working direction. It acquires the soil information S_R at its current working point and binds this soil information S_R with the current position information W. This information is then transmitted to the edge computing unit via the bus at every sampling interval T. The edge computing unit determines whether a complete soil information sampling process has been completed and controls the sowing operation according to the unupdated and updated stages. In this embodiment, if the sampling interval T has not been reached, the soil information S_R(T-1) at sampling time (T-1) and the model M-1 after one-stage fitting are used for a two-stage fitting, with a weight of φ. Conversely, if the sampling interval T has been reached, the edge computing unit performs a two-stage fitting of the soil information S_R(T) at time T and the model M-1 after one-stage fitting, with a weight of φ. The setting of the weight φ depends on the sampling frequency of the airborne real-time soil information monitoring unit used. The higher the sampling frequency, the higher the weight φ is attached to the real-time sampled soil information S_R(T). The actual working control quantity C is calculated using the model weight φ and the pre-control quantity C_P, including seeding amount S, fertilizer application amount F, depth limit wheel position D, downforce N, and compaction pressure R. The edge computing unit transmits the real-time working control quantity C to the operation controller via a high-speed USB bus. After parsing the signal, the operation controller transmits each data of the real-time control quantity to each terminal control node via the CAN bus. Each terminal control node receives the control signal and controls the terminal actuator.

[0051] During the crop harvesting stage, the edge computing unit receives crop yield information and latitude and longitude information of the work plot sent by the crop yield measurement equipment and satellite positioning equipment, binds them to generate yield-location information H_W, and imports the yield-location information H_W into the model update and management center for correction of the soil model M of the work plot.

[0052] Figure 6This illustrates a post-sowing yield correction method, which modifies the field prescription using data from the crop harvest stage. During harvest, the machine continuously collects yield data H from the field and uses an edge computing unit to combine this data with the latitude and longitude data of the location to generate yield-location information H_W. After harvest, the yield-location information H_W is sent to the model update and management center. The model update and management center generates estimated yield information P_H based on the previous actual prescription P, compares the estimated yield information P_H with the yield-location information H_W, and calculates a correction function τ(x). This correction function τ(x) is used to correct the soil model M. Before the next sowing operation, the model update and management center will transmit the corrected soil information M_C.

[0053] Figure 7 The diagram illustrates the model iterative optimization method of this invention. The model set includes: plot soil information model M, pre-sowing random sampled soil information S_D, pre-prescription model P_P, pre-regulation amount C_P, real-time soil information S_R(T), actual working prescription model P, actual working regulation amount C, yield model H, actual yield H_R, corrected regional soil information model M_C, prescription generation network GPN, regulation parameter generation network GCN, yield prediction model GHM, one-stage fitting function η(x), two-stage fitting weight φ, and post-sowing yield model correction function τ(x). The plot soil information model M contains soil characteristic information including latitude and longitude location information, mainly including soil nitrogen content, soil moisture content, organic matter content, organic carbon content, and pH value. The data types of the pre-sowing random sampled soil information S_D, the real-time soil information S_R(T), and the soil information model M are the same, but the specific values ​​are different. The pre-prescription model P_P is obtained by inputting the plot soil information model M and the pre-sowing random sampled soil information S_D into the prescription generation network GPN. The pre-regulation quantity C_P is obtained by inputting the pre-prescription P_P into the regulation parameter generation network GCN. The actual working regulation quantity is obtained by comprehensively calculating the pre-regulation quantity and the weight φ. In the preferred embodiment of the present invention, a data table with a weight corresponding to the current acquisition frequency of the airborne device is used. For example, when the acquisition frequency is 5HZ, 4HZ, 3HZ, 2HZ, 1HZ, and 0.5HZ, the corresponding weight φ is 80%, 70%, 60%, 50%, 40%, and 30%, respectively, with the minimum frequency being 0.5HZ. The actual control amount can be obtained by weighting the data of the frequency-corresponding weights and the pre-control amount. In this invention, the prescription generation network GPN and the control parameter generation network GCN are trained on the basis of existing neural networks.

[0054] The optimization iterative method is as follows: After fitting the randomly sampled data S_D with the regional soil information model M, a first-stage fitting function η(x) is calculated, and this data is fed into the prescription generation network GPN to generate a pre-prescription model P_P. The pre-prescription model P_P is then processed by the regulation parameter generation network GCN to generate a pre-regulation amount C_P. In a preferred embodiment of the invention, the first-stage fitting method involves matching the locations and soil information of five random sampling points in the soil information model M, solving for the fitting function η(x), and globally applying this function to the soil information model M. During operation, a second-stage fitting is performed based on the real-time soil information S_R(T) obtained from actual sampling and the fitted soil information M to obtain the fitting weight φ. This weight and the pre-regulation amount C_P are then combined to generate the actual regulation amount C. After the operation is completed, the edge computing unit transmits the final actual regulation amount C and the actual prescription model P to the model update and management center. The model update and management center will then estimate the yield of the plot based on the yield prediction model GHM. The yield prediction model GHM is obtained through training based on an existing neural network. At harvest, the actual yield data H_R is compared with the estimated yield H of the yield prediction model GHM. A correction function τ(x) is calculated to correct the soil model M, resulting in the corrected soil information model M_C for the plot. In a preferred embodiment of the invention, the ratio a of the actual yield data H_R and the estimated yield H is used as the correction function, and the soil model M is multiplied by a. The soil model M contains multiple parameters and is actually a vector; each parameter is multiplied by a.

[0055] This embodiment is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A variable-operation control system based on edge computing, comprising: Model update and management center, real-time soil information monitoring unit, satellite positioning equipment, crop yield measurement equipment, edge computing unit, operation controller, terminal control node and terminal execution mechanism; The model update and management center transmits information with the real-time soil information monitoring unit, crop yield measurement equipment, and edge computing unit. It is used to send the soil information model of the corresponding plot to the edge computing unit according to the location of the plot, and update the soil information model of the plot according to the yield information of the plot fed back by the crop yield measurement equipment. The control system performs operational control according to the pre-sowing stage, the sowing stage, and the crop harvesting stage. In the pre-sowing stage, the model update and management center sends the soil model M of the plot to be operated to the edge computing unit; the real-time soil information monitoring unit samples the soil of the plot to be operated and transmits the sampled soil information S_D to the edge computing unit, and the edge computing unit generates a pre-prescription model P_P based on the sampled soil information S_D and the soil model M. During the sowing stage, the real-time soil information monitoring unit keeps sampling the soil of the work site in real time and sends the real-time sampled soil information to the edge computing unit. The edge computing unit forms the actual sowing prescription P based on the pre-prescription model P_P and the real-time sampled soil information, and sends it to the work controller to control the terminal execution mechanism to execute it. During the crop harvesting stage, the edge computing unit receives crop yield information and latitude and longitude information of the work plot sent by the crop yield measurement equipment and satellite positioning equipment, binds them to generate yield-location information H_W, and imports the yield-location information H_W into the model update and management center for correction of the soil model M of the work plot. The real-time soil information monitoring unit is used to measure the soil characteristics of the work site and send the measured real-time soil information to the edge computing unit. The satellite positioning device is connected to the edge computing unit and is used to determine the location of the work site; Crop yield measurement equipment is used to measure the crop yield of the work plot and send the measured crop yield information of the work plot to the edge computing unit; The edge computing unit is used to process the information transmitted by each unit and transmit the processed information to the job controller. The job controller connects the edge computing unit and each terminal control node, receives information from the edge computing unit, and generates control information to send to the corresponding terminal control node. Multiple terminal control nodes, each corresponding to an independent terminal execution mechanism, wherein the terminal control node receives control information from the job controller and sends it to the terminal execution mechanism; The terminal actuator performs operations based on the received execution information.

2. The variable operation control system based on edge computing according to claim 1, characterized in that, In the pre-sowing stage, the real-time soil information monitoring unit performs multiple random samplings of the plot to be operated on, and feeds back the collected soil information S_D of the plot to be operated on to the edge computing unit in vector form; the edge computing unit compares the difference between the collected soil information S_D and the received soil model M, and generates a one-stage fitting function η(x) for the soil model M; the edge computing unit combines the soil model M and the fitting function η(x) to calculate the pre-prescription model P_P of the plot to be operated on.

3. The variable operation control system based on edge computing according to claim 2, characterized in that, The pre-prescription model P_P includes sowing decision schemes under different latitude and longitude coordinates within the plot to be operated.

4. A variable-operation control system based on edge computing according to claim 2, characterized in that, The one-stage fitting function η(x) for the soil model M is generated in a linear, piecewise, or polynomial manner.

5. A variable operation control system based on edge computing according to claim 1, characterized in that, During the sowing stage, the edge computing unit uses the pre-prescription model P_P to calculate the pre-regulation amount C_P for each terminal execution structure. The real-time soil information monitoring unit acquires the soil information S_R at the work point in real time during sowing, and binds the soil information S_R and the current work point location information W to form S_R-W information. The S_R-W information is transmitted to the edge computing unit every sampling interval T. The edge computing unit determines whether a complete soil information sampling process has been completed based on whether the sampling interval has been reached. It performs a second-stage fitting based on the soil information S_R and the soil model after the first-stage fitting to obtain the soil model weight φ. The model weight φ and the pre-regulation amount C_P are used to calculate the actual operation regulation amount C, and the actual regulation amount C is transmitted to the operation controller to control the operation of the terminal execution mechanism.

6. A variable operation control system based on edge computing according to claim 5, characterized in that, If the edge computing unit determines that the sampling interval T has not been reached, it performs a second-stage fitting using the soil information S_R(T-1) at sampling time T-1 and the soil model after the first-stage fitting to obtain the soil model weight φ; if the edge computing unit determines that the sampling interval T has been reached, it performs a second-stage fitting using the soil information S_R(T) at time T and the soil model after the first-stage fitting to obtain the soil model weight φ.

7. A variable operation control system based on edge computing according to claim 5 or 6, characterized in that, The weight φ is set according to the sampling frequency of the real-time soil information monitoring unit used. The higher the sampling frequency, the higher the weight φ is attached to the real-time sampled soil information S_R.

8. A variable operation control system based on edge computing according to claim 1, characterized in that, During the crop harvesting stage, the edge computing unit generates yield-location information H_W based on the yield data H of the work plot measured in real time by the yield measuring equipment and the latitude and longitude data of the work plot obtained by the satellite equipment. After the crop harvest is completed, the edge computing unit sends the yield-location information H_W to the model update and management center. The model update and management center generates estimated yield information P_H based on the actual prescription P, compares the estimated yield information P_H with the yield-location information H_W, calculates the correction function τ(x) of the corrected soil model M, and sends the corrected soil model M_C before the next sowing operation begins.

9. A variable-operation control method based on edge computing, wherein the variable-operation control system based on edge computing as described in any one of claims 1-8 is used for control, characterized in that: In the pre-sowing stage, the model update and management center sends the soil model M of the plot to be operated to the edge computing unit; the real-time soil information monitoring unit samples the soil of the plot to be operated and transmits the sampled soil information S_D to the edge computing unit, and the edge computing unit generates a pre-prescription model P_P based on the sampled soil information S_D and the soil model M. During the sowing stage, the real-time soil information monitoring unit keeps sampling the soil of the work site in real time and sends the real-time sampled soil information to the edge computing unit. The edge computing unit forms the actual sowing prescription P based on the pre-prescription model P_P and the real-time sampled soil information, and sends it to the work controller to control the terminal execution mechanism to execute it. During the crop harvesting stage, the edge computing unit receives crop yield information and latitude and longitude information of the work plot sent by the crop yield measurement equipment and satellite positioning equipment, binds them to generate yield-location information H_W, and imports the yield-location information H_W into the model update and management center for correction of the soil model M of the work plot.

Images