Unmanned aerial vehicle cooperative photovoltaic material transportation method and system based on self-evolution constraint

Abstract

The invention discloses an unmanned aerial vehicle cooperative photovoltaic material transportation method and system based on self-evolution constraint, and belongs to the technical field of intelligent unmanned aerial vehicle control and complex environment autonomous transportation. The method comprises the following steps: collecting multi-source sensing data of an operation scene, and constructing a unified state space including terrain, wind field and feasible navigation area constraints based on the data; inputting the unified state space into a meta-policy network, and generating a high-level decision instruction containing a task intention; modeling an unmanned aerial vehicle group into a multi-agent Markov game framework based on the high-level decision instruction, performing distributed solution by adopting a graph reinforcement learning framework, and outputting a cooperative task and path intention sequence allocated to each unmanned aerial vehicle; and taking the path intention sequence as a reference trajectory, combining unmanned aerial vehicle dynamics and an energy constraint model, constructing a model prediction control problem, performing online solving, generating a safe actual control instruction, and driving an unmanned aerial vehicle group to execute a cooperative transportation task of photovoltaic materials.

Description

Technical Field

[0001] This invention belongs to the field of intelligent unmanned aerial vehicle (UAV) control and autonomous transportation in complex environments, specifically relating to a method and system for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints. Background Technology

[0002] In complex terrains such as mountainous areas and high slopes used in photovoltaic projects, material transportation has long relied on manual handling or ground vehicles. In recent years, drones have been gradually introduced for aerial transport. Current mainstream technical solutions mainly include path planning and flight control methods based on classical control theory, such as using PID and sliding mode control to achieve attitude stability, and combining A* and RRT algorithms to generate obstacle avoidance paths. Some systems introduce machine learning methods, using models such as support vector machines and Gaussian process regression to identify wind disturbances or load changes online to improve environmental adaptability. In addition, some research has attempted to use deep neural networks for environmental perception and strategy generation to achieve a certain degree of autonomous navigation and task decision-making. Multi-drone collaboration often adopts a centralized task allocation mechanism, combined with V2V communication to complete formation flight and path coordination.

[0003] However, existing technologies still have significant limitations when dealing with complex dynamic environments: reinforcement learning strategies often ignore physical constraints such as UAV dynamics, battery energy, and payload sway, resulting in the generated control commands being infeasible or causing instability in real flight; model predictive control, while ensuring physical feasibility, lacks self-learning capabilities and is difficult to adapt to unstructured scenarios such as sudden wind speed changes and terrain undulations; multi-UAV collaboration relies heavily on centralized scheduling, resulting in heavy computational burden and poor communication fault tolerance, making it difficult to support decentralized collaboration of large-scale groups; at the same time, the systems generally lack feedback analysis and strategy evolution mechanisms for task execution performance, and cannot dynamically adjust the reward function or optimize the network structure based on historical performance, making it difficult to continuously improve the level of intelligence. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned problems and provide a method and system for unmanned aerial vehicle (UAV) collaborative photovoltaic material transportation based on self-evolutionary constraints, which can solve the technical problems of insufficient intelligence, unstable control and low collaborative efficiency in photovoltaic material transportation in mountainous and complex terrain.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for unmanned aerial vehicle (UAV) collaborative photovoltaic material transportation based on self-evolutionary constraints, comprising the following steps: Collect multi-source sensor data of the operation scenario, and construct a unified state space that includes terrain, wind field and feasible navigation area constraints based on the data; The unified state space is input into the meta-policy network to generate high-level decision instructions containing task intent; Based on the aforementioned high-level decision-making instructions, the drone swarm is modeled as a multi-agent Markov game framework, and a graph reinforcement learning framework is used for distributed solution, outputting the cooperative task and path intention sequence assigned to each drone. Using the path intention sequence as a reference trajectory, and combining it with UAV dynamics and energy constraint models, a model predictive control problem is constructed and solved online to generate safe actual control commands, driving the UAV swarm to perform the collaborative transportation task of photovoltaic materials.

[0006] A further improvement of this invention is that the multi-source sensor data collected in the operational scenario specifically includes: The attitude and angular velocity data of the UAV are collected by the inertial measurement unit (IMU). UAV location data is acquired using GPS or RTK (Real-Time Dynamic Carrier Phase Differential) technology. Near-ground obstacle and terrain elevation data are collected using LiDAR or SLAM (Simultaneous Localization and Mapping) technology. Local wind speed and direction data are collected using anemometers or track extrapolation wind field estimation methods. Load and swing angle data are collected using a hoisting angle sensor and a load sensor.

[0007] A further improvement of the present invention is that the meta-policy network adopts a structure based on Transformer or Graph Attention Network (GAT) to process spatiotemporal and graph topology information; and the meta-policy network is trained and updated through a meta-learning algorithm and has a structural self-evolution mechanism; the structural self-evolution mechanism is configured to: when the performance index is lower than a threshold, activate the evolution operator to adjust the network topology, and retrain on the experience pool to verify the improvement.

[0008] A further improvement of this invention is that the distributed solution is achieved using a graph reinforcement learning framework, specifically as follows: The drone swarm is represented as a graph structure, with each drone as a node in the graph; Each UAV exchanges status information, including ID, location, remaining energy, and intended location, through periodic broadcasts or event-driven methods via a local communication network (V2V / Mesh). By employing a graph attention network (GAT) or a graph convolutional network combined with a policy network, each UAV can update its policy based on local observations and neighborhood information.

[0009] A further improvement of this invention is that the specific steps of constructing the model predictive control problem include: A coupled model incorporating UAV six-degree-of-freedom dynamics, battery energy consumption, load swing, and wind disturbance input is established as a prediction model. The objective function of the constructed optimization problem is in the form of:

[0010] in, To predict the time domain, To control the input sequence, For system status, For reference only. To control the increment, For the remaining energy, This is an energy reference value. , , These are the weight matrices for the corresponding terms; The constraints of the optimization problem include upper limits for thrust / attitude, limits for sling angle, and minimum safe intervals between UAVs.

[0011] A further improvement of this invention is that, after driving the drone swarm to perform the transportation task, it also includes feedback and self-evolution steps: Collect task execution data and calculate key performance indicators including task completion rate, average energy consumption, average transportation time, path deviation, and load swing RMS value. Based on the aforementioned key performance indicators, Bayesian optimization or policy gradient methods are used to fine-tune the reward function weights and meta-learning parameters of the meta-policy network. The trajectory samples from this task are added to the experience pool for offline reinforcement learning or supervised fine-tuning to enhance policy robustness.

[0012] A further improvement of this invention is that the reward function adopts a multi-objective weighted form:

[0013] in, For the total reward, , , , These are weighting coefficients, corresponding to the importance of task completion, energy consumption, stability, and security, respectively. Rewards for completing the task. As an energy reward, As a stability reward, As a security reward.

[0014] Secondly, the present invention provides a drone-coordinated photovoltaic material transportation system based on self-evolutionary constraints, comprising the following modules: The environmental perception and state construction module is used to collect multi-source sensor data of the operation scenario and construct a unified state space based on the data, including terrain, wind field and feasible navigation area constraints. The cognitive self-evolutionary decision-making module is used to input the unified state space into the meta-policy network to generate high-level decision instructions containing task intent; The group collaborative game allocation module is used to model the UAV swarm as a multi-agent Markov game framework based on the high-level decision instructions, and to use a graph reinforcement learning framework for distributed solution, outputting the collaborative task and path intention sequence allocated to each UAV. The physical constraint model predictive control module is used to take the path intention sequence as a reference trajectory, combine the UAV dynamics and energy constraint model, construct the model predictive control problem and solve it online, generate safe actual control commands, and drive the UAV swarm to perform the collaborative transportation task of photovoltaic materials.

[0015] Thirdly, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints.

[0016] Fourthly, the present invention provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a UAV-based collaborative photovoltaic material transportation method based on self-evolutionary constraints. First, by collecting multi-source sensor data and constructing a unified state space, the system gains high-precision, comprehensive environmental perception capabilities, laying a reliable data foundation for subsequent intelligent decision-making and effectively overcoming decision-making blind spots caused by incomplete information in complex mountainous environments. Next, the unified state space is input into a meta-policy network to generate high-level decision commands, endowing the system with the high-level cognitive ability to directly extract effective task intentions from complex, high-dimensional environmental information, giving the decision-making process intelligence and flexibility that surpasses traditional preset rules. Furthermore, a distributed solution is performed based on a multi-agent Markov game framework and a graph reinforcement learning framework. This method transforms the group collaboration problem into a decentralized autonomous decision-making process. Each UAV can form efficient collaborative tasks and path planning based on local information interaction, fundamentally avoiding the computational bottlenecks and single-point failure risks of centralized control, and significantly improving the scalability and collaborative robustness of large-scale clusters. Most importantly, this method submits the generated path intention sequence to the physical constraint model prediction control layer for final decision-making and execution. By deeply integrating intelligent decision-making with precise UAV dynamics, energy constraints, and other physical models and performing online rolling optimization, it ensures that every control command embodies both strategic intent and physical feasibility. This fundamentally guarantees absolute safety in flight execution, high precision in trajectory tracking, and optimal energy consumption. Through the interconnected steps described above, this method ultimately achieves a balance of intelligence, collaboration, and safety for UAV swarms in complex photovoltaic construction scenarios, fulfilling the core objective of efficient and reliable material transportation. Attached Figure Description

[0018] The accompanying drawings described herein are for illustrative purposes only and are not intended to limit the scope of the invention in any way. Furthermore, the shapes and proportions of the components in the drawings are merely illustrative to aid in understanding the invention and do not specifically limit the shapes and proportions of the components of the invention.

[0019] Figure 1 This is a schematic diagram of the process of the UAV-coordinated photovoltaic material transportation method based on self-evolutionary constraints of the present invention; Figure 2 This is a schematic diagram of the process of the UAV-coordinated photovoltaic material transportation system based on self-evolutionary constraints of the present invention; Figure 3 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0025] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0026] The present invention will now be described in further detail with reference to the accompanying drawings: This invention employs key technologies such as cognitive self-evolutionary reinforcement learning, group cooperative game theory, and physical constraint model predictive control (MPC) to achieve efficient, safe, and intelligent transportation of photovoltaic materials in mountainous, high-slope, and wind-turbulent environments. It solves the problems of insufficient intelligence, unstable control, and low collaborative efficiency in existing technologies. The system of this invention mainly consists of three core layers: a cognitive self-evolutionary reinforcement learning layer, a group cooperative game theory layer, and a physical constraint model predictive control layer. In addition, it includes a task feedback and strategy evolution module for continuous system optimization and self-learning.

[0027] Example 1 like Figure 1 As shown, this invention provides a method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints, comprising the following steps: S1, collect multi-source sensor data of the operation scenario, and construct a unified state space including terrain, wind field and feasible navigation area constraints based on the data; S2, input the unified state space into the meta-policy network to generate high-level decision instructions containing task intent; S3. Based on the high-level decision-making instructions, the drone swarm is modeled as a multi-agent Markov game framework, and a graph reinforcement learning framework is used for distributed solution to output the cooperative task and path intention sequence assigned to each drone. S4. Using the path intention sequence as a reference trajectory, and combining the UAV dynamics and energy constraint model, a model predictive control problem is constructed and solved online to generate safe actual control commands to drive the UAV swarm to perform the collaborative transportation task of photovoltaic materials. S5. During and after the execution of the transportation task, execution result data is collected to calculate key performance indicators; and the key performance indicators are fed back to the meta-policy network for optimization of its decision-making strategy, thereby completing the current transportation task and laying the foundation for improving the efficiency of subsequent tasks.

[0028] In the above steps, "multi-source sensor data" is collected in real time by the IMU (MPU-6050), GPS / RTK module (u-blox F9P), LiDAR sensor (Velodyne VLP-16), anemometer (RM Young05103), and mounted angle sensor (Sensitec AG AMR-Sensor) carried by each UAV, with a sampling frequency set to 20Hz. All data are fused using Kalman filtering to construct a unified state vector S=[x, y, z; v x , v y , v z ; θ ; E ; Fw ; L ], where x, y, z are position coordinates, v x v y v z For velocity components, For attitude angle, E Remaining battery power F w For external wind disturbance and L For load quality, the state vector is uploaded to the edge computing node, where it is combined with the digital elevation model (DEM) and Gaussian process regression method to establish a time-varying wind field map, generating a high-resolution dynamic environment map and a feasible navigation area map.

[0029] The "Meta-Policy Network" is deployed at the ground dispatch center. It uses a Transformer-GAT hybrid architecture to process spatiotemporal graph inputs and outputs task assignment probability distributions, target area recommendations, and path direction vectors. This network is trained using the MAML algorithm and has cross-scene transfer capabilities.

[0030] The "multi-agent Markov game framework" treats each drone as an independent agent, whose local observations include its own state and the positions, energies, and path intentions of other drones in its communication neighborhood. The graph structure G = ( V , E In this model, V represents an individual UAV, E represents a communication connection, and the adjacency matrix is ​​dynamically updated based on relative positions. Each UAV broadcasts its ID, location, energy level, and path intention via a V2V / Mesh network every 100ms. The message frames include CRC checksums to enhance anti-interference capabilities.

[0031] The "Model Predictive Control Problem" is run on an embedded controller (NVIDIA Jetson AGX Xavier), predicting in the time domain. H Set to 3 seconds, sampling period The timeframe is 50ms. The optimization problem comprehensively considers the 6-DOF dynamic equations, the battery discharge model, the first-order model of the slewing load (maximum allowable swing angle ±8°), the upper and lower limits of thrust (0-12N), the attitude angle constraint (±30°), and the minimum safety distance (≥5m). The final output is the thrust magnitude, the desired attitude angle, and the adjustment signal of the slewing mechanism, which are then analyzed by the flight control system (PX4 Autopilot) and executed by the drive motor.

[0032] In some embodiments, the multi-source sensor data collected during the operation specifically includes: The attitude and angular velocity data of the UAV are collected by the inertial measurement unit (IMU). UAV location data is acquired using GPS or RTK (Real-Time Dynamic Carrier Phase Differential) technology. Near-ground obstacle and terrain elevation data are collected using LiDAR or SLAM (Simultaneous Localization and Mapping) technology. Local wind speed and direction data are collected using anemometers or track extrapolation wind field estimation methods. Load and swing angle data are collected using a hoisting angle sensor and a load sensor.

[0033] In this embodiment, the IMU uses an MPU-6050 module, integrating a triaxial accelerometer and gyroscope, with a measurement range of ±16g / ±2000° / s, and is transmitted via I... 2 The Type-C interface communicates with the main control board for real-time calculation of roll, pitch, yaw angles, and angular velocities. The GPS / RTK module is a u-blox F9P, offering centimeter-level horizontal positioning accuracy (1cm + 1ppm), meeting the precise navigation requirements in complex terrain. The LiDAR uses a Velodyne VLP-16 with a 10Hz scanning frequency and a vertical field of view of ±15°, used to construct near-ground 3D point clouds and identify terrain undulations and obstacle distribution. The anemometer is an RM Young 05103 with a measurement range of 0~75m / s, estimating local wind fields through a combination of slipstream thrust and UKF filtering. The load angle sensor is a Sensitec AMR-Sensor with a 100Hz response frequency and a measurement accuracy of ±0.5°, used in conjunction with a load cell (model: Zemic H3-C3) to monitor load and sway status in real time.

[0034] In some embodiments, the meta-policy network adopts a structure based on Transformer or Graph Attention Network (GAT) to process spatiotemporal and graph topology information; and the meta-policy network is trained and updated through a meta-learning algorithm and has a structural self-evolution mechanism; the structural self-evolution mechanism is configured to: when the performance index is lower than a threshold, activate the evolution operator to adjust the network topology and retrain on the experience pool to verify the improvement.

[0035] In some embodiments, the distributed solution using a graph reinforcement learning framework specifically includes: The drone swarm is represented as a graph structure, with each drone as a node in the graph; Each UAV exchanges status information, including ID, location, remaining energy, and intended location, through periodic broadcasts or event-driven methods via a local communication network (V2V / Mesh). By employing a graph attention network (GAT) or a graph convolutional network combined with a policy network, each UAV can update its policy based on local observations and neighborhood information.

[0036] In some embodiments, the steps of constructing the model predictive control problem include: A coupled model incorporating UAV six-degree-of-freedom dynamics, battery energy consumption, load swing, and wind disturbance input is established as a prediction model. The objective function of the constructed optimization problem is in the form of:

[0037] in, To predict the time domain, To control the input sequence, for state, For reference only. To control the increment, For the remaining energy, This is an energy reference value. , , These are the weight matrices for the corresponding terms; The constraints of the optimization problem include upper limits for thrust / attitude, limits for sling angle, and minimum safe intervals between UAVs.

[0038] The MPC controller runs on a real-time operating system (RTOS) and predicts the time domain. H Set the step size to 60 steps (i.e., 3 seconds, sampling period 50ms). State vector Includes position, velocity, attitude, angular velocity, and suspension angle, with a dimension of 12; control input. The thrust is generated by four motors. The dynamic model employs simplified 6-DOF rigid body equations and couples them with a first-order suspended oscillation model (suspension length 1.2m, damping coefficient 0.3). The energy model is based on battery discharge curve fitting, with energy consumption per unit time proportional to the square of the total thrust.

[0039] In the cost function, The penalty for deviation from the reference trajectory is initially set to diag([1.0, 1.0, 2.0, 0.5, 0.5, 0.5]); Suppress drastic changes in the control quantity to prevent mechanical shock; the initial value is diag([0.1, 0.1, 0.1, 0.1]). Encourage energy-efficient flights, with weights dynamically adjusted based on remaining mission mileage.

[0040] The optimization problem was transformed into a sparse quadratic programming problem using the OSQP solver (version 0.6.3). The average solution time on the Jetson TX2 platform was 38ms, which met the real-time requirements.

[0041] In some embodiments, after driving the drone swarm to perform the transportation task, a feedback and self-evolution step is also included: Collect task execution data and calculate key performance indicators including task completion rate, average energy consumption, average transportation time, path deviation, and load swing RMS value. Based on the aforementioned key performance indicators, Bayesian optimization or policy gradient methods are used to fine-tune the reward function weights and meta-learning parameters of the meta-policy network. The trajectory samples from this task are added to the experience pool for offline reinforcement learning or supervised fine-tuning to enhance policy robustness.

[0042] In some embodiments, the reward function takes a multi-objective weighted form:

[0043] in, For the total reward, , , , These are weighting coefficients, corresponding to the importance of task completion, energy consumption, stability, and security, respectively. Rewards for completing the task. As an energy reward, As a stability reward, As a security reward.

[0044] The dynamic adjustment of reward function weights and policy network structure includes: when the task completion rate, average energy consumption, or load stability index is lower than a preset threshold, the model optimization mechanism is activated, and a new policy structure is generated by adjusting the topology parameters of the neural network, and then retrained and validated on historical task data.

[0045] The specific judgment logic is as follows: The system continuously monitors the KPIs of the last 5 tasks. If the task completion rate is below 90% for at least 3 of them, the average energy consumption exceeds the baseline value by 15%, or the RMS value of the hoist swing exceeds 2.5°, then the current network structure is determined to be unsuitable, and the evolution process is initiated. The evolution process uses a controllable genetic algorithm: the population size is set to 20, and individuals represent different network topologies (e.g., 4-8 layers, 2-6 attention heads, 128-512 hidden dimensions). The fitness function comprehensively considers task scores and inference latency.

[0046] The pruning strategy adopts the L1-norm criterion, removing the channel with the smallest absolute weight value; the expansion operation randomly adds one or more attention modules. The candidate structure is lightweightly retrained on the historical experience pool, and the scheme that is superior to the original structure is selected and then enters the A / B testing phase. It is run in a real environment for a week to confirm that the performance has been stably improved before it is deployed online.

[0047] In practical applications, the evolutionary operator can also be the NEAT algorithm or other neural architecture search NAS methods, and this application embodiment does not limit this.

[0048] Example 2 like Figure 2 As shown, the present invention also provides a UAV-coordinated photovoltaic material transportation system based on self-evolutionary constraints, comprising the following modules: The environmental perception and modeling module is used to collect multi-source sensor data of the work scenario, construct the system state space, and establish a dynamic environment model. The cognitive self-evolutionary reinforcement learning module is used to generate high-level policy instructions containing task intentions based on the system state space and output the corresponding uncertainty quantification value. The group collaborative game decision-making module is used to model the interaction relationship of multiple UAVs based on graph structure, receive high-level strategy instructions, perform distributed task allocation and path coordination, and output the task path intention of each UAV. The physical constraint model predictive control module is used to take high-level strategy instructions and mission path intentions as reference inputs, combine dynamics, energy and load swing constraints, solve for the optimal control sequence that satisfies physical feasibility, and output execution instructions to drive the UAV to fly. The feedback and evolution module is used to collect execution result data during and after the execution of the transportation task to calculate key performance indicators; and to feed the key performance indicators back to the meta-policy network for optimization of its decision-making strategy.

[0049] The above modules form a multi-layered closed-loop system through communication and data interaction, enabling drone swarms to achieve autonomous perception, intelligent decision-making, collaborative execution, and dynamic optimization in complex terrain.

[0050] Example 3: Please see Figure 3 As shown, the present invention also provides an electronic device 100 for a UAV-coordinated photovoltaic material transportation method based on self-evolution constraints; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0051] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the UAV-coordinated photovoltaic material transportation method based on self-evolutionary constraints described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0052] The at least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor. The processor 102 is the control center of the electronic device 100, connecting various parts of the electronic device 100 via various interfaces and lines.

[0053] The memory 101 in the electronic device 100 stores multiple instructions to implement a UAV-coordinated photovoltaic material transportation method based on self-evolutionary constraints, and the processor 102 can execute the multiple instructions to achieve the following: Collect multi-source sensor data of the operation scenario, and construct a unified state space that includes terrain, wind field and feasible navigation area constraints based on the data; The unified state space is input into the meta-policy network to generate high-level decision instructions containing task intent; Based on the aforementioned high-level decision-making instructions, the drone swarm is modeled as a multi-agent Markov game framework, and a graph reinforcement learning framework is used for distributed solution, outputting the cooperative task and path intention sequence assigned to each drone. Using the path intention sequence as a reference trajectory, and combining it with UAV dynamics and energy constraint models, a model predictive control problem is constructed and solved online to generate safe actual control commands, driving the UAV swarm to perform the collaborative transportation task of photovoltaic materials.

[0054] Example 4: If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0055] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0058] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0060] Many embodiments and applications beyond the examples provided will be apparent to those skilled in the art upon reading the foregoing description. Therefore, the scope of this teaching should not be determined by reference to the foregoing description, but rather by reference to the foregoing claims and the full scope of their equivalents. For purposes of completeness, all articles and references, including patent applications and publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein in the foregoing claims is not intended as a waiver of that subject matter, nor should it be construed as an indication that the applicant has not considered that subject matter as part of the disclosed inventive subject matter.

[0061] The above content provides a further detailed description of the present invention. It should not be construed that the specific embodiments of the present invention are limited to this. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered to fall within the scope of protection of the present invention as defined by the submitted claims.

Claims

1. A method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints, characterized in that, Includes the following steps: Collect multi-source sensor data of the operation scenario, and construct a unified state space that includes terrain, wind field and feasible navigation area constraints based on the data; The unified state space is input into the meta-policy network to generate high-level decision instructions containing task intent; Based on the aforementioned high-level decision-making instructions, the drone swarm is modeled as a multi-agent Markov game framework, and a graph reinforcement learning framework is used for distributed solution, outputting the cooperative task and path intention sequence assigned to each drone. Using the path intention sequence as a reference trajectory, and combining it with UAV dynamics and energy constraint models, a model predictive control problem is constructed and solved online to generate safe actual control commands, driving the UAV swarm to perform the collaborative transportation task of photovoltaic materials.

2. The method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints according to claim 1, characterized in that, The multi-source sensor data collected in the operation scenario specifically includes: The attitude and angular velocity data of the UAV are collected by the inertial measurement unit (IMU). UAV location data is acquired using GPS or RTK (Real-Time Dynamic Carrier Phase Differential) technology. Near-ground obstacle and terrain elevation data are collected using LiDAR or SLAM (Simultaneous Localization and Mapping) technology. Local wind speed and direction data are collected using anemometers or track extrapolation wind field estimation methods. Load and swing angle data are collected using a hoisting angle sensor and a load sensor.

3. The method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints according to claim 1, characterized in that, The meta-policy network adopts a structure based on Transformer or Graph Attention Network (GAT) to process spatiotemporal and graph topology information; furthermore, the meta-policy network is trained and updated through a meta-learning algorithm and has a self-evolutionary structure mechanism; the self-evolutionary structure mechanism is configured to: when the performance index is lower than a threshold, activate the evolution operator to adjust the network topology and retrain on the experience pool to verify the improvement.

4. The method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints according to claim 1, characterized in that, The distributed solution using a graph reinforcement learning framework is as follows: The drone swarm is represented as a graph structure, with each drone as a node in the graph; Each UAV exchanges status information, including ID, location, remaining energy, and intended location, through periodic broadcasts or event-driven methods via a local communication network (V2V / Mesh). By employing a graph attention network (GAT) or a graph convolutional network combined with a policy network, each UAV can update its policy based on local observations and neighborhood information.

5. The method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints according to claim 1, characterized in that, The specific steps for constructing the model predictive control problem include: A coupled model incorporating UAV six-degree-of-freedom dynamics, battery energy consumption, load swing, and wind disturbance input is established as a prediction model. The objective function of the constructed optimization problem is in the form of: in, To predict the time domain, To control the input sequence, For system status, For reference only. To control the increment, For the remaining energy, This is an energy reference value. , , These are the weight matrices for the corresponding terms; The constraints of the optimization problem include upper limits for thrust / attitude, limits for sling angle, and minimum safe intervals between UAVs.

6. The method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints according to claim 1, characterized in that, After driving the drone swarm to perform the transportation mission, the process also includes feedback and self-evolution steps: Collect task execution data and calculate key performance indicators including task completion rate, average energy consumption, average transportation time, path deviation, and load swing RMS value. Based on the aforementioned key performance indicators, Bayesian optimization or policy gradient methods are used to fine-tune the reward function weights and meta-learning parameters of the meta-policy network. The trajectory samples from this task are added to the experience pool for offline reinforcement learning or supervised fine-tuning to enhance policy robustness.

7. A method for UAV-coordinated photovoltaic material transportation based on self-evolutionary constraints according to claim 6, characterized in that, The reward function adopts a multi-objective weighted form: in, For the total reward, , , , These are weighting coefficients, corresponding to the importance of task completion, energy consumption, stability, and security, respectively. Rewards for completing the task. As an energy reward, As a stability reward, As a security reward.

8. A drone-coordinated photovoltaic material transportation system based on self-evolutionary constraints, characterized in that, Includes the following modules: The environmental perception and state construction module is used to collect multi-source sensor data of the operation scenario and construct a unified state space based on the data, including terrain, wind field and feasible navigation area constraints. The cognitive self-evolutionary decision-making module is used to input the unified state space into the meta-policy network to generate high-level decision instructions containing task intent; The group collaborative game allocation module is used to model the UAV swarm as a multi-agent Markov game framework based on the high-level decision instructions, and to use a graph reinforcement learning framework for distributed solution, outputting the collaborative task and path intention sequence allocated to each UAV. The physical constraint model predictive control module is used to take the path intention sequence as a reference trajectory, combine the UAV dynamics and energy constraint model, construct the model predictive control problem and solve it online, generate safe actual control commands, and drive the UAV swarm to perform the collaborative transportation task of photovoltaic materials.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the UAV-coordinated photovoltaic material transportation method based on self-evolutionary constraints as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the UAV-coordinated photovoltaic material transportation method based on self-evolutionary constraints as described in any one of claims 1 to 7.

Images