Unmanned aerial vehicle cluster energy efficiency optimization control method and system based on input-output analysis

Abstract

The application discloses a UAV cluster energy efficiency optimization control method and system based on input-output analysis, and belongs to the technical field of UAV energy management. The method collects energy consumption and task execution data of each UAV in real time, constructs a dynamic input-output table, and depicts the correlation between energy flow and task output. The neural network model is used to dynamically identify energy coupling strength coefficients and task contribution weights. These parameters are used as optimization basis, and the linear programming algorithm is used to generate optimal energy distribution instruction set with the minimum unit task energy consumption as the target. The UAV state is adjusted in real time through the instruction distribution and closed-loop optimization mechanism. The system includes data acquisition, dynamic table construction, parameter identification, instruction generation and distribution modules. The application can realize global energy efficiency optimization of the UAV cluster in a complex task environment, and improve energy utilization efficiency and task execution capability.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) energy management technology, and in particular to a method and system for optimizing the energy efficiency of UAV swarms based on input-output analysis. Background Technology

[0002] With the widespread application of unmanned aerial vehicle (UAV) swarm technology in military reconnaissance, disaster relief, agricultural plant protection, and power line inspection, the efficiency and energy sustainability of its collaborative mission execution are becoming increasingly prominent issues. UAV swarm systems typically consist of multiple heterogeneous or homogeneous UAVs, which need to collaboratively complete tasks such as search, monitoring, and data transmission in complex and dynamic environments. However, swarm systems face severe energy constraints in actual operation, particularly limited battery capacity, lack of coordination between energy use and mission execution, and dynamic environmental interference, which seriously restrict their ability to operate over long periods and large areas.

[0003] Traditional energy management methods primarily focus on individual drone energy monitoring and static power allocation, failing to adequately consider the coupling effects of energy flow between drones and differences in task contributions. This results in low overall swarm energy efficiency and unstable task completion quality. Furthermore, existing methods typically rely on preset models and fixed parameters, making them ill-suited to handle changes in task priorities, environmental disturbances, and unforeseen circumstances. They also lack the ability to dynamically identify and optimize the energy-task relationship in real time. Therefore, there is an urgent need for an energy efficiency control method for UAV swarms that can model the relationship between energy input and task output, optimize energy allocation in real time, and support closed-loop dynamic adjustments to improve overall energy utilization efficiency and task execution reliability. Summary of the Invention

[0004] In order to overcome the above-mentioned deficiencies of the prior art, embodiments of this application provide an energy efficiency optimization control method for UAV swarms based on input-output analysis to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this application provides a method for optimizing the energy efficiency of unmanned aerial vehicle (UAV) swarms based on input-output analysis, comprising:

[0006] By deploying energy monitoring sensors on each drone in the cluster, energy consumption data and task execution progress data of each drone are collected in real time when performing collaborative tasks;

[0007] Based on the energy consumption data and the task execution progress data, a dynamic input-output table describing the relationship between energy flow and task output within the UAV cluster is constructed.

[0008] Based on the dynamic input-output table, a neural network model trained on historical data is used to dynamically identify and update the energy coupling strength coefficient and task contribution weight among the drones in the cluster.

[0009] Using the energy coupling strength coefficient and the task contribution weight as optimization parameters, and taking the maximization of overall cluster energy efficiency as the objective function, an optimal energy allocation instruction set is generated.

[0010] The optimal energy allocation instruction set is distributed to each UAV in the cluster to adjust its flight attitude, propulsion power, or mission execution sequence in order to perform the cooperative mission.

[0011] To address the aforementioned issues, this application also provides an energy efficiency optimization control system for drone swarms based on input-output analysis, the system comprising:

[0012] The data acquisition module is used to collect energy consumption data and task execution progress data of each drone in real time when performing collaborative tasks by using energy monitoring sensors deployed on each drone in the cluster.

[0013] The dynamic input-output table construction module is used to construct a dynamic input-output table describing the relationship between energy flow and task output within the UAV cluster based on the energy consumption data and the task execution progress data.

[0014] The coupling coefficient and contribution weight identification module is used to dynamically identify and update the energy coupling strength coefficient and task contribution weight between each UAV in the cluster based on the dynamic input-output table and a neural network model trained on historical data.

[0015] The optimal energy allocation instruction generation module is used to generate an optimal energy allocation instruction set with the energy coupling strength coefficient and the task contribution weight as optimization parameters and the overall energy efficiency of the cluster as the objective function.

[0016] The instruction distribution and UAV adjustment module is used to distribute the optimal energy allocation instruction set to each UAV in the cluster, and adjust its flight attitude, propulsion power or task execution sequence to perform the cooperative task.

[0017] This application significantly improves the energy utilization efficiency and task adaptability of a cluster by constructing an energy efficiency optimization method that integrates dynamic modeling, parameter identification, and optimized control. First, it utilizes multiple sensors to collect energy consumption and task progress data in real time and constructs a dynamic input-output table, achieving a structured description of the relationship between energy flow and task output, overcoming the problems of separate data collection and lagging model updates in traditional methods. Second, a neural network model trained based on historical data dynamically identifies and updates the energy coupling strength coefficient and task contribution weight, enabling the system to have strong generalization and real-time response capabilities, adapting to task changes and environmental interference. Then, using the dynamic input-output table as constraints and the identified energy coupling and task contribution parameters as core optimization variables, an optimization function is constructed with the goal of minimizing unit task energy consumption. A linear programming algorithm is used to solve for the globally optimal energy allocation strategy, maximizing cluster energy efficiency while meeting task requirements. Finally, through command distribution and closed-loop optimization mechanisms, the theoretical optimization results are transformed into specific control commands for each UAV, and data is continuously collected, the model is updated, and the strategy is adjusted during task execution. This system not only significantly improves the overall energy efficiency of drone swarms, but also enhances their robustness and mission completion capabilities in complex mission environments. Attached Figure Description

[0018] Figure 1 A flowchart illustrating an embodiment of the energy efficiency optimization control method for UAV swarms based on input-output analysis provided in this application;

[0019] Figure 2 A functional block diagram of an energy efficiency optimization control system for unmanned aerial vehicle swarms based on input-output analysis, provided as an embodiment of this application;

[0020] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0022] This application provides a method for optimizing the energy efficiency of a drone swarm based on input-output analysis. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and big data and artificial intelligence platforms.

[0023] Reference Figure 1 The diagram shown is a flowchart illustrating an embodiment of the UAV swarm energy efficiency optimization control method based on input-output analysis provided in this application. In this embodiment, the UAV swarm energy efficiency optimization control method based on input-output analysis includes:

[0024] S1. By deploying energy monitoring sensors on each drone in the cluster, real-time data on energy consumption and task execution progress of each drone during collaborative tasks are collected.

[0025] In this embodiment, the energy monitoring sensor is a hardware combination deployed on each drone in the drone swarm to collect data related to energy consumption and task execution during the drone's collaborative task execution in real time. This combination includes a current sensor, a voltage sensor, a battery management system, an onboard GPS module, and a task payload status sensor.

[0026] In this embodiment, the energy consumption data is a set of data reflecting the energy usage of the UAV when performing collaborative tasks. This set of data consists of the UAV's operating current, operating voltage, and remaining battery power data. The task execution progress data is a set of data reflecting the progress of the UAV's collaborative tasks. This set of data consists of the UAV's location information and task execution status information.

[0027] In some embodiments, the step of collecting energy consumption data and task execution progress data of each drone in real time during the execution of collaborative tasks through energy monitoring sensors deployed on each drone in the cluster includes:

[0028] The operating current and operating voltage of each UAV are collected in real time through the current sensor and voltage sensor in the energy monitoring sensor.

[0029] The battery management system in the energy monitoring sensor collects the remaining power data of each drone, and uses the operating voltage, operating current and remaining power data as the energy consumption data of the drone.

[0030] The location information and mission execution status information of each UAV are collected by the airborne GPS module and mission payload status sensor in the energy monitoring sensor as mission execution progress data.

[0031] In this embodiment, the current sensor is a component of the energy monitoring sensor, specifically used to measure the current value of core components such as motors during the operation of the UAV in real time, providing basic current parameters for calculating energy consumption; the voltage sensor is a component of the energy monitoring sensor, specifically used to measure the voltage value output by the UAV battery in real time, providing basic voltage parameters for calculating energy consumption.

[0032] In this embodiment, the battery management system is a component of the energy monitoring sensor, used to monitor the remaining power and charging / discharging status of the drone battery in real time, and can directly output the current remaining power data of the drone.

[0033] In this embodiment, the airborne GPS module is a component of the energy monitoring sensor. It receives satellite signals to obtain the latitude and longitude of the UAV in real time, providing a location basis for judging the progress of the mission. The mission payload status sensor is a component of the energy monitoring sensor. It is used to monitor the working status of the UAV mission payload (such as inspection camera, detection equipment) in real time and output mission execution status information such as whether the mission is being executed and whether the execution has been completed.

[0034] In this embodiment, the energy monitoring sensor is first deployed. For each drone in the drone swarm, the various components of the energy monitoring sensor are fixed at appropriate locations on its fuselage (such as near the battery, at the payload installation location, or on the top of the fuselage): a current sensor is connected in series in the power supply circuit of the drone motor to ensure accurate acquisition of the motor's operating current; a voltage sensor is connected in parallel between the positive and negative terminals of the drone battery to acquire the battery's output voltage; the battery management system is connected to the drone battery through a dedicated interface to achieve real-time monitoring of the remaining power; the onboard GPS module is installed on the top of the drone fuselage in an unobstructed location to ensure stable satellite signal reception; and the payload status sensor is connected to the drone's payload (such as an inspection camera) through a data interface to obtain the payload's operating status in real time.

[0035] In this embodiment, after sensor deployment is completed, the sensor acquisition function is activated to start collecting energy consumption data in real time: the deployed current sensor collects the operating current of each UAV motor in real time at a sampling frequency of 1Hz. The collected current signal is converted into a digital signal after A / D conversion and transmitted to the UAV's onboard processor; the deployed voltage sensor collects the operating voltage of the UAV battery in real time at a sampling frequency of 1Hz, and the voltage signal is transmitted to the onboard processor; the deployed battery management system collects the remaining power data of each UAV battery in real time at a sampling frequency of 0.5Hz, and this data is directly transmitted to the onboard processor in digital signal form; the onboard processor associates and stores the operating voltage, operating current, and remaining power data received at the same time to form the energy consumption data of each UAV. For example, at a certain moment, the operating current of UAV A is 8A, the operating voltage is 24V, and the remaining power is 380Wh. These three sets of data constitute the energy consumption data of UAV A at that moment.

[0036] In this embodiment, while collecting energy consumption data, task execution progress data is also collected in real time: The deployed airborne GPS module receives satellite signals in real time at a sampling frequency of 1Hz, calculates and outputs the latitude and longitude position information of each UAV. This position information is transmitted to the airborne processor, and combined with the preset path of the collaborative task (such as the preset route of an inspection task), it can be determined whether the current position of the UAV meets the task path requirements, thus reflecting the positional progress of the task. The deployed task payload status sensor collects the working status of the UAV's task payload in real time at a sampling frequency of 0.5Hz, for example, during an inspection task. During the mission, the mission payload status sensor can collect mission execution status information such as whether the inspection camera is in shooting mode and whether a target (such as a line defect) has been identified. This information is also transmitted to the airborne processor. The airborne processor associates and stores the received location information with the mission execution status information to form the mission execution progress data of each UAV. For example, at a certain moment, the location information of UAV B is (116.3°E, 39.9°N) (on the preset inspection route) and the mission execution status information is "the inspection camera is shooting, no defect has been identified". These two sets of data constitute the mission execution progress data of UAV B at that moment.

[0037] In this embodiment, to ensure the validity of the collected data, the collected energy consumption data and task execution progress data need to be verified in real time: For energy consumption data, it is determined whether the values ​​of the operating current and operating voltage are within the normal operating range of the UAV (e.g., the normal operating current range of a certain model of UAV is 5-15A, and the normal operating voltage range is 22-26V). If they exceed the range, they are marked as abnormal data and a re-collection is triggered; For task execution progress data, it is determined whether the location information output by the airborne GPS module is valid (e.g., whether the number of satellite positioning points is ≥4, and the positioning accuracy is ≤10m). If the location information is invalid, it is marked as abnormal data. At the same time, it is determined whether the task execution status information is a preset valid status (e.g., "not executed", "in execution", "execution completed"). If it is an invalid status, it is also marked as abnormal data. Abnormal data will be stored separately and processed during subsequent data cleaning; For valid data that passes the verification, it is transmitted in real time to the ground control center of the UAV cluster through the UAV's wireless communication module (e.g., 4G module) to provide basic data for the subsequent construction of the dynamic input-output table.

[0038] In this embodiment of the application, by collecting energy consumption data and task execution progress data in real time, the problem of "separate collection of energy data and task data and poor data timeliness" in the traditional UAV swarm energy efficiency management is solved. This enables the swarm management system to simultaneously obtain the energy usage and task progress of the UAVs, providing data support for subsequent analysis of the correlation between energy consumption and task output. It also avoids the problem of being unable to accurately determine whether energy consumption matches task execution due to data separation.

[0039] S2. Based on the energy consumption data and the task execution progress data, construct a dynamic input-output table describing the relationship between energy flow and task output within the UAV cluster.

[0040] In this embodiment of the application, the dynamic input-output table is a table constructed based on the energy consumption data and task execution progress data of the UAV cluster. This table can describe the energy flow pattern among the UAVs in the cluster and the correspondence between energy input and task output, and it will be updated in real time as the task execution process progresses.

[0041] In some embodiments, constructing a dynamic input-output table describing the relationship between energy flow and task output within a drone swarm based on the energy consumption data and the task execution progress data includes:

[0042] The energy consumption data is arranged into energy input items according to the time series.

[0043] The task execution progress data is arranged into task outputs by time series.

[0044] Calculate the direct consumption coefficient among the drones based on the energy input item;

[0045] A correspondence matrix between the energy input item and the task output item is established using the direct consumption coefficient, thereby forming the dynamic input-output table.

[0046] In this embodiment, the time series is the discrete data collected in chronological order of the time of data generation; the energy input item is a data set formed by arranging the energy consumption data of each UAV in a time series, which can reflect the energy input of each UAV at different time nodes; the task output item is a data set formed by arranging the task execution progress data of each UAV in a time series, which can reflect the task output of each UAV at different time nodes.

[0047] In this embodiment of the application, the direct consumption coefficient is a parameter that quantifies the proportion of energy consumed by one UAV from another UAV in order to complete a task within the cluster. This parameter can reflect the direct energy dependence between UAVs. The correspondence matrix is ​​a matrix constructed with the direct consumption coefficient as the core element. This matrix can clearly present the numerical correspondence between energy input items and task output items, and is a core component of the dynamic input-output table.

[0048] In this embodiment of the application, the energy input item is first constructed. Taking the power inspection task performed by the drone swarm as an example, the time step of the time series is determined to be 5 minutes (based on the frequency of change of drone energy consumption and task progress in the inspection task, 5 minutes can accurately reflect the data changes without causing data redundancy due to the short time). Retrieve the energy consumption data (including operating current, operating voltage, and remaining battery power) of each UAV collected and verified in step S1 from the ground control center. Calculate the total energy consumption value of each UAV within each time step: Total energy consumption value = Operating voltage × Operating current × Time step (time step converted to seconds). For example, UAV A has an operating voltage of 24V and an operating current of 8A in the first 5 minutes, and its total energy consumption value is 24V × 8A × 300s = 57600J = 57.6kJ. Arrange the total energy consumption values ​​of each UAV in each time step in chronological order to form energy input items. For example, the energy input items of UAV A are [57.6kJ (0-5min), 60kJ (5-10min), 58kJ (10-15min)...], and the energy input items of UAV B are [54kJ (0-5min), 55kJ (5-10min), 53kJ (10-15min)...], and so on to complete the construction of energy input items for all UAVs.

[0049] In this embodiment, the construction of task output items is carried out simultaneously, also with a time step of 5 minutes, and the task execution progress data (including location information and task execution status information) of each UAV collected in step S1 is retrieved. For location information, combined with the total mileage of the preset inspection route, the inspection mileage completed by each UAV in each time step is calculated: for example, if the total mileage of the preset inspection route is 10km, UAV A flies from the starting position (116.28°E, 39.88°N) to (116.3°E, 39.9°N) in the first 5 minutes, and the flight distance is calculated to be 0.8km using GPS coordinates, that is, the inspection mileage completed in this time step is 0.8km; for task execution status information, the task results completed by each UAV in each time step are counted, such as the number of line defects identified in the inspection task. UAV A did not identify any defects in the first 5 minutes, so the number of defects is 0, and it identified 1 defect in the second 5 minutes, so the number of defects is 1. Arrange the "inspection mileage + number of defects" of each drone in chronological order within each time step to form task output items. For example, the task output items of drone A are [0.8km + 0 defects (0-5min), 0.7km + 1 defect (5-10min), 0.5km + 1 defect (10-15min)...], and the task output items of drone B are [0.9km + 0 defects (0-5min), 0.8km + 1 defect (5-10min), 0.5km + 0 defects (10-15min)...]. Complete the construction of task output items for all drones.

[0050] In this embodiment, the direct consumption coefficient between each UAV is calculated based on the energy input item, using the "energy dependency analysis method": First, the total energy consumption value of each UAV in each time step is determined (i.e., the value in the energy input item, such as UAV A's total energy consumption of 57.6kJ in 0-5min); then, the additional energy consumption of a certain UAV due to receiving data support from another UAV (such as position correction data, task coordination instructions) within that time step is calculated. This additional energy consumption is obtained by comparing the difference between "energy consumption of UAV during single-unit flight" and "energy consumption of UAV during coordinated flight". For example, UAV A's energy consumption during single-unit flight is 55kJ in 0-5min, and its energy consumption during coordinated flight is... The energy consumption was 57.6 kJ, with an additional energy consumption of 2.6 kJ. Data tracing determined that this additional energy consumption was due to receiving position correction data from UAV B, meaning UAV A consumed 2.6 kJ of energy from UAV B. Using the direct consumption coefficient calculation formula (direct consumption coefficient = energy consumed by one UAV from another UAV / total energy consumption of that UAV), the direct consumption coefficient of UAV A to UAV B is calculated as: 2.6 kJ / 57.6 kJ ≈ 0.045. Similarly, the direct consumption coefficients between all UAVs are calculated at each time step, forming a set of direct consumption coefficients. For example, within 0-5 minutes, the coefficient of UAV A to B is 0.045, and the coefficient of UAV B to A is 0.038, etc.

[0051] In this embodiment, a corresponding relationship matrix is ​​established based on the direct consumption coefficient, forming a dynamic input-output table. First, the rows and columns of the matrix are determined: the rows represent "UAV-time step" combinations (e.g., UAV A-0-5min, UAV A-5-10min, UAV B-0-5min...), and the columns contain "energy input value (kJ)," "task output value (km + number of defects)," and "direct consumption coefficient (corresponding to other UAVs)." The total energy consumption value from the previously constructed energy input item is filled into the "energy input value" column, the "inspection mileage + number of defects" from the task output item is filled into the "task output value" column, and the calculated direct consumption coefficient is filled into the "time step" column. The "Direct Consumption Coefficient" column; for example, the values ​​corresponding to the "Drone A-0-5min" row in the matrix are: energy input value 57.6kJ, task output value 0.8km+0 units, direct consumption coefficient for Drone B 0.045, direct consumption coefficient for Drone C 0.012, etc.; after filling the corresponding data of all "Drone-Time Step" combinations into the matrix, a complete dynamic input-output table is formed. This table will be updated in real time as the task is executed. After each time step of data collection and calculation is completed, the new "Drone-Time Step" row data will be added to the table.

[0052] In this embodiment, by constructing a dynamic input-output table, the problem of "inability to quantify the correlation between energy flow and task output" in traditional UAV swarm energy efficiency management is solved. The scattered energy consumption data and task execution progress data are transformed into a structured table, which clearly presents the correspondence between the energy input and task output of each UAV and the energy dependence between UAVs. This provides a calculable basic model for the subsequent dynamic identification of energy coupling strength coefficient and task contribution weight, and the generation of the optimal energy allocation instruction set, avoiding the problem of unclear energy efficiency optimization direction due to the lack of correlation between energy and task data.

[0053] S3. Based on the dynamic input-output table, and using a neural network model trained on historical data, dynamically identify and update the energy coupling strength coefficient and task contribution weight among the drones in the cluster.

[0054] In this embodiment, the neural network model is a machine learning model trained on historical data to dynamically identify the energy coupling strength coefficient and task contribution weight in a drone swarm. The model predicts parameters by learning the mapping relationship between data. The energy coupling strength coefficient is a parameter that reflects the degree of energy dependence between a drone and other drones in the drone swarm. The larger the coefficient, the stronger the energy influence or demand of the drone on other drones. The task contribution weight is a parameter that reflects the contribution of a drone to the overall collaborative task completion in the drone swarm. The larger the weight, the more significant the role of the drone in the total task output.

[0055] In some embodiments, the step of dynamically identifying and updating the energy coupling strength coefficient and task contribution weight among the drones in the cluster based on the dynamic input-output table and a neural network model trained on historical data includes:

[0056] The ideal energy coupling strength coefficient and task contribution weight corresponding to each group of data in the dynamic input-output table are fitted using the least squares method, and used as training labels for the neural network model to be trained.

[0057] The historical energy input items and task output items in the dynamic input-output table, along with their corresponding training labels, are used together as training samples to perform supervised training on the neural network model to be trained.

[0058] Input the data from the current dynamic input-output table into the trained neural network model;

[0059] The forward propagation calculation of the trained neural network model outputs updated energy coupling strength coefficients and task contribution weights.

[0060] In this embodiment, the least squares method is a mathematical optimization method that fits the ideal parameters (energy coupling strength coefficient, task contribution weight) corresponding to the data in the dynamic input-output table by minimizing the sum of squares between the predicted value and the true value. The training labels are the ideal energy coupling strength coefficient and task contribution weight obtained by fitting the least squares method, which are used to guide the supervised training of the neural network model.

[0061] In this embodiment of the application, the historical energy input item is a set of energy consumption data of each UAV arranged in time sequence over a period of time in the dynamic input-output table; the historical task output item is a set of task execution progress data of each UAV arranged in time sequence over a period of time in the dynamic input-output table.

[0062] In this embodiment, supervised training is a neural network training method that uses historical energy input items and historical task output items as inputs and training labels as expected outputs. The model parameters are iteratively adjusted to make the predicted output close to the labels. The dynamic input-output table data at the current moment is the data of energy input items and task output items corresponding to the latest time step in the dynamic input-output table of the UAV cluster at the current task stage. Forward propagation calculation is the reasoning process of the trained neural network model, which passes the current moment data from the input layer to the hidden layer and the output layer, and calculates the final parameters (updated energy coupling strength coefficient and task contribution weight) through linear transformation and activation function.

[0063] In some embodiments, when performing supervised training on the neural network model to be trained, the loss function used is the mean square error function between the predicted value and the true value. The calculation formula is as follows:

[0064]

[0065] in, The number of training samples. and The neural network model is for the first The predicted values ​​of the energy coupling strength coefficient and task contribution weight for each sample. and For the first The true values ​​of the energy coupling strength coefficient and the task contribution weight are obtained by fitting historical data of each sample.

[0066] In this embodiment, the mean squared error function is a loss function used to measure the error between the predicted value of the neural network model and the training label (true value). The model prediction accuracy is quantified by calculating the mean of the squared error.

[0067] In this embodiment, the least squares method is first used to fit the training labels. Taking a drone swarm performing a power line inspection task (a total of 5 drones, numbered U1-U5) as an example, historical data of the past 30 inspection tasks are retrieved from the dynamic input-output table (each task contains 3 time steps, a total of 90 sets of sample data, each set of samples contains a data pair of "historical energy input item - historical task output item"). For each set of samples, the ideal energy coupling strength coefficient is defined as "the normalized value of the sum of the direct consumption coefficients of the drone in this sample and all other drones" (the direct consumption coefficient comes from the calculation result of step S2). For example, the set of direct consumption coefficients of a certain sample U1 is [0.045 (for U2), 0.012 (for U3), 0.008 (for U4), 0.005 (for U5)], and the sum of the coefficients is 0.07. The maximum sum of the coefficients of all drones in this sample is 0.1, so the ideal energy coupling strength coefficient of sample U1 is 0.07. The ideal energy coupling strength coefficient is 0.07 / 0.1=0.7. The ideal task contribution weight is defined as "the proportion of the UAV task output value in this sample to the total task output value of the cluster" (task output value = 0.6 × inspection mileage + 0.4 × number of defects). For example, the task output value of sample U1 is (0.6 × 1km + 0.4 × 0.5) = 0.8, and the total task output value of the cluster is 3.2. Therefore, the ideal task contribution weight of sample U1 is 0.8 / 3.2=0.25. The least squares method is used to minimize "the sum of squares of the ideal parameters of all samples and the initial predicted parameters" to finally determine the ideal energy coupling strength coefficient and task contribution weight corresponding to 90 sets of samples, which are used as training labels for the neural network model.

[0068] In this embodiment, the neural network model is then constructed and trained. First, the model structure is determined: the input layer has 2 nodes (corresponding to the total energy consumption value of historical energy input items and the task output value of historical task output items, respectively); the hidden layer has 2 layers, with the first layer having 12 nodes (using the ReLU activation function to extract non-linear features of the data) and the second layer having 8 nodes (using the ReLU activation function to further optimize feature mapping); the output layer has 2 nodes (linear activation functions, outputting the energy coupling strength coefficient and task contribution weight, with values ​​ranging from 0 to 1). Then, the training samples are divided: 90 sets of "historical energy input items - historical task output items - training labels" data are divided into a training set (63 sets) and a validation set (27 sets) in a 7:3 ratio. The training set is used for model parameter learning, and the validation set is used to evaluate the model's generalization ability. During training, the mean squared error function is used to calculate the loss. For example, in a certain training round, the number of training samples is 63. The predicted value of the energy coupling strength coefficient of a certain sample is 0.68, the true value is 0.7 (squared difference 0.0004), the predicted value of the task contribution weight is 0.24, the true value is 0.25 (squared difference 0.0001), and the loss term for this sample is 0.0005. The total loss term for 63 samples is 0.063, so the loss value is (1 / 63) × 0.063 = 0.001. The Adam optimizer (learning rate 0.001) is used for iterative training. After each training round, the loss is evaluated using a validation set. When the iteration reaches 150 rounds, the loss on the validation set stabilizes below 0.001, training stops, and the trained neural network model is obtained.

[0069] In this embodiment of the application, the input data for the current moment is then prepared. From the dynamic input-output table updated in step S2, the dynamic data of each UAV at the current time step (such as 10-15 min of this inspection task) is extracted: taking U1 as an example, the total energy consumption value of the energy input item at the current moment is 58kJ, and the task output value of the task output item is (0.6×0.5km+0.4×1)=0.7. These two data are combined to form the input vector (58kJ, 0.7); similarly, the energy input value and task output value of U2-U5 at the current moment are extracted to form their respective input vectors.

[0070] In this embodiment, the output update parameters are finally calculated through forward propagation. The input vector of U1 (58kJ, 0.7) is input into the trained neural network model: the input vector first enters the input layer, undergoes a linear transformation (input value × input layer weight + bias) and is then passed to the first hidden layer. It is processed by the ReLU activation function (positive values ​​are retained, negative values ​​are set to 0) to obtain 12 feature values; these feature values ​​are then passed to the second hidden layer, undergo another linear transformation and ReLU activation function processing to obtain 8 optimized feature values; finally, the 8... Each feature value is passed to the output layer. After a linear transformation (feature value × output layer weight + bias), the updated energy coupling strength coefficient of output U1 is 0.72 and the task contribution weight is 0.26. Using the same process, the updated energy coupling strength coefficients of U2-U5 (U2: 0.68, U3: 0.55, U4: 0.52, U5: 0.48) and task contribution weights (U2: 0.24, U3: 0.20, U4: 0.18, U5: 0.16) are calculated respectively, completing the dynamic update of all UAV parameters in the cluster.

[0071] In this embodiment, by dynamically identifying and updating the energy coupling strength coefficient and task contribution weight, the problem of "static parameter settings that cannot adapt to environmental changes (such as wind speed fluctuations and task priority adjustments)" in traditional UAV swarm energy efficiency management is solved. The neural network model learns data patterns based on historical data and outputs parameters in real time by combining dynamic data at the current moment. This ensures that the parameters accurately reflect the current energy dependence and task contribution of the swarm, providing precise optimization parameters for the subsequent generation of the optimal energy allocation instruction set, and avoiding the problem of poor energy efficiency optimization due to parameter lag.

[0072] S4. Using the energy coupling strength coefficient and the task contribution weight as optimization parameters, and taking the maximization of overall cluster energy efficiency as the objective function, generate the optimal energy allocation instruction set.

[0073] In this embodiment, the constraints are based on the limiting rules determined by the dynamic input-output table, used to limit the range of energy allocation and ensure that the energy allocation conforms to the energy flow law within the cluster and meets the task output requirements. The optimization parameters are a collective term for the energy coupling strength coefficient and the task contribution weight. These two parameters are used to quantify the energy dependence relationship and task contribution degree among UAVs and are the core basis for optimizing energy allocation. The objective function is a mathematical expression used to measure the overall energy efficiency of the cluster. In this application, the goal is to maximize the overall energy efficiency of the cluster, specifically to minimize the energy consumption per unit task output. The optimal energy allocation instruction set is an energy allocation scheme obtained by solving the objective function. This scheme specifies the energy usage quota of each UAV in the form of instructions to guide the UAVs to adjust their operating status.

[0074] In some embodiments, generating an optimal energy allocation instruction set, using the energy coupling strength coefficient and the task contribution weight as optimization parameters and maximizing the overall energy efficiency of the cluster as the objective function, includes:

[0075] An energy flow constraint equation is constructed based on the energy coupling strength coefficient, and a task output constraint equation is constructed based on the task contribution weight.

[0076] Construct an optimization objective function with the goal of minimizing energy consumption per unit task output. The optimization objective function is the ratio of total energy consumption of the cluster to total task output of the cluster.

[0077] The optimal solution of the objective function is obtained by using a linear programming algorithm under the constraints of the energy flow equation and the task output, thereby generating the optimal energy allocation instruction set.

[0078] In this embodiment, the energy flow constraint equation is a mathematical equation constructed based on the energy coupling strength coefficient, used to limit unreasonable energy flow within the cluster and avoid affecting the normal operation of associated drones due to insufficient energy allocation of a certain drone; the task output constraint equation is a mathematical equation constructed based on the task contribution weight, used to ensure that the energy allocation can support the cluster to complete the minimum task objective and avoid task failure due to insufficient total energy input.

[0079] In this application embodiment, the energy consumption per unit task output is the ratio of the total energy consumption of the cluster to the total task output of the cluster. The smaller the ratio, the higher the overall energy efficiency of the cluster. The linear programming algorithm is a mathematical algorithm for solving the optimal solution of a linear objective function under linear constraints. In this application, it is used to find an energy allocation scheme that minimizes the energy consumption per unit task output under the constraints of energy flow and task output.

[0080] In some embodiments, the constraint set of the optimization objective function is constructed based on the energy flow constraint equation and the task output constraint equation, and its mathematical expression is as follows:

[0081]

[0082] in, This indicates the total energy consumption of the cluster. Indicates the total task output of the cluster. For the number of drones, For the first The energy coupling strength coefficient of the drone For the first The mission contribution weight corresponding to the energy coupling strength coefficient of the UAV. Assign the part to be optimized to the first The energy value of the drone and The first The lower and upper thresholds for energy allocation of drones.

[0083] In this embodiment, the constraint set is a collection of energy flow constraint equations, task output constraint equations, and upper and lower limits of single UAV energy allocation constraints, used to comprehensively limit the value range of optimization variables; the upper and lower limits of energy allocation are energy allocation boundary values ​​determined based on the characteristics of the UAV battery, with the lower limit being 30% of the UAV's remaining power (to avoid over-discharge) and the upper limit being the UAV's current remaining power (to avoid energy waste).

[0084] In this embodiment of the application, a constraint set is first constructed. Taking the execution of a power inspection task by a drone swarm (a total of 5 drones, numbered U1-U5, with energy coupling strength coefficients updated in step S3 as k1=0.72, k2=0.68, k3=0.55, k4=0.52, k5=0.48 respectively; task contribution weights as w1=0.26, w2=0.24, w3=0.20, w4=0.18, w5=0.16 respectively) as an example, an energy flow constraint equation is first constructed: According to the mathematical expression of the constraint set, the energy flow constraint equation means that the sum of the products of the energy allocation value of all drones and the corresponding energy coupling strength coefficients shall not exceed the total available energy of the swarm. The total available energy of the cluster is the sum of the remaining power of each drone collected in step S1, i.e., U1 has 380Wh remaining, U2 has 400Wh remaining, U3 has 390Wh remaining, U4 has 370Wh remaining, and U5 has 360Wh remaining, with a total available energy of 380+400+390+370+360=1900Wh. The energy allocation values ​​of each drone to be optimized are E1, E2, E3, E4, and E5. Substituting them into the energy coupling strength coefficient, the energy flow constraint equation is 0.72E1+0.68E2+0.55E3+0.52E4+0.48E5≤1900Wh.

[0085] In this embodiment, a task output constraint equation is then constructed. This equation states that the sum of the products of the energy allocation values ​​of all UAVs and their corresponding task contribution weights must not be less than the minimum energy contribution value required to complete the current task. The remaining inspection mileage for the current task is 8km. Based on historical data, the minimum energy contribution value required to complete a 1km inspection task is 45Wh. Therefore, the total minimum energy contribution value is 8 × 45 = 360Wh. Substituting the task contribution weights, the task output constraint equation is 0.26E1 + 0.24E2 + 0.20E3 + 0.18E4 + 0.16E5 ≥ 360Wh.

[0086] In this embodiment, the upper and lower limits of single UAV energy allocation are then determined: based on battery protection requirements, the lower limit threshold is 30% of the remaining power, and the upper limit threshold is the current remaining power. Therefore, the upper and lower limits of U1 are E1∈[114Wh (380×30%), 380Wh], U2 is E2∈[120Wh (400×30%), 400Wh], U3 is E3∈[117Wh (390×30%), 390Wh], U4 is E4∈[111Wh (370×30%), 370Wh], and U5 is E5∈[108Wh (360×30%), 360Wh]. The energy flow constraint equation, the task output constraint equation, and the upper and lower limit constraints are integrated to form a complete constraint set.

[0087] In this embodiment, an optimization objective function is then constructed: maximizing the overall energy efficiency of the cluster is equivalent to minimizing the energy consumption per unit task output. ,that is F needs to be minimized. Wherein, the total cluster task output... The total inspection mileage is 0.6 × total inspection mileage + 0.4 × total number of defects identified. The total inspection mileage and total number of defects identified are determined by the energy allocation value and task contribution weight of each UAV. That is, the total task output = (0.26E1 + 0.24E2 + 0.20E3 + 0.18E4 + 0.16E5) / 45 (since 45Wh corresponds to the task output of 1km inspection task). After substituting, the objective function can be simplified to F = (E1 + E2 + E3 + E4 + E5) / [(0.26E1 + 0.24E2 + 0.20E3 + 0.18E4 + 0.16E5) / 45] = 45 × (E1 + E2 + E3 + E4 + E5) / (0.26E1 + 0.24E2 + 0.20E3 + 0.18E4 + 0.16E5). Minimizing F can achieve the minimum energy consumption per unit task output.

[0088] In this embodiment, a linear programming algorithm is finally used to solve for the optimal solution: the constraint set and objective function are input into a linear programming solver (such as the linprog function in MATLAB), and the number of iterations is set to 1000. During the solution process, the algorithm first generates an initial energy allocation scheme (such as E1=200Wh, E2=200Wh, E3=200Wh, E4=200Wh, E5=200Wh) within the feasible region defined by the constraint set, and verifies whether the scheme satisfies the constraints: the energy flow constraint is 0.72×200+0.68×200+0.55×200+0.52×200+0.48×200=590≤1900 (satisfied), and the task output constraint is 0.26×200+0.24×200+0.20×200+0.18×200+0.16×200=208<360 (not satisfied). The algorithm then adjusts the energy allocation values. After multiple iterations, when E1=320Wh, E2=350Wh, E3=310Wh, E4=290Wh, and E5=280Wh, the energy flow constraint is 0.72×320+0.68×350+0.55×310+0.52×290+0.48×280=924.1≤1900 (satisfied), and the task output constraint is 0.26×320+0.24×350+0.20×310+0.18×290+0.16×280326.2. If the target of 360Wh is still not met, the energy allocation is adjusted further to E1=380Wh (upper limit), E2=400Wh (upper limit), E3=390Wh (upper limit), E4=370Wh (upper limit), and E5=360Wh (upper limit). At this point, the task output constraint is 0.26×380+0.24×400+0.20×390+0.18×370+0.16×360=397≥360 (satisfied). The objective function F=1900 / 397≈4.78Wh / unit task output is minimized. Based on this optimal energy allocation value, an optimal energy allocation instruction set is generated, such as U1's "Energy allocation 380Wh, matching propulsion power 1100W" and U2's "Energy allocation 400Wh, matching propulsion power 1150W".

[0089] In this embodiment, by constructing a constraint set and an objective function and solving for the optimal solution, the problem of "energy allocation focusing only on the needs of individual drones and ignoring the global correlation of the cluster" in traditional drone swarm energy efficiency management is solved. The constraint set ensures that energy allocation conforms to physical laws and task requirements, the objective function clarifies the direction of energy efficiency optimization, the linear programming algorithm achieves globally optimal allocation, and the final generated instruction set can minimize the energy consumption per unit task output while completing the task, directly promoting the overall energy efficiency improvement of the cluster.

[0090] In this embodiment, the dynamic input-output table constructed in step S2 is the basis for constructing the constraints. The energy flow pattern and task output data in the dynamic input-output table determine the specific form of the energy flow constraint equation and the task output constraint equation. The energy coupling strength coefficient and task contribution weight updated in step S3 are the core optimization parameters, which directly participate in the construction of the constraint equation and the objective function. The accuracy of the parameters determines the reliability of the optimization results.

[0091] S5. Distribute the optimal energy allocation instruction set to each UAV in the cluster, and adjust their flight attitude, propulsion power or task execution sequence to execute the cooperative task.

[0092] In some embodiments, distributing the optimal energy allocation instruction set to each UAV in the cluster and adjusting its flight attitude, propulsion power, or task execution sequence to perform the cooperative task includes:

[0093] Flight attitude adjustment commands, including speed and altitude commands, are distributed to each UAV through the cluster communication link;

[0094] The propulsion power adjustment command, which includes motor power control commands, is distributed to each UAV through the cluster communication link.

[0095] The task execution sequence adjustment instruction, which includes the task priority adjustment instruction, is distributed to each UAV through the cluster communication link.

[0096] In this embodiment, the cluster communication link is a communication channel used for transmitting instructions and data between the UAV cluster ground control center and each UAV. It features low latency and high reliability and can support multiple UAVs receiving instructions simultaneously.

[0097] In this embodiment, the flight attitude adjustment command is a subset of commands including speed and altitude commands, used to adjust the flight state of the UAV and reduce unnecessary energy consumption by optimizing the flight attitude; the speed command is a component of the flight attitude adjustment command and is used to specify the target flight speed of the UAV, which needs to match the power requirement corresponding to the optimal energy allocation value; the altitude command is a component of the flight attitude adjustment command and is used to specify the target flight altitude of the UAV, which needs to avoid obstacles to reduce the additional energy consumption caused by path correction.

[0098] In this embodiment, the propulsion power adjustment command is a subset of commands that includes the motor power control command. It is used to adjust the output power of the UAV motor to ensure that the motor power matches the optimal energy allocation value. The motor power control command is the core of the propulsion power adjustment command. It is used to specify the target output power of the UAV motor and directly determines the energy consumption rate of the UAV.

[0099] In this embodiment, the task execution sequence adjustment instruction is a subset of instructions that includes the task priority adjustment instruction. It is used to adjust the order in which the UAV performs tasks, thereby improving task output efficiency by prioritizing high-value tasks. The task priority adjustment instruction is the core content of the task execution sequence adjustment instruction, which is used to clarify the priority level of different task stages and ensure that energy is used preferentially for critical task stages.

[0100] In this embodiment, a cluster communication link is first established. Taking a drone cluster performing a power line inspection task (a total of 5 drones, numbered U1-U5) as an example, the ground control center deploys a 4G private network base station (coverage radius ≥5km, supporting simultaneous connection of ≥10 drones). Each drone is equipped with a 4G communication module (supporting TD-LTE protocol), and a two-way communication link is established between the base station and the ground control center. The communication link uses the AES-256 encryption algorithm to ensure command security. The link latency is tested to be ≤100ms, meeting the real-time control requirements and providing a stable channel for subsequent command distribution.

[0101] In this embodiment, flight attitude adjustment commands are then distributed, and the energy allocation values ​​of each UAV are extracted from the optimal energy allocation command set: U1's energy allocation value is 380Wh, corresponding to a remaining mission duration of approximately 20 minutes (1200 seconds), and the average power is 380Wh × 3600s / h ÷ 1200s ≈ 1140W. Based on the formula relating UAV power to flight speed (power = 0.5 × air density × frontal area × speed³ × drag coefficient, air density is 1.225 kg / m³, frontal area is 0.2 m², drag coefficient is 0.8), the target flight speed is calculated to be 6.8 m / s by substituting the power of 1140W; considering the distribution of obstacles around the inspection route (maximum obstacle 60m), the target flight altitude is set to 80m (20m above obstacles to avoid path correction). The ground control center sends a flight attitude adjustment command of "speed 6.8m / s, altitude 80m" to U1 via the trunking communication link; similarly, U2 has an energy allocation value of 400Wh, calculates the target speed of 7.0m / s and altitude of 85m, and sends the corresponding command; U3-U5 calculate the target speed and altitude according to their respective energy allocation values ​​and complete the distribution of flight attitude adjustment commands.

[0102] In this embodiment, a propulsion power adjustment command is then distributed. The target motor power is calculated based on the energy allocation value and remaining mission duration of each UAV: ​​U1 has a remaining mission duration of 20 minutes, an energy allocation value of 380Wh, and a target motor power of 380Wh ÷ (20 ÷ 60) h ​​= 1140W. The ground control center sends a propulsion power adjustment command of "motor power 1140W" to U1 through the cluster communication link; U2 has a remaining mission duration of 19 minutes, an energy allocation value of 400Wh, and a target motor power of 400Wh ÷ (19 ÷ 60) h ​​≈ 1263W. The corresponding command is then issued; U3-U5 calculate the target motor power according to the same logic and distribute the command to ensure that the motor power of each UAV matches the energy allocation value, avoiding excessive power leading to energy waste or excessive power affecting mission progress.

[0103] In this embodiment, a task execution sequence adjustment instruction is then distributed. Based on the task execution progress data collected in step S1 (U1 has completed 2km of inspection, and 3km of the remaining 8km is a defect-intensive section), a task priority adjustment instruction is issued to U1, which is "prioritize the inspection of the defect-intensive section, and then execute the regular inspection section", thus forming a task execution sequence adjustment instruction. U2 has completed 2.2km of inspection, and 2.5km of the remaining 7.8km is a collaborative inspection section (which needs to cooperate with U1). A task execution sequence adjustment instruction is issued to U2, which is "prioritize cooperating with U1 to complete the collaborative inspection section, and then execute the independent inspection section". U3-U5 are issued execution sequence instructions containing task priorities according to task relevance and defect distribution, thereby improving task output efficiency.

[0104] In some embodiments, during the execution of the collaborative task by the cluster, a new round of energy consumption data and task execution progress data are collected in real time.

[0105] Based on the new round of energy consumption data and task execution progress data, update the dynamic input-output table;

[0106] Based on the updated dynamic input-output table, the energy coupling strength coefficient and the task contribution weight are dynamically identified and updated again through the trained neural network model.

[0107] Using the updated dynamic input-output table, the energy coupling strength coefficient, and the task contribution weight, a new optimal energy allocation instruction set is generated to achieve continuous closed-loop optimization of cluster energy efficiency.

[0108] In this embodiment of the application, continuous closed-loop optimization is a mechanism that continuously adjusts the energy allocation strategy of the UAV swarm through a cyclical process of "real-time acquisition of new data - model update - re-optimization - execution of new instructions" to ensure that the swarm energy efficiency is always in the optimal state.

[0109] In this embodiment, continuous closed-loop optimization is finally achieved. During the execution of collaborative tasks in the cluster, a new round of data is collected in real time at 15-minute intervals: through the energy monitoring sensor deployed in step S1, the following data are collected: U1's new round of operating current 7.5A, voltage 24V, and remaining power 320Wh (energy consumption data); U1's location (116.35°E, 39.93°N) and the number of defects identified (4) (task execution progress data); U2-U5 are collected in the same way. Based on this data, the dynamic input-output table is updated according to the construction logic in step S2. The new round of energy consumption data of U1 is added to the energy input item according to the time series, and the task execution progress data is added to the task output item. The direct consumption coefficient is recalculated (as the frequency of collaboration between U1 and U2 increases, the direct consumption coefficient is updated from 0.045 to 0.052). The updated dynamic input-output table is input into the neural network model trained in step S3. The energy coupling strength coefficient (U1 updated from 0.72 to 0.75) and task contribution weight (U1 updated from 0.26 to 0.28) are calculated and updated through forward propagation. Based on the updated dynamic input-output table, coefficients, and weights, a new optimal energy allocation instruction set (U1 new energy allocation value 320Wh) is generated according to the optimization logic of step S4. This set is then redistributed to each UAV to adjust its status, forming a closed loop and continuously optimizing the cluster's energy efficiency.

[0110] In this embodiment, the problems of "optimization instructions cannot be implemented" and "energy efficiency optimization lacks dynamic adjustment" in traditional UAV swarm energy efficiency management are solved by instruction distribution and continuous closed-loop optimization. Instruction distribution transforms the optimal energy allocation strategy into an operation that the UAV can execute, ensuring that the optimization plan is implemented. Continuous closed-loop optimization can adapt to environmental changes (such as wind speed fluctuations and task priority adjustments) in real time during the task process, avoid the energy efficiency decline caused by static optimization, and always maintain the overall energy efficiency of the swarm at a high level.

[0111] like Figure 2 The diagram shown is a functional block diagram of an energy efficiency optimization control system for unmanned aerial vehicle (UAV) swarms based on input-output analysis, provided in an embodiment of this application.

[0112] The UAV swarm energy efficiency optimization control system 100 based on input-output analysis described in this application can be installed in an electronic device. Depending on the functions implemented, the UAV swarm energy efficiency optimization control system 100 may include a data acquisition module 101, a dynamic input-output table construction module 102, a coupling coefficient and contribution weight identification module 103, an optimal energy allocation instruction generation module 104, and an instruction distribution and UAV adjustment module 105. The module described in this application can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0113] In this embodiment, the functions of each module / unit are as follows:

[0114] Data acquisition module 101 is used to collect energy consumption data and task execution progress data of each UAV in real time when performing collaborative tasks by energy monitoring sensors deployed on each UAV in the cluster.

[0115] The dynamic input-output table construction module 102 is used to construct a dynamic input-output table describing the relationship between energy flow and task output within the UAV cluster based on the energy consumption data and the task execution progress data.

[0116] The coupling coefficient and contribution weight identification module 103 is used to dynamically identify and update the energy coupling strength coefficient and task contribution weight between each UAV in the cluster based on the dynamic input-output table and a neural network model trained on historical data.

[0117] The optimal energy allocation instruction generation module 104 is used to generate an optimal energy allocation instruction set with the energy coupling strength coefficient and the task contribution weight as optimization parameters and the overall energy efficiency of the cluster as the objective function.

[0118] The instruction distribution and UAV adjustment module 105 is used to distribute the optimal energy allocation instruction set to each UAV in the cluster, and adjust its flight attitude, propulsion power or task execution sequence to perform the cooperative task.

[0119] In the embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0120] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0122] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application.

[0123] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application.

Claims

1. A method for optimizing the energy efficiency control of UAV swarms based on input-output analysis, characterized in that, The method includes: By deploying energy monitoring sensors on each drone in the cluster, energy consumption data and task execution progress data of each drone are collected in real time when performing collaborative tasks; Based on the energy consumption data and the task execution progress data, a dynamic input-output table describing the relationship between energy flow and task output within the UAV cluster is constructed, including: arranging the energy consumption data into energy input items by time series; arranging the task execution progress data into task output items by time series; calculating the direct consumption coefficient between each UAV based on the energy input items; and establishing a correspondence matrix between the energy input items and the task output items using the direct consumption coefficient, thereby forming the dynamic input-output table. Based on the dynamic input-output table, a neural network model trained on historical data is used to dynamically identify and update the energy coupling strength coefficient and task contribution weight among the drones in the cluster. Using the energy coupling strength coefficient and the task contribution weight as optimization parameters, and taking the maximization of overall cluster energy efficiency as the objective function, an optimal energy allocation instruction set is generated. The optimal energy allocation instruction set is distributed to each UAV in the cluster to adjust its flight attitude, propulsion power, or mission execution sequence in order to perform the cooperative mission.

2. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 1, characterized in that, The energy monitoring sensors deployed on each drone in the cluster collect real-time energy consumption data and task execution progress data of each drone during collaborative tasks, including: The operating current and operating voltage of each UAV are collected in real time through the current sensor and voltage sensor in the energy monitoring sensor. The battery management system in the energy monitoring sensor collects the remaining power data of each drone, and uses the operating voltage, operating current and remaining power data as the energy consumption data of the drone. The location information and mission execution status information of each UAV are collected by the airborne GPS module and mission payload status sensor in the energy monitoring sensor as mission execution progress data.

3. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 1, characterized in that, The step of dynamically identifying and updating the energy coupling strength coefficient and task contribution weight among the drones in the cluster based on the dynamic input-output table and a neural network model trained on historical data includes: The ideal energy coupling strength coefficient and task contribution weight corresponding to each group of data in the dynamic input-output table are fitted using the least squares method, and used as training labels for the neural network model to be trained. The historical energy input items and task output items in the dynamic input-output table, along with their corresponding training labels, are used together as training samples to perform supervised training on the neural network model to be trained. Input the data from the current dynamic input-output table into the trained neural network model; The forward propagation calculation of the trained neural network model outputs updated energy coupling strength coefficients and task contribution weights.

4. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 1, characterized in that, The process of generating an optimal energy allocation instruction set, using the energy coupling strength coefficient and the task contribution weight as optimization parameters and maximizing the overall energy efficiency of the cluster as the objective function, includes: An energy flow constraint equation is constructed based on the energy coupling strength coefficient, and a task output constraint equation is constructed based on the task contribution weight. An optimization objective function is constructed with the goal of minimizing energy consumption per unit of task output. The optimization objective function is the ratio of total energy consumption of the cluster to total task output of the cluster. The optimal solution of the objective function is obtained by using a linear programming algorithm under the constraints of the energy flow equation and the task output, thereby generating the optimal energy allocation instruction set.

5. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 1, characterized in that, The step of distributing the optimal energy allocation instruction set to each UAV in the cluster and adjusting its flight attitude, propulsion power, or mission execution sequence to execute the cooperative task includes: Flight attitude adjustment commands, including speed and altitude commands, are distributed to each UAV through the cluster communication link; The propulsion power adjustment command, which includes motor power control commands, is distributed to each UAV through the cluster communication link. The task execution sequence adjustment instruction, which includes the task priority adjustment instruction, is distributed to each UAV through the cluster communication link.

6. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 3, characterized in that, When performing supervised training on the neural network model to be trained, the loss function used is the mean squared error function between the predicted value and the true value. The calculation formula is as follows: in, The number of training samples. and These are the neural network models for the first... The predicted values ​​of the energy coupling strength coefficient and task contribution weight for each sample. and For the first The true values ​​of the energy coupling strength coefficient and the task contribution weight are obtained by fitting historical data of each sample.

7. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 4, characterized in that, The constraint set for the optimization objective function is constructed based on the energy flow constraint equation and the task output constraint equation, and its mathematical expression is as follows: in, This indicates the total energy consumption of the cluster. Indicates the total task output of the cluster. For the number of drones, For the first The energy coupling strength coefficient of the drone For the first The mission contribution weight corresponding to the energy coupling strength coefficient of the UAV. Assign the part to be optimized to the first The energy value of the drone and The first The lower and upper thresholds for energy allocation of drones.

8. The method for optimizing the energy efficiency control of UAV swarms based on input-output analysis as described in claim 3, characterized in that, The method further includes: During the execution of the collaborative task in the cluster, a new round of energy consumption data and task execution progress data are collected in real time. Based on the new round of energy consumption data and task execution progress data, update the dynamic input-output table; Based on the updated dynamic input-output table, the energy coupling strength coefficient and the task contribution weight are dynamically identified and updated again through the trained neural network model. Using the updated dynamic input-output table, the energy coupling strength coefficient, and the task contribution weight, a new optimal energy allocation instruction set is generated to achieve continuous closed-loop optimization of cluster energy efficiency.

9. An energy efficiency optimization control system for unmanned aerial vehicle (UAV) swarms based on input-output analysis, characterized in that, The system includes: The data acquisition module is used to collect energy consumption data and task execution progress data of each drone in real time when performing collaborative tasks by using energy monitoring sensors deployed on each drone in the cluster. The dynamic input-output table construction module is used to construct a dynamic input-output table describing the relationship between energy flow and task output within a drone cluster based on the energy consumption data and the task execution progress data. This includes: arranging the energy consumption data into energy input items by time series; arranging the task execution progress data into task output items by time series; calculating the direct consumption coefficient between each drone based on the energy input items; and establishing a correspondence matrix between the energy input items and the task output items using the direct consumption coefficient, thereby forming the dynamic input-output table. The coupling coefficient and contribution weight identification module is used to dynamically identify and update the energy coupling strength coefficient and task contribution weight between each UAV in the cluster based on the dynamic input-output table and a neural network model trained on historical data. The optimal energy allocation instruction generation module is used to generate an optimal energy allocation instruction set with the energy coupling strength coefficient and the task contribution weight as optimization parameters and the overall energy efficiency of the cluster as the objective function. The instruction distribution and UAV adjustment module is used to distribute the optimal energy allocation instruction set to each UAV in the cluster, and adjust its flight attitude, propulsion power or task execution sequence to perform the cooperative task.

Images