Charging scheduling method and system for autonomous tracking of unmanned aerial vehicle

Abstract

The invention relates to the technical field of unmanned aerial vehicle energy management, and discloses an unmanned aerial vehicle autonomous tracking charging scheduling method and system, and the method comprises the steps: carrying out the environment feature perception of an operation region, and building a multi-mode sensor monitoring network; collecting operation parameters of the unmanned aerial vehicle and generating parameter dynamic characteristic data; constructing a parameter-demand response relation model, and analyzing electric energy consumption characteristics and charging demands; identifying a charging influence area based on the endurance guarantee dynamic characteristics, and generating demand effect coupling data; and dynamically planning a charging path by using a machine learning model, and optimizing operation parameters. The system comprises an environment sensing module, a state acquisition module, a scheduling prediction module, an influence evaluation module and a control feedback module. According to the invention, the dynamic and intelligent charging scheduling of the unmanned aerial vehicle is realized, the cruising ability and the operation efficiency are improved, and the method is suitable for the autonomous tracking task of the unmanned aerial vehicle in a complex environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle energy management, in particular to a method and system for autonomous tracking and charging scheduling of unmanned aerial vehicles. BACKGROUND

[0002] In the current era of rapid technological development, unmanned aerial vehicles (UAVs) have been widely and deeply applied in many fields due to their characteristics such as maneuverability, flexibility, and the ability to operate in complex areas. In the field of agriculture, UAV plant protection enables precise pesticide spraying, improving work efficiency and crop yield. In the logistics industry, UAV delivery can break through traffic restrictions and provide fast material delivery services for remote areas. In power inspection, UAVs can conduct comprehensive detection of high-voltage lines to promptly identify potential fault hazards. In disaster monitoring scenarios, UAVs can quickly reach disaster areas to provide real-time and accurate disaster information for rescue decision-making. However, the problem of insufficient battery endurance of UAVs is like a Damocles sword, severely restricting the full play of their functions and the further expansion of application scenarios.

[0003] Current common UAV charging scheduling solutions on the market have many drawbacks that need to be addressed. In terms of environmental perception, most methods rely on a single type of sensor, such as using only GPS positioning and simple height sensors, which provides very limited environmental information. When performing inspection tasks in mountainous areas, UAVs are not only prone to collision accidents due to improper path planning, but also have difficulty in reasonably arranging charging nodes in complex terrain, resulting in low charging efficiency and even forced task interruption due to power depletion. In terms of charging demand assessment and path planning, traditional solutions mostly use static and empirical strategies. For example, in logistics delivery, the load weight, flight route, and environmental conditions of UAVs vary significantly in different delivery tasks, and their power consumption characteristics also change accordingly. However, traditional methods cannot capture these dynamic changes in real time, cannot accurately assess charging demand, and thus cannot plan energy-saving and efficient charging paths, resulting in waste of time and energy.

[0004] From the system level, the existing charging scheduling system has obvious short board in anti-interference ability and data processing ability. In the complex urban electromagnetic environment area, the sensor is easy to be disturbed, and the accuracy and reliability of the collected data are greatly discounted, so that the subsequent charging scheduling decision lacks scientific basis. At the same time, in the face of the massive data generated in the process of unmanned aerial vehicle operation, the traditional data processing technology is difficult to dig out the potential law behind the data, and it is difficult to establish the precise parameter and charging demand response relationship model, and it is difficult to realize the fine management and optimization scheduling of the charging process of unmanned aerial vehicle. Therefore, the research and development of an unmanned aerial vehicle autonomous tracking charging scheduling method and system which can comprehensively perceive the working environment, accurately evaluate the charging demand, efficiently plan the charging path, and has strong anti-interference and data processing ability, has become the key to promote the sustainable development of unmanned aerial vehicle industry. SUMMARY

[0005] The purpose of the present application is to provide an unmanned aerial vehicle autonomous tracking charging scheduling method and system to solve the problems raised in the above background.

[0006] To achieve the above purpose, the present application provides the following technical scheme: an unmanned aerial vehicle autonomous tracking charging scheduling method, the method comprising:

[0007] Step 1: perceive the environmental characteristics of the working area, obtain the spatial topology data; determine the key charging nodes according to the spatial topology data, and establish a multi-mode sensor monitoring network;

[0008] Step 2: anti-interference enhancement processing is performed on the multi-mode sensor monitoring network, and real-time collection of unmanned aerial vehicle operation parameters is performed to obtain real-time operation state data; the real-time operation state data is processed based on the time sequence evolution of the endurance mileage, the moving speed and the remaining power to generate parameter dynamic characteristic data;

[0009] Step 3: establish a parameter-demand response relationship model according to the parameter dynamic characteristic data and the spatial topology data; analyze the power consumption characteristics according to the parameter-demand response relationship model, and evaluate the charging demand related to the distance to generate power scheduling dynamic data;

[0010] Step 4: perform unmanned aerial vehicle endurance correlation processing according to the power scheduling dynamic data and the real-time operation state data to generate endurance guarantee dynamic characteristic data; identify the charging influence area based on the endurance guarantee dynamic characteristic data to generate node influence range data; perform demand mode coupling based on the moving speed, the remaining power and the scheduling efficiency according to the node influence range data to generate demand action coupling data;

[0011] Step 5: build a machine learning scheduling model according to the coupling data, and use the machine learning scheduling model to dynamically plan the charging path to generate path planning data; based on the path planning data, real-time optimization of operating parameters is performed to obtain unmanned aerial vehicle tracking control feedback data.

[0012] Preferably, step 1 comprises the following steps:

[0013] Step 11: multi-view remote sensing surveying and mapping of the work area is performed to obtain environmental spatial sample data;

[0014] Step 12: the type, distribution height and coverage range of the terrain of the environmental spatial sample data are identified to obtain spatial structure data;

[0015] Step 13: environmental characteristic analysis based on obstacle density, path passability and light intensity is performed on the environmental spatial sample data according to the spatial structure data to obtain regional environmental characteristic data;

[0016] Step 14: electromagnetic environment feature analysis is performed according to the regional environmental characteristic data to obtain electromagnetic environment feature data, wherein the electromagnetic environment feature analysis includes signal strength monitoring around the work area, interference source positioning of the environmental spatial sample data, and communication quality analysis based on signal strength and interference source;

[0017] Step 15: the electromagnetic environment feature data, the regional environmental characteristic data and the spatial structure data are merged into spatial topology data;

[0018] Step 16: key charging nodes are determined according to the spatial topology data, and a multi-mode sensor monitoring network is established.

[0019] Preferably, step 16 comprises the following steps:

[0020] Step 161: path mutation point data is obtained by performing mutation analysis on the path connection features according to the spatial topology data, wherein the mutation analysis specifically identifies positions where the terrain type and passability change significantly;

[0021] Step 162: load sensitive area data is obtained by performing sensitivity analysis on the regional load state according to the spatial topology data to identify areas with large changes in power demand, wherein the load sensitivity analysis includes demand intensity evaluation based on endurance mileage and remaining power, pass efficiency evaluation based on moving speed and obstacle density, and communication stability evaluation based on signal strength and interference source;

[0022] Step 163: the spatial distribution of key charging nodes is determined according to the preset work task demand and system constraint data, the path mutation point data and the load sensitive area data to obtain node arrangement data;

[0023] Step 164: Sensor type determination is made according to node arrangement data, and a multi-mode sensor monitoring network including position sensors, power sensors, speed sensors, and signal sensors is established.

[0024] Preferably, step 2 includes the following steps:

[0025] Step 21: Anti-interference enhancement processing is performed on the multi-mode sensor monitoring network based on electromagnetic shielding and vibration isolation, and real-time parameters during the operation of the UAV are collected to obtain real-time operation state data.

[0026] Step 22: The remaining power in the real-time operation state data is segmented based on time series to establish a power-time relationship curve.

[0027] Step 23: Frequency characteristics and amplitude distribution of power fluctuations are calculated based on the power-time relationship curve, and correlation analysis of power fluctuations and distance is performed to obtain power-distance evolution characteristic data.

[0028] Step 24: A relationship curve between moving speed and distance is established based on the real-time operation state data, and the speed change rate of different distance segments is calculated to obtain speed time series evolution characteristic data.

[0029] Step 25: Real-time changes in the endurance distance in the real-time operation state data are recorded, and correlation characteristic analysis of endurance distance and distance is performed to obtain endurance time series evolution characteristic data.

[0030] Step 26: Comprehensive influence analysis of parameter combinations on charging demand is performed based on the power-distance evolution characteristic data, speed time series evolution characteristic data, and endurance time series evolution characteristic data, and critical values and adaptive intervals of key parameter combinations are identified to generate time series evolution characteristic data of parameter combinations.

[0031] Step 27: A dynamic evolution model of a three-dimensional parameter space is established based on the time series evolution characteristic data, and parameter change characteristics are extracted to obtain parameter dynamic characteristic data.

[0032] Preferably, step 21 includes the following steps:

[0033] Anti-interference enhancement processing is performed on the multi-mode sensor monitoring network based on electromagnetic shielding and vibration isolation, and real-time parameters during the operation of the UAV are collected to obtain real-time operation state data; wherein the electromagnetic shielding includes metal shell packaging of the sensors, shielding layer grounding of the signal cables, and planning of the sensor grounding points, and the vibration isolation includes setting shock absorbing supports, using flexible connecting pieces, and performing vibration resistance optimization of the sensor installation positions.

[0034] Preferably, step 27 includes the following steps:

[0035] Step 271: Construct a three-dimensional parameter space with remaining power, moving speed and endurance mileage as coordinate axes according to the time evolution characteristic data, so as to obtain parameter space coordinate data;

[0036] Step 272: Establish a parameter change trajectory based on the parameter space coordinate data, and construct a parameter motion trajectory curve through space mapping of time sequence sampling points, so as to obtain parameter trajectory data;

[0037] Step 273: Perform distance-based hierarchical processing on the parameter trajectory data, identify parameter change characteristics including change trend, change rate and mutual correlation in different distance segments, so as to obtain hierarchical feature data;

[0038] Step 274: Construct a dynamic model of the three-dimensional parameter space according to the hierarchical feature data, and establish a continuous expression of parameter change through space interpolation and numerical fitting, so as to obtain dynamic evolution model data;

[0039] Step 275: Perform feature extraction on the dynamic evolution model data based on gradient features, curvature features and speed features of parameter change, so as to obtain model feature data;

[0040] Step 276: Perform parameter correlation analysis based on the coupling relationship and mutual feedback mechanism between parameters based on the model feature data, so as to obtain parameter correlation data;

[0041] Step 277: Extract an index representing the dynamic change rule of the parameters according to the model feature data and the parameter correlation data, so as to generate parameter dynamic feature data.

[0042] Preferably, step 3 comprises the following steps:

[0043] Step 31: Establish a parameter-demand characteristic correspondence table including the mapping relationship between parameter change and charging demand in different regions according to the parameter dynamic feature data and space topology data, so as to obtain parameter response data;

[0044] Step 32: Perform data standardization and feature dimension reduction processing on the parameter response data, and establish a mathematical model of parameter-demand response, so as to obtain response model data;

[0045] Step 33: Perform machine learning training based on the response model data, and construct a nonlinear mapping relationship between parameters and demand response, so as to obtain a parameter-demand response relationship model;

[0046] Step 34: Perform electric energy consumption characteristic analysis including power consumption, path loss and charging efficiency according to the parameter-demand response relationship model, so as to obtain consumption characteristic data;

[0047] Step 35: Distance-related demand range calculation based on consumption characteristic data to obtain charging demand data, wherein the demand range calculation includes analysis of the influence of power consumption decay law and regional load intensity;

[0048] Step 36: Identification of key features and mutation points in the power scheduling process based on scheduling dynamic data to obtain power scheduling dynamic data.

[0049] Preferably, step 4 includes the following steps:

[0050] Step 41: Analysis of the influence law of different parameter combinations on the endurance guarantee effect according to the power scheduling dynamic data and real-time running state data to obtain guarantee influence data;

[0051] Step 42: Guarantee performance evaluation based on endurance maintenance time, charging completion time characteristics and path deviation characteristics of guarantee influence data to obtain performance evaluation data;

[0052] Step 43: Analysis of the evolution characteristics of guarantee effect over time of performance evaluation data to obtain endurance guarantee dynamic characteristic data;

[0053] Step 44: Numerical simulation of the charging influence area according to the endurance guarantee dynamic characteristic data, and establishment of a multi-field coupled analysis model including demand field, path field and efficiency field to obtain influence area data;

[0054] Step 45: Boundary identification and spatial partitioning of influence area data to obtain node influence range data;

[0055] Step 46: Demand mode coupling based on mobile speed, remaining power and scheduling efficiency according to node influence range data, and identification of demand field evolution law and critical state characteristics to generate demand action coupling data.

[0056] Preferably, step 5 includes the following steps:

[0057] Step 51: Feature extraction of demand action coupling data and construction of training sample set of machine learning model to obtain training sample data, wherein the training sample data includes input features and path labels;

[0058] Step 52: Construction of machine learning network structure based on hybrid architecture of decision tree algorithm and reinforcement learning according to training sample data, and model training to obtain path planning model;

[0059] Step 53: Model prediction accuracy optimization of path planning model based on cross-validation method to obtain scheduling model data;

[0060] Step 54: Dynamic programming of the charging path is performed using the scheduling model data, so as to obtain path planning data, wherein the path planning data includes lateral offset, longitudinal extension, and spatial coverage;

[0061] Step 55: Key stages and path mutation characteristics in the charging process are analyzed for the path planning data, so as to obtain path evolution data, wherein the key stages include an initial departure stage, a demand growth stage, a charging execution stage, a range recovery stage, a return adjustment stage, and a multi-machine coordination stage;

[0062] Step 56: Real-time optimization of operating parameters for each stage is performed according to the path evolution data, so as to obtain unmanned aerial vehicle tracking control feedback data.

[0063] Preferably, the present application further includes an unmanned aerial vehicle autonomous tracking charging scheduling system for performing the unmanned aerial vehicle autonomous tracking charging scheduling method as described above, and the unmanned aerial vehicle autonomous tracking charging scheduling system includes:

[0064] An environment perception module for perceiving environmental features of a work area and obtaining spatial topology data; determining key charging nodes according to the spatial topology data and establishing a multi-mode sensor monitoring network;

[0065] A state acquisition module for performing anti-interference enhancement processing on the multi-mode sensor monitoring network and real-time acquisition of unmanned aerial vehicle operating parameters, so as to obtain real-time operating state data; performing time sequence evolution processing of the real-time operating state data based on range, moving speed, and remaining power, so as to generate parameter dynamic characteristic data;

[0066] A scheduling prediction module for establishing a parameter-demand response relationship model according to the parameter dynamic characteristic data and the spatial topology data; performing electric energy consumption characteristic analysis according to the parameter-demand response relationship model and distance-related charging demand evaluation, so as to generate electric energy scheduling dynamic data;

[0067] An influence evaluation module for performing unmanned aerial vehicle range correlation processing according to the electric energy scheduling dynamic data and the real-time operating state data, so as to generate range guarantee dynamic characteristic data; performing charging influence area identification based on the range guarantee dynamic characteristic data, so as to generate node influence range data; performing demand mode coupling based on moving speed, remaining power, and scheduling efficiency according to the node influence range data, so as to generate demand action coupling data;

[0068] A control feedback module for constructing a machine learning scheduling model according to the demand action coupling data and performing charging path dynamic programming using the machine learning scheduling model, so as to generate path planning data; performing real-time optimization of operating parameters based on the path planning data, so as to obtain unmanned aerial vehicle tracking control feedback data.

[0069] Compared with the prior art, the present application has the beneficial effects of:

[0070] In the environmental perception and charging node layout link, comprehensive environmental space sample data is obtained through multi-view remote sensing surveying and mapping, and multiple factors such as terrain, obstacles, illumination, and electromagnetic environment are deeply analyzed to generate accurate spatial topology data. On this basis, considering the path connection mutation point, regional load sensitive area and operation task demand, the spatial distribution of key charging nodes is accurately determined, and a monitoring network covering multiple sensors is established. This accurate layout method ensures that the charging nodes can fully adapt to complex and variable working environments. Whether in complex terrains such as mountainous areas and urban high-rise groups, or in areas with strong electromagnetic interference, the unmanned aerial vehicle can quickly find a suitable charging point, improving the utilization rate of charging facilities and reducing the construction and operation costs caused by unreasonable layout.

[0071] In terms of data acquisition and processing, the multi-mode sensor monitoring network is subjected to dual anti-interference enhancement processing of electromagnetic shielding and vibration isolation, from sensor packaging, cable grounding to installation position optimization, to comprehensively guarantee the accuracy and reliability of data acquisition. At the same time, the unmanned aerial vehicle operation parameters are subjected to time sequence evolution processing based on endurance mileage, moving speed and remaining power, a three-dimensional parameter space dynamic evolution model is constructed, the parameter change law is deeply mined, and accurate parameter dynamic characteristic data is generated. This fine data processing method can reflect the subtle changes of the unmanned aerial vehicle operation state in real time and accurately, providing solid data support for subsequent charging scheduling decision-making, and effectively avoiding decision-making errors caused by data errors.

[0072] In the charging demand evaluation and path planning stage, based on the parameter dynamic characteristic data and the spatial topological data, a parameter-demand response relationship model is constructed, the electric energy consumption characteristics are comprehensively analyzed, and the charging demand under different distances is accurately evaluated. Combined with the unmanned aerial vehicle endurance correlation processing, the charging influence area identification and the demand mode coupling, the machine learning scheduling model is used to realize the dynamic planning of the charging path and the real-time optimization of the operation parameters. In actual operation, whether it is facing sudden weather changes, task route adjustments, or unmanned aerial vehicle state fluctuations, the system can quickly respond, predict the charging demand in advance, plan the shortest, safest and most energy-saving charging path, and real-time adjust the unmanned aerial vehicle operation parameters, improve the endurance and operation efficiency of the unmanned aerial vehicle, and reduce the time and energy waste caused by unreasonable charging arrangement. The charging scheduling system of the present application adopts modular design, and each functional module has clear division and close cooperation, forming a complete and efficient closed-loop management system. From environmental perception, state acquisition, to scheduling prediction, influence evaluation, to control feedback, the whole process of intelligent management of unmanned aerial vehicle charging scheduling is realized. This not only improves the stability and reliability of the system, but also greatly reduces the cost of manual intervention, provides a strong guarantee for long-time and large-scale operation of unmanned aerial vehicles in agriculture, logistics, inspection and other fields, and has broad market application prospect and great economic value. BRIEF DESCRIPTION OF DRAWINGS

[0073] Fig. 1 The working principle diagram of the unmanned aerial vehicle autonomous tracking charging scheduling method described in the present application;

[0074] Fig. 2 The design diagram for environmental feature perception and spatial topological data generation;

[0075] Fig. 3 The design diagram for three-dimensional parameter space dynamic model construction. DETAILED DESCRIPTION

[0076] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0077] Please refer to Figs. 1-3 The unmanned aerial vehicle autonomous tracking charging scheduling method according to the present application specifically realizes the following steps:

[0078] Step 1: environmental feature perception is performed on the operation area to obtain spatial topological data; key charging nodes are determined according to the spatial topological data, and a multi-mode sensor monitoring network is established;

[0079] Step 2: Anti-interference enhancement processing is performed on the multi-mode sensor monitoring network, and real-time collection of unmanned aerial vehicle operation parameters is performed to obtain real-time operation state data; time sequence evolution processing is performed on the real-time operation state data based on endurance mileage, moving speed, and remaining power to generate parameter dynamic characteristic data;

[0080] Step 3: A parameter-demand response relationship model is established according to the parameter dynamic characteristic data and the spatial topology data; electric energy consumption characteristic analysis is performed according to the parameter-demand response relationship model, and distance-related charging demand evaluation is performed to generate electric energy scheduling dynamic data;

[0081] Step 4: Endurance correlation processing is performed on the unmanned aerial vehicle according to the electric energy scheduling dynamic data and the real-time operation state data to generate endurance guarantee dynamic characteristic data; charging influence area identification is performed based on the endurance guarantee dynamic characteristic data to generate node influence range data; demand mode coupling is performed based on moving speed, remaining power, and scheduling efficiency according to the node influence range data to generate demand action coupling data;

[0082] Step 5: A machine learning scheduling model is constructed according to the demand action coupling data, and charging path dynamic planning is performed using the machine learning scheduling model to generate path planning data; real-time optimization of operation parameters is performed based on the path planning data to obtain unmanned aerial vehicle tracking control feedback data.

[0083] In step 1, when the environment characteristic of the work area is perceived, multi-angle remote sensing mapping needs to be performed by multiple unmanned aerial vehicles flying at different heights and angles, and combined with the scanning data of the ground fixed sensor to obtain environment space sample data including visible light image, infrared thermal imaging data and laser point cloud data. After obtaining the environment space sample data, a deep learning algorithm is used for terrain recognition, which specifically performs pixel-level segmentation and labeling on terrain types such as mountains, plains, and water areas, as well as distribution height and coverage range, to obtain accurate spatial structure data.

[0084] In the environmental characteristic analysis link, the number, type and distribution density of obstacles per unit area are calculated, and the path passability algorithm is used to evaluate the passability difficulty of each region. At the same time, light intensity data is collected to analyze the possible impact of solar charging on unmanned aerial vehicles, and then regional environmental characteristic data is formed. In terms of electromagnetic environment characteristics, electromagnetic signal monitoring equipment is deployed to monitor the signal strength around the work area in real time, time difference positioning method is used to accurately locate the interference source, and based on the signal strength attenuation model and the location information of the interference source, the degree of influence on communication quality is analyzed, and finally the electromagnetic environment characteristic data is obtained.

[0085] After that, the electromagnetic environment feature data, regional environment characteristic data, and spatial structure data are fused to form spatial topology data. The spatial topology data is stored in a three-dimensional grid form, and each grid cell contains multi-dimensional information such as terrain, obstacles, lighting, and electromagnetic signals.

[0086] In determining the key charging nodes, first, the path connection features are analyzed for abruptness. Specifically, by calculating the terrain type change rate and trafficability index difference of adjacent grid cells, positions with significant changes in terrain type and trafficability, such as steep slopes and dense building groups, are identified, thereby obtaining path abruptness point data. Next, the regional load state is analyzed for sensitivity. This analysis includes calculating the electrical energy demand intensity of each region based on the range and remaining power, evaluating the traffic efficiency based on the moving speed and obstacle density, and analyzing the communication stability based on the signal strength and interference sources, thereby identifying regions with large changes in electrical energy demand and obtaining load-sensitive zone data.

[0087] Then, in combination with the pre-set task requirements and system constraint data, such as task period and UAV endurance, the spatial distribution of key charging nodes near the path abruptness points and load-sensitive zones is determined, and node arrangement data is obtained. Finally, according to the node arrangement data, the sensor types are determined, and a multi-mode sensor monitoring network is established, including position sensors, power sensors, speed sensors, and signal sensors, to achieve comprehensive monitoring of the work area.

[0088] In the multi-view remote sensing surveying and mapping process, the flight height of different flights of UAVs needs to be adjusted according to the complexity of the terrain of the work area. For complex terrain such as mountainous areas, the flight height of the UAV is relatively low, generally between 100 meters and 200 meters, in order to obtain more detailed terrain information; while in relatively flat areas such as plains, the flight height of the UAV can be appropriately increased, such as between 300 meters and 500 meters, to expand the surveying and mapping range and improve the surveying and mapping efficiency. The flight angle also needs to be diversified, including vertical downward, 45-degree oblique, and other angles, to ensure the acquisition of omnidirectional images and data.

[0089] The arrangement of ground fixed sensors also needs to be reasonably planned according to the size and shape of the work area. In the edge and key positions of the work area, such as entrances and exits, near obstacles, etc., the sensors need to be densely arranged to ensure the integrity and accuracy of the data. The scanning frequency of the sensors also needs to be set according to the actual demand, generally 1 to 5 times per second, to obtain real-time change data of the environment.

[0090] When using deep learning algorithms for terrain recognition, the algorithm needs to be trained first. The training data uses a large number of labeled terrain images and data, and by continuously adjusting the parameters of the algorithm, the recognition accuracy of the algorithm is improved. In practical application, the algorithm can classify each pixel in the environmental space sample data, accurately identify different terrain types, and mark their distribution height and coverage range.

[0091] When calculating the density of obstacles, the obstacles need to be identified and classified first. The identification of obstacles can be achieved through image recognition techniques such as edge detection, shape analysis, etc. to separate obstacles from the environment. Then, the number of obstacles per unit area is calculated, and the type of obstacle is determined according to its size and shape, such as trees, buildings, vehicles, etc. The path accessibility algorithm needs to consider the distribution of obstacles, the slope of the terrain, the intensity of light, etc. to evaluate the difficulty of passing through each region, providing a reference for the path planning of the UAV.

[0092] The deployment of electromagnetic signal monitoring equipment needs to consider the size of the work area and the complexity of the electromagnetic environment. In areas with complex electromagnetic environments, such as near communication base stations, high-voltage power lines, etc., the number of monitoring equipment needs to be increased to monitor the changes in signal strength in real time. The positioning accuracy of the time difference positioning method for the interference source is closely related to the arrangement position and number of monitoring equipment, so it is necessary to arrange the monitoring equipment reasonably to ensure that the positioning accuracy meets the requirements.

[0093] When constructing the three-dimensional grid of spatial topology data, the size of the grid needs to be set according to the precision requirements of the work area. For areas with high precision requirements, the grid size can be set to 1m x 1m x 1m; for areas with low precision requirements, the grid size can be appropriately increased, such as 5m x 5m x 5m. The multi-dimensional information in each grid cell needs to be updated in real time to reflect the changes in the environment.

[0094] When determining the spatial distribution of key charging nodes, the pre-set work task requirements and system constraint data are important reference. For example, if the work task requires the UAV to fly for 4 hours, and the endurance of the UAV is 2 hours, then charging nodes need to be arranged at regular intervals on the UAV flight path to ensure that the UAV can be charged in time to complete the work task. At the same time, the positions of the path mutation points and the load sensitive areas need to be considered, and charging nodes need to be arranged near these areas to improve the efficiency and reliability of charging.

[0095] The various sensors in the multi-mode sensor monitoring network need to have high precision and reliability. The position sensor can use satellite positioning systems such as GPS and Beidou, combined with inertial navigation equipment, to ensure the position measurement accuracy of the unmanned aerial vehicle; the power sensor needs to monitor the remaining power of the unmanned aerial vehicle battery in real time, and the precision needs to reach within 1%; the speed sensor can use Doppler radar or visual speed measurement method to measure the flight speed of the unmanned aerial vehicle; the signal sensor needs to monitor the electromagnetic signal strength of the operation area in real time to provide support for the communication and charging scheduling of the unmanned aerial vehicle.

[0096] In addition, after establishing the multi-mode sensor monitoring network, the sensors need to be calibrated and debugged to ensure the accuracy and reliability of the sensor measurement data. At the same time, a sensor data transmission and processing system needs to be established to transmit the sensor collected data to the data processing center in real time for analysis and processing to provide data support for subsequent charging scheduling.

[0097] In step 2, when the multi-mode sensor monitoring network is subjected to anti-interference enhancement processing, electromagnetic shielding operation needs to be implemented from three aspects of sensor packaging, cable processing and grounding point planning. High permeability metal material is selected for the sensor shell for full-sealed packaging to block the interference of external electromagnetic signals; the signal cable adopts a double-shielded structure, the inner shielding layer suppresses high-frequency interference, and the outer shielding layer deals with low-frequency interference, and the shielding layer is reliably grounded by welding, and the grounding resistance needs to be controlled at a low level to ensure the electromagnetic shielding effect. At the same time, the finite element analysis software is used to simulate and optimize the sensor grounding point position, analyze the ground loop current distribution under different grounding point arrangements, and select the position with the smallest ground potential difference as the grounding point to reduce the influence of ground loop interference on sensor data acquisition.

[0098] In the vibration isolation aspect, a rubber shock pad is provided between the sensor and the mounting bracket, and the hardness and thickness of the shock pad need to be selected according to the vibration frequency and amplitude of the unmanned aerial vehicle during flight, generally with a hardness of 50 to 70 degrees of Shore and a thickness of 5 to 10 millimeters to effectively absorb vibration energy. A flexible cable connector is used to connect the sensor and the unmanned aerial vehicle body, and the length and flexibility of the connector need to meet the displacement requirements of the sensor during the flight of the unmanned aerial vehicle to reduce vibration transmission. The modal analysis software is used to calculate the vibration mode of the unmanned aerial vehicle body to determine the best installation position of the sensor, avoiding the resonance frequency points of the body, such as installing the sensor at the non-node position of the unmanned aerial vehicle arm to avoid resonance with the body.

[0099] When collecting real-time parameters during the operation of the UAV, the data such as the remaining power, the moving speed, the endurance mileage and the like need to be collected synchronously, and the sampling frequency is adjusted according to the flight state of the UAV. During the dynamic flight stage such as acceleration and deceleration of the UAV, the sampling frequency is set to 10 Hz to 20 Hz to capture the rapid change of the parameters. During the cruising stage, the sampling frequency can be reduced to 5 Hz to 10 Hz to reduce the data storage amount. After the real-time operation state data is collected, the remaining power is processed in sections based on time sequence, and the power data is divided into multiple continuous time intervals in chronological order, with the length of each time interval being determined according to the characteristics of the flight task. For example, when performing a long-time inspection task, the time interval can be set to 10 minutes to 15 minutes, and a power-time relationship curve is established.

[0100] When analyzing the power fluctuation according to the power-time relationship curve, the power fluctuation signal in the time domain is converted into a frequency domain signal through a Fourier transform algorithm, the frequency characteristics of the power fluctuation are calculated, and the main fluctuation frequency component is identified. For example, when the UAV is passing through the cloud layer, the power fluctuation can present a specific low-frequency characteristic. At the same time, the power fluctuation amplitude distribution of different flight stages such as the take-off stage, the cruising stage and the landing stage is analyzed, and the maximum value, the minimum value and the average value of the power fluctuation of each stage are counted. The correlation coefficient calculation method is used to analyze the correlation between the power fluctuation and the flight distance, to determine the correlation degree between the power consumption and the flight distance, so as to obtain the power-distance evolution characteristic data.

[0101] For the moving speed parameter, a relationship curve of the moving speed and the flight distance is established, the curve is divided according to each kilometer or each half kilometer of distance, the speed change rate in each distance section, i.e. the change amount of the speed per unit distance, is calculated, the speed change characteristics of the UAV in different motion states such as acceleration, deceleration and uniform speed are identified, and the speed time sequence evolution characteristic data is obtained. For example, when the UAV approaches the charging node, the speed change rate can present a negative value, indicating that the UAV is decelerating to prepare for docking charging.

[0102] When recording the real-time change of the endurance mileage, the remaining endurance mileage needs to be calculated in real time in combination with the flight trajectory of the UAV and the battery consumption model. When analyzing the correlation characteristics of the endurance mileage and the flight distance, the influence of environmental factors such as the terrain and the wind speed is considered. For example, when flying against the wind, the descending rate of the endurance mileage can be accelerated. Through the establishment of a multiple regression model, the influence degree of each factor on the endurance mileage is analyzed, and the endurance time sequence evolution characteristic data is obtained.

[0103] The power-distance evolution characteristic data, the speed time sequence evolution characteristic data and the endurance time sequence evolution characteristic data are input into a comprehensive analysis model, the model is constructed based on a rule engine, the comprehensive influence of parameter combination on charging demand is analyzed by setting different threshold conditions. For example, when the remaining power is lower than a set threshold (such as 25%) and the moving speed is greater than a certain speed value (such as 15 m / s), it is determined that the high charging demand state is determined, the critical value and the adaptive interval of the key parameter combination are identified, and the time sequence evolution characteristic data of the parameter combination is generated.

[0104] When constructing the three-dimensional parameter space, the remaining power is taken as the Z axis, the moving speed is taken as the X axis, and the endurance mileage is taken as the Y axis to establish a rectangular coordinate system. The values of the remaining power, the moving speed and the endurance mileage of each time sampling point are mapped to coordinate points in the three-dimensional space, the coordinate points are smoothed by a trajectory fitting algorithm such as a cubic spline interpolation algorithm, a parameter change trajectory curve is established, and the change trend of the parameters with time is intuitively displayed.

[0105] When performing distance-based hierarchical processing on the parameter trajectory data, the trajectory curve is divided into multiple levels according to a hierarchical interval of every 5 kilometers or 10 kilometers of flight distance. In each level, the change trend of the parameters is analyzed, such as whether the remaining power decreases linearly or nonlinearly in the distance segment; the change rate of the parameters is calculated, such as the amount of decrease of the remaining power per kilometer of distance; and the mutual correlation between the parameters is analyzed, such as whether the change of the moving speed will affect the consumption rate of the remaining power, thereby obtaining hierarchical characteristic data.

[0106] When constructing a dynamic model of the three-dimensional parameter space according to the hierarchical characteristic data, a Kriging interpolation method is used to perform interpolation calculation on unknown points in the space, a continuous expression of parameter change is established through spatial interpolation and numerical fitting, and dynamic evolution model data is obtained. When performing feature extraction on the dynamic evolution model data, the gradient feature of parameter change, i.e., the change rate of the parameters in the space, is calculated; the curvature feature is analyzed to describe the bending degree of the parameter change trajectory; and the speed feature is extracted to reflect the speed of change of the parameters with time, thereby obtaining model characteristic data.

[0107] When performing parameter correlation analysis based on the model characteristic data, the coupling relationship and mutual feedback mechanism between the parameters are considered, such as the decrease of the remaining power causing the increase of the output power of the unmanned aerial vehicle power system, thereby affecting the moving speed, and the change of the moving speed in turn affecting the consumption rate of the remaining power. By establishing a causal relationship diagram, the mutual influence path and degree between the parameters are analyzed, and parameter correlation data is obtained. Finally, according to the model characteristic data and the parameter correlation data, indexes capable of representing the dynamic change law of the parameters, such as the remaining power consumption rate index and the moving speed fluctuation coefficient, are extracted, and parameter dynamic characteristic data is generated, thereby providing data support for subsequent charging scheduling.

[0108] Throughout the implementation process, the anti-interference processing effect of the sensor needs to be tested. The sensor data acquisition experiment can be carried out in an electromagnetic interference environment and a vibration environment. The data noise level before and after processing is compared to ensure that the anti-interference enhancement processing achieves the expected effect. At the same time, the generation process of the parameter dynamic characteristic data is verified. Through comparison and analysis with historical data, the accuracy and reliability of the data are ensured. In addition, a data quality control mechanism needs to be established to detect and process abnormal values of the collected real-time running state data, such as identifying and eliminating abnormal data points through the 3σ principle to ensure the accuracy of subsequent analysis.

[0109] In step 3, when establishing the parameter-demand response relationship model, the parameter dynamic characteristic data needs to be first associated and integrated with the spatial topology data. The parameter dynamic characteristic data includes the dynamic change law of parameters such as residual power, moving speed, and cruising range, while the spatial topology data covers information such as the terrain, obstacle distribution, and electromagnetic environment of the operation area. In specific operation, taking a three-dimensional grid as the basic unit, the parameter dynamic characteristic data (such as the average residual power change rate and typical moving speed range of the unmanned aerial vehicle in this area) in each grid is corresponded with the spatial topology data (such as terrain type, obstacle density, and signal strength) one by one to construct a parameter-demand characteristic correspondence table. Each entry in the table records the mapping relationship between parameter change and charging demand in a specific area in detail, for example, in a mountain grid area, when the residual power is less than 30% and the moving speed is greater than 12 m / s, the system automatically determines that it is in a high-demand state that needs to be prioritized for charging.

[0110] After completing the construction of the parameter-demand characteristic correspondence table, the parameter response data needs to be standardized and feature dimensionality reduced. In the standardization process, for different dimension parameters (such as residual power in percentage and moving speed in m / s), the minimum-maximum standardization method is used to map all parameters to the unified interval of [0, 1], eliminating the influence of dimension difference on subsequent analysis. Feature dimensionality reduction uses the principal component analysis (PCA) method to calculate the covariance matrix between parameters to determine the principal components that can represent the main trend of data change. In actual operation, set a cumulative contribution rate threshold (such as 95%), and only keep the first few principal components that contribute most to data variance, so as to compress the high-dimensional parameter space to a low-dimensional space, reduce the calculation complexity, and avoid excessive loss of information.

[0111] In establishing the mathematical model of parameter-demand response, based on the reduced parameter data, a linear regression model is first attempted to be constructed, assuming a linear relationship between the parameters and charging demand, and the model parameters are fitted by the least squares method. If the fitting effect of the linear model is not good (such as low goodness of fit R²), a polynomial model or other nonlinear model, such as a quadratic polynomial model, is further constructed to capture the nonlinear mapping relationship between the parameters and the demand. During the model construction process, historical flight data and charging records are used as training samples, and the model hyperparameters are adjusted by cross-validation method to ensure the generalization ability of the model.

[0112] When constructing the parameter-demand response relationship model, machine learning algorithms are used to optimize and train the mathematical model. Specifically, the random forest algorithm or the neural network algorithm is selected, where the random forest algorithm can effectively handle nonlinear relationships and feature interactions by constructing multiple decision trees and integrating their prediction results; the neural network algorithm can automatically learn the complex mapping relationship between the parameters and the demand by using a multi-layer perceptron structure. During the training process, the historical data is divided into training set, validation set and test set, the training set is used for model parameter update, the validation set is used for hyperparameter adjustment, and the test set is used for model performance evaluation. During the training process, the loss function value of the model on the validation set needs to be monitored to avoid overfitting.

[0113] When analyzing the electric energy consumption characteristics based on the parameter-demand response relationship model, three dimensions of electric energy consumption, path loss and charging efficiency need to be considered. The electric energy consumption analysis needs to combine with the unmanned aerial vehicle power system model to calculate the electric energy consumption rate under different flight states (such as cruising, climbing, descending); the path loss analysis considers factors such as obstacle blocking and terrain undulation affecting the unmanned aerial vehicle flight path, and evaluates the additional energy consumption caused by path detour or additional power output; the charging efficiency analysis needs to consider factors such as charging mode (such as fast charging, slow charging), battery state (such as remaining electric energy, temperature) affecting the charging energy conversion rate. By integrating these factors, detailed consumption characteristic data is generated to provide a basis for charging demand assessment.

[0114] When performing distance-related charging demand assessment, first, an electric energy consumption decay model is established based on the consumption characteristic data. This model considers the impact of distance on electric energy consumption. Generally, the farther the unmanned aerial vehicle is from the charging node, the higher the round-trip energy consumption to reach the charging node, and therefore the higher the urgency of the charging demand. At the same time, the impact of regional load intensity needs to be analyzed. For high-load areas (such as areas where multiple unmanned aerial vehicles are operating simultaneously), due to fierce competition for charging resources, the charging demand range of the area needs to be appropriately expanded (such as increasing a certain percentage of the range based on the original distance) to ensure that unmanned aerial vehicles can obtain charging resources in a timely manner. By considering the electric energy consumption decay law and the regional load intensity, the charging demand data of each region is calculated to determine the charging urgency level at different locations.

[0115] When generating the dynamic data of power scheduling, the key features and mutation points in the entire scheduling process need to be identified. Key features include peak intervals of scheduling efficiency, high and low changes of charging resource utilization, etc. These features can be extracted through sliding window statistical methods, such as calculating the average scheduling efficiency in each time window to identify time periods with higher efficiency. Mutation point identification uses wavelet transform method to decompose the scheduling data into different frequency domains to detect abnormal fluctuation points in the data sequence, such as moments of sudden surge in charging demand or scheduling abnormal points caused by charging node failure. By accurately identifying these key features and mutation points, dynamic data of power scheduling that can reflect the dynamic changes of power scheduling process is generated, providing decision support for subsequent endurance association processing and charging path planning.

[0116] In the implementation process, the construction of the parameter-demand characteristic correspondence table needs to ensure the accuracy and integrity of the data. The mapping relationship in the table can be verified through a combination of manual review and automatic verification. For example, for the high demand determination threshold of mountain grid areas, it needs to be adjusted in combination with the actual flight test data of unmanned aerial vehicles in this terrain to ensure the rationality of the threshold setting. When performing feature dimensionality reduction, the mapping relationship between the principal components and the original parameters needs to be saved to restore the actual meaning of the parameters in subsequent analysis. The training of the machine learning model needs to ensure the diversity and representativeness of the historical data, covering different operation scenarios, weather conditions, and unmanned aerial vehicle models to improve the generalization ability of the model. The various models involved in the analysis of power consumption characteristics (such as power system models and charging efficiency models) need to be customized and adjusted according to the specific unmanned aerial vehicle model and charging equipment parameters to ensure the accuracy of the analysis results. The distance-related charging demand assessment needs to be updated regularly according to changes in the operation area (such as the addition of obstacles and the adjustment of charging node locations) to adapt to the dynamic changes of the environment. The generation of dynamic data of power scheduling needs to establish a real-time monitoring mechanism to analyze the scheduling process in real time through streaming data processing technology to ensure the timeliness and accuracy of mutation point identification. In addition, a complete log recording system needs to be established to record the input data, processing parameters, and output results of each processing step for subsequent tracing and optimization.

[0117] In step 4, when performing the UAV endurance correlation processing, the power scheduling dynamic data and the real-time running state data need to be input into the influence analysis model. Taking the UAV distribution scenario in a certain logistics park as an example, when the power scheduling dynamic data shows that the current load of a certain charging node is low, and the real-time running state data shows that a certain UAV has a remaining power of 22%, a moving speed of 14 m / s, and is currently located 3 kilometers away from the charging node, the model will analyze the influence of different charging schemes on the endurance under this parameter combination. For example, if the UAV chooses to go directly to the charging node, it needs to evaluate whether the energy consumption of flying to the charging node will cause the power to be lower than the safety threshold; if it continues to work, it needs to analyze whether the remaining power can support it to the next charging node. By comparing the endurance maintenance effects of different schemes, the guarantee influence data is obtained, such as the endurance time extension amount corresponding to different charging times.

[0118] When performing performance evaluation on the guarantee influence data, three dimensions are expanded: endurance maintenance time, charging completion time characteristics, and path deviation characteristics. Still taking the logistics park as an example, if a certain charging scheme makes the UAV endurance maintenance time increase by 50 minutes, the charging completion time is 28 minutes (fast charging mode), and the deviation angle of the charging path from the original distribution path is 15 degrees, then the three dimensions are quantitatively scored respectively. The endurance maintenance time score can be based on the task demand to set a benchmark value, and the part exceeding the benchmark value is given a higher score; the charging completion time characteristic score considers whether it meets the task time window requirement; the path deviation characteristic score is determined according to the influence degree of the deviation angle on subsequent work, thereby obtaining the performance evaluation data.

[0119] When analyzing the evolution characteristics of the guarantee effect over time, a time series model needs to be established. For example, in a certain inspection task, within 1 hour after charging, the UAV has a good endurance guarantee effect because the battery is in the high-efficiency discharge stage; after 1 hour, the battery performance gradually decays, and the guarantee effect decreases accordingly. By statistically analyzing the guarantee effect indicators at different time points, such as the remaining power maintenance rate, an evolution curve is drawn, thereby obtaining the endurance guarantee dynamic characteristic data, which provides a time dimension reference for subsequent charging influence area identification.

[0120] When performing charging influence area identification, taking a certain industrial park as an example, the endurance guarantee dynamic characteristic data is numerically simulated using the finite element method. A multi-field coupled analysis model of demand field, path field, and efficiency field is established: the demand field reflects the charging demand intensity at different positions, such as the position near the work-intensive area having a higher demand intensity; the path field simulates the optimal path probability distribution of the UAV from different positions to the charging node, considering factors such as terrain and obstacles affecting the path; the efficiency field represents the charging efficiency distribution in different areas, such as the center area near the charging node having a higher efficiency, and the edge area having a lower efficiency. By solving the partial differential equations of the multi-field coupling, the influence area data is obtained, which directly shows the influence range of the charging node in space.

[0121] When performing boundary recognition and spatial partitioning on impact area data, a contour extraction algorithm is used to determine the effective impact range of the charging node. Taking a charging node in an industrial park as an example, its impact range may be a circular area centered on the node with a radius of 2.5 kilometers, but due to building obstruction, the actual impact range will be reduced to 1.8 kilometers in the direction of building density and expanded to 3 kilometers in the open direction. According to the complexity of the terrain and the distribution of obstacles, the impact range is divided into different areas, such as open areas, hilly areas, and building-dense areas, and node impact range data is obtained, with each partition corresponding to different charging scheduling strategies.

[0122] When coupling demand patterns, taking the agricultural plant protection unmanned aerial vehicle operation scenario as an example, the moving speed, remaining power, and scheduling efficiency are used as input variables to construct the coupling model. When the remaining power of the unmanned aerial vehicle is less than 15% and the moving speed is 8 m / s, to avoid failure to return due to insufficient power, the system will automatically increase the scheduling efficiency, preferentially assigning the nearest charging node to the unmanned aerial vehicle, and adjusting its moving speed to 6 m / s to reduce energy consumption. In this process, the evolution law of the demand field is identified, such as the unmanned aerial vehicle gradually moving to the edge of the farmland as the operation time increases, and the center of the demand field also shifts; meanwhile, the critical state characteristics are identified, such as when the remaining power is less than 10%, it is in an emergency charging state, which requires triggering the highest priority scheduling, thereby generating demand action coupling data to guide subsequent charging path planning.

[0123] In specific implementation, the impact analysis model needs to be customized with parameters according to different operation scenarios. For example, in the logistics park scenario, the delivery time window and path planning efficiency need to be considered; in the agricultural plant protection scenario, the terrain of the operation area and the progress of pesticide spraying tasks need to be considered. The dimension weight of performance evaluation can be adjusted according to actual demand, such as the weight of the endurance maintenance time in emergency rescue tasks being higher than the path deviation characteristics. The establishment of multi-scene coupling analysis model needs to accurately input spatial topology data and parameter dynamic characteristic data to ensure that the simulation results meet the actual scene. The threshold setting (such as the remaining power critical value) in demand pattern coupling needs to be determined through a large amount of historical data statistical analysis, combined with factors such as unmanned aerial vehicle model and battery performance for adjustment.

[0124] In addition, the endurance correlation process needs to establish a real-time data interaction mechanism to ensure the synchronous update of dynamic data of power scheduling and real-time operation state data. The performance evaluation results need to be fed back to the scheduling system for optimizing the subsequent charging scheme. The identification of the charging influence area needs to update the model parameters regularly according to environmental changes (such as new buildings, terrain reconstruction) to ensure the accuracy of the influence range. The demand mode coupling model needs to have self-learning ability, continuously optimize the mapping relationship between parameter combination and scheduling strategy through accumulation of historical scheduling data, and improve the accuracy of coupling analysis. The entire implementation process needs to record key data nodes, such as guaranteeing influence data in different scenarios, intermediate results of multi-scene coupling simulation, etc., in order to trace back and optimize the model later.

[0125] In step 5, when building the machine learning scheduling model, taking the city express drone delivery scenario as an example, first, the demand action coupling data is extracted. For example, when the remaining power of the drone is 18%, the moving speed is 12 m / s, and it is in a high-load area of the commercial district, the demand action coupling data will include feature parameters such as moving speed change rate (such as the rate from 15 m / s to 12 m / s), remaining power gradient (5% decrease per kilometer), scheduling efficiency index (waiting time and distance comprehensive score of the current charging node), etc. These features are used as input features, and the corresponding historical optimal charging path (such as the path from the current position to the charging node through 3 intersections and right turn) is used as the path label to build a training sample set containing thousands of samples, which covers path data under different combinations of power, speed, and load area.

[0126] When building the machine learning network structure according to the training sample data, a hybrid architecture of decision tree algorithm and reinforcement learning is adopted. The decision tree part is used to process discrete features such as area load level (high / medium / low), and learns path selection rules under different loads by dividing nodes, for example, preferentially selecting the nearest charging node in a high-load area. The reinforcement learning part sets the state space as the current position, power, speed, and surrounding charging node state of the drone, the action space as the path selection to different charging nodes, and the reward function as the weighted value of the time and energy consumption to reach the charging node. By constantly trying and failing in the simulation environment, the model learns the optimal strategy, for example, when the power is lower than 20%, even if a charging node is slightly far away but has shorter waiting time, the model will tend to choose that node.

[0127] When cross-validating the path planning model, the training sample set is divided into 5 parts, and each time 4 parts are trained and 1 part is validated, repeated 5 times. During the validation process, the maximum depth of the decision tree (such as from 5 layers to 8 layers), the learning rate of reinforcement learning (such as from 0.01 to 0.005) and other hyperparameters are adjusted to gradually improve the path prediction accuracy of the model on the validation set. For example, when the depth of the decision tree is 7 layers and the learning rate is 0.008, the path prediction accuracy of the model for high-load areas is significantly improved, and the optimized scheduling model data is finally obtained.

[0128] When using the scheduling model data to dynamically plan the charging path, taking a certain express drone as an example, the current location is at the A intersection of the city trunk road, the remaining battery capacity is 20%, and the surrounding 3 charging nodes are located at the B intersection (1.2 kilometers away, with an estimated waiting time of 5 minutes), the C intersection (0.8 kilometers away, with an estimated waiting time of 10 minutes), and the D intersection (1.5 kilometers away, with an estimated waiting time of 2 minutes). The model calculates the lateral offset (such as needing to deviate from the original delivery route by 300 meters to go to the D intersection), the vertical extension (needing to climb 10 meters to avoid tall buildings), and the spatial coverage (covering 2 delivery points along the way), generates path planning data, and finally recommends the path to the D intersection, although the distance is slightly farther, but the waiting time is short, and the overall time consumption is less.

[0129] When analyzing the key stages of the path planning data, taking the multi-drone collaborative charging scenario as an example, in the initial departure stage, multiple drones simultaneously depart from different work points, and the path planning aims to avoid air collisions and maintain a safe distance; in the demand growth stage, when the battery level of a certain drone drops to 15%, the model automatically adjusts its path direction and preferentially drives to the nearest available charging node; in the charging execution stage, the planning accurately interfaces the descent path of the charging node, such as vertically descending along the specified flight line at a speed of 5 m / s; in the endurance recovery stage, after charging is completed, the planning returns to the shortest path to the work area; in the return adjustment stage, the path is fine-tuned according to the remaining battery capacity and new work tasks; in the multi-drone collaboration stage, a conflict detection mechanism is established, such as setting a safe flight corridor for each drone to avoid path intersection of multiple drones.

[0130] When optimizing the running parameters in real time according to the path evolution data, taking the charging execution stage as an example, when the drone is 100 meters away from the charging node, the path evolution data shows that it needs to enter the precise docking stage, at this time the flight speed is optimized from 8 m / s to 5 m / s in real time, the flight speed is reduced to improve the docking accuracy; at the same time, the flight altitude is adjusted from 50 meters to 20 meters, the visual navigation system is turned on, and the docking identifier of the charging node is identified; when the distance to the charging node is 10 meters, the speed is further reduced to 2 m / s, the hovering mode is started, and the charging interface is docked. These optimized parameters are transmitted to the drone flight control system through the data bus, generating tracking control feedback data to ensure that the drone accurately completes the charging docking.

[0131] In practice, the construction of training sample sets must cover a variety of scenarios, such as daytime / nighttime, sunny / rainy, and data under varying traffic flows, to avoid model overfitting in specific scenarios. Training of hybrid architectures requires balancing the weights of decision trees and reinforcement learning. This can be achieved by setting different loss function coefficients, for example, a 40% loss for the decision tree component and a 60% loss for the reinforcement learning component. Lateral offset calculations in path planning should incorporate urban building distribution data to avoid planning routes that cross no-fly zones. The division of key phases should be adjusted based on the drone model and charging equipment characteristics. For example, the charging execution phase distance for small drones can be set at 50 meters, while for large drones, it can be set at 100 meters. The thresholds for real-time optimization parameters (such as the distance nodes for speed adjustment) should be determined through extensive flight testing, taking into account the drone's braking performance and sensor accuracy to ensure appropriate adjustment timing. Furthermore, a regular model update mechanism should be established. After accumulating a sufficient amount of new scheduling data, the model should be retrained to adapt to environmental changes. For example, when new charging nodes are added in the city or traffic regulations change, the path planning strategy should be updated promptly. The entire implementation process needs to record the iterative process of model training, the intermediate results of path planning, and the historical data of parameter optimization to facilitate subsequent analysis of model performance and optimization of scheduling strategies.

[0132] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0133] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A charging scheduling method for autonomous tracking of a UAV, characterized in that: The following steps are involved: Step 1: Perceive the environmental characteristics of the work area and obtain spatial topology data; Determine key charging nodes based on spatial topology data and establish a multi-mode sensor monitoring network; Step 2: Perform anti-interference enhancement processing on the multi-mode sensor monitoring network and collect UAV operating parameters in real time to obtain real-time operating status data; The real-time operating status data is processed based on the time series evolution of the driving range, moving speed and remaining power to generate parameter dynamic characteristic data; Step 3: Establish a parameter-demand response relationship model based on the parameter dynamic characteristic data and spatial topology data; analyze the power consumption characteristics based on the parameter-demand response relationship model, conduct distance-related charging demand assessment, and generate dynamic power scheduling data; Step 4: Perform drone endurance correlation processing based on power dispatch dynamic data and real-time operating status data to generate endurance guarantee dynamic characteristic data; identify charging impact areas based on endurance guarantee dynamic characteristic data to generate node impact range data; Based on the node influence range data, demand patterns based on movement speed, remaining power, and scheduling efficiency are coupled to generate demand action coupling data; Step 5: Build a machine learning scheduling model based on the demand-action coupling data, and use the machine learning scheduling model to dynamically plan the charging path and generate path planning data; Based on the path planning data, the operating parameters are optimized in real time to obtain the UAV tracking control feedback data.

2. The charging scheduling method for autonomous tracking of a UAV according to claim 1, characterized in that: Step 1 includes the following steps: Step 11: Conduct multi-view remote sensing mapping of the operation area to obtain environmental space sample data; Step 12: Identify the type, distribution height, and coverage of the terrain of the environmental space sample data to obtain spatial structure data; Step 13: Perform environmental characteristic analysis on the environmental space sample data based on obstacle density, path passability, and light intensity according to the spatial structure data, thereby obtaining regional environmental characteristic data; Step 14: Perform electromagnetic environment feature analysis based on the regional environmental characteristic data to obtain electromagnetic environment feature data. The electromagnetic environment feature analysis includes signal strength monitoring around the operating area, interference source location based on environmental space sample data, and communication quality analysis based on signal strength and interference sources. Step 15: Merge the electromagnetic environment characteristic data, regional environment characteristic data, and spatial structure data into spatial topology data; Step 16: Determine key charging nodes based on spatial topology data and establish a multi-mode sensor monitoring network.

3. The charging scheduling method for autonomous tracking of a UAV according to claim 2, characterized in that: Step 16 includes the following steps: Step 161: performing a mutation analysis on the path connection characteristics based on the spatial topology data to obtain path mutation point data, wherein the mutation analysis specifically identifies locations where the terrain type and traversability change significantly; Step 162: Perform a sensitivity analysis on the regional load status based on the spatial topology data to identify areas with significant variations in power demand and obtain load-sensitive area data. The load sensitivity analysis includes a demand intensity assessment based on range and remaining battery life, a traffic efficiency assessment based on moving speed and obstacle density, and a communication stability assessment based on signal strength and interference sources. Step 163: Determine the spatial distribution of key charging nodes based on preset task requirements and system constraint data, path mutation point data, and load-sensitive area data to obtain node layout data; Step 164: Determine the sensor type based on the node layout data, and establish a multi-mode sensor monitoring network including a position sensor, a power sensor, a speed sensor, and a signal sensor.

4. The charging scheduling method for autonomous tracking of a UAV according to claim 3, characterized in that: Step 2 includes the following steps: Step 21: Perform anti-interference enhancement processing based on electromagnetic shielding and vibration isolation on the multi-mode sensor monitoring network, and collect real-time parameters during the operation of the UAV to obtain real-time operation status data; Step 22: Segment the remaining power in the real-time operating status data based on the time series, thereby establishing a power-time relationship curve; Step 23: Calculate the frequency characteristics and amplitude distribution of the power fluctuation based on the power-time relationship curve, and perform a correlation analysis between the power fluctuation and the distance to obtain power-distance evolution characteristic data; Step 24: Establish a relationship curve between moving speed and distance based on the real-time running status data, and calculate the speed change rate of different distance segments to obtain speed time series evolution characteristic data; Step 25: Record the real-time changes in the cruising range in the real-time operating status data, and perform correlation feature analysis between the cruising range and the distance, thereby obtaining cruising range time series evolution feature data; Step 26: Analyze the comprehensive impact of parameter combinations on charging demand based on the power-distance evolution characteristic data, the speed time-series evolution characteristic data, and the endurance time-series evolution characteristic data, and identify the critical values ​​and adaptation intervals of key parameter combinations, thereby generating time-series evolution characteristic data of the parameter combinations; Step 27: Establish a dynamic evolution model of the three-dimensional parameter space based on the time series evolution feature data, and extract parameter change features to obtain parameter dynamic feature data.

5. The charging scheduling method for autonomous tracking of a UAV according to claim 4, characterized in that: Step 21 includes the following steps: The multi-mode sensor monitoring network is subjected to anti-interference enhancement processing based on electromagnetic shielding and vibration isolation, and real-time parameters of the drone during operation are collected to obtain real-time operation status data; electromagnetic shielding includes metal shell packaging of the sensor, grounding of the shielding layer of the signal cable, and planning of the sensor grounding point; vibration isolation includes setting up shock-absorbing brackets, using flexible connectors, and optimizing the anti-vibration of the sensor installation position.

6. The charging scheduling method for autonomous tracking of a UAV according to claim 5, characterized in that: Step 27 includes the following steps: Step 271: constructing a three-dimensional parameter space with remaining power, moving speed, and cruising range as coordinate axes based on the time-series evolution feature data, thereby obtaining parameter space coordinate data; Step 272: establishing a parameter change trajectory based on the parameter space coordinate data, and constructing a parameter motion trajectory curve through spatial mapping of the time series sampling points, thereby obtaining parameter trajectory data; Step 273: performing distance-based layered processing on the parameter trajectory data to identify parameter change characteristics including change trends, change rates, and mutual correlations in different distance segments, thereby obtaining layered feature data; Step 274: constructing a dynamic model of the three-dimensional parameter space based on the hierarchical feature data, and establishing a continuous expression of parameter changes through spatial interpolation and numerical fitting, thereby obtaining dynamic evolution model data; Step 275: extracting the gradient feature, curvature feature, and velocity feature of the dynamic evolution model data based on parameter changes, thereby obtaining model feature data; Step 276: performing parameter correlation analysis based on the coupling relationship and mutual feedback mechanism between parameters based on the model characteristic data, thereby obtaining parameter correlation data; Step 277: extracting indicators that characterize the dynamic change patterns of parameters based on the model feature data and the parameter association data, thereby generating parameter dynamic feature data.

7. The charging scheduling method for autonomous tracking of a UAV according to claim 6, characterized in that: Step 3 includes the following steps: Step 31: Establishing a parameter-demand characteristic correspondence table including a mapping relationship between parameter changes and charging demands in different regions based on the parameter dynamic characteristic data and the spatial topology data, thereby obtaining parameter response data; Step 32: Perform data standardization and feature dimension reduction processing on the parameter response data, and establish a mathematical model of parameter-demand response to obtain response model data; Step 33: Perform machine learning training based on the response model data to construct a nonlinear mapping relationship between parameters and demand response, thereby obtaining a parameter-demand response relationship model; Step 34: Analyze the power consumption characteristics including power consumption, path loss, and charging efficiency based on the parameter-demand response relationship model to obtain consumption characteristic data; Step 35: Calculate the distance-related demand range based on the consumption characteristic data to obtain charging demand data, where the demand range calculation includes analysis of the influence of power consumption attenuation law and regional load intensity; Step 36: Identify key features and mutation points in the power dispatching process of the dispatching dynamic data, thereby obtaining the power dispatching dynamic data.

8. The charging scheduling method for autonomous tracking of a UAV according to claim 7, characterized in that: Step 4 includes the following steps: Step 41: Analyze the influence of different parameter combinations on the endurance guarantee effect based on the power dispatch dynamic data and the real-time operation status data, thereby obtaining guarantee impact data; Step 42: Performing a guarantee performance evaluation on the guarantee impact data based on the endurance maintenance duration, charging completion time characteristics, and path deviation characteristics, thereby obtaining performance evaluation data; Step 43: Analyze the evolution characteristics of the performance evaluation data over time to obtain dynamic characteristic data of the endurance guarantee; Step 44: Perform numerical simulation of the charging impact area based on the endurance guarantee dynamic characteristic data, and establish a multi-field coupling analysis model including the demand field, path field, and efficiency field to obtain impact area data; Step 45: Perform boundary identification and spatial partitioning on the impact area data to obtain node impact range data; Step 46: Demand pattern coupling based on movement speed, remaining power, and scheduling efficiency is performed according to the node influence range data, and the demand field evolution law and critical state characteristics are identified to generate demand action coupling data.

9. The charging scheduling method for autonomous tracking of a UAV according to claim 8, characterized in that: Step 5 includes the following steps: Step 51: Extract features from the demand-action coupling data and construct a training sample set for the machine learning model, thereby obtaining training sample data, wherein the training sample data includes input features and path labels; Step 52: Construct a machine learning network structure based on a hybrid architecture of a decision tree algorithm and reinforcement learning according to the training sample data, and perform model training to obtain a path planning model; Step 53: Optimize the model prediction accuracy of the path planning model based on a cross-validation method to obtain scheduling model data; Step 54: Dynamically plan the charging path using the scheduling model data to obtain path planning data, where the path planning data includes lateral offset, longitudinal extension, and spatial coverage; Step 55: Analyze the key stages and path mutation characteristics of the charging process of the path planning data to obtain path evolution data. The key stages include the initial departure stage, the demand growth stage, the charging execution stage, the range recovery stage, the return adjustment stage, and the multi-machine coordination stage. Step 56: Optimize the operating parameters of each stage in real time based on the path evolution data to obtain the UAV tracking control feedback data.

10. A charging scheduling system for autonomous tracking of drones, characterized in that: The charging scheduling method for autonomous tracking of a UAV according to claim 1 is used to execute the charging scheduling method for autonomous tracking of a UAV, wherein the charging scheduling system for autonomous tracking of a UAV comprises: The environmental perception module is used to perceive the environmental characteristics of the operating area and obtain spatial topology data; based on the spatial topology data, it determines the key charging nodes and establishes a multi-mode sensor monitoring network; The state acquisition module is used to enhance anti-interference processing of the multi-mode sensor monitoring network and collect UAV operating parameters in real time to obtain real-time operating status data. The real-time operating status data is processed based on the time series evolution of the cruising range, moving speed, and remaining power to generate parameter dynamic characteristic data. The scheduling prediction module is used to establish a parameter-demand response relationship model based on parameter dynamic characteristic data and spatial topology data; analyze the power consumption characteristics based on the parameter-demand response relationship model, and conduct distance-related charging demand assessment to generate dynamic power scheduling data; The impact assessment module is used to process the drone's endurance correlation based on the dynamic data of power scheduling and real-time operating status data to generate dynamic characteristic data of endurance assurance; identify the charging impact area based on the dynamic characteristic data of endurance assurance to generate node impact range data; and couple the demand pattern based on the moving speed, remaining power and scheduling efficiency based on the node impact range data to generate demand-action coupling data; The control feedback module is used to build a machine learning scheduling model based on the demand-action coupling data, and use the machine learning scheduling model to dynamically plan the charging path to generate path planning data; based on the path planning data, the operating parameters are optimized in real time to obtain the UAV tracking control feedback data.

Images