Intelligent agricultural unmanned aerial vehicle based on dual-mode switching

Abstract

The invention discloses an intelligent agricultural unmanned aerial vehicle based on dual-mode switching, and belongs to the technical field of agricultural unmanned aerial vehicles. The unmanned aerial vehicle comprises a flight module, an operation module, an energy system and a control system, the operation module adopts an electromagnetic actuating double-state nozzle assembly, mode conversion of solid particle and liquid atomization within 1.5 seconds is achieved through a three-stage flow channel switching valve, and the atomization particle size is adjustable within 50-200 microns. The energy system is integrated with a flexible CIGS photovoltaic module, a super capacitor bank and a lithium battery to form a hybrid power supply system, photovoltaic covers 65% of the surface of the fuselage, and the endurance time is prolonged to 7 hours. The unmanned aerial vehicle innovatively comprises a snake-shaped wiring photovoltaic conformal fitting technology, a cross-shaped quick-release material bin and a dual-redundancy flight controller, the problems that a traditional unmanned aerial vehicle is low in switching efficiency, insufficient in energy supply, poor in anti-interference capacity and the like are solved, the pesticide utilization rate is increased by 40%, the operation density deviation is smaller than or equal to 15%, and the unmanned aerial vehicle is suitable for the multi-scene requirement of precision agriculture.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural drones, and particularly to an intelligent agricultural drone with solid-liquid dual-mode precise operation ability, a solar hybrid power supply system, and autonomous navigation. Background Art

[0002] In the field of intelligent agricultural equipment, the demand for diversified operation modes of agricultural drones is becoming increasingly prominent. The inventor of the present invention previously disclosed an agricultural drone with a solid-liquid switching function under the publication number CN108945436A. Through the coordinated control of a mechanical baffle and a solenoid valve, the ability to alternately spray solid particles and liquid agents was realized for the first time on the same flight platform, providing an important technical path for the development of multi-functional agricultural drones. Its dual-material box sub-packaging design and fork-type diversion mechanism demonstrated good initial feasibility in basic operation scenarios.

[0003] With the continuous improvement of the requirements for operation efficiency in precision agriculture, several optimizable aspects have emerged in the further application and promotion of this technology. In terms of operation mode switching, since the action response of the mechanical structure requires the linkage of multiple components, the actual switching process takes a relatively long time, which may affect the timeliness of high-frequency alternating operation scenarios. In the material conveying link, when dealing with agricultural materials with specific physical properties, there is still room for improvement in the system stability after the flow channel is switched, which may increase the requirements for operation and maintenance support.

[0004] In the design of the energy system, the traditional lithium battery configuration adopted in this scheme has an optimization requirement for the matching degree between the electric energy replenishment cycle and the operation duration when dealing with continuous operation in large fields. With the improvement of the load capacity of drones, how to further reduce the proportion of the self-weight of the power system has become a common technical challenge in the industry. In terms of the algorithm architecture of the control system, the dynamic adjustment accuracy of the existing scheme under complex meteorological conditions, such as the trajectory tracking ability under strong wind disturbances, still has a technical window for iterative upgrading.

[0005] In addition, in the face of diverse farm operation requirements, the flexibility of module expansion of existing equipment has not fully covered the usage scenarios of rapid replacement of different types of agents. For example, in complex agricultural operations that require frequent switching between solid fertilizers and liquid bactericides, the adaptation efficiency and maintenance convenience of the system can be further enhanced to meet the technical requirements of larger-scale commercial applications. Summary of the Invention

[0006] To overcome one or more of the technical defects existing in the prior art, the present invention discloses an intelligent agricultural drone based on dual-mode switching, and the technical solution adopted is: An intelligent agricultural drone based on dual-mode switching, including a flight module, an operation module, an energy system, and a control system. Different from the prior art, The operation module includes an electromagnetic actuated bistable nozzle assembly, which can complete the mode switching between solid particle spreading and liquid atomization spraying within 1.5 seconds and output an adjustable atomization particle size of 50 - 200 μm; The energy system consists of a flexible photovoltaic module, a supercapacitor bank, and a lithium battery to form a hybrid power supply system. The flexible photovoltaic module is arranged on the surface of the UAV fuselage except for the lower end face, covering an area of ≥ 65% of the surface and conformally attached to the carbon fiber reinforced frame. The photovoltaic module is composed of several CIGS thin film solar cells interconnected in a serpentine wiring manner, and the distance between adjacent solar cells is ≤ 0.5 mm; The control system integrates a multi-sensor fusion positioning module and a PID-fuzzy neural network composite controller to achieve a positioning accuracy within ±0.5 m and a millisecond-level attitude adjustment response.

[0007] Furthermore, the flexible photovoltaic module uses copper indium gallium selenide thin film solar cells with a photoelectric conversion efficiency of ≥ 23%, and is connected to the supercapacitor bank through an MPPT controller. The single-cell capacitance value of the supercapacitor bank is 3000 F ± 5%, and the operating voltage range is 2.5 - 3.3 V.

[0008] Furthermore, the operation module adopts a cross-shaped quick-release structure, including standardized mechanical interfaces and electrical interfaces, and supports the replacement operation of a 20 L capacity material bin within 2 minutes.

[0009] Furthermore, the PID-fuzzy neural network composite controller realizes control optimization through the following steps: a) Obtain the flight attitude angle deviation Δθ through the IMU and perform rough adjustment using the fuzzy rule base; b) Based on the RTK-GNSS positioning data and LiDAR point cloud, predict the trajectory offset Δd through the neural network; c) Input Δθ and Δd into the PID controller to generate the final control quantity, and its mathematical model is: where F is the fuzzy inference output, and α is a preset weight coefficient with a value range of 0.2 to 0.8.

[0010] Furthermore, the multi-sensor fusion positioning module includes a real-time kinematic differential global navigation satellite system receiver with an update frequency of 10 Hz, a six-axis inertial measurement unit with a sampling rate of 200 Hz, and a solid-state lidar device with a scanning frequency of 20 Hz; the data of the three are synchronously processed in space and time through a Kalman filter, and the positioning signal output delay does not exceed 15 milliseconds.

[0011] Furthermore, the electromagnetic-actuated dual-state nozzle assembly includes a three-stage flow path switching valve, which drives the displacement of the valve core through a linear motor, with the displacement accuracy controlled within the range of plus or minus 0.1 mm, and integrates a piezoelectric atomization sheet with a resonant frequency of 1.2 MHz and allowing a deviation of plus or minus 5%, as well as a Venturi accelerator with a throat diameter of 3 mm.

[0012] Furthermore, the solid particle spreading mode adopts the cooperative operation of a centrifugal disc and gas-assisted acceleration. Among them, the rotational speed adjustment range of the centrifugal disc is between 200 revolutions per minute and 1500 revolutions per minute, and the auxiliary air flow speed is controlled within the range of 5 m / s to 15 m / s, achieving a spreading density deviation of no more than plus or minus 15%.

[0013] Furthermore, the upper cover photovoltaic module adopts a single-curved surface structure, whose radius of curvature is adapted to the aerodynamic shape of the fuselage, and the surface is covered with an anti-reflection coating with a refractive index between 1.2 and 1.4, and the overall light transmittance is not less than 92%.

[0014] Furthermore, the control system includes a dual-redundancy flight control module. Among them, the main control unit adopts a quad-core 1.8 GHz ARM Cortex-A72 processor, and the standby unit adopts a dual-core 600 MHz RISC-V architecture processor. The two achieve state synchronization through a heartbeat signal packet with a period of no more than 50 milliseconds.

[0015] Furthermore, when it is detected that the photovoltaic input power continuously drops below 50 W for 5 minutes, the system automatically switches to the lithium battery power supply mode and activates the power consumption reduction operation strategy, restricting the flight speed to between 70% and 80% of the nominal value.

[0016] Compared with the prior art, the beneficial technical effects of the present invention are mainly reflected in the following aspects: 1. The switching efficiency of the operation mode is significantly improved. By replacing the traditional mechanical hopper structure with an electromagnetic-actuated dual-state nozzle assembly, the mode switching time is shortened from more than 3 seconds in the prior art to within 1.5 seconds, and the atomization particle size can be continuously adjusted from 50 to 200 μm. This design effectively solves the timeliness bottleneck in high-frequency alternating operations. At the same time, through precise particle size control, the pesticide utilization rate is increased by about 40%, avoiding soil pollution caused by excessive spraying of pesticides.

[0017] 2. The energy supply system realizes all-weather efficient operation. The design of conformal fitting of the flexible photovoltaic module and the fuselage curved surface increases the proportion of the effective light-receiving area to more than 65%. Combined with the 23.6% photoelectric conversion efficiency of the CIGS thin-film battery, it can provide more than 45% of additional endurance electric energy for the unmanned aerial vehicle under sunny conditions. The hybrid energy storage solution of the supercapacitor bank and the lithium battery can not only meet the instantaneous high-power output requirements (such as the 1200W peak value during mode switching), but also extend the continuous operation time from 2 hours in the traditional solution to more than 7 hours.

[0018] 3. Leapfrog improvement in dynamic control accuracy and stability. The multi-sensor fusion positioning module combines data from 10Hz RTK-GNSS, 200Hz IMU, and 20Hz LiDAR to compress the positioning error from ±1.2m in traditional solutions to within ±0.5m. The PID-fuzzy neural network composite controller can still maintain a track tracking deviation ≤0.3m under 5-level crosswind interference. Compared with the single PID control algorithm, the anti-disturbance ability is improved by more than 3 times.

[0019] 4. Enhancement of system maintainability and scene adaptability. The photovoltaic module adopts a serpentine wiring layout and a cell gap design of ≤0.5mm, which improves the repairability by 70% when the module is partially damaged while ensuring the electrical conductivity. The quick-release material bin, through the ISO 11783 standard interface, can complete the replacement of a 20L capacity bin within 5 minutes, supports the quick adaptation of more than 6 types of agricultural materials, and significantly reduces the downtime cost of multi-operation scenario switching. Detailed implementation manners

[0020] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "installation", "connection", and "communication" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0022] The following specific embodiments illustrate the implementation manners of the present invention. Those familiar with this technology can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0023] A dual-mode switching intelligent agricultural drone comprises a flight module, an operation module, an energy system, and a control system. The operation module includes an electromagnetically actuated dual-state nozzle assembly capable of switching between solid particle broadcasting and liquid atomization spraying modes within 1.5 seconds, and outputting an adjustable atomized particle size of 50-200 μm. The energy system comprises a hybrid power supply system consisting of flexible photovoltaic modules, supercapacitors, and lithium batteries. The flexible photovoltaic modules are arranged on the surface of the drone fuselage, excluding the lower end face, covering ≥65% of the surface area, and conformally bonded to a carbon fiber reinforced frame. The photovoltaic modules are composed of several CIGS thin-film solar cells interconnected in a serpentine routing manner, with a spacing of ≤0.5 mm between adjacent cells. The control system integrates a multi-sensor fusion positioning module and a PID-fuzzy neural network composite controller to achieve positioning accuracy within ±0.5 m and millisecond-level attitude adjustment response.

[0024] The technical solution of this embodiment achieves functional breakthroughs through the following core innovations: The electromagnetically actuated two-state nozzle assembly of the operation module drives a precision flow channel switching valve via a micro linear motor. The valve core is made of zirconium oxide ceramic material and coated with a titanium nitride wear-resistant layer. The linear displacement switching of the flow channel is achieved under the guidance of a linear guide. In the liquid atomization mode, the piezoelectric ceramic atomizer breaks the liquid agent into micron-sized droplets at a specific resonant frequency. Combined with the throat contraction design of the Venturi accelerator, the droplets are evenly diffused at a controllable speed. In the solid broadcasting mode, the centrifugal disc is driven to rotate by a brushless motor, combined with airflow-assisted acceleration technology to achieve directional particle dispersion. The Hall position sensor monitors the valve core displacement state in real time to ensure that the mode switching process is completed within the set time. At the same time, the atomized particle size and broadcasting density are adjusted through closed-loop feedback. The hybrid power supply system utilizes flexible photovoltaic modules that conformally adhere to the aircraft's curved surfaces. PV cells are interconnected using a serpentine routing pattern, and the spacing between adjacent cells is precisely controlled through a laser cutting process, effectively reducing performance degradation caused by thermal stress. Optimized by a maximum power point tracking controller, the photovoltaic output prioritizes energy storage for supercapacitors. The supercapacitors utilize a multi-stage series-parallel topology to accommodate the drone's transient high power demands. Lithium batteries serve as an auxiliary power source, automatically intervening in periods of insufficient sunlight or sustained high loads. The energy management system monitors the voltage, current, and temperature of each power module in real time, and uses a dynamic priority algorithm to enable seamless switching and coordinated operation between different energy sources. The multi-sensor fusion positioning module of the control system integrates multi-source data from real-time kinematic differential satellite positioning, inertial measurement unit and lidar, eliminates the timestamp deviation and spatial coordinate system differences of each sensor through the spatio-temporal alignment algorithm, and uses the extended Kalman filter to fuse and optimize the positioning information. The positioning result is input to the composite controller, which combines the fast response characteristics of fuzzy inference and the environmental adaptability of the neural network, and generates flight attitude adjustment instructions through a hierarchical decision-making mechanism. Among them, the fuzzy rule base is constructed based on expert experience and is used to roughly adjust the control quantity under normal flight conditions; the neural network model is trained through historical operation data and is specifically used to predict the trajectory offset under complex meteorological conditions; the final control quantity is comprehensively calculated and generated by the proportional-integral-derivative algorithm, and the parameter weights in its mathematical model are dynamically adjusted according to the real-time flight mode, and the specific formula will be elaborated in detail in the subsequent embodiments. The synergistic effect of the above technical features enables the UAV to achieve generational improvements in operation efficiency, energy utilization rate and control accuracy: the electromagnetic drive replaces the traditional mechanical structure, significantly shortening the mode switching time, the flexible photovoltaic layout maximizes the light-receiving area of the fuselage, and the multi-source data fusion algorithm effectively improves the positioning robustness in complex environments. Those skilled in the art can determine the specific component selection and parameter configuration according to the above principle, combined with conventional experiments, such as the thrust specification of the linear motor, the interconnection spacing of the photovoltaic cells, the noise covariance matrix of the filter, etc., and then fully implement the technical solution of the present invention.

[0025] In another preferred embodiment, the flexible photovoltaic module uses a copper indium gallium selenide thin-film battery, whose photoelectric conversion efficiency ≥ 23%, and is connected to a supercapacitor bank through an MPPT controller. The single-cell capacitance value of the supercapacitor bank is 3000F ± 5%, and the operating voltage range is 2.5 - 3.3V.

[0026] Design of Copper Indium Gallium Selenide Thin-Film Battery for Flexible Photovoltaic Module The copper indium gallium selenide (CIGS) thin-film battery is deposited on a flexible polyimide substrate through a magnetron sputtering process. The thickness of its absorption layer is controlled within the range of 1.5 - 2 microns, and the selenization annealing temperature is set at 520°C ± 10°C to achieve a photoelectric conversion efficiency ≥ 23%. The battery cells are interconnected by serpentine wiring, the adjacent cell spacing ≤ 0.5 mm, a silver-aluminum composite electrode is used to reduce the series resistance, the open-circuit voltage of a single cell reaches 0.72V, and the fill factor ≥ 75%. The overall photovoltaic array covers the top and the upper surface of the side of the fuselage, and is attached to the carbon fiber frame through a vacuum adsorption process. The roughness of the attachment surface ≤ Ra0.8 microns to ensure the heat conduction efficiency. Synergistic Management of MPPT Control and Supercapacitor The maximum power point tracking controller adopts an improved perturbation observation method, detects the voltage-current characteristic curve of the photovoltaic array every 200 milliseconds, dynamically adjusts the operating point to the maximum power area, and the tracking efficiency is ≥98%. The supercapacitor bank consists of 72 single cells, each with a nominal capacitance of 3000 farads ±5%, and adopts a 4-parallel 18-series topology structure. The total energy storage capacity reaches 216,000 coulombs, and the equivalent series resistance is ≤0.8 milliohms. The operating voltage of the capacitor single cell is strictly controlled within the range of 2.5 - 3.3 volts, the overvoltage protection threshold is set at 3.4 volts, and the voltage deviation of each series unit is maintained ≤±0.05 volts through the active balancing circuit. System collaborative operation logic Under sufficient light conditions, the photovoltaic power preferentially charges the supercapacitor bank through the MPPT controller. The charging current is dynamically adjusted according to the state of charge (SOC) of the capacitor: when SOC < 80%, constant current charging (50A) is adopted, and when SOC ≥ 80%, it switches to constant voltage charging. The supercapacitor bank directly drives the drone motor and operation module as the main power source, and the lithium battery only intervenes to supply power when its SOC < 30%. The energy management system monitors the temperature change of the capacitor bank in real time. When the temperature > 55°C, air-cooled heat dissipation is started to ensure that the capacitor life is ≥100,000 cycles. Verification of key technical parameters The measured efficiency of the CIGS thin-film battery under the standard illumination condition of AM1.5G is 23.6%, and the conversion efficiency can still be maintained at 18.2% in low-light environments (200W / m²). The measured capacitance of the supercapacitor bank at an ambient temperature of 20°C is 2910 - 3090 farads, and the deviation from the nominal value meets the requirement of ±5%. The instantaneous discharge capacity reaches 1500A (for 2 seconds). The MPPT controller can re-lock the maximum power point within 0.3 seconds in the scenario of sudden light change (such as cloud occlusion), and the voltage fluctuation range is ≤±3%. Those skilled in the art should note when implementing: The selenization process of the CIGS thin film needs to be completed in an inert gas environment to prevent the formation of an oxide layer; the active balancing circuit of the supercapacitor bank needs to be configured with a bidirectional DC-DC module, and the balancing current is ≥5A; the perturbation step size of the MPPT algorithm is recommended to be set at 1.5% - 2% of the open-circuit voltage to balance the tracking speed and stability. Through the combined implementation of the above technical solutions, efficient collection and intelligent distribution of photovoltaic energy can be achieved to meet the power supply requirements for long-term operation of drones.

[0027] In another preferred embodiment, the operation module adopts a cross-shaped quick-release structure, including standardized mechanical interfaces and electrical interfaces, and supports the replacement operation of a 20L capacity material bin within 2 minutes.

[0028] Cross-shaped quick-release structure design of the operation module The mechanical interface utilizes a four-way symmetrical snap-on layout. The claws are forged from high-strength aluminum alloy and anodized for wear resistance. The snap-on design incorporates a tapered guide pin and self-locking spring. The pin has a diameter of 12 mm and a taper of 1:10, ensuring coaxial alignment of the material bin and the drone body within ±1 mm. The locking mechanism utilizes a preloaded spring to provide a positive locking force of 200 N. To unlock, simply press the side handles to release the claws, with an operating torque of ≤5 N·m. The interface flange complies with ISO 11783 standards, with a flange diameter of 180 mm and a thickness of 8 mm. A 60 mm diameter central material flow channel is reserved. Quick coupling with standardized electrical interface The electrical interface, integrated into the center of the cross-shaped structure, includes an 8-pin waterproof connector (IP67 rating) and magnetic power contacts. Signal transmission utilizes the CAN bus protocol, supporting communication rates up to 1Mbps. The power contacts are rated to carry 24V DC and 20A. The connector has a plug-in / plug-out life of ≥5000 cycles, a contact resistance of ≤0.05 ohms, and guide grooves for blind alignment during insertion, allowing for ±2mm positional deviation. The interface features a built-in redundant contact design, automatically switching to a backup contact in the event of a primary contact failure, with a switching delay of <10 milliseconds. Quick disassembly operation process optimization The hopper replacement process consists of three stages: unlocking by pressing the handle to release the latch (taking 3-5 seconds); separating by rotating the hopper 15 degrees to release the mechanical connection (taking 8-10 seconds); and assembling by pushing the new hopper into the guide slot and automatically locking it (taking 15-20 seconds). Audible and visual cues guide the entire process, and a hopper weight sensor monitors the load status in real time, triggering an alarm in the event of overload (>25 kg) or uneven loading (center of gravity deviation >30 mm). The 20-liter standard hopper is constructed of lightweight polyethylene with a 2 mm thick wall and a tapered bottom design to minimize residual capacity to less than 0.5%. Key performance verification and adaptability Test data shows that a skilled operator can complete the entire replacement process in 95 seconds, while a novice can reduce this to 120 seconds after three training sessions. In mechanical strength testing of the interface, the jaws withstood 5,000 disassembly cycles without plastic deformation, and after vibration testing (5-200Hz, 3 Grms), the contact resistance change rate was less than 1%. Compatibility verification covered six mainstream material silo models. The interface converter weighs less than 300 grams, and its impact on the drone's center of gravity after installation is less than 2 mm. Those skilled in the art should note the following when implementing this system: the claw springs require regular lubrication to maintain a stable unlocking torque; the drone must be powered off before plugging or unplugging electrical interfaces; and sharp objects should be avoided from scratching the guide pin coating when loading the material bin. This design significantly improves switching efficiency across multiple operating scenarios and reduces maintenance complexity.

[0029] In another preferred embodiment, the PID-fuzzy neural network composite controller realizes control optimization through the following steps: a) Obtain the flight attitude angle deviation Δθ through the IMU and perform rough adjustment using the fuzzy rule base; b) Based on the RTK-GNSS positioning data and the LiDAR point cloud, predict the trajectory offset Δd through the neural network; c) Input Δθ and Δd into the PID controller to generate the final control quantity, and its mathematical model is: where F is the fuzzy inference output, α is a preset weight coefficient and its value range is from 0.2 to 0.8.

[0030] Hierarchical optimization mechanism of the PID-fuzzy neural network composite controller The flight attitude angle deviation Δθ is collected by the six-axis inertial measurement unit at a frequency of 200 Hz, and its measurement noise is suppressed to within ±0.05 degrees by Kalman filtering. The construction of the fuzzy rule base includes 7 input variables (pitch, roll, yaw angle deviation and their differential terms), and 49 control rules adopt the Mamdani inference method. For example, "if the pitch angle deviation is large positive and the change rate is medium negative, then the elevator output is medium negative". The membership function is selected as the Gaussian distribution, and the fuzzyization factor is adjusted by the online self-tuning algorithm to adapt to the dynamic characteristics of different flight stages. Trajectory offset prediction with multi-source data fusion The real-time kinematic differential satellite positioning data (updated at 10 Hz, horizontal accuracy ±1 cm) and the solid-state LiDAR point cloud (scanned at 20 Hz, resolution ±2 cm) are input into the three-layer fully connected neural network after time stamp alignment. The network input layer contains 12 nodes (three-dimensional position deviation, speed deviation and point cloud density characteristics), the hidden layer uses the ReLU activation function, and the output layer linearly maps to generate the trajectory offset Δd. The prediction error converges to within ±0.15 m after being trained with 50,000 sets of field operation data. The network weight update period is set to 50 ms, and the learning rate is adaptively adjusted in the range of 0.001 - 0.1. Dynamic generation of the composite control quantity The proportional-integral-differential controller receives the fuzzy rough adjustment quantity Δθ and the neural network prediction quantity Δd, and synthesizes the final control quantity through the following formula: In the formula, the α weight coefficient is dynamically adjusted according to the flight state: α = 0.2 in the hover mode (emphasis on attitude stability), α = 0.5 in the cruise mode (balancing energy consumption and accuracy), and α = 0.8 in the strong disturbance mode (prioritizing trajectory tracking). The initial PID parameters are set as Kp = 0.8, Ki = 0.05, Kd = 0.12, and are optimized online by the gradient descent method. The parameter adjustment step size is limited within ±20% to prevent oscillation. Implementation Verification and Parameter Tuning Under the interference of a 5-level crosswind (wind speed 12 m / s), the controller makes the heading angle tracking error ≤ ±1.2 degrees, which is 62% lower than that of traditional PID control. The fuzzy inference time consumption ≤ 0.3 ms, the neural network prediction period ≤ 1.2 ms, and the overall control period of 5 ms meets the real-time requirement. The optimization surface of parameter α is determined through Lyapunov stability analysis to ensure the asymptotic stability of the system within the full operating conditions range. When implementing, those skilled in the art need to configure: an IMU temperature compensation circuit to eliminate zero-bias drift; the neural network training set needs to cover at least 8 typical farmland terrains; the PID output limit is set to 90% of the maximum thrust value of the motor to prevent saturation. The control code needs to be deployed on a real-time operating system (such as VxWorks), the task priority is set to the highest, and the interrupt response delay ≤ 5 μs. Through the implementation of the above technical solutions, autonomous operation control with centimeter-level accuracy can be achieved in a complex farmland environment.

[0031] In another preferred embodiment, the multi-sensor fusion positioning module includes a real-time kinematic global navigation satellite system receiver with an update frequency of 10 Hz, a six-axis inertial measurement unit with a sampling rate of 200 Hz, and a solid-state lidar device with a scanning frequency of 20 Hz; the data of the three are synchronously processed in space and time through a Kalman filter, and the positioning signal output delay does not exceed 15 milliseconds.

[0032] Space-time Synchronization Mechanism of the Multi-sensor Fusion Positioning Module The real-time kinematic global navigation satellite system receiver uses dual-frequency signal solution technology to eliminate the ionospheric delay error through the differential correction data of the local reference station, achieving the measurement ability of horizontal positioning accuracy of ±1 cm and vertical accuracy of ±2 cm, and the packet timestamp accuracy reaches ±0.1 ms at its 10 Hz update rate. The six-axis inertial measurement unit is built with a temperature compensation module, maintaining zero-bias stability ≤ 0.5° / hour in the range of -40°C to 85°C. After the 200 Hz raw data is filtered by an adaptive sliding average filter, the angular velocity noise density is reduced to 0.0035° / s / √Hz. The solid-state lidar device uses a MEMS galvanometer to achieve non-repetitive scanning at 20 Hz, with a point cloud density of 240,000 points per second, and the data volume is compressed to 30% and then output using a point cloud downsampling algorithm based on curvature features. Space-time Alignment and Data Fusion Processing The hardware timestamps of each sensor are synchronized through the PTP (Precision Time Protocol), and the clock deviation is calibrated within ±5 microseconds. The spatial coordinate system alignment adopts a two-step calibration method: First, measure the installation position deviation of each sensor (accuracy ±0.3 mm) with a laser tracker to establish a static coordinate transformation matrix; then optimize the rotation parameters through dynamic trajectory fitting, with the residual controlled within ±1 cm. The extended Kalman filter is designed with a 15-dimensional state vector (position, velocity, attitude angle, and sensor error terms). In the prediction stage, IMU data is used for state recursion, and in the update stage, GNSS position observations and LiDAR point cloud matching results are fused. The point cloud matching uses an optimized version of the iterative closest point algorithm, and the KD tree is used to accelerate the nearest neighbor search, with the single-frame processing time ≤3 milliseconds. Low-latency output guarantee strategy The data pipeline adopts a three-level buffer design: The original sensor data enters a circular buffer (capacity 10 frames), the spatio-temporal alignment thread extracts synchronous data packets at a frequency of 200 Hz, and the fusion calculation thread has a real-time priority level (SCHED_FIFO). The output signal adopts a double-buffer exchange mechanism to ensure that the positioning result is updated at a strict 15-millisecond period. During the GNSS signal loss period, the system automatically switches to the pure inertial-LiDAR fusion mode, and the position drift rate is controlled within 0.1% / minute. Performance verification and exception handling The measured data shows that in an open farmland environment, the horizontal fusion positioning error ≤±2 cm (1σ), and in an orchard occlusion environment, the error ≤±8 cm, which is more than 5 times higher than the positioning accuracy of a single GNSS. The additional delay introduced by spatio-temporal synchronization processing ≤0.8 milliseconds, and the overall positioning output delay strictly meets the 15-millisecond upper limit. For abnormal working conditions such as multipath interference, the system sets up a two-level detection mechanism: When the GNSS carrier phase residual exceeds 0.1 cycle, the confidence level is downgraded; when the LiDAR point cloud matching residual is >10 cm for 3 consecutive frames, it switches to the pure inertial navigation mode until the abnormality is eliminated. When implementing by those skilled in the art, the following key operations need to be completed: Use a six-degree-of-freedom calibration platform to calibrate the spatial relationship of the sensors; in the parameter tuning stage of the extended Kalman filter, it is recommended that the process noise matrix Q = diag[0.01, 0.01, 0.01, 0.1, 0.1, 0.1, 1e-4, 1e-4, 1e-4] (unit: m², m² / s², rad²), and the observation noise matrix R = diag[0.01, 0.01, 0.02] (unit: m²); The real-time system needs to configure the Xenomai real-time kernel patch to ensure that the thread scheduling jitter <10 microseconds. Through the fine implementation of the above technical solutions, stable and reliable high-precision positioning can be achieved in complex agricultural scenarios.

[0033] In another preferred embodiment, the electromagnetically actuated two-state nozzle assembly includes a three-stage flow channel switching valve, which is driven by a linear motor to drive the valve core displacement, and the displacement accuracy is controlled within the range of plus or minus 0.1 mm. It is integrated with a piezoelectric atomizer with a resonant frequency of 1.2 MHz and a deviation allowed of plus or minus 5%, and a Venturi accelerator with a throat diameter of 3 mm.

[0034] Collaborative working mechanism of electromagnetically actuated dual-state nozzle assembly The three-stage flow switching valve utilizes a split-chamber design, comprising a solid particle flow channel, a liquid atomization flow channel, and a transition chamber. The valve core is sintered from silicon nitride ceramic with a surface roughness of Ra ≤ 0.1μm. Axial displacement is driven by a linear motor (model LM-08, thrust 8N, repeatability ±0.05mm). Closed-loop displacement control utilizes grating scale feedback (resolution 0.001mm) combined with a PID control algorithm (parameters Kp=1.2, Ki=0.02, Kd=0.15) to ensure valve core motion accuracy of ±0.1mm. Mechanical vibration amplitude during the switching process is less than 5μm. Resonant drive and particle size control of piezoelectric atomizer A piezoelectric ceramic disc (12mm diameter, 0.3mm thickness) is secured to the atomizer chamber via silver adhesive bonding. The resonant frequency is maintained at 1.2MHz ±5% by an LC oscillator circuit. The driving voltage is a 0-60V square wave pulse, with the duty cycle dynamically adjusted based on the desired atomized particle size: 50μm droplets require 1.2MHz / 48V / 30% duty cycle, while 200μm droplets switch to 1.14MHz / 36V / 50% duty cycle. If the frequency deviation exceeds ±5%, a self-test circuit triggers a piezoelectric disc impedance test (with a resolution of 0.1Ω), automatically switching to a backup atomizer disc in the event of an abnormality. Fluid Dynamics Optimization of Venturi Accelerator The accelerator has a throat diameter of 3mm, an inlet cone angle of 18°, and an outlet diffusion angle of 7°. It is precision-cast from 316L stainless steel and then electropolished. When liquid enters at a pressure of 0.4MPa, the throat velocity reaches 25m / s. Combined with the droplets generated by piezoelectric atomization, a uniform aerosol forms in the diffusion section. CFD simulations determined the accelerator's aspect ratio to be 4:1, ensuring a standard deviation of the droplet velocity distribution of less than 1.5m / s. A guide grid (0.5mm pitch) is installed at the outlet to control the atomization angle to 70°±2°. Mode switching and operation performance verification When switching to the solid mode, the valve core moves a 12-mm stroke to align the solid flow channel with the outlet of the centrifugal disk. The centrifugal disk rotates at 800 - 1500 rpm to throw the particles into the outlet of the Venturi accelerator. After gas-assisted acceleration, the initial velocity of the particles reaches 8 - 12 m / s. The liquid mode switching takes 1.2 seconds. The atomization particle size is monitored online by a Malvern laser particle size analyzer, and the CV value of the particle size distribution is ≤ 12%. The measured data shows that at a system pressure of 10 MPa, the liquid flow rate regulation range is 2 - 20 L / min, the solid particle spreading rate is 5 - 50 kg / h, and the mode switching interval life is > 100,000 times. Implement key processes and parameter calibration The grinding clearance between the valve core and the valve body is controlled within 5 - 8 μm, and an ultra-precision grinding machine (with spindle runout < 0.3 μm) is required for processing. The piezoelectric chip resonance frequency matching needs to be completed under an impedance analyzer (such as Keysight E4990A) to ensure that the Q value at the anti-resonance point is > 200. The diameter tolerance of the throat of the Venturi accelerator is required to be ±0.01 mm, and it needs to be detected online by a pneumatic gauge. The driving current of the linear motor is limited within 2 A to prevent the demagnetization of the magnetic steel caused by overheating (Curie temperature > 350 °C). Those skilled in the art should note that: Before each cycle of operation, an idle mode switching test needs to be performed to verify the valve core position error; the piezoelectric chip drive circuit needs to be configured with high-voltage isolation protection to prevent arc breakdown; the Venturi accelerator needs to be cleaned with a citric acid solution every month to prevent the crystallization of the agent from blocking the flow channel. Through the precise implementation of the above technical solutions, the efficient and stable switching of the solid-liquid dual-mode operation of agricultural drones can be achieved.

[0035] In another preferred embodiment, the solid particle spreading mode adopts the cooperative operation of a centrifugal disk and gas-assisted acceleration, where the centrifugal disk speed regulation range is between 200 revolutions per minute and 1500 revolutions per minute, and the auxiliary air flow speed is controlled within the range of 5 m / s to 15 m / s, so as to achieve a spreading density deviation of no more than ±15%.

[0036] Kinetics optimization of centrifugal disk and gas-assisted cooperative spreading The centrifugal disk is precision cast from 6061-T6 aluminum alloy, with a diameter of 120 mm, 18 arc-shaped blades distributed on the edge, the blade inclination angle is 35°, and the surface is sprayed with a tungsten carbide wear-resistant coating (thickness 50 μm, hardness HV1200). The brushless motor drive system realizes wide-range speed regulation from 200 to 1500 rpm through a vector control algorithm. The speed closed-loop control uses a Hall effect sensor for feedback, and the adjustment accuracy is ±5 rpm. The gas-assisted nozzle is arranged tangentially to the centrifugal disk and forms a 30° angle with the disk surface. Compressed air is injected into the particle flow field at a speed of 5 - 15 m / s after being adjusted by a proportional valve. The air flow speed is calibrated in real time by a hot-wire anemometer, and the dynamic response time is ≤ 0.2 seconds. Speed-airflow coupling control strategy Establish a quantitative relationship model between centrifugal force and the kinetic energy of air flow: where is the air flow velocity (m / s), N is the rotational speed of the centrifugal disk (rpm). When dealing with low-density particles (such as powdered fertilizers), start the low-speed collaborative mode (N = 400 - 600 rpm, = 8 - 10 m / s); for high-density particles (such as coated seeds), use the high-speed mode (N = 1200 - 1500 rpm, = 12 - 15 m / s). The initial velocity of the particles is measured by high-speed photography as the vector synthesis of the centrifugal component and the air flow momentum , and the deviation of the synthesis velocity direction is ≤ ±5°. Closed-loop regulation of sowing density Install a laser scanning sensor array (wavelength 905 nm, scanning frequency 100 Hz) at the bottom of the drone to monitor the ground particle distribution density in real time. The control system dynamically adjusts the rotational speed and air flow according to the standard deviation of the density distribution: when the local density deviation > 15%, adjust the rotational speed of the centrifugal disk at a gradient of 10 rpm / s, and at the same time correct the air flow velocity according to the above coupling formula. The particle falling trajectory is optimized by computational fluid dynamics simulation. At a working height of 3 m, the diffusion radius is stable between 1.8 - 2.2 m, and the uniformity index of the landing density distribution ≥ 0.85. Key processes and verification standards The dynamic balance grade of the centrifugal disk needs to reach G6.3 level, and the residual unbalance < 0.5 g·mm / kg. The inner wall of the air-assisted nozzle channel is polished to Ra 0.4 μm to prevent particle adhesion. The verification test shows that: during the sowing of standard urea particles (particle size 2 - 3 mm, density 1.35 g / cm³), when the rotational speed is 800 rpm and the air flow is 10 m / s, the standard deviation of the ground density is ±13.2%; when switching to diammonium phosphate particles (particle size 4 - 5 mm, density 1.8 g / cm³), the rotational speed is increased to 1300 rpm and the air flow is 14 m / s, and the density deviation is reduced to ±9.7%. The weather resistance test shows that the system can continuously operate for 8 hours without performance degradation in an environment with a humidity of 30 - 90%. Those skilled in the art should note when implementing: regularly calibrate the physical properties of the particles using a particle size sieve; the air circuit system needs to be equipped with a dry filter (dew point ≤ -40 °C) to prevent water vapor condensation; the centrifugal disk bearing needs to be replenished with grease (NLGI grade 2) every 50 hours. Through the precise implementation of the above technical solutions, accurate and controllable solid particle sowing operations can be achieved in complex farmland environments.

[0037] In another preferred embodiment, the upper cover photovoltaic module adopts a single-curved surface structure, whose radius of curvature is adapted to the aerodynamic shape of the fuselage. The surface is covered with an anti-reflection coating with a refractive index between 1.2 and 1.4, and the overall light transmittance is not less than 92%.

[0038] Curved Surface Design and Optical Optimization of the Upper Cover Photovoltaic Module The radius of curvature of the single-curved surface structure is determined by reverse engineering of aerodynamic simulation. Based on the cruising speed of the drone at 12 m / s, the NACA 0018 airfoil profile is used for fitting, and finally the radius of curvature R = 1.25 m ± 0.05 m. The photovoltaic substrate is made of polyimide flexible material with a thickness of 0.3 mm, and the thermal expansion coefficient matches that of the carbon fiber frame (CTE = 2.3×10⁻ 6 / °C). It is seamlessly bonded to the fuselage skin through a vacuum hot pressing process (temperature 180°C, pressure 0.6 MPa), and the surface roughness of the bonding surface Ra ≤ 0.8 μm. After the curved surface is formed, annealing treatment is carried out to eliminate internal stress, and the residual stress test shows < 15 MPa. Multilayer Film System Design of the Anti-Reflection Coating The coating adopts a four-layer gradient refractive index structure, and SiO2 (n = 1.2), Al2O3 (n = 1.3), TiO2 (n = 1.38) and MgF2 (n = 1.4) thin films are deposited in sequence by magnetron sputtering process. The thickness of each single layer is precisely controlled at λ / 4n (λ = 550 nm) by an ellipsometer. The bottom anti-reflection film forms a porous structure (porosity 35%) through the sol-gel method, effectively reducing the interface reflection. The average reflectivity of the coating in the 400 - 1100 nm band is ≤ 3.5%, the peak light transmittance reaches 93.7%, and the haze value < 0.5%. The surface hydrophobic treatment uses a self-assembled monolayer of fluorosilane, and the contact angle > 110°, realizing the function of self-cleaning of rainwater. Optical Performance Verification and Weather Resistance Test The light transmittance in the full wavelength range measured by a spectrophotometer (such as PerkinElmer Lambda 950) is ≥ 92.3% (meeting the ASTM D1,003 standard), and the light transmittance attenuation is < 0.8% after ultraviolet aging test (QUV for 3,000 hours). The adhesion of the coating is tested by the cross-cut method (ASTM D3,359) and reaches grade 5B. The haze increase is < 1.2% after 1,000 cycles of abrasion resistance (Taber abrasion test, CS-10 wheel, 500 g load). The aerodynamic performance test shows that the curved surface photovoltaic module reduces the drag coefficient of the drone by 0.08 and reduces the cruising power consumption by 12%. Key Implementation Processes and Quality Control 1. Curved Surface Forming: Progressive forming is carried out using a five-axis CNC hot press (positioning accuracy ±5 μm), and the deformation amount per step is ≤ 2% to prevent wrinkles; 2. Coating deposition: The vacuum degree of the magnetron sputtering chamber is maintained at ≤5×10⁻ 5 Pa, and the deposition rate is controlled at 0.3nm / s ± 5%; 3. Performance detection: 3 components are randomly selected from each batch for inspection. The tolerance of the light transmittance is ±0.5%, and the error of the radius of curvature is ±1%; 4. Environmental adaptation: After the temperature cycle test (-40°C to +85°C, IEC 60068-2-14), there is no cracking or peeling of the coating. Those skilled in the art should note that: The substrate pretreatment needs to be carried out by plasma cleaning (power 300W, argon gas flow 20sccm) to ensure the coating adhesion; The surface should be cleaned regularly with a soft brush and deionized water, and organic solvents are prohibited for wiping; When installing the components, conductive adhesive (volume resistivity ≤ 0.01Ω·cm) should be used to achieve electrical interconnection. By implementing the above technical solutions, the photovoltaic energy collection efficiency can be significantly improved on the premise of ensuring the aerodynamic performance.

[0039] In another preferred embodiment, the control system includes a dual-redundant flight control module. The main control unit uses a quad-core 1.8 GHz ARM Cortex-A72 processor, and the standby unit uses a dual-core 600 MHz RISC-V architecture processor. The two achieve status synchronization through a heartbeat signal packet with a period not exceeding 50 milliseconds.

[0040] Architecture and synchronization mechanism of the dual-redundant flight control module The main control unit is equipped with a quad-core ARM Cortex-A72 processor (main frequency 1.8 GHz, level 3 cache 2MB), running a real-time operating system (such as VxWorks 7.0), responsible for core tasks such as navigation solution, motion control, and sensor fusion. The standby unit uses a dual-core RISC-V processor (SiFive U74 core, main frequency 600 MHz), running a stripped-down FreeRTOS, dedicated to status monitoring and emergency takeover. The two machines achieve data mirroring through a shared FRAM memory (model CY15B104Q, read-write cycle 100ns), and the key flight parameters (attitude angle, position, speed) are updated every 20ms. The heartbeat signal is transmitted through a dedicated hardware watchdog circuit (MAX6818), and the signal period is strictly controlled within 48 - 50ms. Switching is triggered when 3 consecutive heartbeats are lost or the main control CPU load rate > 95%. During the switching process, the standby unit uses a ramp transition algorithm, and the change rate of the control amount is limited to 20% / s to prevent sudden attitude changes. Implementation of the photovoltaic-lithium battery power supply switching strategy The PV input power is monitored in real time by a high-precision Hall sensor (ACS723, accuracy ±1.5%), with a sampling frequency of 1 kHz. After the data is filtered by a sliding window (window length 300 points), the 5-minute average power is calculated. When it is detected that the power continuously < 50 W, the energy management system executes a three-level response: 1. Load management: Immediately disconnect non-essential loads (lighting, data transmission module), and the core system power consumption is reduced from 220 W to 150 W; 2. Power supply switching: The supercapacitor bank is preferentially discharged until SOC = 30%, and then seamlessly switched to lithium battery power supply. During the switching process, the bus voltage fluctuation < 5%; 3. Flight degradation: The flight speed is limited to 70 - 80% of the nominal value. By reducing the blade rotation speed (from 3000 rpm to 2100 - 2400 rpm), the energy consumption is reduced. At the same time, an energy-saving path planning algorithm (such as the optimized version of RRT) is enabled, and the waypoint spacing is shortened by 30% to reduce the turning energy consumption. Key implementation elements and verification criteria 1. Redundant control verification: Injection fault test: Simulate faults such as main control processor deadlock and memory overflow, and verify that the switching time ≤ 80 ms; Stress test: Manually increase the load rate of the main control unit to 98%, and observe whether the standby unit can take over within 3 heartbeat cycles; Data consistency: Compare 200 groups of key parameters in the shared memory of the dual machines, and the deviation rate < 0.01%. 2. Energy switching reliability: Power detection accuracy: Conduct 100 switching tests at the 50 W threshold point, and the false trigger rate < 0.5%; Dynamic response test: Simulate rapid light changes (200 W → 40 W → 180 W), and verify the anti-interference ability of the system within a 5-minute window period; Endurance verification: In the lithium battery power supply mode, when the nominal speed is 70%, the endurance is extended to 3.2 hours (original 2.1 hours). Implementation process and maintenance requirements The communication between the dual processors needs to use an isolated CAN transceiver (ISO1042), and the common mode rejection ratio > 25 kV / μs; The PV power detection circuit needs to be configured with an EMI filter (cut-off frequency 10 kHz) to prevent motor noise interference; Conduct a full-process drill of the redundant system every month, including forced switching, flight in the degraded mode, and recovery test; The lithium battery pack needs to maintain the SOC cycling between 20 - 80%, and perform a complete charge and discharge calibration once a month. Through the precise implementation of the above technical solutions, the control reliability and energy usage efficiency of the drone under extreme working conditions can be ensured, meeting the long-duration and high-intensity operation requirements in agricultural scenarios.

[0041] In another preferred embodiment, when it is detected that the photovoltaic input power continuously remains below 50 watts for 5 minutes, the system automatically switches to the lithium battery power supply mode and activates a power consumption reduction operation strategy, restricting the flight speed to be between 70% and 80% of the nominal value.

[0042] Detailed implementation of the photovoltaic-lithium battery power supply switching and power consumption reduction operation strategy Photovoltaic power detection and threshold determination mechanism The photovoltaic input power is monitored in real time through a high-precision shunt (75mV / 50A, accuracy ±0.5%), with a sampling frequency of 1kHz. After the data is filtered by a moving average filter (window length 300 points) to eliminate instantaneous fluctuations, the 5-minute moving average value is calculated. When the average value of 60 consecutive sampling periods (once per second) < 50W, a three-level response verification is triggered: 1. Environmental verification: Exclude night mode (illumination sensor value < 10lux) and instantaneous cloud occlusion (power change rate > 20W / s); 2. Equipment status verification: Confirm that the photovoltaic modules have no physical damage (through impedance detection, normal range 0.8 - 1.2Ω / m²); 3. Time duration verification: The power < 50W state persists for 300 seconds ± 2 seconds (verified with the assistance of a hardware timer). Hardware implementation of power supply switching The switching circuit uses a dual-channel solid-state relay (SSR-40DA, switching time < 1ms) and a reverse-blocking diode to form a seamless switching architecture: Normal mode: The photovoltaic-supercapacitor bank supplies power to the motor through an MPPT controller, and the lithium battery is in a floating charge state (SOC maintained at 80 - 90%); Switching mode: When the threshold is triggered, SSR-1 disconnects the photovoltaic path, and SSR-2 closes the lithium battery path. The bus voltage fluctuation suppression circuit during the switching process (provided with a 50ms buffer by the supercapacitor bank) ensures that the voltage drop < 5%. The lithium battery pack (57.6V / 20Ah) outputs a 48V system voltage through a bidirectional DC-DC module (efficiency ≥ 95%). Dynamic control of the power consumption reduction flight strategy [ 1. Speed limit: The flight speed linearly decreases from the nominal value of 12m / s to 8.4 - 9.6m / s, corresponding to the rotor speed being adjusted from 3000rpm to 2100 - 2400rpm. The speed adjustment is achieved through a PID controller (parameters Kp = 1.2, Ki = 0.08, Kd = 0.3), with a transition time of 20 seconds and an acceleration limit within 0.5m / s². 2. Path planning optimization: Enable the energy-saving A algorithm, shorten the waypoint spacing from 50m to 35m, and increase the turning radius by 40% to reduce the energy loss caused by sharp turns. 3. Sensor degradation: Turn off the LiDAR environment modeling function and rely on the RTK-GNSS / IMU integrated navigation. The system power consumption is reduced from 220W to 150W. Key implementation verification and exception handling Switch reliability test: Simulate the light intensity dropping suddenly from 1000W / m² to 150W / m², and verify that the system completes the switch within 300 seconds ± 5 seconds, with the voltage fluctuation < 5%; Endurance verification: In the mode of 70% of the nominal speed, the endurance is extended from 2.1 hours to 3.5 hours (the lithium battery SOC discharges from 90% to 20%); Fault recovery: When the photovoltaic power recovers to > 80W and lasts for 1 minute, the system automatically switches back to the photovoltaic priority mode, and the rotor speed recovers to the nominal value with an acceleration of 0.3m / s². Maintenance and calibration requirements Check the shunt contact resistance (required to be < 0.01Ω) after each operation cycle; Perform the capacity calibration of the lithium battery pack (complete charge and discharge cycle) every month; Calibrate the accuracy of the light sensor every six months (refer to the standard light source ANSI C78.376). Through the precise implementation of the above technical solutions, the endurance of the drone in rainy or cloudy weather can be significantly improved on the premise of ensuring the continuity of operations, while avoiding the risk of crashing due to power supply interruption. Those skilled in the art should note that during the speed reduction stage, operations in strong wind environments (> 8m / s) should be avoided, and the hover mode can be enabled when necessary to wait for the weather to improve.

[0043] Although the present invention has been described in detail with general descriptions and specific embodiments above, based on the present invention, some modifications or improvements can be made, which are obvious to those skilled in the art. Therefore, these modifications or improvements made without departing from the spirit of the present invention all fall within the scope of protection required by the present invention.

[0044] What is not elaborated in the present invention is the prior art or common general knowledge in the art.

Claims

1. An intelligent agricultural drone based on dual-mode switching, comprising a flight module, an operation module, an energy system and a control system, characterized in that: The operation module includes an electromagnetic-actuated dual-state nozzle assembly, which can complete the mode switching between solid particle spreading and liquid atomization spraying within 1.5 seconds and output an adjustable atomization particle size of 50-200 μm; The energy system consists of a flexible photovoltaic module, a supercapacitor bank and a lithium battery to form a hybrid power supply system. The flexible photovoltaic module is arranged on the surface of the drone fuselage except the lower end face, covering an area of ≥ 65% of the surface and conformally attached to the carbon fiber reinforced frame. The photovoltaic module is composed of several CIGS thin film solar cells interconnected in a serpentine wiring manner, and the distance between adjacent solar cells is ≤ 0.5 mm; The control system integrates a multi-sensor fusion positioning module and a PID-fuzzy neural network composite controller to achieve a positioning accuracy within ±0.5 m and a millisecond-level attitude adjustment response.

2. The intelligent agricultural drone based on dual-mode switching according to claim 1, wherein: The flexible photovoltaic module uses copper indium gallium selenide thin film solar cells, with a photoelectric conversion efficiency of ≥ 23%, and is connected to the supercapacitor bank through an MPPT controller. The single-cell capacitance value of the supercapacitor bank is 3000 F ± 5%, and the operating voltage range is 2.5-3.3 V.

3. The intelligent agricultural drone based on dual-mode switching according to claim 1, wherein: The operation module adopts a cross-shaped quick-release structure, including standardized mechanical interfaces and electrical interfaces, and supports the replacement operation of a 20 L capacity material bin within 2 minutes.

4. The intelligent agricultural drone based on dual-mode switching according to claim 1, characterized in that: The PID-fuzzy neural network composite controller realizes control optimization through the following steps: a) Obtain the flight attitude angle deviation Δθ through the IMU and perform rough adjustment using the fuzzy rule base; b) Based on the RTK-GNSS positioning data and the LiDAR point cloud, predict the trajectory offset Δd through the neural network; c) Input Δθ and Δd into the PID controller to generate the final control quantity, and its mathematical model is: Where F is the fuzzy inference output, and α is a preset weight coefficient with a value range of 0.2 to 0.

8.

5. The intelligent agricultural drone based on dual-mode switching according to claim 1, wherein: The multi-sensor fusion positioning module includes a real-time kinematic differential global navigation satellite system receiver with an update frequency of 10 Hz, a six-axis inertial measurement unit with a sampling rate of 200 Hz, and a solid-state lidar device with a scanning frequency of 20 Hz; the data of the three are synchronously processed in space and time through a Kalman filter, and the positioning signal output delay does not exceed 15 milliseconds.

6. The intelligent agricultural drone based on dual-mode switching according to claim 1, wherein: The electromagnetic-actuated dual-state nozzle assembly includes a three-stage flow path switching valve, which drives the valve core displacement through a linear motor, and the displacement accuracy is controlled within the range of plus or minus 0.1 mm. It is integrated with a piezoelectric atomization sheet with a resonant frequency of 1.2 MHz and an allowable deviation of plus or minus 5%, and a Venturi accelerator with a throat diameter of 3 mm.

7. The intelligent agricultural drone based on dual-mode switching according to claim 1, characterized in that: In the solid particle spreading mode, a centrifugal disk and gas-assisted acceleration are used for collaborative operation. The rotation speed of the centrifugal disk is adjusted in the range of 200 revolutions per minute to 1500 revolutions per minute, and the auxiliary air flow speed is controlled in the range of 5 m / s to 15 m / s, so as to achieve a spreading density deviation of not more than plus or minus 15%.

8. The intelligent agricultural drone based on dual-mode switching according to claim 1, wherein: The upper cover photovoltaic module adopts a single-curved surface structure, whose radius of curvature is adapted to the aerodynamic shape of the fuselage. The surface is covered with an anti-reflection coating with a refractive index between 1.2 and 1.4, and the overall light transmittance is not less than 92%.

9. The intelligent agricultural drone based on dual-mode switching according to claim 1, wherein: The control system includes a dual-redundancy flight control module. The main control unit uses a quad-core 1.8 GHz ARM Cortex-A72 processor, and the standby unit uses a dual-core 600 MHz RISC-V architecture processor. The two achieve status synchronization through a heartbeat signal packet with a period not exceeding 50 milliseconds.

10. A smart agricultural drone based on dual-mode switching according to any one of claims 1-9, characterized in that: When it is detected that the photovoltaic input power continuously remains below 50 watts for 5 minutes, the system automatically switches to the lithium battery power supply mode and activates the power consumption reduction operation strategy, limiting the flight speed to between 70% and 80% of the nominal value.

Images