Multi-rotor unmanned aerial vehicle collection route optimization system for water area monitoring

Abstract

The application discloses a multi-rotor unmanned aerial vehicle collection route optimization system for water area monitoring and relates to the technical field of unmanned aerial vehicle route planning.The system comprises a skeleton generation module, a task reconstruction module, a flight section construction module and a cooperative scheduling module.The skeleton generation module extracts a surface floating film mask band based on multi-source sensing data and generates a dynamic boundary skeleton line.The task reconstruction module reconstructs three-dimensional coordinates of an original data collection task with a safety lateral offset based on the skeleton line.The flight section construction module determines a flight section transmembrane type, converts a transmembrane anti-interference operation and a timing dynamic load into equivalent energy consumption, and calculates a total floating film coupling route cost.The cooperative scheduling module generates a multi-machine control scheme through iterative optimization under multi-dimensional constraints.The application effectively overcomes the defect that static planning is not sensitive to water surface dynamic characteristics, eliminates the problems of invalid sampling samples and spectral data distortion, and realizes precise control of multi-machine energy consumption and route optimization under complex weather.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) route planning technology, specifically to a multi-rotor UAV data acquisition route optimization system for water area monitoring. Background Technology

[0002] In water environment monitoring scenarios such as tidal river sections, estuaries, and lake / reservoir inlets, stormwater runoff or sudden pollutant leaks often form dynamically spreading pollution plumes due to hydrodynamic processes such as tides and wind. To accurately obtain the spatial distribution morphology and water quality parameters of these pollution plumes, multi-rotor UAVs are commonly used in engineering applications to perform combined tasks such as multi-point water quality sampling, low-altitude image acquisition, and multi-band spectral observation. During operations, the multi-rotor UAV needs to hover above the water surface to complete operations such as lowering the sampling sling, pumping water from the target body, and performing multimodal optical imaging.

[0003] Existing UAV route planning algorithms typically treat pre-defined sampling and observation points as static spatial coordinates, and use minimizing the spatial distance of the flight path or minimizing the expected flight time as the main planning basis. This conventional route planning has some applicability in open, clean water. However, in dynamically spreading pollution plumes, tidal fronts and the physical boundaries of the pollution plume often form aggregation zones of surface-active substances, oil films, or floating debris due to water flow convergence. Existing route planning lacks a mechanism to perceive and constrain these dynamic water surface characteristics, directly guiding the UAV to the original fixed coordinates. This results in water sampling points easily falling into areas of surface contaminant accumulation, and the extracted shallow water samples being mixed with surface impurities, thus losing their representativeness of water quality. Simultaneously, spectral observation points often experience data distortion because the field of view is directly facing an area of ​​abnormal reflection from the water surface film. In addition, when multiple drones work together to perform sampling tasks, the existing algorithm fails to effectively account for the dramatic increase in physical load variables before and after the sampling operation, and ignores the hovering and anti-interference costs caused by the frequent shuttle of drones across the water surface boundary under complex weather conditions. This results in the actual energy consumption of the operation far exceeding the static prediction of the system, which can easily lead to the safety hazard of drones running out of power.

[0004] Therefore, the technical problem that urgently needs to be solved in this field is: how to overcome the shortcomings of existing static route planning in being insensitive to the dynamic characteristics of the water surface under complex meteorological and hydrodynamic conditions, and effectively incorporate dynamic execution loads and environmental disturbance costs into the route optimization dimension while ensuring the validity of spectral data and the representativeness of water samples, so as to achieve safe relocation of multi-task acquisition locations and accurate planning of multi-aircraft collaborative routes. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a multi-rotor UAV data acquisition route optimization system for water area monitoring. This system solves the problems that existing static route planning methods are insensitive to dynamic water surface characteristics under complex meteorological and hydrodynamic conditions, leading to sample failure, spectral data distortion, and uncontrolled energy consumption in multi-UAV collaborative operation.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-rotor UAV flight path optimization system for water area monitoring, comprising:

[0007] The skeleton generation module is used to access multi-source sensor data and meteorological window set, extract the tidal surface convergence floating film masking zone based on multi-source sensor data, and output the floating film boundary skeleton line, skeleton point position, floating film zone half width and lateral direction normal vector. Based on the meteorological window set, unexecutable skeleton segments are removed.

[0008] The task reconstruction module is used to calculate the distance between the initial nominal position of the original data acquisition task and the skeleton point position to determine the position of the associated skeleton point. It calculates the lateral offset distance based on the half-width of the floating membrane strip, and generates the three-dimensional coordinates of the lateral task point by combining the lateral direction normal vector, the target side and target execution height corresponding to the original data acquisition task, and the lateral offset distance. It outputs the reconstructed set of lateral data acquisition tasks.

[0009] The segment construction module is used to construct candidate route segments and determine the route segment type from the reconstructed set of lateral data acquisition tasks. It uses dynamic load as a delay calculation variable, calculates the total cost of the floating membrane coupled route based on the equivalent energy consumption matching the route segment type, and outputs a set of floating membrane coupled route segments.

[0010] The collaborative scheduling module is used to combine the set of floating membrane coupled route segments, the reconstructed set of lateral data acquisition tasks, and the set of UAV resources to generate a task execution sequence, calculate the total cost of the floating membrane coupled route according to the time sequence, and perform iterative optimization to output a set of multi-UAV collaborative route control schemes.

[0011] Compared with existing technologies, it has the following advantages:

[0012] This proposed multi-rotor UAV flight path optimization system for water monitoring effectively overcomes the shortcomings of conventional flight path planning algorithms in avoiding floating agglomerates on the water surface during dynamic water environment monitoring. By fusing multi-source sensor data, the system accurately extracts convergent agglomerates of surface pollutants and reduces their dimensionality to a dynamic boundary skeleton line to guide flight path generation. Based on the generated skeleton line model, the system automatically performs lateral safety offset repositioning of the original data acquisition location according to real-time wind field conditions and sensor field-of-view parameters. This mechanism ensures that sampling points are free from interference from surface contaminants, guarantees the true representativeness of water quality samples, and effectively avoids spectral distortion caused by abnormal reflections from the water surface film, significantly improving the accuracy and effectiveness of data acquisition from the spatial layout source.

[0013] To address the risk of energy consumption runaway during multi-drone collaborative operations in disturbed environments, this solution deeply incorporates disturbance-resistant operation losses and temporal load variables into the cost assessment and scheduling model. The system can accurately identify flight trajectories of UAVs crossing or extending boundaries, quantify the additional energy consumption caused by surface operations into equivalent energy consumption values, and remove the changes in total weight of the aircraft due to sampling as a delay calculation variable. In the global scheduling optimization phase, the system strictly uses safe power and load limits as boundary constraints, relying on iterative algorithms to optimize the execution sequence of multi-drone tasks, and synchronously updates the total global scheduling cost with each time the sequence is updated. This completely eliminates the serious errors caused by conventional algorithms that estimate energy consumption solely based on spatial distance, significantly reduces the frequency of invalid fleet turnarounds, and ensures safe long-endurance operation of UAVs under complex weather constraints. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the system framework of the present invention.

[0015] Figure 2 This is a schematic diagram of the system execution flow of the present invention. Detailed Implementation

[0016] 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.

[0017] Please see Figures 1 to 2 This application provides a multi-rotor UAV data acquisition route optimization system for water area monitoring, including a skeleton generation module, a task reconstruction module, a flight segment construction module, and a collaborative scheduling module;

[0018] The skeleton generation module is configured to extract the tidal surface convergence floating film masking zone based on meteorological windows and multi-source water surface observation data, and convert it into dynamic boundary skeleton lines, skeleton point positions, floating film zone half-width, and lateral direction normal vectors.

[0019] This module integrates multi-source sensor data from the target water area. Input parameters include a low-altitude preview image data matrix, a low-altitude spectral preview data matrix, a surface flow velocity vector, pollution source locations, and a set of meteorological windows. Using the acquisition timestamps of the low-altitude preview image data matrix as a reference, the module performs temporal interpolation alignment (e.g., linear interpolation alignment) on the low-altitude spectral preview data matrix and the surface flow velocity vector. Subsequently, based on the pose parameters of the airborne equipment and camera intrinsic parameters, the time-aligned data is uniformly orthophoto-projected and transformed into the local coordinate system plane of the target water area, thereby constructing a discrete spatial grid covering the entire region.

[0020] Specifically, the low-altitude preview image data matrix represents a two-dimensional pixel array in the visible light band of the water surface, the low-altitude spectral preview data matrix represents a reflectivity array containing multiple invisible light bands, the surface flow velocity vector represents the hydrodynamic vector of the water surface driven by tides or wind, and the axes of the local coordinate system of the target water area constitute a three-dimensional Cartesian space reference. By forcibly aligning sensor detection results with different spatial resolutions, acquisition angles, and sampling frequencies to a unified physical spatiotemporal reference plane, the misalignment deviations in time delay and spatial projection caused by heterogeneous sensors are suppressed, ensuring that the optical and hydrodynamic features extracted subsequently for the same water surface grid cell correspond to the same source in terms of spatial location and time stamp.

[0021] On the constructed discrete spatial grid, for each grid cell, its neighborhood set is defined according to a preset topological expansion radius, which is preferably 3 to 5 grid cells. This module calculates the image anomaly response, spectral anomaly response, and surface convergence response of the grid cell respectively.

[0022] The formula for calculating the anomaly response in an image is:

[0023]

[0024] In the formula, This indicates the image anomaly response of the grid cell. This represents the image value of the grid cell. This represents the image mean of the neighborhood set corresponding to the grid cell.

[0025] The formula for calculating the spectral anomaly response is:

[0026]

[0027] In the formula, This represents the spectral anomaly response of the grid cells. This represents the multi-band spectral vector of the grid cell. This represents the spectral mean vector of the neighborhood set corresponding to the grid cell. This indicates the magnitude of the orientation quantity.

[0028] The formula for calculating the surface convergence response is:

[0029]

[0030] In the formula, This represents the dimensionless surface convergence response of the grid cells. The divergence operator represents the surface velocity vector at the grid cell. This indicates that the operation takes the positive value. This represents the maximum absolute value of the negative divergence of the surface velocity vector field within the target water area, or a preset reference convergence divergence constant.

[0031] Specifically, a grid cell is a basic spatial computational unit divided according to a preset grid scale. The image anomaly response, through the difference between the central value and the surrounding mean, and after mean normalization, characterizes the degree of brightness or texture abrupt changes on the local water surface under visible light, highlighting the visual boundaries of floating objects. The spectral anomaly response quantifies distance abrupt changes in the high-dimensional spectral vector space, characterizing the magnitude of the vector Euclidean distance between the local water surface and the surrounding background in multi-band reflectivity, used to identify accumulations of chemical substances invisible to the naked eye. In engineering calculations of the above image and spectral anomaly responses, if the water surface appears completely black or has no reflection, resulting in a zero mean for the corresponding neighborhood set, the system automatically appends a very small positive real number to the denominator to avoid division-by-zero overflow errors. The surface convergence response characterizes the inward convergence tendency of matter generated by the hydrodynamic field in that local area. When the divergence of the surface velocity vector field is negative, it indicates that the water flow converges at that location. Introducing a divergence normalization denominator maps the divergence value, which has the physical dimension of the reciprocal of time, to a dimensionless scalar between 0 and 1. The complex physical phenomenon of tidal fronts is decomposed into three independent and dimensionally unified scalars: high visual contrast, high spectral heterogeneity, and high hydrodynamic convergence. This overcomes the technical deficiency of conventional UAVs, which rely solely on visible light to identify obstacles on the water surface and are easily interfered with by the reflection of sunglass surfaces.

[0032] This module fuses the three independent scalar fields mentioned above to calculate the overall response of the floating film masking of the grid cells:

[0033]

[0034] In the formula, This represents the composite response of the floating film masking of the grid cells. This represents the weighting coefficient corresponding to the image anomaly response. This represents the weighting coefficient corresponding to the spectral anomaly response. Represents the weight coefficients corresponding to the surface convergence response, and satisfies .

[0035] This module extracts a grid set as the background water sample area outside a radius centered on the pollution source location. It calculates the mean and standard deviation of the overall response to the floating film masking within this background water sample area. The mean is then added to the standard deviation by a preset multiple, which is set as the background adaptive threshold. This preset multiple is preferably 2 to 3 to cover the vast majority of background environmental noise. Subsequently, grid sets with overall response values ​​higher than this background adaptive threshold and spatially meeting connectivity requirements are selected and filled into the candidate floating film masking area.

[0036] Specifically, the integrated response of the floating film masking system is a global scalar field that combines optical and hydrodynamic characteristics. The weighting coefficients are allocation factors used to balance the real-time data quality of different sensors; dynamically adjusting these coefficients allows for fault-tolerant synergy among heterogeneous sensors. The background adaptive threshold is a dynamic segmentation boundary generated based on real-time statistical patterns of clean water extracted from spatial prior rules. By extracting real-time background statistical parameters to generate the adaptive threshold, it avoids the problem of floating film leakage caused by drastic changes in global illumination or seasonal changes in overall water quality.

[0037] Within the candidate floating film shielding area, this module combines the pollution source location and surface velocity vector to generate a macroscopic baseline indicating the main diffusion direction of the pollution plume. Multiple cross-sections are then laid out along the normal to this macroscopic baseline. Local maxima grid cells representing the integrated shielding response of the floating film on each cross-section are extracted. These local maxima grid cells are sequentially connected along the main diffusion direction to generate a smooth floating film boundary skeleton line. The skeleton line is parameterized using normalized parameters to determine the location of the core skeleton points on the skeleton line.

[0038] The module searches outwards step by step from the skeleton point position along the normal directions on both sides perpendicular to the tangent of the skeleton line, calculates the horizontal span (specifically the horizontal straight line span) when the overall response of the floating film masking shows a decaying trend until it is less than the background adaptive threshold, and records it as the half width of the floating film band.

[0039] To establish the lateral direction normal vector pointing inwards from the plume, this module calculates the cosine of the dot product between two opposing candidate normal unit vectors and the reference vector, and selects the candidate normal unit vector with the largest dot product cosine as the output lateral direction normal vector.

[0040]

[0041] In the formula, This represents the lateral direction normal vector pointing inwards from the skeleton point. Let be the candidate normal unit vector for any one side. To and Opposite candidate normal unit vector, Let these be the variables in the two opposite candidate normal unit vectors. Indicates the location of the pollution source. This represents the surface velocity vector at the location of the skeleton point. Indicates the location of the skeleton points. The time scale parameter is preset. The reference vector is obtained by multiplying the surface velocity vector at the pollution source location point along the skeleton point location by the time scale parameter, and then subtracting the coordinates of the endpoints from the coordinates of the skeleton point location.

[0042] Specifically, the macroscopic baseline characterizes the overall physical direction of pollutant transport. By introducing the macroscopic baseline for dimensionality reduction constraints, it ensures that the skeleton orientation conforms to the hydrodynamic diffusion trend. The cross-section is an artificially assisted calculation line segment used to reduce the dimensionality of the two-dimensional search. The floating film boundary skeleton line is a topological one-dimensional continuous curve that runs through the core of the high-response region. The half-width of the floating film zone characterizes the lateral boundary distance from the center of the skeleton outwards where physical disturbances rapidly decay to the level of normal water bodies. By probing the potential field gradient along the normal until the potential field falls to the adaptive background threshold, the half-width parameter is closed-loop bound to the dynamically calculated background features. The time scale parameter characterizes the theoretical time step of the flow particle's movement along the velocity direction, transforming the velocity vector into an extended boundary with spatial displacement dimensions. Direction discrimination is achieved by constructing a reference baseline vector representing the macroscopic flow trend of the pollution plume. The angle between the candidate normal vector and the reference vector is compared using inner product operations, and the normal vector pointing into the depth of the pollution plume is rigorously selected. The large-area unstructured two-dimensional planar floating membrane anomaly region is reduced in dimension and reconstructed into an ordered one-dimensional data chain with clear topological extension boundaries and internal and external spatial semantic attributes, providing mathematical coordinate framework support for the precise layout of lateral water quality sampling points and spectral observation points.

[0043] This module performs a set intersection operation on the calculated skeleton lines and the three-dimensional spatial coordinates of all skeleton point locations, and on the executable three-dimensional geospatial envelope generated by mapping real-time wind field and rainfall parameters in the currently updated meteorological window set. Specifically, this executable three-dimensional geospatial envelope is generated by performing a difference operation on a wind shear safety threshold plane and a polygonal column indicating no-fly zones due to rainfall or dense fog. In practical applications, the wind shear safety threshold plane can be a flight altitude upper limit plane set based on a wind shear safety threshold, the polygonal column can be a polygonal column indicating no-fly zones due to rainfall or dense fog based on radar reflectivity, and the difference operation can specifically be a spatial Boolean difference operation. Skeleton segments that do not fall within the intersection range are removed and marked as non-executable skeleton segments for the current window.

[0044] Specifically, the meteorological window period verification is a Boolean logic judgment process that verifies the safety constraints of the generated theoretical geometric sites under real-world environmental conditions. It ensures that the topological skeleton generated by the underlying algorithm remains entirely within the safety envelope of the UAV's current electromechanical wind resistance and optical sensor visibility requirements. It preemptively truncates unexecutable path nodes caused by excessive local gusts or sudden rainfall / fog, preventing the system from generating flight path instructions with flight hazards or collecting invalid data. Finally, this module structurally encapsulates and outputs the cleaned and verified skeleton lines, skeleton point positions, floating membrane half-width, and lateral direction normal vectors.

[0045] The task reconstruction module is configured to reconstruct the position of the original data acquisition task based on the boundary skeleton line of the floating membrane and its lateral direction normal vector, and output the reconstructed lateral data acquisition task set.

[0046] This module receives the floating membrane boundary skeleton line, skeleton point positions, floating membrane strip half-width, and lateral direction normal vector output by the skeleton generation module, and simultaneously reads the raw data acquisition task set from the input system. This raw data acquisition task set contains several raw data acquisition tasks, each with an initial nominal position, task type, execution priority, required hardware device module, task target side, target execution height, and data gain benchmark value.

[0047] This module calculates the horizontal plane distance between the initial nominal position of each raw data acquisition task and all skeleton point positions on the floating membrane boundary skeleton line, and selects the skeleton point position with the smallest horizontal plane distance as the associated skeleton point position for that task. If the minimum horizontal plane distance is greater than a preset task association threshold, the raw data acquisition task is determined to be a non-floating membrane associated task and its initial nominal position remains unchanged; if the minimum horizontal plane distance is not greater than the preset task association threshold, it is determined to be a floating membrane associated task and enters the subsequent lateral position reconstruction stage. The preset task association threshold is preferably half the width of the floating membrane strip at the associated skeleton point position plus a preset spatial tolerance distance. This spatial tolerance distance is set based on the combined extreme value of the UAV satellite positioning error and the orthophoto projection error of the airborne camera, and is preferably 5 to 10 meters.

[0048] Specifically, the initial nominal location represents the fixed latitude and longitude or local planar coordinates issued by the demand side, while the task target side indicates the system's monitoring tendency between the pollution plume side and the background water side. By introducing a task association threshold dynamically calculated based on the half-width of the floating membrane zone and spatial tolerance, the system automatically filters and eliminates ordinary open water tasks far from the floating membrane shielding zone, avoiding the waste of computing power and topological disorder caused by forced global task association.

[0049] For water sampling tasks identified as being associated with a floating film, this module calculates the lateral offset distance based on target laterality and physical constraints. This lateral offset distance is equal to the sum of the half-width of the floating film zone at the associated skeleton point, the rotor downwash influence radius, the horizontal sway of the sampling device, and a preset safety interval distance. The rotor downwash influence radius is pre-configured based on airflow disturbance test data when a multi-rotor UAV hovers above the water surface. The horizontal sway of the sampling device is calculated by a preset wind deflection physical model using the physical length of the sampling sling and the current wind speed in the meteorological window set, or taken from a pre-calibrated maximum sway constant. By laterally expanding the sampling point from the center of the floating film shielding zone, it ensures that the sampling hose or water collection container directly contacts the underlying effective water body, preventing near-surface water samples from being distorted by floating film enrichment, while also avoiding the UAV downwash airflow from disrupting the natural distribution of the frontal floating film.

[0050] For low-altitude spectral acquisition tasks identified as being associated with floating film, this module calculates the lateral offset distance based on sensor hardware parameters. This lateral offset distance is configured to be no less than the sum of the half-width of the floating film strip and a preset safety interval, and no greater than the half-width of the spectral sensor's field of view. The half-width of the field of view is determined by the target execution altitude and the spectral sensor's field of view angle through trigonometric function projection. Setting upper and lower limits for this offset distance ensures that the edge of the spectral sensor's field of view accurately covers the boundary of the pollution plume, while its main acquisition angle avoids the central region of the floating film, preventing multi-band reflectivity distortion caused by abnormal specular reflection from the water surface and reflection from the organic film layer. In engineering calculations, if no lateral offset distance satisfying the above constraints exists at the current target execution altitude, the system gradually increases the target execution altitude according to a preset altitude step to expand the half-width of the field of view; this altitude step is preferably 1 to 2 meters. If the constraints are still not met when the altitude is increased to the maximum permissible flight altitude of the aircraft, the task is downgraded to a lower-level observation task or marked as unexecutable.

[0051] For the visible light boundary recognition task in the floating membrane associated task, the lateral offset distance is calculated based on the field of view and shooting height of the visible light camera, so that the camera's field of view covers the floating membrane boundary skeleton line and the UAV body is not located directly above the center of the floating membrane.

[0052] After determining the lateral offset distance of each floating membrane associated task, the module uses the associated skeleton point position, lateral direction normal vector, target side, lateral offset distance and target execution height to generate the reconstructed 3D coordinates of the lateral task point.

[0053] The formula for calculating the three-dimensional coordinates of the lateral mission point is:

[0054]

[0055] In the formula, This represents the reconstructed 3D coordinates of the lateral mission point. This indicates the three-dimensional coordinates of the associated skeleton point location on the water surface datum elevation corresponding to the task. This indicates the target side of the task and takes only a positive 1 or a negative 1. This indicates the lateral offset distance required for the task. This represents the lateral direction normal vector at the location of the associated skeleton point, pointing inwards from the plume, parallel to the water surface, and with its vertical component being zero. Indicates the target execution altitude of the task. This represents a unit vector perpendicular to the water surface and pointing upwards.

[0056] Specifically, the location of the associated skeleton point represents a two-dimensional coordinate on the water surface reference elevation. After calculating the horizontal projected coordinate offset using a formula, the system generates a final three-dimensional coordinate with unified dimensions and a closed physical dimension by introducing a unit vector in the vertical reference direction and the target execution height. A positive 1 is assigned to the target side to indicate the pollution plume side, and a negative 1 is assigned to the background water side. The original fixed point monitoring task is dimensionally reduced and mapped to the floating membrane skeleton coordinate system and derived along the normal direction, constructing a lateral execution coordinate system attached to the dynamic boundary skeleton.

[0057] This module performs a safety check between the generated 3D coordinates of the lateral task point and the executable 3D geospatial envelope defined by the meteorological window set. If the 3D coordinates of the lateral task point fall outside this envelope, the system searches for an alternative skeleton point along the corresponding floating membrane boundary skeleton line within a preset skeleton search range and recalculates the alternative lateral task point. This skeleton search range is preferably 5% to 10% of the total length of the skeleton line. If no valid coordinate point satisfying the meteorological window set conditions is found after traversing the search range, the system marks the task as unexecutable during the current window. A secondary closed-loop check of physical space and meteorological conditions is performed before outputting the final reconstructed task, intercepting anomalies caused by lateral offsets that result in new coordinates unexpectedly falling into wind shear excess areas or rainband coverage areas, ensuring the flight safety of the underlying planning data in the real environment.

[0058] Finally, this module encapsulates the updated coordinates of the task structure and outputs a reconstructed set of lateral data acquisition tasks. This set provides fundamental task nodes with clearly defined lateral information, offset distances, and valid spatial coordinates for downstream route cost calculations.

[0059] The segment construction module is configured to construct floating membrane coupled route segments based on the reconstructed set of lateral data acquisition tasks and environmental energy consumption parameters, and output a set of floating membrane coupled route segments.

[0060] This module receives the reconstructed lateral data acquisition task set output by the task reconstruction module, and simultaneously acquires the wind speed vector, basic UAV flight energy consumption parameters, and task execution power parameters within the meteorological window. The basic UAV flight energy consumption parameters include energy consumption coefficient per unit distance, energy consumption coefficient per unit altitude climb, wind resistance energy consumption coefficient, and energy consumption coefficient per unit payload. These parameters are obtained based on factory calibration test data from multi-rotor UAVs during unloaded horizontal flight, climb flight, wind resistance flight, and flights with different payloads. The task execution power parameters include hovering power, sampling device power, and spectral sensor power, configured by the rated power parameters of the corresponding airborne equipment. All cost metrics in the above calculation process are uniformly converted into equivalent energy consumption units, eliminating dimensional conflicts caused by directly adding different dimensions such as distance, time, and operational costs.

[0061] This module establishes candidate route segments for any two lateral data acquisition tasks in the reconstructed lateral data acquisition task set, and determines the segment type of the candidate route segment based on the execution side of these two lateral data acquisition tasks and the associated floating membrane boundary skeleton line. If the starting and ending tasks are both floating membrane-related tasks and are located on the same floating membrane boundary skeleton line with the same execution side, the candidate route segment is classified as a same-side along-membrane segment; if the starting and ending tasks are both floating membrane-related tasks and are located on different execution sides on the same floating membrane boundary skeleton line, the candidate route segment is classified as a cross-membrane switching segment; if the starting or ending node of the candidate route segment contains at least one takeoff point, return point, or non-floating membrane-related task, it is classified as an off-membrane transfer segment. This directly maps spatial topology relationships to the control logic of UAVs crossing environmental barriers, providing a complete classification basis without overlap or omissions for subsequent refined calculations of additional energy consumption.

[0062] For each candidate route segment, this module calculates the basic flight energy consumption and mission execution energy consumption baseline, and sums them to obtain the conventional route cost.

[0063] The formula for calculating the cost of a conventional air route is:

[0064]

[0065] In the formula, This represents the conventional route cost of the candidate route segment. This indicates the basic flight energy consumption of the candidate route segment. This represents the base value of the execution energy consumption for the final task.

[0066] The formula for calculating basic flight energy consumption is:

[0067]

[0068] In the formula, Indicates basic flight energy consumption. This represents the energy consumption coefficient per unit distance for flight, which has corresponding physical dimensions. This represents the three-dimensional flight distance of the candidate route segment. This represents the energy consumption coefficient per unit altitude climbed, with corresponding physical dimensions. Indicates the final mission objective execution height. Indicates the starting point of the task objective execution height. This represents the drag energy consumption coefficient with corresponding dimensions. This represents the scalar component of wind resistance for the candidate flight path segment. This represents the energy consumption coefficient per unit load with corresponding physical dimensions. This indicates the dynamic payload of the drone when it is executing this flight segment.

[0069] Specifically, the three-dimensional flight distance represents the straight-line length in space between two lateral mission points. The positive value portion is used to filter out the potential energy work done during the UAV's descent, thus aligning with the actual power consumption characteristics of multi-rotor aircraft, which struggle to effectively recover potential energy. (Wind drag scalar component) The energy consumption interference of headwinds and crosswinds on the flight attitude of multi-rotor UAVs can be comprehensively characterized by taking the negative positive value of the inner product of the wind speed vector in the area where the flight segment is located and the unit direction vector of the candidate flight segment, or by taking the magnitude of the cross product of the wind speed vector and the unit direction vector.

[0070] To pre-calculate route segment costs, this module decouples conventional route costs into static baseline costs and dynamic payload costs. Since the current dynamic payload of the UAV cannot be determined in advance due to its separation from the global scheduling sequence, this module treats the dynamic payload as a delayed calculation variable when calculating basic flight energy consumption. It only generates static baseline costs from distance, altitude changes, and wind resistance components that are fixedly related to the route segment. These static baseline costs are then dynamically updated by the downstream scheduling module based on the actual sampling sequence.

[0071] Based on the conventional route cost, this module introduces energy consumption penalties and route benefits caused by surface shielding features to calculate the total cost of the floating membrane coupled route.

[0072] The formula for calculating the total cost of the floating membrane coupled route is:

[0073]

[0074] In the formula, This represents the total cost of the floating membrane coupled route for the candidate route segment. This represents the conventional route cost of the candidate route segment. This indicates the additional costs incurred by switching between membrane segments. This indicates the additional costs incurred due to operations near the floating membrane. This indicates that the revenue item is continuously executed on the same side of the membrane.

[0075] Specifically, the additional cost of cross-membrane switching is only included when the candidate route segment is a cross-membrane switching segment, and the additional cost of floating membrane proximity operation represents the increase in operation time caused by abnormal water surface reflection and downwash airflow interference when the UAV performs hovering operations near the floating membrane boundary. The above-mentioned additional costs of cross-membrane switching and floating membrane proximity operation are determined based on the product of the corresponding pre-calibrated operation additional time and the UAV hovering power parameter, thereby converting them into equivalent energy consumption. The benefit term of continuous execution along the same side of the membrane is only included when the candidate route segment is a along the same side of the membrane, representing the equivalent energy consumption saved by avoiding cross-membrane turning back. To establish a safety boundary when the dynamic load is unknown, the benefit term of continuous execution along the same side of the membrane is forcibly constrained to no more than a preset proportion of the sum of the static baseline cost and the additional cost, which is preferably 30% to 50%. The environmental interference factor is converted into an equivalent energy consumption unit of the same dimension, and the physical interference of the complex surface environment is reduced in dimension and mapped to the energy consumption cost characteristics of the underlying layer of route planning.

[0076] Finally, the module encapsulates the starting and ending nodes, segment types, and total cost model of the floating membrane coupled route containing dynamic load variables for all candidate route segments in a structured manner, and outputs a set of floating membrane coupled route segments, providing a directed graph topology network with environmental penalty weights for multi-UAV collaborative scheduling.

[0077] The collaborative scheduling module is configured to generate a multi-UAV allocation and timing execution sequence based on the set of floating membrane coupled route segments, the reconstructed set of lateral data acquisition tasks, and the set of UAV resources, and output a set of multi-UAV collaborative route control schemes.

[0078] This module receives the set of floating membrane coupled flight segments output by the flight segment construction module and the reconstructed set of lateral data acquisition tasks output by the task reconstruction module, while also reading the UAV resource set configured by the system. This UAV resource set includes all multi-rotor UAVs in the system, and each multi-rotor UAV has its takeoff point position, current remaining battery power, return-to-home safety reserve battery power, maximum allowable payload, and executable hardware modules on board.

[0079] The module first performs a capability matching check between the set of lateral data acquisition tasks and the multi-rotor UAV. If the hardware modules required for a certain lateral data acquisition task are completely included by the executable hardware modules carried by the corresponding multi-rotor UAV, a preliminary usability association is established. Subsequently, the module executes a lateral coordination strategy, dividing the tasks into a polluted plume side task group, a background water side task group, and a boundary identification supplementary task group according to the target side of the task, and prioritizing the allocation of tasks on the same target side to the same multi-rotor UAV, so as to reduce the probability of cross-membrane switching segments from the global allocation starting point.

[0080] Based on capability matching checks and a side-by-side coordination strategy, this module generates an initial task execution sequence for each multi-rotor UAV and constructs a total cost function model for multi-UAV collaborative scheduling.

[0081] The formula for calculating the total cost of multi-UAV cooperative scheduling is:

[0082]

[0083] In the formula, M represents the total cost of multi-drone collaborative scheduling, where M represents the total number of multi-rotor drones in the drone resource set. This represents the task execution sequence assigned to the m-th multirotor UAV. This indicates the floating membrane coupled route segment in the mission execution sequence. This represents the total cost of the floating membrane coupled route for the corresponding floating membrane coupled route segment. This represents the return energy consumption of the m-th multi-rotor UAV. This represents the set of data acquisition tasks contained within the corresponding task execution sequence. This represents the baseline value of data revenue obtained by completing the corresponding data collection task, converted into equivalent energy consumption units.

[0084] Specifically, to analyze the dynamic load, which serves as a delay calculation variable in the flight segment construction module, this module accumulates the increased load after each water sampling task is executed point by point according to the time sequence in the initial task execution sequence, calculating the actual dynamic load of the multi-rotor UAV in each candidate flight segment. This actual dynamic load is then substituted into the basic flight energy consumption equation to update the specific value of the total cost of the floating membrane coupled flight path. The return energy consumption is calculated from the three-dimensional spatial distance, altitude change, wind resistance component between the last lateral task point in the current task execution sequence and the corresponding multi-rotor UAV takeoff point or preset recovery point, as well as the final dynamic load after all tasks are completed, and is updated synchronously each time a test sequence is generated. The pre-calculation of the spatial physical characteristics of the flight segment is strictly decoupled from the dynamic solution of the time sequence load, eliminating the uncertainty of the load state caused by the undetermined task order. During the cost aggregation process, the virtual data value index is converted into an equivalent energy consumption unit with the same dimension as flight energy consumption, ensuring that the addition and subtraction operations are consistent in both physical dimension and mathematical logic.

[0085] This module applies multi-dimensional dynamic constraint verification to the initial task execution sequence after refreshing the cost values. The system determines whether the estimated total flight energy consumption of each multi-rotor UAV is less than or equal to the difference between the current remaining battery power and the safe return battery power reserve, whether the actual dynamic load of each flight segment is less than or equal to the maximum allowable load of the corresponding aircraft type, and whether the spatiotemporal interaction nodes of each task are completely within the envelope range defined by the meteorological window set. If any constraint is broken, the system will trigger the cross-aircraft reallocation of the violating task or remove it from the current task execution sequence.

[0086] Under the premise of satisfying all constraints, this module performs local iterative optimization on the initial task execution sequence. The system generates test sequences by exchanging adjacent task sequences within the same multi-rotor UAV or by exchanging tasks on the same side across different UAVs. If the total cost of multi-UAV cooperative scheduling corresponding to the test sequence is lower than that of the current sequence and satisfies the above-mentioned multi-dimensional dynamic constraints, the test sequence is used for iteration until the total cost of multi-UAV cooperative scheduling no longer decreases, or a suboptimal solution is accepted based on a preset annealing probability to escape the local optimum, or a preset upper limit for the number of iterations is reached. This preset upper limit for the number of iterations is preferably 500 to 1000 times.

[0087] The formula for calculating the annealing probability is:

[0088]

[0089] In the formula, This represents the annealing probability following the Metropolis criterion. The increment of cost for the test sequence relative to the current sequence is represented by T, which represents the virtual temperature parameter of the system that decays exponentially with the number of iterations.

[0090] For dynamic changes in the floating membrane boundary skeleton line caused by updates to meteorological windows or multi-source sensor data (such as surface observation data) during mission execution, this module performs local incremental replanning. During local incremental replanning, the system locks the current real-time 3D physical coordinates, remaining battery power, and current physical payload of the affected multirotor UAV as the virtual starting node for incremental replanning, ensuring a smooth transition between newly generated control commands and the current aerial state of the multirotor UAV. The system only extracts tasks whose lateral mission points have changed geographical coordinates due to boundary displacement, flight segments whose lateral attributes have changed due to these changes, and unexecuted flight segments directly adjacent to the aforementioned nodes for recalculation. Completed historical nodes and unaffected spatial segments maintain their original control commands. By converting global optimization into local heuristic iteration with constraint verification and limited-range incremental replanning, the system's replanning efficiency in dynamically disturbed environments is improved.

[0091] Finally, this module encapsulates the optimized task execution sequence and its corresponding flight and maneuver commands in a structured manner, outputting a set of multi-UAV cooperative route control schemes. This set of multi-UAV cooperative route control schemes includes the takeoff position, task execution sequence, three-dimensional lateral mission points, mission execution altitude, segment flight speed, and return path of each multi-rotor UAV, which is used to directly send data to the underlying flight control system to perform physical data acquisition operations.

[0092] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A multi-rotor UAV data acquisition route optimization system for water area monitoring, characterized in that, include: The skeleton generation module is used to access multi-source sensor data and meteorological window set, extract the tidal surface convergence floating film masking zone based on multi-source sensor data, and output the floating film boundary skeleton line, skeleton point position, floating film zone half width and lateral direction normal vector. Based on the meteorological window set, unexecutable skeleton segments are removed. The task reconstruction module is used to calculate the distance between the initial nominal position of the original data acquisition task and the skeleton point position to determine the position of the associated skeleton point. It calculates the lateral offset distance based on the half-width of the floating membrane strip, and generates the three-dimensional coordinates of the lateral task point by combining the lateral direction normal vector, the target side and target execution height corresponding to the original data acquisition task, and the lateral offset distance. It outputs the reconstructed set of lateral data acquisition tasks. The segment construction module is used to construct candidate route segments and determine the route segment type from the reconstructed set of lateral data acquisition tasks. It uses dynamic load as a delay calculation variable, calculates the total cost of the floating membrane coupled route based on the equivalent energy consumption matching the route segment type, and outputs a set of floating membrane coupled route segments. The collaborative scheduling module is used to combine the set of floating membrane coupled route segments, the reconstructed set of lateral data acquisition tasks, and the set of UAV resources to generate a task execution sequence, calculate the total cost of the floating membrane coupled route according to the time sequence, and perform iterative optimization to output a set of multi-UAV collaborative route control schemes.

2. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 1, characterized in that, The skeleton generation module extracts the tidal surface convergence masking zone based on multi-source sensor data, including: The low-altitude preview image data matrix, low-altitude spectral preview data matrix and surface flow velocity vector contained in the multi-source sensor data are interpolated and aligned in the time dimension and orthophoto projected, and then transformed into a discrete spatial grid in the local coordinate system plane of the target water area. The image anomaly response, spectral anomaly response, and surface convergence response generated based on the negative divergence of the surface velocity vector were calculated for the grid cells within the discrete spatial grid. The overall response of the floating film masking of the grid cells is calculated by weighted fusion of image anomaly response, spectral anomaly response and surface convergence response. A grid set is extracted outside the range centered on the pollution source location and with a preset safe distance as the radius as the background water sample area. The statistical parameters of the comprehensive response of the floating film masking in the background water sample area are calculated and the background adaptive threshold is set. The grid set with the comprehensive response of the floating film masking higher than the background adaptive threshold and meeting the spatial connectivity requirements is filled into the candidate floating film masking area to extract the tidal surface convergent floating film masking zone.

3. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 2, characterized in that, The skeleton generation module generates skeleton point positions, floating membrane half-width, and lateral direction normal vectors. Based on the meteorological window set, it removes unexecutable skeleton segments, including: A macroscopic baseline indicating the main diffusion direction of the pollution plume is generated by combining the location of the pollution source with the surface velocity vector. Cross-sections are laid out along the normal to the macroscopic baseline, and the local maxima grid cells of the integrated response of the floating film on the cross-section are connected in sequence to generate the floating film boundary skeleton line. The skeleton point position is determined by parameter normalization. Search outwards along the normal directions on both sides from the skeleton point position, and record the lateral span when the integrated response of the floating film masking decays to less than the background adaptive threshold as the half width of the floating film band. Based on the surface velocity vector at the location of the pollution source along the skeleton point, the reference reference vector is constructed by subtracting the coordinates of the skeleton point from the endpoint of the extension according to the time scale parameter. The candidate normal unit vector with the largest cosine value of the dot product with the reference reference vector is selected as the lateral direction normal vector. The three-dimensional spatial coordinates corresponding to the boundary skeleton lines and skeleton point positions of the floating membrane are intersected with the executable three-dimensional geospatial envelope generated by the difference operation of the plane based on the wind shear safety threshold and the polygonal column of the no-fly zone due to rain or dense fog. The corresponding skeleton segments outside the intersection range are then removed.

4. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 1, characterized in that, The task reconstruction module calculates the lateral offset distance based on the half-width of the floating membrane strip, including: For water sampling tasks, the following calculations are performed: half width of the floating film, radius of influence of the rotor downwash, horizontal sway of the sampling device calculated based on the physical length of the sampling sling and the current wind speed in the meteorological window set combined with the wind deflection physical model, and the sum of the preset safety interval distances are used as the lateral offset distance. For low-altitude spectral acquisition tasks, the lateral offset distance is configured to be no less than the sum of the half-width of the floating film strip and the preset safety interval distance, and no greater than the half-width of the field of view coverage determined by the target execution height and the field of view projection of the spectral sensor. When there is no lateral offset distance that meets the constraints at the target execution height, the target execution height is gradually increased according to the preset height step size to expand the half-width coverage of the field of view.

5. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 4, characterized in that, The task reconstruction module determines the location of associated skeleton points and, combined with the lateral direction normal vector, target lateral orientation, target execution height, and lateral offset distance, generates the 3D coordinates of the lateral task points, including: Calculate the horizontal plane distance between the initial nominal position of the original data acquisition task and the positions of all skeleton points on the floating membrane boundary skeleton line. If the minimum horizontal plane distance is greater than the task association threshold composed of the sum of the half width of the floating membrane strip and the preset spatial tolerance distance, then the original data acquisition task is determined to be a non-floating membrane associated task. If the minimum horizontal plane distance is not greater than the task association threshold, then the original data acquisition task is determined as a floating film association task, and the skeleton point position corresponding to the minimum horizontal plane distance is determined as the association skeleton point position. Based on the three-dimensional coordinates of the associated skeleton point located on the water surface reference elevation, a horizontal projection coordinate offset consisting of the target side, the lateral direction normal vector, and the lateral offset distance is added, and a height vector consisting of the target execution height and the unit vector pointing upwards perpendicular to the water surface is superimposed to generate the three-dimensional coordinates of the side mission point. The three-dimensional coordinates of the lateral task point are verified against the executable three-dimensional geospatial envelope. If the three-dimensional coordinates of the lateral task point fall outside the executable three-dimensional geospatial envelope, an alternative skeleton point is searched along the floating membrane boundary skeleton line within the preset skeleton search range, and the alternative lateral task point is recalculated.

6. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 5, characterized in that, The flight segment construction module determines the flight segment type and uses dynamic load as a delay calculation variable, including: When the starting and ending tasks of a candidate route segment are both related to the floating membrane and are located on the same execution side of the same floating membrane boundary skeleton line corresponding to the related skeleton point, the candidate route segment is determined to be a same-side parallel membrane route segment. When the starting and ending tasks of a candidate route segment are both floating membrane-related tasks and are located on different execution sides of the same floating membrane boundary skeleton line corresponding to the associated skeleton point, the candidate route segment is determined to be a cross-membrane switching route segment. When the starting or ending mission of a candidate route segment is determined to include at least a non-floating membrane-related mission, a takeoff point, or a return point, the candidate route segment is determined to be an off-membrane transfer segment. When calculating the basic flight energy consumption of candidate route segments, the three-dimensional flight distance and altitude changes that are fixedly related to the candidate route segments are integrated. The wind resistance scalar component is obtained by performing an inner product operation with the wind speed vector contained in the meteorological window set and the unit direction vector of the candidate route segment, or by performing a cross product operation to obtain the modulus, and a static baseline cost is generated. The dynamic load generated by the candidate route segment is used as a delay calculation variable and handed over to the collaborative scheduling module to refresh the total cost of the floating membrane coupled route.

7. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 6, characterized in that, The segment construction module calculates the total cost of the floating membrane coupled route based on the segment type and equivalent energy consumption, including: The cost of a regular route is generated by summing the base flight energy consumption of the candidate route segment with the base energy consumption of the destination mission. For cross-membrane switching segments, additional costs for cross-membrane switching and additional costs for adjacent floating membrane operations are included in addition to the regular route costs. Among them, the additional cost of cross-membrane switching and the additional cost of floating membrane adjacent operation are determined based on the product of the pre-calibrated additional operation time and the UAV hovering power parameter; For the same-side parallel membrane segment, the revenue item for continuous execution of the same-side parallel membrane is subtracted from the cost of the regular route, and the revenue item for continuous execution of the same-side parallel membrane is constrained to not exceed a preset ratio of the sum of the static baseline cost and the additional cost. All indicators involved in the cost calculation are uniformly converted into equivalent energy consumption units to calculate the total cost of the floating membrane coupled route.

8. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 1, characterized in that, The collaborative scheduling module calculates the total cost of refreshing the floating membrane coupled route based on the actual dynamic load according to the time sequence, including: In the initial mission execution sequence, the load added after the water sampling mission is executed is accumulated point by point according to the time node to analyze the actual dynamic load. The actual dynamic load is substituted into the basic flight energy consumption equation to refresh the specific value of the total cost of the floating membrane coupled route. Apply multi-dimensional dynamic constraint verification to the initial task execution sequence after refreshing the cost value to determine whether the estimated total flight energy consumption of the multi-rotor UAV exceeds the difference between the current remaining power and the power reserved for safe return, whether the actual dynamic load exceeds the maximum allowable load, and whether the spatiotemporal interaction nodes are completely within the envelope range defined by the meteorological window period set. If any constraint is broken, the violation task will be reassigned across machines or removed from the task execution sequence.

9. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 8, characterized in that, The collaborative scheduling module performs iterative optimizations, including: Test sequences are generated by exchanging adjacent task sequences within the same multirotor UAV or by exchanging tasks on the same side across different UAVs. The total cost of the floating membrane coupled route and the total return energy consumption within each multi-rotor UAV mission execution sequence are aggregated, and the total data benefit benchmark value corresponding to the data acquisition task included in the corresponding mission execution sequence and converted into equivalent energy consumption units is subtracted to construct a multi-UAV cooperative scheduling total cost function model. For test sequences that satisfy the multidimensional dynamic constraint verification, if the total cost of multi-UAV collaborative scheduling corresponding to the test sequence is lower than that of the current sequence, test sequence iteration is adopted. If the total cost of multi-UAV collaborative scheduling corresponding to the test sequence is not reduced, the annealing probability is calculated based on the cost increment of the test sequence relative to the current sequence and the system virtual temperature parameter that decays exponentially with the number of iterations. The suboptimal solution is accepted based on the annealing probability to escape the local optimum. The iteration continues until the total cost of multi-drone collaborative scheduling no longer decreases or reaches the preset limit for the number of iterations.

10. The multi-rotor UAV data acquisition route optimization system for water area monitoring according to claim 8, characterized in that, The collaborative scheduling module performs local incremental replanning, including: In response to the dynamic changes in the boundary skeleton line of the floating membrane caused by the update of meteorological window period collection or the refresh of multi-source sensor data, the real-time three-dimensional physical coordinates, real-time remaining power and current physical load of the affected multi-rotor UAV are locked as the virtual starting node for incremental replanning. Extract tasks whose geographical coordinates have changed due to boundary displacement, route segments whose lateral attributes have changed, and unexecuted segments directly adjacent to affected nodes for recalculation, while maintaining the original control commands for completed historical nodes and unaffected space segments.

Images