An adaptive charging location optimization method for intelligent streetlight drones using electromagnetic coupling

Abstract

This invention provides an adaptive charging position optimization method for intelligent street light drones with electromagnetic coupling, relating to the field of drone charging. The method includes acquiring street light location and drone battery information; constructing a spiral detection grid to collect electromagnetic field data and generate a field strength distribution map; selecting the region with the highest field strength to set pose combinations; evaluating the ratio of charging efficiency to hovering power consumption; determining the optimal pose and planning the energy-optimal path; and controlling the drone to adjust deviations in real time to establish a charging connection. This invention improves the charging efficiency of drones in electromagnetically coupled environments and extends their flight time.

Description

Technical Field

[0001] This invention relates to drone charging technology, and more particularly to an adaptive charging location optimization method for electromagnetically coupled smart street light drones. Background Technology

[0002] With the widespread application of drones in logistics, environmental monitoring, and security patrols, their insufficient battery life has become increasingly prominent. Traditional drone charging methods mainly rely on wired charging or battery replacement, which not only require manual intervention but also suffer from low charging efficiency and poor flexibility. Therefore, wireless charging technology based on electromagnetic coupling has gradually become a research hotspot. This technology enables contactless energy transfer between the drone and the charging device, reducing human intervention and improving the drone's mission continuity and operational efficiency.

[0003] In practical applications, the efficiency of electromagnetic coupling is often unstable due to the susceptibility of drones to wind, airflow disturbances, and attitude changes while hovering. Furthermore, the spatial distribution of the electromagnetic field is significantly influenced by the structure, location, and power parameters of the charging device. Therefore, determining the optimal charging location and attitude for the drone is a crucial factor affecting wireless charging efficiency. Existing research largely focuses on improving coil structure or enhancing power control, but lacks adaptive optimization strategies that combine drone flight characteristics with the spatial electromagnetic field distribution. In addition, in complex urban or outdoor environments, obstacles and uneven electromagnetic environments may exist around the charging location. Failure to effectively plan the drone's charging path and hovering attitude can easily lead to excessive energy loss and even increase flight safety hazards.

[0004] Therefore, how to dynamically optimize the charging location and attitude and generate the most energy-efficient flight path by combining the spatial distribution of electromagnetic fields and the energy consumption characteristics of drones during mission execution has become an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides an adaptive charging location optimization method for an electromagnetically coupled intelligent street light drone, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides an adaptive charging position optimization method for an electromagnetically coupled intelligent street light drone, comprising:

[0007] The system obtains the location coordinates and output power of the smart street light, as well as the remaining battery power and charging request of the drone. A spiral detection grid is constructed around the smart street light, and the sampling interval is adaptively adjusted according to the location of the detection points. Electromagnetic field intensity data of each detection point is collected, and surface fitting is performed by combining the data of adjacent points to generate an electromagnetic field intensity distribution map.

[0008] Based on the electromagnetic field intensity distribution map, the three regions with the strongest field intensity are selected as candidate charging regions. Position and attitude sampling combinations are set in the candidate charging regions, and charging efficiency and hovering power consumption data under each combination are collected. The hovering power consumption ratio corresponding to the unit charging efficiency of each sampling combination is calculated as a comprehensive scoring index.

[0009] The optimal target position and attitude are determined based on the comprehensive scoring index, and a path planning space is established in combination with the distribution of environmental obstacles to generate a flight path that takes energy consumption into account.

[0010] The drone is controlled to execute the flight path, and its position and attitude deviations are adjusted in real time. After reaching the target position, an electromagnetic coupling charging connection is established.

[0011] In one alternative embodiment,

[0012] A spiral detection grid is constructed around the smart streetlight. The sampling interval is adaptively adjusted according to the location of the detection points to collect electromagnetic field intensity data at each detection point. The data from adjacent points are then combined to perform surface fitting, generating an electromagnetic field intensity distribution map, including:

[0013] A polar coordinate system is established with the location of the smart street light as the origin. The initial sampling radius is determined according to the power level of the street light, and an initial spiral detection grid is constructed within the initial sampling radius.

[0014] Calculate the electromagnetic field strength difference between adjacent detection points in the initial spiral detection grid. When the difference exceeds the preset strength difference threshold, determine the sampling interval adjustment coefficient in the current detection area based on the radiation directionality of the smart street light and the distribution characteristics of the obstruction, and construct a denser spiral detection grid.

[0015] Electromagnetic field intensity data of the encrypted spiral detection grid are obtained. An outlier detection method based on spatial correlation is adopted. By calculating the covariance of the electromagnetic field distribution between the detection point and its corresponding neighboring detection points, outlier detection points that do not conform to the spatial continuity characteristics are identified.

[0016] The abnormal detection points were repeatedly measured and the data was corrected by combining the radiation characteristics of the smart street light. The radial basis function considering the electromagnetic field attenuation law was used to perform surface fitting on all electromagnetic field intensity data of the initial spiral detection grid and the densified spiral detection grid to generate an electromagnetic field intensity distribution map.

[0017] In one alternative embodiment,

[0018] Repeated measurements were taken at abnormal detection points, and data correction was performed based on the radiation characteristics of the smart streetlights. A radial basis function, considering the electromagnetic field attenuation law, was used to perform surface fitting on all electromagnetic field intensity data from both the initial and refined spiral detection grids, generating an electromagnetic field intensity distribution map, including:

[0019] Obtain the location information of the abnormal detection points, perform multiple repeated measurements at the abnormal detection point locations, calculate the standard deviation and mean of the multiple measurement data for each abnormal detection point, remove the measurement values ​​that exceed the preset standard deviation range, and calculate the average value of the remaining measurement data as the preliminary correction data for the abnormal detection points;

[0020] The distance attenuation coefficient from the abnormal detection point to the intelligent street light is calculated based on the power level of the intelligent street light, and the direction attenuation coefficient is calculated based on the azimuth angle of the abnormal detection point in the polar coordinate system. The preliminary correction data is multiplied by the distance attenuation coefficient and the direction attenuation coefficient respectively to obtain the theoretical correction value considering the dual attenuation characteristics.

[0021] The ratio of the preliminary corrected data to the theoretical corrected value is calculated as the correction ratio. Based on the correction ratio, the detection point data in the grid area around the anomaly detection point is linearly interpolated to generate the final corrected data of the anomaly detection point.

[0022] The final corrected data of abnormal detection points is merged with the normal detection point data in the initial spiral detection grid and the densified spiral detection grid to form a complete electromagnetic field intensity dataset.

[0023] In the polar coordinate system, the detection point is divided into several sectors. The local correlation coefficient of the detection point in each sector is calculated. The local correlation coefficient is used as the weight coefficient for piecewise fitting. The complete electromagnetic field intensity dataset is then fitted to the surface in different regions to generate an electromagnetic field intensity distribution map.

[0024] In one alternative embodiment,

[0025] Based on the electromagnetic field intensity distribution map, the three regions with the strongest field intensity were selected as candidate charging regions. Within these candidate charging regions, position and attitude sampling combinations were set up, and charging efficiency and hovering power consumption data were collected for each combination. The hovering power consumption ratio corresponding to the unit charging efficiency of each sampling combination was calculated as a comprehensive scoring index, including:

[0026] The field strength values ​​in the electromagnetic field intensity distribution map are sorted in descending order, and the regions with the largest field strength values ​​are selected as initial candidate regions. From the initial candidate regions, regions that meet the preset minimum area and the distance between regions meets the preset minimum interval requirement are selected as candidate charging regions, and the field strength distribution data of the candidate charging regions are recorded.

[0027] Based on the field strength distribution data of the candidate charging areas, an adaptive position sampling grid is established in each candidate charging area. Multiple sets of horizontal yaw angle parameters and vertical pitch angle parameters are set at each sampling point of the adaptive position sampling grid to form a combined position-attitude sampling scheme.

[0028] Collect induced voltage and charging current data at each sampling combination location in the combined sampling scheme, calculate the charging efficiency value, and record the hovering power consumption data of the UAV at the sampling combination location;

[0029] Based on the charging efficiency value and hovering power consumption data, the ratio of charging efficiency to hovering power consumption under each position-attitude combination sampling scheme is calculated, and the ratio of charging efficiency to hovering power consumption is used as a comprehensive scoring index.

[0030] In one alternative embodiment,

[0031] The optimal target position and attitude are determined based on the comprehensive scoring index. A path planning space is established by considering the distribution of environmental obstacles, and a flight path with optimal energy consumption is generated, including:

[0032] All candidate charging positions are sorted according to the comprehensive score index of position-attitude combination, and the position with the highest comprehensive score index is selected as the target charging position. The attitude parameters corresponding to the target charging position are taken as the target charging attitude.

[0033] Acquire spatial distribution information of static and dynamic obstacles in the environment, combine it with the current remaining battery power information of the drone to determine the maximum flight distance constraint, mark the area exceeding the maximum flight distance constraint and the spatial distribution area of ​​obstacles as non-flyable areas, and generate a passable path planning space;

[0034] Within the passable path planning space, a path search range is set with the current position of the drone as the starting point and the target charging position as the ending point. A sampling point grid with a preset density is generated within the path search range, and the path energy consumption value between adjacent sampling points is calculated.

[0035] Based on the path energy consumption value, the adjacent sampling points with the lowest energy consumption are selected to construct the flight path. The flight path is then smoothed and verified to meet the constraints of the path planning space, generating the final flight path.

[0036] In one alternative embodiment,

[0037] Based on the path energy consumption value, the adjacent sampling points with the lowest energy consumption are selected to construct the flight path. The flight path is then smoothed and verified to meet the constraints of the path planning space. The final flight path is generated as follows:

[0038] Based on the path energy consumption value, the optimal neighboring sampling point of the current sampling point is selected. Among them, the comprehensive path energy consumption value of the selected neighboring sampling point is the smallest and the change in heading angle is less than the maximum turning angle of the UAV. The neighboring sampling points of the selected sampling point are sorted according to the path energy consumption value and added to the candidate sequence. The selection process is repeated until the target position is found. The selected sampling point sequence is connected to construct the initial flight path.

[0039] Extract the turning points in the initial flight path where the heading angle change is greater than a preset threshold, establish a local adjustment area centered on the turning point, determine the size of the adjustment area based on the minimum turning radius of the UAV, generate transition path candidate points within the adjustment area, calculate the connection angle and path length between each candidate point and the original path, select the candidate point sequence that satisfies the turning radius constraint and has the smallest path increment as the transition path, and replace the original turning point with the transition path to obtain a smooth flight path.

[0040] Based on the path planning space calculation, the minimum distance from the sampling point to the obstacle boundary on the smooth flight path is calculated. When the minimum distance is less than the safety distance threshold, the adjustment area of ​​the corresponding turning position is expanded to regenerate the transition path until the smooth flight path meets the safety distance requirement. The smooth flight path that meets all constraints is taken as the final flight path.

[0041] In one alternative embodiment,

[0042] Controlling the drone to execute the flight path, adjusting its position and attitude deviations in real time, and establishing an electromagnetic coupling charging connection upon reaching the target position includes:

[0043] A reference trajectory is established by obtaining a sequence of sampling points based on the planned flight path. A sliding prediction window is selected on the reference trajectory. The prediction trajectory within the prediction window is calculated based on the current position and attitude of the UAV. The flight path deviation is calculated between the prediction trajectory and the reference trajectory. The control gain is determined based on the flight path deviation and the real-time wind speed.

[0044] Based on the control gain, a trajectory adjustment command is generated, and the UAV is controlled to execute the trajectory adjustment command to eliminate flight path deviation. The actual position and attitude after adjustment are obtained, the prediction window is updated, and the trajectory adjustment is repeated. When the UAV reaches the preset distance range of the target charging position, the charging seat features are obtained to establish a relative position relationship, the alignment deviation of the UAV relative to the charging seat is calculated, and an alignment command is generated based on the alignment deviation.

[0045] The alignment command is executed to approach the charging base, contact force data is acquired, the approach speed and contact posture are adjusted according to the contact force data, the position is maintained after a stable contact is established, electromagnetic field data is acquired, and the coupling position is adjusted according to the electromagnetic field data to establish an electromagnetic coupling charging connection.

[0046] Real-time monitoring of wind speed and charging status; when wind speed exceeds a safety threshold or charging status is abnormal, control the drone to disconnect the electromagnetic coupling charging connection and move away from the target charging location.

[0047] A second aspect of the present invention provides an electronic device, comprising:

[0048] processor;

[0049] Memory used to store processor-executable instructions;

[0050] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0051] A third aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0052] In this embodiment, an electromagnetic field intensity distribution map is accurately obtained by using a spiral detection grid around the smart streetlight and adaptive sampling interval technology. This effectively solves the problem of insufficient positioning accuracy under traditional fixed sampling methods and improves the optimization accuracy of the drone charging location. By comprehensively considering the ratio of charging efficiency to hovering power consumption as a scoring index, this invention achieves intelligent selection of the optimal charging position and attitude, significantly improving drone charging efficiency, reducing hovering energy consumption, and extending the drone's endurance and operational capabilities. The invention also constructs an energy-optimal flight path based on the distribution of environmental obstacles and realizes real-time drone attitude adjustment and adaptive charging connection functions, enhancing the system's adaptability and stability in complex environments and improving the safety and reliability of the drone charging process. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the adaptive charging position optimization method for an electromagnetically coupled intelligent street light drone according to an embodiment of the present invention.

[0054] Figure 2 This is a flowchart of an automatic charging system for unmanned aerial vehicles according to an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the adaptive charging position optimization method for an electromagnetically coupled intelligent street light drone according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0058] The system obtains the location coordinates and output power of the smart street light, as well as the remaining battery power and charging request of the drone. A spiral detection grid is constructed around the smart street light, and the sampling interval is adaptively adjusted according to the location of the detection points. Electromagnetic field intensity data of each detection point is collected, and surface fitting is performed by combining the data of adjacent points to generate an electromagnetic field intensity distribution map.

[0059] Based on the electromagnetic field intensity distribution map, the three regions with the strongest field intensity are selected as candidate charging regions. Position and attitude sampling combinations are set in the candidate charging regions, and charging efficiency and hovering power consumption data under each combination are collected. The hovering power consumption ratio corresponding to the unit charging efficiency of each sampling combination is calculated as a comprehensive scoring index.

[0060] The optimal target position and attitude are determined based on the comprehensive scoring index, and a path planning space is established in combination with the distribution of environmental obstacles to generate a flight path that takes energy consumption into account.

[0061] The drone is controlled to execute the flight path, and its position and attitude deviations are adjusted in real time. After reaching the target position, an electromagnetic coupling charging connection is established.

[0062] In one optional implementation, a spiral detection grid is constructed around the smart street light. The sampling interval is adaptively adjusted according to the location of the detection points. Electromagnetic field intensity data is collected at each detection point, and surface fitting is performed by combining the data from adjacent points to generate an electromagnetic field intensity distribution map, including:

[0063] A polar coordinate system is established with the location of the smart street light as the origin. The initial sampling radius is determined according to the power level of the street light, and an initial spiral detection grid is constructed within the initial sampling radius.

[0064] Calculate the electromagnetic field strength difference between adjacent detection points in the initial spiral detection grid. When the difference exceeds the preset strength difference threshold, determine the sampling interval adjustment coefficient in the current detection area based on the radiation directionality of the smart street light and the distribution characteristics of the obstruction, and construct a denser spiral detection grid.

[0065] Electromagnetic field intensity data of the encrypted spiral detection grid are obtained. An outlier detection method based on spatial correlation is adopted. By calculating the covariance of the electromagnetic field distribution between the detection point and its corresponding neighboring detection points, outlier detection points that do not conform to the spatial continuity characteristics are identified.

[0066] The abnormal detection points were repeatedly measured and the data was corrected by combining the radiation characteristics of the smart street light. The radial basis function considering the electromagnetic field attenuation law was used to perform surface fitting on all electromagnetic field intensity data of the initial spiral detection grid and the densified spiral detection grid to generate an electromagnetic field intensity distribution map.

[0067] When constructing a spiral detection grid around a smart street light, adaptive sampling technology can be used to accurately measure the electromagnetic field intensity distribution. A polar coordinate system is established with the smart street light's location as the origin, and the initial sampling radius is determined based on the street light's power rating. For example, for a 500W smart street light, the initial sampling radius is set to 20 meters; for an 800W smart street light, it is set to 25 meters; and for a 1000W smart street light, it is set to 30 meters. Within the determined initial sampling radius, an initial spiral detection grid is constructed. This spiral grid expands outwards from the center point of the street light, with an initial sampling angle interval of 15 degrees and a radial distance of 2 meters between adjacent spiral rings.

[0068] After the initial spiral detection grid is constructed, the difference in electromagnetic field strength between adjacent detection points needs to be calculated. Specifically, a portable electromagnetic field strength meter can be used to collect data at each detection point and record the electromagnetic field strength value. For any two adjacent points, if the difference in electromagnetic field strength exceeds a preset threshold (usually set to 10% of the electromagnetic field strength measurement range), it indicates that the electromagnetic field gradient in that area is large, requiring grid refinement. After determining the areas requiring refinement, a sampling interval adjustment coefficient is calculated based on the radiation directionality of the smart streetlights and the distribution characteristics of obstructions in the surrounding environment. For example, the sampling interval adjustment coefficient is set to 0.5 in unobstructed areas, 0.3 in areas with metallic obstructions, and 0.4 in areas with non-metallic obstructions. Multiplying the original sampling interval by the corresponding adjustment coefficient yields the new sampling interval for the refined area, thus constructing the refined spiral detection grid.

[0069] After the encrypted spiral detection grid is constructed, electromagnetic field strength is measured at each detection point in the grid to obtain a complete electromagnetic field strength dataset. Due to potential errors during measurement, an outlier detection method based on spatial correlation is needed to filter the data. This method is based on the principle that the electromagnetic field should have a certain continuity in spatial distribution, and identifies outliers by calculating the covariance of the electromagnetic field distribution between a detection point and its neighboring points. Specifically, for any detection point, all detection points within a radius r around it are selected, and the electromagnetic field strength covariance matrix between that point and its surrounding points is calculated. The eigenvalues ​​of the covariance matrix reflect the distribution characteristics of the data. When the ratio of the largest eigenvalue to the smallest eigenvalue exceeds a preset threshold (usually set to 10), the detection point is determined to be an outlier. In practical applications, the detection radius r is usually set to 5% of the initial sampling radius; for an initial sampling radius of 20 meters, r is 1 meter.

[0070] For identified abnormal detection points, three repeated measurements are performed, and the average value is taken. Data correction is then performed based on the radiation characteristics of the smart streetlight. The radiation characteristics of smart streetlights typically exhibit a decreasing trend with increasing distance, and may differ in different directions. Data correction uses a weighted average method, with the weighting coefficient determined based on the distance between the measurement point and the streetlight, as well as the azimuth angle. For example, for a measurement point d meters away from the streetlight with an azimuth angle of θ, the weighting coefficient can be set as w = 1 / (1+0.1d) × (1+0.05×cos(θ)).

[0071] After data processing, a radial basis function considering the electromagnetic field attenuation law is used to perform surface fitting on all electromagnetic field intensity data of the initial spiral detection grid and the refined spiral detection grid. The electromagnetic field intensity attenuates with distance according to an inverse square law; therefore, a radial basis function with a distance term is selected for fitting. Specifically, for any point at position (r, θ), its electromagnetic field intensity value can be calculated by weighting and summing all measurement points. The weighting function is a Gaussian radial basis function, and its characteristic width parameter is adaptively determined based on the data distribution characteristics. For example, in electromagnetic field measurements around a 500W smart street light, the initial spiral grid contains 120 detection points. 15 areas requiring refinement are identified, and after refinement, a total of 85 detection points are added, ultimately forming a detection grid with 205 effective detection points. Measurement data shows that the electromagnetic field intensity is highest directly below the street light, reaching 120 μT, and gradually decreases with increasing distance, dropping below 2 μT at 20 meters. The electromagnetic field intensity distribution map generated by surface fitting clearly shows the intensity variation trend. The electromagnetic field intensity shows obvious asymmetric attenuation in the northeast direction due to the presence of metal facilities, while the attenuation in the southwest direction is more uniform.

[0072] This electromagnetic field intensity distribution map can be used for adaptive charging location optimization of drones. By identifying the optimal location of the electromagnetic field intensity within the charging efficiency threshold range, and considering factors such as charging stability and minimization of electromagnetic interference, efficient electromagnetic coupling charging between drones and smart streetlights can be achieved.

[0073] By constructing a spiral detection grid around smart streetlights and combining it with adaptive sampling, high-precision modeling of the spatial electromagnetic field distribution was achieved. Dynamically adjusting the sampling interval based on the difference between detection points enables local densification in areas of drastic field strength changes, improving the refinement of electromagnetic field data and avoiding blind spots caused by insufficient sampling. Spatial correlation is used for outlier identification and correction, effectively eliminating data deviations caused by environmental interference or measurement errors, ensuring data continuity and reliability. Combining radial basis functions and incorporating electromagnetic field attenuation laws for surface fitting more realistically reflects the spatial distribution characteristics of the electromagnetic field. The resulting electromagnetic field intensity distribution map is not only high-resolution but also highly accurate, providing a reliable decision-making basis for UAVs to select the optimal charging position and attitude during electromagnetic coupling charging, thereby significantly improving charging efficiency and flight energy utilization.

[0074] In one optional implementation, the abnormal detection points are repeatedly measured and the data is corrected by combining the radiation characteristics of the smart street light. A radial basis function considering the electromagnetic field attenuation law is used to perform surface fitting on all electromagnetic field intensity data of the initial spiral detection grid and the refined spiral detection grid to generate an electromagnetic field intensity distribution map, including:

[0075] Obtain the location information of the abnormal detection points, perform multiple repeated measurements at the abnormal detection point locations, calculate the standard deviation and mean of the multiple measurement data for each abnormal detection point, remove the measurement values ​​that exceed the preset standard deviation range, and calculate the average value of the remaining measurement data as the preliminary correction data for the abnormal detection points;

[0076] The distance attenuation coefficient from the abnormal detection point to the intelligent street light is calculated based on the power level of the intelligent street light, and the direction attenuation coefficient is calculated based on the azimuth angle of the abnormal detection point in the polar coordinate system. The preliminary correction data is multiplied by the distance attenuation coefficient and the direction attenuation coefficient respectively to obtain the theoretical correction value considering the dual attenuation characteristics.

[0077] The ratio of the preliminary corrected data to the theoretical corrected value is calculated as the correction ratio. Based on the correction ratio, the detection point data in the grid area around the anomaly detection point is linearly interpolated to generate the final corrected data of the anomaly detection point.

[0078] The final corrected data of abnormal detection points is merged with the normal detection point data in the initial spiral detection grid and the densified spiral detection grid to form a complete electromagnetic field intensity dataset.

[0079] In the polar coordinate system, the detection point is divided into several sectors. The local correlation coefficient of the detection point in each sector is calculated. The local correlation coefficient is used as the weight coefficient for piecewise fitting. The complete electromagnetic field intensity dataset is then fitted to the surface in different regions to generate an electromagnetic field intensity distribution map.

[0080] For example, when repeatedly measuring anomaly detection points and correcting the data based on the radiation characteristics of the smart streetlight, it is necessary to obtain the precise location information of the anomaly detection points, including their radial distance and angle values ​​in the polar coordinate system. At each identified anomaly detection point location, five independent repeated measurements are performed using a high-precision electromagnetic field measurement device, with a 30-second interval between each measurement to eliminate environmental interference. The standard deviation and mean of the five sets of collected data are calculated. If a measurement value deviates from the mean by more than twice the standard deviation, it is considered an outlier and is discarded. In practical applications, for example, for an anomaly detection point located 15 meters from the smart streetlight at an azimuth angle of 45 degrees, the electromagnetic field strength values ​​obtained from the five measurements are 26.5 μT, 25.8 μT, 32.1 μT, 27.2 μT, and 26.9 μT, respectively, with a calculated mean of 27.7 μT and a standard deviation of 2.45 μT. Since the deviation of 32.1 μT from the mean exceeded twice the standard deviation (4.9 μT), it was removed, and the average of the remaining four measurements, 26.6 μT, was used as the preliminary correction data for this abnormal detection point.

[0081] When calculating the distance attenuation coefficient of anomaly detection points based on the power level of smart streetlights, the physical law of electromagnetic field attenuation with distance must be considered. For a smart streetlight with a power of P watts, the distance attenuation coefficient at a distance of d meters can be calculated by normalizing the reference power and reference distance. Specifically, 500W is selected as the reference power and 10 meters as the reference distance. For a smart streetlight with a power of P, the distance attenuation coefficient at a distance of d meters is the square of the reference distance multiplied by the actual power divided by the ratio of the reference power to the square of the actual distance. Meanwhile, the radiation characteristics of smart streetlights also differ in azimuth. Generally, due to the design of the lamp structure, the radiation is strongest directly below (0 degrees), gradually weakening as the azimuth angle increases. The directional attenuation coefficient can be approximated by a cosine function, specifically 1 minus 0.2 multiplied by the ratio of the azimuth angle to 360 degrees. For a point with an azimuth angle of 45 degrees, the directional attenuation coefficient is 1 - 0.2 × (45 / 360) ≈ 0.975. Multiplying the initial corrected value of 26.6 μT by the distance attenuation coefficient of 0.71 and the directional attenuation coefficient of 0.975 respectively, we obtain the theoretical correction value considering the dual attenuation characteristics: 26.6 × 0.71 × 0.975 ≈ 18.4 μT.

[0082] When calculating the correction ratio, the preliminary correction data is divided by the theoretical correction value to obtain the correction ratio. In the above case, the correction ratio is 26.6 / 18.4 ≈ 1.45, indicating that the actual measured value is about 45% higher than the theoretical prediction. This deviation may be due to local electromagnetic field enhancement effects or systematic errors in the measurement equipment. To ensure the spatial continuity of the data, it is necessary to correct the detection point data within the grid area surrounding the anomaly detection point. Specifically, an influence radius (usually twice the initial sampling interval) is defined centered on the anomaly detection point, and a linearly decaying correction coefficient is applied to all detection point data within this range. The correction coefficient decreases linearly with the distance from the point to the anomaly detection point, from complete correction at the anomaly detection point (correction ratio of 1.45) to no correction at the edge of the influence radius (correction ratio of 1). The difference between the corrected data at the anomaly detection point and the initial data is multiplied by the correction coefficient and added to the initial data to obtain the corrected data. For a detection point 1.5 meters away from the abnormal detection point and with an initial electromagnetic field strength of 25.0 μT, if the influence radius is 4 meters, the correction coefficient is (4-1.5) / 4×(1.45-1)+1≈1.17, and the corrected electromagnetic field strength is 25.0×1.17≈29.25 μT.

[0083] The final corrected data from all abnormal detection points were merged with the normal detection point data from the initial and refined spiral detection grids to form a complete electromagnetic field strength dataset. In a typical smart street light electromagnetic field measurement case, the initial spiral detection grid contained 150 detection points, and the refined spiral detection grid added 65 detection points, of which 12 abnormal detection points were identified and corrected, ultimately forming a complete dataset containing 215 valid detection points. To improve the accuracy of surface fitting, the entire measurement area was divided into 8 equal-angle sectors in polar coordinates, each covering a 45-degree angle range. Within each sector, the local correlation coefficient between detection points was calculated, reflecting the correlation strength of the data within the region. The local correlation coefficient was obtained by calculating the Pearson correlation coefficient of electromagnetic field strength values ​​between all point pairs within the sector; the closer the value is to 1, the stronger the correlation. For example, the local correlation coefficient calculated in the 0-45 degree sector is 0.92, in the 45-90 degree sector it is 0.85, and in other sectors it is 0.88, 0.90, 0.86, 0.91, 0.89 and 0.87 respectively.

[0084] The local correlation coefficients of each sector are used as weighting coefficients for piecewise fitting, and regional surface fitting is performed on the complete electromagnetic field intensity dataset. During the fitting process, for data points at sector boundaries, the influence of two adjacent sectors is considered simultaneously, and a weighted average is used to ensure the continuity of the surface at the boundaries. Specifically, a multivariate radial basis function is used for fitting, with a Gaussian radial basis function as the kernel function, and its width parameter is adaptively adjusted according to the density of data distribution within each sector. The electromagnetic field intensity distribution map generated by surface fitting visually displays the electromagnetic field distribution around the smart street light. This accurate electromagnetic field distribution map provides an important reference for UAVs to adaptively select the optimal charging location. By identifying areas with suitable and stable electromagnetic field intensity, UAVs can achieve high-efficiency and high-stability electromagnetic coupling charging.

[0085] In this embodiment, by repeatedly measuring abnormal detection points, eliminating outliers, and correcting the data using the radiation characteristics of smart streetlights, measurement noise and environmental interference are effectively eliminated, improving the reliability of electromagnetic field data. The introduction of dual corrections for distance attenuation and direction attenuation makes the corrected data for abnormal points more consistent with actual electromagnetic radiation patterns. Furthermore, linear interpolation is used to correct the surrounding grid, achieving continuity and smoothness of the spatial data. Merging the corrected abnormal point data with the normal detection point data and employing regionally weighted surface fitting accurately reflects the distribution characteristics of the electromagnetic field in different spatial regions. The overall effect is a more complete, accurate, and high-resolution electromagnetic field intensity distribution map, providing a reliable basis for UAV charging location selection, attitude optimization, and energy efficiency assessment, while significantly improving the scientific rigor and efficiency of charging decisions.

[0086] In one optional implementation, based on the electromagnetic field intensity distribution map, the three regions with the highest field intensity are selected as candidate charging regions. Within these candidate charging regions, position and attitude sampling combinations are set up, and charging efficiency and hovering power consumption data for each combination are collected. The hovering power consumption ratio corresponding to the unit charging efficiency of each sampling combination is calculated as a comprehensive scoring index, including:

[0087] The field strength values ​​in the electromagnetic field intensity distribution map are sorted in descending order, and the regions with the largest field strength values ​​are selected as initial candidate regions. From the initial candidate regions, regions that meet the preset minimum area and the distance between regions meets the preset minimum interval requirement are selected as candidate charging regions, and the field strength distribution data of the candidate charging regions are recorded.

[0088] Based on the field strength distribution data of the candidate charging areas, an adaptive position sampling grid is established in each candidate charging area. Multiple sets of horizontal yaw angle parameters and vertical pitch angle parameters are set at each sampling point of the adaptive position sampling grid to form a combined position-attitude sampling scheme.

[0089] Collect induced voltage and charging current data at each sampling combination location in the combined sampling scheme, calculate the charging efficiency value, and record the hovering power consumption data of the UAV at the sampling combination location;

[0090] Based on the charging efficiency value and hovering power consumption data, the ratio of charging efficiency to hovering power consumption under each position-attitude combination sampling scheme is calculated, and the ratio of charging efficiency to hovering power consumption is used as a comprehensive scoring index.

[0091] In this embodiment, based on the generated electromagnetic field intensity distribution map, the field strength values ​​in the distribution map are sorted in descending order to select the optimal charging area. First, the data matrix of the electromagnetic field intensity distribution map is imported into the processing program, the field strength value of each grid point is extracted, and a field strength value-location index table is constructed. All grid points are arranged in descending order of field strength value, and the top 15% of grid points are selected as initial candidate areas. For a 300W smart street light, the generated electromagnetic field intensity distribution map contains approximately 5000 grid points, and the initial candidate areas contain approximately 750 grid points. The initial candidate areas are often discontinuous, requiring region clustering to group grid points with adjacent distances less than 0.5 meters into the same area. After clustering, areas that meet a preset minimum area and a preset minimum interval requirement are selected as candidate charging areas. The preset minimum area is typically set to 0.8 square meters, sufficient to accommodate the hovering range of a medium-sized drone; the preset minimum interval requirement is set to 2 meters to ensure minimal interference between different charging areas. After screening, the three areas with the highest electric field strength were selected as candidate charging areas: 4.5 meters directly below the smart streetlight (Area A), 8.2 meters to the southeast (Area B), and 7.5 meters to the northwest (Area C). The average electric field strength in Area A was 45 μT, in Area B it was 32 μT, and in Area C it was 28 μT.

[0092] Based on the field strength distribution data of the candidate charging regions, an adaptive position sampling grid is established within each candidate charging region. The sampling grid density is proportional to the field strength gradient within the region; the larger the gradient, the denser the sampling points. The standard deviation of the field strength within each candidate region is calculated as a gradient evaluation index. The standard deviation of the field strength in region A is 3.2 μT, in region B it is 2.4 μT, and in region C it is 2.1 μT. Based on the standard deviation results, the sampling interval for region A is set to 0.3 meters, and the sampling interval for regions B and C is set to 0.4 meters. Based on the formed sampling grid, multiple sets of attitude parameter combinations are set at each sampling point. The horizontal yaw angle parameter ranges from 0 degrees to 345 degrees, with an interval of 45 degrees, for a total of 8 yaw angle values; the vertical pitch angle parameter ranges from -30 degrees to 30 degrees, with an interval of 15 degrees, for a total of 5 pitch angle values. Each sampling point forms 8 × 5 = 40 position-attitude combinations. Taking region A as an example, the resulting sampling grid contains 25 sampling points, and a total of 25 × 40 = 1000 position-attitude combinations need to be tested.

[0093] When collecting induced voltage and charging current data at each sampling combination location in the combined sampling scheme, a UAV test platform equipped with an electromagnetic receiving coil was used. The UAV flew and hovered sequentially according to preset position-attitude combinations, staying at each sampling combination location for 10 seconds, and the induced voltage and charging current data were recorded through the data acquisition system. At the same time, the power consumption data of the UAV while hovering at that location was recorded, including the sum of motor power and control system power. For a certain sampling location in area A (X=4.2 meters, Y=0.3 meters, Z=4.5 meters), under an attitude of 0 degrees yaw angle in the horizontal plane and 0 degrees pitch angle in the vertical plane, the induced voltage was measured to be 12.5V, the charging current was 1.6A, and the charging power was 20.0W; the UAV hovering power consumption was 120W. At the same location, when the horizontal yaw angle was adjusted to 45 degrees and the vertical pitch angle was adjusted to 15 degrees, the measured induced voltage dropped to 11.2V, the charging current dropped to 1.4A, and the charging power was 15.7W; the drone's hovering power consumption increased to 132W, which is because the drone needs more power to maintain hovering stability due to the deviation from the balance attitude.

[0094] Based on the collected charging efficiency and hovering power consumption data, a comprehensive score index for each position-attitude combination is calculated. The charging efficiency value is equal to the ratio of the actual charging power to the theoretical maximum charging power, where the theoretical maximum charging power is calculated based on the maximum field strength of the current area and the parameters of the UAV receiving coil. For area A, the theoretical maximum charging power is 30W, the measured highest charging power is 20.0W, and the charging efficiency is 66.7%. For area B, the theoretical maximum charging power is 22W, the measured highest charging power is 14.5W, and the charging efficiency is 65.9%. For area C, the theoretical maximum charging power is 20W, the measured highest charging power is 13.2W, and the charging efficiency is 66.0%. The ratio of charging efficiency to hovering power consumption is obtained by dividing the charging efficiency value by the hovering power consumption, expressed as a percentage per watt. In area A, the charging efficiency of a certain sampling combination is 66.7%, the hovering power consumption is 120W, and the ratio of charging efficiency to hovering power consumption is 0.556% / W. In region B, the optimal sampling combination achieves a charging efficiency of 65.9% and a hovering power consumption of 115W, with a charging efficiency to hovering power consumption ratio of 0.573% / W. In region C, the optimal sampling combination achieves a charging efficiency of 66.0% and a hovering power consumption of 118W, with a charging efficiency to hovering power consumption ratio of 0.559% / W.

[0095] The comprehensive score index of all position-attitude combinations was ranked, and the combination with the highest score was selected as the optimal charging position and attitude. In the test case, the combination of position coordinates (X=7.8 m, Y=2.6 m, Z=5.2 m), horizontal yaw angle of 90 degrees, and vertical pitch angle of 0 degrees in region B achieved the highest score of 0.573% / W. This result indicates that although region A has a higher absolute field strength and charging power, region B has better overall charging efficiency due to more stable drone hovering and lower power consumption. The selection of the optimal charging position and attitude fully considers the balance between charging efficiency and energy consumption, rather than simply pursuing the maximum field strength. Experimental data also shows that the horizontal yaw angle has a significant impact on charging efficiency, and the efficiency is usually highest when it is aligned with the polarization direction of the electromagnetic field; while the vertical pitch angle has a relatively small impact on charging efficiency within ±15 degrees, but a large impact on hovering power consumption. Through this comprehensive evaluation method, the drone can adaptively select the optimal charging location and attitude, forming an efficient electromagnetic coupling charging system around the smart streetlights. This greatly improves charging efficiency and reduces energy loss, providing a reliable energy replenishment solution for the drone to perform tasks for a long time and with high efficiency.

[0096] In this embodiment, by analyzing the electromagnetic field intensity distribution map, the region with the strongest field intensity is selected as the candidate charging region. A position-attitude combination sampling grid is then established within this region to achieve refined optimization of the UAV's charging position and attitude. By collecting induced voltage, charging current, and hovering power consumption data for each combination, the hovering power consumption ratio corresponding to unit charging efficiency is calculated as a comprehensive scoring index, enabling objective evaluation of charging performance under different positions and attitudes. The overall effect is to accurately identify the optimal charging region and attitude combination, balancing charging efficiency and energy consumption, enabling the UAV to achieve efficient and low-energy charging during electromagnetic coupling charging, while simultaneously improving the system's reliability and adaptability, providing a scientific basis for autonomous charging of the UAV.

[0097] In one optional implementation, the optimal target position and attitude are determined based on a comprehensive scoring index, and a path planning space is established in conjunction with the distribution of environmental obstacles to generate a flight path that optimizes energy consumption, including:

[0098] All candidate charging positions are sorted according to the comprehensive score index of position-attitude combination, and the position with the highest comprehensive score index is selected as the target charging position. The attitude parameters corresponding to the target charging position are taken as the target charging attitude.

[0099] Acquire spatial distribution information of static and dynamic obstacles in the environment, combine it with the current remaining battery power information of the drone to determine the maximum flight distance constraint, mark the area exceeding the maximum flight distance constraint and the spatial distribution area of ​​obstacles as non-flyable areas, and generate a passable path planning space;

[0100] Within the passable path planning space, a path search range is set with the current position of the drone as the starting point and the target charging position as the ending point. A sampling point grid with a preset density is generated within the path search range, and the path energy consumption value between adjacent sampling points is calculated.

[0101] Based on the path energy consumption value, the adjacent sampling points with the lowest energy consumption are selected to construct the flight path. The flight path is then smoothed and verified to meet the constraints of the path planning space, generating the final flight path.

[0102] In this embodiment, when sorting all candidate charging locations based on the comprehensive score index of position-attitude combination, a multi-level screening strategy is adopted to ensure the selection of the optimal charging location. The comprehensive score index of all position-attitude combinations within each candidate charging area is imported into the data processing module and sorted from high to low score. For the three candidate regions in the aforementioned test, totaling approximately 3000 position-attitude combinations, the top ten combinations with the highest scores after sorting were as follows: Region B (X=7.8m, Y=2.6m, Z=5.2m), horizontal yaw angle 90 degrees, vertical pitch angle 0 degrees, score 0.573% / W; Region A (X=4.5m, Y=0.2m, Z=4.5m), horizontal yaw angle 0 degrees, vertical pitch angle -15 degrees, score 0.571% / W; Region B (X=7.6m, Y=2.8m, Z=5.2m), horizontal yaw angle 90 degrees, vertical pitch angle 0 degrees, score 0.570% / W; and so on. The position with the highest score (X=7.8m, Y=2.6m, Z=5.2m) was selected as the target charging position, and the corresponding attitude parameters of horizontal yaw angle 90 degrees and vertical pitch angle 0 degrees were selected as the target charging attitude. To verify the stability of the selection results, five consecutive tests were conducted at this location, with charging efficiencies of 65.9%, 66.1%, 65.7%, 65.8%, and 66.0%, and hovering power consumption of 115W, 114W, 116W, 115W, and 114W, respectively. The corresponding comprehensive scores were all between 0.570% / W and 0.580% / W, indicating that the selection results have good stability.

[0103] When acquiring spatial distribution information of obstacles in the environment, multi-sensor fusion technology is combined to achieve all-round environmental perception. Three-dimensional point cloud data of the environment is acquired using lidar, visual cameras, and ultrasonic sensors mounted on a drone to identify static obstacles such as buildings, trees, and utility poles. Simultaneously, visual target tracking algorithms are used to detect dynamic obstacles such as moving vehicles and pedestrians. In a test case of the environment around a smart street light, the identified static obstacles included: a 6.5-meter-tall tree located at coordinates (X=5.2m, Y=-3.8m); a utility pole located at coordinates (X=-2.5m, Y=4.0m); and a fence located within the coordinate range (X=8.0m to X=12.0m, Y=-5.0m to Y=5.0m). Dynamic obstacles included the detected pedestrian activity area, mainly concentrated within the coordinate range (X=-3.0m to X=3.0m, Y=-8.0m to Y=-3.0m). The maximum flight distance constraint is determined by combining the current remaining battery power information of the drone. When the drone's battery power is 35%, the maximum safe flight distance calculated according to the energy consumption model is 250 meters. The area exceeding the maximum flight distance constraint and the aforementioned obstacle spatial distribution area are uniformly marked as a non-flying area. An obstacle expansion model is constructed in three-dimensional space, with the safe expansion radius of static obstacles set at 2 meters and the safe expansion radius of dynamic obstacles set at 3 meters, generating a passable path planning space.

[0104] When setting the path search range within the traversable path planning space, energy consumption optimization constraints are considered. Starting from the UAV's current position (X=-15.0 m, Y=10.0 m, Z=8.0 m) and ending at the target charging position (X=7.8 m, Y=2.6 m, Z=5.2 m), the straight-line distance between the two points is calculated to be 25.7 meters. To balance search efficiency and path quality, the path search range is set as an ellipsoid with the line connecting the start and end points as its central axis. The major axis is 1.5 times the distance between the start and end points, and the minor axis is 0.8 times the distance between the start and end points. A sampling point grid with a preset density is generated within this search range, with a sampling interval of 2 meters, resulting in approximately 420 effective sampling points. When calculating the path energy consumption between adjacent sampling points, three factors are comprehensively considered: horizontal flight energy consumption, vertical climb energy consumption, and attitude adjustment energy consumption. Horizontal flight energy consumption is directly proportional to flight distance and air resistance; vertical climb energy consumption is directly proportional to changes in altitude and gravitational potential energy; and attitude adjustment energy consumption is directly proportional to changes in attitude angle. For any two adjacent sampling points i and j, calculate the combined energy consumption from point i to point j. For example, the horizontal flight energy consumption from sampling point (X=-10.0 m, Y=8.0 m, Z=7.5 m) to sampling point (X=-8.0 m, Y=7.0 m, Z=7.0 m) is 2.8 J, the vertical descent energy consumption is -0.5 J (negative values ​​indicate potential energy release), the attitude adjustment energy consumption is 0.2 J, and the combined energy consumption is 2.5 J.

[0105] When constructing a flight path by selecting the combination of adjacent sampling points with the lowest energy consumption based on the path energy consumption value, an improved A* algorithm is used to achieve a globally energy-optimal path search. The evaluation function of the traditional A* algorithm is modified into a comprehensive energy consumption evaluation function, including two parts: known path energy consumption and estimated remaining energy consumption. The algorithm starts from the starting point and expands by selecting the node with the lowest energy consumption evaluation each time until a complete path to the destination is found. In the test case, after 82 iterations, the algorithm generated an initial path consisting of 14 key points with a total energy consumption of 98.5 J. This initial path is then smoothed using cubic spline interpolation to eliminate sharp turns while ensuring that the interpolation points remain within the passable space. The smoothed path contains 35 smooth transition points, with a slightly increased total energy consumption of 102.3 J, but significantly improved flight stability. Collision detection is performed on the smoothed path to verify that all points and their connections on the path satisfy the constraints of the path planning space. After successful verification, this path is determined as the final flight path. The generated flight path is as follows: Starting from the starting point (X=-15.0 m, Y=10.0 m, Z=8.0 m), the drone first descends to the southeast, avoiding obstacles such as utility poles. It then passes key points including the intermediate points (X=-8.5 m, Y=5.2 m, Z=6.8 m) and (X=-3.2 m, Y=1.8 m, Z=6.0 m), and finally approaches the target charging location from the northwest (X=7.8 m, Y=2.6 m, Z=5.2 m), simultaneously adjusting its attitude to a horizontal yaw angle of 90 degrees and a vertical pitch angle of 0 degrees to complete the entire flight. This path comprehensively considers obstacle avoidance, energy minimization, and smooth transition, ensuring the drone can safely and efficiently reach the optimal charging location and maximize electromagnetic coupling charging efficiency.

[0106] In this embodiment, the optimal charging location and attitude are selected through a comprehensive scoring index, and a path planning space is established by combining the distribution of environmental obstacles and the remaining battery power of the UAV, thus realizing intelligent optimization of the UAV charging path. By generating a sampling point grid and calculating the energy consumption of adjacent points, the flight energy consumption of different paths can be scientifically evaluated, the path with the lowest energy consumption is selected and smoothed, ensuring that the path is safe, feasible, and efficient. The overall effect is that the UAV can autonomously plan the optimal flight route from its current position to the charging target, reducing energy consumption while avoiding static and dynamic obstacles, improving endurance utilization and charging efficiency, and enhancing flight safety and autonomous decision-making capabilities, providing reliable support for electromagnetic coupling charging of UAVs in complex environments.

[0107] In one alternative embodiment,

[0108] Based on the path energy consumption value, the adjacent sampling points with the lowest energy consumption are selected to construct the flight path. The flight path is then smoothed and verified to meet the constraints of the path planning space. The final flight path is generated as follows:

[0109] Based on the path energy consumption value, the optimal neighboring sampling point of the current sampling point is selected. Among them, the comprehensive path energy consumption value of the selected neighboring sampling point is the smallest and the change in heading angle is less than the maximum turning angle of the UAV. The neighboring sampling points of the selected sampling point are sorted according to the path energy consumption value and added to the candidate sequence. The selection process is repeated until the target position is found. The selected sampling point sequence is connected to construct the initial flight path.

[0110] Extract the turning points in the initial flight path where the heading angle change is greater than a preset threshold, establish a local adjustment area centered on the turning point, determine the size of the adjustment area based on the minimum turning radius of the UAV, generate transition path candidate points within the adjustment area, calculate the connection angle and path length between each candidate point and the original path, select the candidate point sequence that satisfies the turning radius constraint and has the smallest path increment as the transition path, and replace the original turning point with the transition path to obtain a smooth flight path.

[0111] Based on the path planning space calculation, the minimum distance from the sampling point to the obstacle boundary on the smooth flight path is calculated. When the minimum distance is less than the safety distance threshold, the adjustment area of ​​the corresponding turning position is expanded to regenerate the transition path until the smooth flight path meets the safety distance requirement. The smooth flight path that meets all constraints is taken as the final flight path.

[0112] In this embodiment, when selecting the optimal neighboring sampling point based on the path energy consumption value, a constrained greedy search strategy is used to construct the energy-optimal path. Starting from the current position of the UAV as the initial sampling point, all sampling points adjacent to this point are detected, and the comprehensive energy consumption value for flying to each adjacent point is calculated. The comprehensive energy consumption value includes three parts: horizontal displacement energy consumption, vertical displacement energy consumption, and heading adjustment energy consumption. When selecting the next optimal neighboring point, a heading angle change constraint is set to ensure that the heading angle change between two adjacent path segments does not exceed the UAV's maximum turning angle, typically set to 45 degrees. For example, if the current UAV is at the starting point with a heading angle of 90 degrees, the comprehensive energy consumption values ​​calculated for its five adjacent sampling points are 5.2J, 6.8J, 7.1J, 7.8J, and 6.5J, respectively. Simultaneously, the heading angle changes are calculated to be 0 degrees, 33 degrees, 25 degrees, 18 degrees, and 40 degrees, all of which do not exceed the maximum turning angle of 45 degrees. The point with the lowest energy consumption is selected as the next path point, and the neighboring sampling points of this point are added to the candidate sequence in order of path energy consumption value. This selection process is repeated, each time choosing the sampling point with the minimum overall energy consumption and satisfying the heading constraint from the candidate sequence and adding it to the path, until the target position is found. The final initial flight path consists of 16 sampling points, with a total energy consumption of 102.6 J.

[0113] When extracting turning points in the initial flight path where the heading angle change exceeds a preset threshold, the heading angle change threshold is set to 30 degrees. The heading angle change is calculated for the two path segments formed by three adjacent sampling points in the initial path. When the change exceeds 30 degrees, the midpoint is marked as a turning point. In this case, three turning points were detected, with corresponding heading angle changes of 38 degrees, 42 degrees, and 35 degrees, respectively. A local adjustment region is established centered on each turning point, with the size determined based on the minimum turning radius of the UAV. For small UAVs, the minimum turning radius is typically set to 3 meters; therefore, the radius of the adjustment region is calculated to be 1.5 times the minimum turning radius, i.e., 4.5 meters. Within each adjustment region, a uniformly distributed grid of transition path candidate points is generated at 0.5-meter intervals. Taking the first turning point as an example, approximately 250 candidate points are generated within a 4.5-meter radius around this point. The connection angle and path length increment between each candidate point and its adjacent points before and after the original path turning point are calculated. The candidate point sequence that satisfies the turning radius constraint and has the smallest path increment is selected as the transition path from all candidate points. For the first turning point, three candidate points were selected to form a transition path. The maximum heading angle change of this transition path is 18 degrees, satisfying the turning radius constraint, and the path length increment is 0.8 meters. This transition path replaces the original turning point, resulting in a locally smoothed path. The same smoothing process is performed on the remaining two turning points, ultimately yielding a complete smoothed flight path containing a total of 22 path points, increasing the total energy consumption to 105.2 J.

[0114] When calculating the minimum distance from sampling points to obstacle boundaries on a smooth flight path based on path planning space, spatial distance field technology is used to improve computational efficiency. A distance field is constructed for each obstacle in the environment, recording the shortest distance from each point in space to the obstacle surface. For each sampling point on the smooth flight path, its value in the distance field is queried to obtain the distance to the nearest obstacle. The safe distance threshold is set to 2.5 meters to ensure sufficient safe distance between the drone and obstacles. During the inspection, it was found that the distance from the smooth path point near the second turning point to the utility pole obstacle was 2.2 meters, which is less than the safe distance threshold of 2.5 meters. To address this, the adjustment area at the corresponding turning point was expanded, increasing the adjustment radius from 4.5 meters to 6.0 meters, and candidate transition path points were regenerated and selected. After expanding the adjustment area, a new sequence of transition path points was selected, increasing the minimum distance between this path and the utility pole obstacle to 3.1 meters, meeting the safe distance requirement. Collision detection was performed again on the adjusted complete smooth path to confirm that the minimum distance from all path points to obstacles is greater than the safe distance threshold. The final flight path consisted of 24 key points, with a total energy consumption of 107.5J, which was about 4.8% higher than the initial path, but significantly improved flight stability and safety.

[0115] The optimized flight path starts from the origin, passes through a series of smoothly transitioning key points, and avoids obstacles such as utility poles. Near trees, the path maintains a safe distance of at least 3.1 meters to ensure flight safety. Turns in the path are smoothly curved using multiple transition points, allowing the drone to maintain stable flight and reduce energy loss during changes in heading. In the final stage near the target location, the drone gradually descends and adjusts its attitude to the optimal charging angle, preparing for electromagnetic coupling charging. The entire path design achieves overall energy optimization while meeting safety requirements, enabling the drone to efficiently reach the optimal charging location around the smart streetlights.

[0116] In this embodiment, an initial flight path is constructed by selecting the optimal adjacent sampling points based on the path energy consumption value, ensuring minimal changes in the UAV's heading and the lowest energy consumption, thereby improving the energy efficiency and stability of path planning. Based on the initial path, local adjustments are made to turning points with large changes in heading angle. By considering the UAV's minimum turning radius, a transition path is generated, making the flight path smoother and reducing energy consumption and attitude adjustment load caused by sharp turns. By calculating the minimum distance between the smooth path and obstacles and dynamically adjusting the turning areas, the flight path is ensured to meet safety distance constraints. The overall effect is that the UAV can autonomously generate a safe, smooth, and energy-optimized flight path in complex environments, improving endurance efficiency and charging reliability while simultaneously ensuring safety and flight stability.

[0117] like Figure 2The diagram illustrates the process of the drone automatic charging system in this embodiment.

[0118] In one optional implementation, controlling the drone to execute the flight path, adjusting position and attitude deviations in real time, and establishing an electromagnetic coupling charging connection after reaching the target position includes:

[0119] A reference trajectory is established by obtaining a sequence of sampling points based on the planned flight path. A sliding prediction window is selected on the reference trajectory. The prediction trajectory within the prediction window is calculated based on the current position and attitude of the UAV. The flight path deviation is calculated between the prediction trajectory and the reference trajectory. The control gain is determined based on the flight path deviation and the real-time wind speed.

[0120] Based on the control gain, a trajectory adjustment command is generated, and the UAV is controlled to execute the trajectory adjustment command to eliminate flight path deviation. The actual position and attitude after adjustment are obtained, the prediction window is updated, and the trajectory adjustment is repeated. When the UAV reaches the preset distance range of the target charging position, the charging seat features are obtained to establish a relative position relationship, the alignment deviation of the UAV relative to the charging seat is calculated, and an alignment command is generated based on the alignment deviation.

[0121] The alignment command is executed to approach the charging base, contact force data is acquired, the approach speed and contact posture are adjusted according to the contact force data, the position is maintained after a stable contact is established, electromagnetic field data is acquired, and the coupling position is adjusted according to the electromagnetic field data to establish an electromagnetic coupling charging connection.

[0122] Real-time monitoring of wind speed and charging status; when wind speed exceeds a safety threshold or charging status is abnormal, control the drone to disconnect the electromagnetic coupling charging connection and move away from the target charging location.

[0123] In this embodiment, when controlling the UAV to execute a flight path, a reference trajectory is established based on a sequence of sampling points obtained from the planned flight path. The reference trajectory consists of 24 key points, each containing three-dimensional coordinates and timestamp information. A sliding prediction window is selected on the reference trajectory, with a window length typically set to 5 seconds, corresponding to the distance the UAV can fly at its current speed. Based on the UAV's current position and attitude, the predicted trajectory within the prediction window is calculated using the UAV's dynamics model. The dynamics model considers parameters such as the UAV's mass, moment of inertia, thrust coefficient, and torque coefficient, and obtains the predicted position and attitude values ​​of the UAV within the future time window by solving the differential equations of motion. The predicted trajectory is compared with the reference trajectory to calculate the flight path deviation, including position and attitude deviations. Simultaneously, real-time wind speed is measured using an onboard wind speed sensor; in the test scenario, the measured wind speed is 2.3 m / s, and the wind direction is northeast. The control gain is determined based on the flight path deviation and real-time wind speed. The control gain matrix includes two parts: position control gain and attitude control gain. When the wind speed is high, the control gain is appropriately increased to improve anti-interference capability; when the wind speed is low, the control gain is appropriately decreased to improve control stability. In this test, due to the moderate wind speed, the position control gain was set to 0.8 and the attitude control gain was set to 1.2.

[0124] The trajectory adjustment command is generated based on the control gain, and includes three-dimensional position correction and three-axis attitude correction. The UAV is controlled to execute the trajectory adjustment command to eliminate flight path deviation, with an execution frequency of 50Hz. Taking a certain moment in the flight path as an example, the coordinates of the reference trajectory point are (X=-6.5m, Y=3.2m, Z=6.4m), and the coordinates of the predicted trajectory point are (X=-6.7m, Y=3.4m, Z=6.5m). The calculated position deviation is (ΔX=-0.2m, ΔY=0.2m, ΔZ=0.1m). Multiplying the position deviation by the position control gain of 0.8, the position correction is obtained as (ΔX'=-0.16m, ΔY'=0.16m, ΔZ'=0.08m). Similarly, the attitude correction is calculated, and the trajectory adjustment command is generated. After executing the trajectory adjustment command, the UAV acquires its adjusted actual position and attitude. The actual position, measured by the onboard positioning system, is (X = -6.58 m, Y = 3.25 m, Z = 6.42 m), with a position control accuracy of ±0.1 m. The prediction window is updated, and the trajectory adjustment is repeated until the UAV flies to the vicinity of the target charging location. When the UAV reaches a preset distance range (usually set to 2 meters) from the target charging location, the charging dock features are acquired through a visual recognition system to establish a relative positional relationship. The charging dock features include specific visual and infrared markers to facilitate precise positioning of the UAV. The alignment deviation of the UAV relative to the charging dock is calculated, including position and attitude deviations. In the test, the measured position deviations were (ΔX = 0.18 m, ΔY = -0.12 m, ΔZ = 0.05 m), and the attitude deviations were (Δyaw angle = 3.5 degrees, Δpitch angle = 2.1 degrees, Δroll angle = 1.8 degrees). Alignment commands are generated based on the alignment deviations, containing precise position and attitude adjustments.

[0125] When the drone approaches the charging dock after executing the alignment command, a segmented control strategy is employed. When the drone is far from the charging dock (above 0.5 meters), it approaches at a faster speed; when it is close to the charging dock (within 0.5 meters), the speed is reduced to achieve precise alignment. When the drone makes contact with the charging dock, contact force data is acquired via a contact force sensor. This contact force data includes force components in three directions and torque components along three axes. The measured contact force data is (Fx=0.8N, Fy=-0.5N, Fz=2.2N), and the torque data is (Mx=0.15N·m, My=0.09N·m, Mz=-0.12N·m). The approach speed and contact attitude are adjusted based on the contact force data to ensure that the contact force remains within a safe range (typically 2-5N) and that the contact surfaces are parallel. Once a stable contact is established, the drone maintains its current position and attitude, at which point the contact force stabilizes at (Fx=0.3N, Fy=0.2N, Fz=3.5N). Electromagnetic field data is acquired using an electromagnetic field sensor built into the charging dock to measure the coupling state between the charging coils. The electromagnetic field data includes magnetic field strength and distribution information; the measured magnetic field strength is 38 μT, and the magnetic field distribution uniformity is 92%. Fine-tuning is performed based on the electromagnetic field data to achieve the optimal coupling position between the drone's receiving coil and the charging dock's transmitting coil. After adjustment, the magnetic field strength is increased to 42 μT, and the magnetic field distribution uniformity is increased to 96%, establishing a stable electromagnetic coupling charging connection. The charging system begins operation, with a measured charging voltage of 12.6V, a charging current of 2.1A, a charging power of 26.5W, and a charging efficiency of 65%.

[0126] During real-time monitoring of wind speed and charging status, the wind speed monitoring frequency is 1Hz, and the charging status monitoring frequency is 10Hz. The wind speed safety threshold is set at 8m / s; exceeding this value may prevent the drone from maintaining stable hovering. Charging status monitoring includes four parameters: charging voltage, charging current, charging power, and battery temperature. During normal charging, the voltage should be within the range of 11-14V, the current within the range of 1.5-3A, the charging power within the range of 20-40W, and the battery temperature should be below 45℃. In one test, the wind speed suddenly increased to 7.5m / s, approaching the safety threshold, and the system issued a warning signal. At the same time, the charging current fluctuated more, from a stable 2.1A to between 1.8-2.4A. The system determined that the charging environment had become unstable, but had not yet reached the emergency disconnection standard, so it increased the control gain to improve wind resistance and continued monitoring. Ten seconds later, the wind speed further increased to 8.3m / s, exceeding the safety threshold, and the system immediately executed the safety disconnection procedure. The drone is controlled to slowly reduce the contact force. When the contact force drops below 0.5N, it moves vertically away from the charging base by 0.5 meters at a speed of 0.2m / s, and then horizontally away from the charging base by 2 meters at a speed of 0.5m / s, entering a safe hovering state. The entire disconnection process takes approximately 5 seconds, ensuring the safety of both the drone and the charging equipment. The system continues to monitor wind speed, and when the wind speed drops to a safe range (below 6m / s) and remains so for 30 seconds, the charging docking procedure can be re-executed. This dynamic response mechanism ensures the reliable operation of the electromagnetically coupled smart street light drone charging system under various environmental conditions, improving the system's safety and stability.

[0127] In this embodiment, by calculating flight deviations in real time based on a reference trajectory and a prediction window, and dynamically adjusting the control gain in conjunction with environmental factors such as wind speed, the UAV can accurately correct its position and attitude during flight, ensuring path tracking accuracy. Upon approaching the charging dock, by acquiring alignment deviation and contact force data, the approach speed and attitude are adjusted, enabling the UAV to establish a robust and precise electromagnetic coupling connection. The solution also monitors charging status and wind speed in real time, automatically disconnecting and safely detaching in case of abnormalities or exceeding limits, ensuring the safety and reliability of the UAV charging process. The overall effect is that the UAV can autonomously complete the entire process from flight to precise alignment to stable electromagnetic coupling charging, improving charging efficiency, energy utilization, and operational safety, while also possessing real-time adaptability to environmental disturbances.

[0128] A second aspect of the present invention provides an electronic device, comprising:

[0129] processor;

[0130] Memory used to store processor-executable instructions;

[0131] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0132] A third aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0133] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive charging position optimization method for electromagnetically coupled intelligent street light drones, characterized in that, include: The system obtains the location coordinates and output power of the smart street light, as well as the remaining battery power and charging request of the drone. A spiral detection grid is constructed around the smart street light, and the sampling interval is adaptively adjusted according to the location of the detection points. Electromagnetic field intensity data of each detection point is collected, and surface fitting is performed by combining the data of adjacent points to generate an electromagnetic field intensity distribution map. Based on the electromagnetic field intensity distribution map, the three regions with the strongest field intensity are selected as candidate charging regions. Position and attitude sampling combinations are set in the candidate charging regions, and charging efficiency and hovering power consumption data under each combination are collected. The hovering power consumption ratio corresponding to the unit charging efficiency of each sampling combination is calculated as a comprehensive scoring index. The optimal target position and attitude are determined based on the comprehensive scoring index, and a path planning space is established in combination with the distribution of environmental obstacles to generate a flight path that takes energy consumption into account. The drone is controlled to execute the flight path, and its position and attitude deviations are adjusted in real time. After reaching the target position, an electromagnetic coupling charging connection is established.

2. The method according to claim 1, characterized in that, A spiral detection grid is constructed around the smart streetlight. The sampling interval is adaptively adjusted according to the location of the detection points to collect electromagnetic field intensity data at each detection point. The data from adjacent points are then combined to perform surface fitting, generating an electromagnetic field intensity distribution map, including: A polar coordinate system is established with the location of the smart street light as the origin. The initial sampling radius is determined according to the power level of the street light, and an initial spiral detection grid is constructed within the initial sampling radius. Calculate the electromagnetic field strength difference between adjacent detection points in the initial spiral detection grid. When the difference exceeds the preset strength difference threshold, determine the sampling interval adjustment coefficient in the current detection area based on the radiation directionality of the smart street light and the distribution characteristics of the obstruction, and construct a denser spiral detection grid. Electromagnetic field intensity data of the encrypted spiral detection grid are obtained. An outlier detection method based on spatial correlation is adopted. By calculating the covariance of the electromagnetic field distribution between the detection point and its corresponding neighboring detection points, outlier detection points that do not conform to the spatial continuity characteristics are identified. The abnormal detection points were repeatedly measured and the data was corrected by combining the radiation characteristics of the smart street light. The radial basis function considering the electromagnetic field attenuation law was used to perform surface fitting on all electromagnetic field intensity data of the initial spiral detection grid and the densified spiral detection grid to generate an electromagnetic field intensity distribution map.

3. The method according to claim 2, characterized in that, Repeated measurements were taken at abnormal detection points, and data correction was performed based on the radiation characteristics of the smart streetlights. A radial basis function, considering the electromagnetic field attenuation law, was used to perform surface fitting on all electromagnetic field intensity data from both the initial and refined spiral detection grids, generating an electromagnetic field intensity distribution map, including: Obtain the location information of the abnormal detection points, perform multiple repeated measurements at the abnormal detection point locations, calculate the standard deviation and mean of the multiple measurement data for each abnormal detection point, remove the measurement values ​​that exceed the preset standard deviation range, and calculate the average value of the remaining measurement data as the preliminary correction data for the abnormal detection points; The distance attenuation coefficient from the abnormal detection point to the intelligent street light is calculated based on the power level of the intelligent street light, and the direction attenuation coefficient is calculated based on the azimuth angle of the abnormal detection point in the polar coordinate system. The preliminary correction data is multiplied by the distance attenuation coefficient and the direction attenuation coefficient respectively to obtain the theoretical correction value considering the dual attenuation characteristics. The ratio of the preliminary corrected data to the theoretical corrected value is calculated as the correction ratio. Based on the correction ratio, the detection point data in the grid area around the anomaly detection point is linearly interpolated to generate the final corrected data of the anomaly detection point. The final corrected data of abnormal detection points is merged with the normal detection point data in the initial spiral detection grid and the densified spiral detection grid to form a complete electromagnetic field intensity dataset. In the polar coordinate system, the detection point is divided into several sectors. The local correlation coefficient of the detection point in each sector is calculated. The local correlation coefficient is used as the weight coefficient for piecewise fitting. The complete electromagnetic field intensity dataset is then fitted to the surface in different regions to generate an electromagnetic field intensity distribution map.

4. The method according to claim 1, characterized in that, Based on the electromagnetic field intensity distribution map, the three regions with the strongest field intensity were selected as candidate charging regions. Within these candidate charging regions, position and attitude sampling combinations were set up, and charging efficiency and hovering power consumption data were collected for each combination. The hovering power consumption ratio corresponding to the unit charging efficiency of each sampling combination was calculated as a comprehensive scoring index, including: The field strength values ​​in the electromagnetic field intensity distribution map are sorted in descending order, and the regions with the largest field strength values ​​are selected as initial candidate regions. From the initial candidate regions, regions that meet the preset minimum area and the distance between regions meets the preset minimum interval requirement are selected as candidate charging regions, and the field strength distribution data of the candidate charging regions are recorded. Based on the field strength distribution data of the candidate charging areas, an adaptive position sampling grid is established in each candidate charging area. Multiple sets of horizontal yaw angle parameters and vertical pitch angle parameters are set at each sampling point of the adaptive position sampling grid to form a combined position-attitude sampling scheme. Collect induced voltage and charging current data at each sampling combination location in the combined sampling scheme, calculate the charging efficiency value, and record the hovering power consumption data of the UAV at the sampling combination location; Based on the charging efficiency value and hovering power consumption data, the ratio of charging efficiency to hovering power consumption under each position-attitude combination sampling scheme is calculated, and the ratio of charging efficiency to hovering power consumption is used as a comprehensive scoring index.

5. The method according to claim 1, characterized in that, The optimal target position and attitude are determined based on the comprehensive scoring index. A path planning space is established by considering the distribution of environmental obstacles, and a flight path with optimal energy consumption is generated, including: All candidate charging positions are sorted according to the comprehensive score index of position-attitude combination, and the position with the highest comprehensive score index is selected as the target charging position. The attitude parameters corresponding to the target charging position are taken as the target charging attitude. Acquire spatial distribution information of static and dynamic obstacles in the environment, combine it with the current remaining battery power information of the drone to determine the maximum flight distance constraint, mark the area exceeding the maximum flight distance constraint and the spatial distribution area of ​​obstacles as non-flyable areas, and generate a passable path planning space; Within the passable path planning space, a path search range is set with the current position of the drone as the starting point and the target charging position as the ending point. A sampling point grid with a preset density is generated within the path search range, and the path energy consumption value between adjacent sampling points is calculated. Based on the path energy consumption value, the adjacent sampling points with the lowest energy consumption are selected to construct the flight path. The flight path is then smoothed and verified to meet the constraints of the path planning space, generating the final flight path.

6. The method according to claim 5, characterized in that, Based on the path energy consumption value, the adjacent sampling points with the lowest energy consumption are selected to construct the flight path. The flight path is then smoothed and verified to meet the constraints of the path planning space. The final flight path is generated as follows: Based on the path energy consumption value, the optimal neighboring sampling point of the current sampling point is selected. Among them, the comprehensive path energy consumption value of the selected neighboring sampling point is the smallest and the change in heading angle is less than the maximum turning angle of the UAV. The neighboring sampling points of the selected sampling point are sorted according to the path energy consumption value and added to the candidate sequence. The selection process is repeated until the target position is found. The selected sampling point sequence is connected to construct the initial flight path. Extract the turning points in the initial flight path where the heading angle change is greater than a preset threshold, establish a local adjustment area centered on the turning point, determine the size of the adjustment area based on the minimum turning radius of the UAV, generate transition path candidate points within the adjustment area, calculate the connection angle and path length between each candidate point and the original path, select the candidate point sequence that satisfies the turning radius constraint and has the smallest path increment as the transition path, and replace the original turning point with the transition path to obtain a smooth flight path. Based on the path planning space calculation, the minimum distance from the sampling point to the obstacle boundary on the smooth flight path is calculated. When the minimum distance is less than the safety distance threshold, the adjustment area of ​​the corresponding turning position is expanded to regenerate the transition path until the smooth flight path meets the safety distance requirement. The smooth flight path that meets all constraints is taken as the final flight path.

7. The method according to claim 1, characterized in that, Controlling the drone to execute the flight path, adjusting its position and attitude deviations in real time, and establishing an electromagnetic coupling charging connection upon reaching the target position includes: A reference trajectory is established by obtaining a sequence of sampling points based on the planned flight path. A sliding prediction window is selected on the reference trajectory. The prediction trajectory within the prediction window is calculated based on the current position and attitude of the UAV. The flight path deviation is calculated between the prediction trajectory and the reference trajectory. The control gain is determined based on the flight path deviation and the real-time wind speed. Based on the control gain, a trajectory adjustment command is generated, and the UAV is controlled to execute the trajectory adjustment command to eliminate flight path deviation. The actual position and attitude after adjustment are obtained, the prediction window is updated, and the trajectory adjustment is repeated. When the UAV reaches the preset distance range of the target charging position, the charging seat features are obtained to establish a relative position relationship, the alignment deviation of the UAV relative to the charging seat is calculated, and an alignment command is generated based on the alignment deviation. The alignment command is executed to approach the charging base, contact force data is acquired, the approach speed and contact posture are adjusted according to the contact force data, the position is maintained after a stable contact is established, electromagnetic field data is acquired, and the coupling position is adjusted according to the electromagnetic field data to establish an electromagnetic coupling charging connection. Real-time monitoring of wind speed and charging status; when wind speed exceeds a safety threshold or charging status is abnormal, control the drone to disconnect the electromagnetic coupling charging connection and move away from the target charging location.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images