An anti-interference unmanned aerial vehicle swarm scheduling method and system

Abstract

The present disclosure provides an anti-interference UAV swarm scheduling method and system, comprising: constructing a spatial direction interference characterization model based on time sequence observation data of a UAV swarm; constructing an interference cone identification model composed of interference cones generated by mapping a concentrated interference area; constructing an anti-interference task allocation model to generate a task sequence sorted by UAVs for the UAV swarm; constructing a path planning model based on interference cone avoidance based on the interference cone and the task sequence to generate a flyable path that can effectively avoid the interference cone and meet kinematic constraints; and constructing a dynamic rescheduling mechanism based on interference mutations to monitor changes in the interference environment in real time, update the interference cone, and locally redistribute and re-plan paths for affected tasks and UAVs. This method can identify interference direction structures, depict interference boundary mutation characteristics, and achieve anti-interference path avoidance, solving the problems of communication interruption and scheduling failure.

Description

Technical Field

[0001] This disclosure relates to the fields of unmanned aerial vehicle (UAV) swarm control and electromagnetic interference suppression technology, and in particular to an interference-resistant UAV swarm scheduling method and system. Background Technology

[0002] In live-line inspection scenarios of busbar areas in high-voltage substations, the busbar conductors and their auxiliary equipment form a distinctly directional electromagnetic radiation structure during long-term operation. Especially at busbar corners, downlead connections, and near large reactor equipment, electromagnetic energy does not diffuse uniformly but rather forms concentrated radiation areas resembling fans or cones along specific spatial directions. When a drone flies within a range of 30 to 80 meters from the busbar according to a preset inspection path, its communication signal strength remains relatively stable in most of the spatial area, and the corresponding interference level is also relatively mild.

[0003] However, when a UAV flies into the main radiation region corresponding to the extension direction of the busbar conductor, the electromagnetic field intensity in this region increases significantly within a very short time. This causes a rapid decrease in the signal-to-noise ratio of the communication link and a sharp increase in packet loss, resulting in a sudden change in the interference level from low to high in space. This sudden change has a clear directional boundary, rather than a continuous transition. Because existing scheduling methods treat interference as a continuously distributed and slowly changing field in space, they are usually estimated through neighborhood smoothing or interpolation, and therefore cannot identify this interference region with strong directional concentration and abrupt boundary changes. During path planning, this region is still incorrectly identified as passable space. When the UAV actually flies into this region, the communication link will be momentarily interrupted, causing scheduling commands to be unreceived or unexecuted, thus triggering flight control anomalies or even mission interruption. Therefore, in engineering scenarios with strong directional electromagnetic radiation, simply modeling interference as a continuously changing space cannot accurately reflect its spatial structural characteristics, leading to the failure of path planning and scheduling decisions in critical areas.

[0004] Therefore, there is an urgent need for a drone swarm scheduling method that can identify the direction structure of interference, characterize the abrupt changes in interference boundaries, and achieve anti-interference path avoidance, in order to solve the problem of communication interruption and scheduling failure caused by abrupt changes in interference in strong directional electromagnetic radiation scenarios such as busbar areas of high-voltage substations. Summary of the Invention

[0005] In view of this, in order to solve the problems caused by the existing technology, this application provides an anti-interference drone swarm scheduling method and system.

[0006] In a first aspect, this disclosure provides an anti-interference method for scheduling unmanned aerial vehicle (UAV) swarms, the method comprising:

[0007] S1: Based on the time-series observation data of UAV swarms, construct a spatial directional interference characterization model that includes the directional interference intensity and directional interference gradient on each directional sector;

[0008] S2: Based on the spatial directional interference characterization model, construct an interference cone identification model composed of interference cones generated by mapping the interference concentration region. The interference cone identification model includes the geometric parameters and risk level of the interference cones.

[0009] S3: Based on the interference cone, construct an anti-interference task allocation model, quantify the risk of crossing the interference cone into a task allocation cost, and generate a task sequence for the drone swarm after merging and organizing the drones.

[0010] S4: Based on the interference cone and the task sequence, construct a path planning model based on interference cone avoidance, and generate a flyable path for the UAVs assigned tasks in the task sequence that can effectively avoid the interference cone and satisfy kinematic constraints.

[0011] S5: During the process of the UAV performing the mission according to the flightable path, a dynamic rescheduling mechanism based on interference mutation is constructed to monitor the changes in the interference environment in real time. When interference mutation is detected, the interference cone is updated, and the affected missions and UAVs are locally reassigned and replanned based on the updated interference cone and the mission sequence.

[0012] Optionally, S1 includes:

[0013] The time-series observation data of the UAV swarm are processed by time axis alignment and dimension normalization to generate a standard observation dataset that is consistent in time and space.

[0014] Centered on the current location of the drone, the omnidirectional space is discretized into multiple directional sectors;

[0015] Based on the standard observation dataset, the directional interference intensity in each directional sector is calculated, and a robust directional interference intensity sequence is formed by fusion through a time window.

[0016] Calculate the directional interference gradient based on the directional interference intensity of adjacent directional sectors;

[0017] The directional interference intensity and the directional interference gradient on the same directional sector are fused to generate a directional interference characterization unit;

[0018] The spatial directional interference characterization model is composed of directional interference characterization units of the UAV swarm at all sampling times and in all directional sectors.

[0019] Optionally, S2 includes:

[0020] Based on the directional interference gradient, potential interference mutation directions are identified through adaptive threshold determination.

[0021] The potential interference mutation directions are continuously confirmed over a time window, and transient jitters are eliminated while confirmed mutation directions are retained;

[0022] Centered on the confirmed mutation direction, an interference concentration region with a continuous angular range is constructed by continuously expanding in adjacent directions;

[0023] The interference concentration area is mapped as an interference cone in three-dimensional space, and its geometric parameters and risk level are estimated. The generated interference cone and its geometric parameters and risk level constitute the interference cone identification model.

[0024] Optionally, S3 includes:

[0025] Calculate the basic cost elements between each UAV and the task to be assigned. The basic cost elements include the straight-line distance between the UAV and the task pair, the total time it takes for the UAV to fly to the task point and perform the operation, the estimated energy consumption, and the feasibility marker.

[0026] Assess the risk exposure of each UAV-mission pair's flight path crossing the interference cone;

[0027] The basic cost elements are integrated with the risk exposure to construct a comprehensive task cost;

[0028] Under the task allocation constraint, the solution is performed with the goal of minimizing the comprehensive task cost, generating a task allocation set, and then the task allocation set is merged and organized into a task sequence according to the UAV. The task sequence includes the task point coordinates, expected operation time and comprehensive task cost of each task.

[0029] Optionally, S4 includes:

[0030] The interference cone is transformed into a high-cost occupied area in path planning, and a determination function is established to determine whether a spatial point falls into the cone's influence range.

[0031] The flight path is discretized into path segments, and the interference penalty cost for each path segment is calculated according to the determination function.

[0032] The interference penalty cost is fused with the path segment length cost and boundary proximity penalty to construct the comprehensive cost of each path segment;

[0033] Based on the comprehensive cost of the path segment, an initial avoidance path from the starting point to the target point is generated using a graph search method;

[0034] The initial evasion path is subjected to curvature constraints and smoothing to generate an executable evasion path that satisfies the kinematic constraints of the UAV.

[0035] Optionally, the determination function considers both the radial distance of the spatial point relative to the vertex of the cone and the angle relative to the axis of the cone, and introduces a safety expansion boundary.

[0036] During the flight of the UAV, the comprehensive interference intensity index and its rate of change are calculated based on the directional interference characterization unit collected in real time, and interference mutation events are triggered by continuous determination.

[0037] Based on the directional data at the moment of triggering mutation, the set of interference cones is incrementally updated to generate new interference cones or correct the parameters of existing interference cones;

[0038] Based on the updated interference cone, the affected UAVs that are about to enter the high-risk area in the remaining flight path are identified, and their associated affected tasks are collected to form a rescheduling trigger set;

[0039] Local reassignment is performed on the affected tasks within the rescheduling trigger set, and new evasion paths are planned for the reassigned UAVs.

[0040] Secondly, this disclosure provides an anti-interference drone swarm scheduling system applied to the method of the first aspect, the system comprising:

[0041] The spatial directional interference characterization model construction module is used to construct a spatial directional interference characterization model based on the time-series observation data of UAV swarms, which includes the directional interference intensity and directional interference gradient in each directional sector.

[0042] The interference cone identification model construction module is used to construct an interference cone identification model based on the spatial directional interference characterization model, which is composed of interference cones generated by mapping the interference concentration region. The interference cone identification model includes the geometric parameters and risk level of the interference cones.

[0043] The anti-interference task allocation and task sequence generation module is used to construct an anti-interference task allocation model based on the interference cone, quantify the risk of crossing the interference cone into a task allocation cost, and generate a task sequence for the drone swarm after being merged and organized by drones.

[0044] The interference cone avoidance path planning module is used to construct a path planning model based on interference cone avoidance based on the interference cone and the task sequence, and generate a flyable path that can effectively avoid the interference cone and meet the kinematic constraints for the UAVs that have been assigned tasks in the task sequence.

[0045] The dynamic rescheduling module based on interference mutation is used to construct a dynamic rescheduling mechanism based on interference mutation during the process of UAVs executing tasks according to the flightable path. It monitors changes in the interference environment in real time, updates the interference cone when interference mutation is detected, and performs local reallocation and path replanning for the affected tasks and UAVs based on the updated interference cone and the task sequence.

[0046] Thirdly, this disclosure provides an electronic device including a memory and at least one processor, the memory storing a computer program, and the processor executing the computer program to implement the method of the first aspect described above.

[0047] Fourthly, this disclosure provides a computer storage medium storing a computer program that, when executed, implements the method described in the first aspect.

[0048] The beneficial effects of this disclosure are that, compared with the prior art, this disclosure has the following advantages:

[0049] 1) To address the problem that existing technologies simply model interference as a continuously changing spatial field and fail to identify interference regions with directional concentration and abrupt boundary changes, a spatial directional interference characterization model is constructed. This model aligns the time axis and discretizes the directional units of the heterogeneous time-series observation data reported by UAV swarms, calculating the interference intensity and gradient in each direction. This achieves an accurate characterization of the spatial directional structure of interference and its abrupt boundary changes. Based on this, an interference cone identification model is constructed, transforming the directional interference characterization into a geometric interference cone with vertex position, axial direction, expansion angle, effective radius, and risk score. This solves the technical deficiency of traditional methods in being unable to quantify the directionality and abrupt boundary changes of interference, providing accurate environmental perception information for subsequent scheduling decisions.

[0050] 2) To address the problem of communication interruption and scheduling failure after UAVs fly into strong directional electromagnetic radiation scenarios due to the misjudgment of interference abrupt change areas as passable spaces by existing scheduling methods, an anti-interference task allocation model is constructed. The risk of interference cones is explicitly incorporated into the task allocation decision, and the crossing risk exposure of UAV-task pairs is calculated and included in the comprehensive cost. This enables the swarm to actively avoid task combinations that require crossing high-risk cones during the task allocation phase, thereby reducing the probability of UAVs entering strong directional radiation areas from the source. Furthermore, by constructing a path planning model based on interference cone avoidance, interference cones are mapped as plannable high-cost occupied areas, generating flyable avoidance paths that meet curvature continuity constraints. This ensures that UAVs can effectively avoid interference abrupt change areas during flight, guaranteeing stable communication links and continuous task execution.

[0051] 3) To address the issue of new interference cones suddenly appearing or existing interference intensifying in a dynamically changing electromagnetic environment, a dynamic rescheduling mechanism based on interference mutation is constructed. During UAV flight, the interference change rate is monitored in real time and continuously assessed. Once an interference mutation is triggered, the interference structure set is incrementally updated, the affected UAVs are identified, and local reassignment and path replanning are performed on the affected tasks. This enables the swarm to quickly adapt to the dynamic electromagnetic environment, avoids batch task interruptions caused by environmental mutations, and significantly improves the robustness and reliability of the UAV swarm scheduling system in complex and strong directional electromagnetic radiation scenarios. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0053] Figure 1 A flowchart of an anti-interference drone swarm scheduling method provided in an embodiment of this disclosure is shown;

[0054] Figure 2 A flowchart illustrating the anti-interference task allocation and task sequence generation provided in an embodiment of this disclosure is shown.

[0055] Figure 3 A flowchart illustrating the interference cone avoidance path planning provided in an embodiment of this disclosure is shown.

[0056] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0057] The present disclosure will be further described below with reference to the accompanying drawings. The following embodiments are only used to illustrate the technical solutions of the present disclosure more clearly, and should not be used to limit the scope of protection of the present disclosure.

[0058] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0059] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of the invention, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0060] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0061] This disclosure provides an anti-interference drone swarm scheduling method. The method identifies the spatial directional structure of interference by constructing a spatial directional interference characterization model, then builds an interference cone identification model based on this model, and performs anti-interference task allocation, path planning based on interference cone avoidance, and dynamic rescheduling based on interference mutations, thereby achieving reliable scheduling of drone swarms in strong directional electromagnetic radiation scenarios. The technical solution provided by this disclosure is described in detail below with reference to the accompanying drawings.

[0062] Figure 1 A flowchart of the anti-interference drone swarm scheduling method provided in this disclosure embodiment is shown below. Figure 1 As shown, the process may include the following steps:

[0063] S1: Based on the time-series observation data of UAV swarms, a spatial directional interference characterization model is constructed, which includes the directional interference intensity and directional interference gradient on each directional sector.

[0064] This step is used to construct a spatial directional interference characterization model. By aligning the time axis, discretizing the directional units, and calculating the interference intensity and gradient of the heterogeneous time-series observation data reported by the swarm of UAVs, the original communication quality data is transformed into an interference characterization with directional resolution capabilities, providing input data for subsequent identification of interference cones. This is specifically achieved through the following sub-steps.

[0065] S1.1: By aligning the time axis and normalizing the dimensions of the heterogeneous time-series observation data of the UAV swarm, a spatiotemporally consistent standard observation dataset is generated.

[0066] The system collects real-time airborne data reported by each UAV in the swarm during flight at a fixed sampling period. This sampling period can be set based on the actual communication bandwidth and computing resources, typically between 0.05 and 0.5 seconds, to ensure that rapidly changing interference characteristics are captured without introducing excessive data processing burden. For each UAV, the system obtains its timestamp, spatial coordinates, communication signal strength, and link packet loss rate at each raw sampling moment. The spatial coordinates are provided by the UAV's positioning system, including longitude, latitude, and altitude information, or converted to three-dimensional coordinates in the local Cartesian coordinate system. The communication signal strength is usually given in decibels and milliwatts, reflecting the received signal power level. The link packet loss rate characterizes the proportion of data packets lost per unit time and is obtained statistically by the airborne communication module.

[0067] Because the data reporting times of different UAVs may differ, and the timestamps of different observations from the same UAV may also be out of sync, the system needs to resample and align all the original observation data. Specifically, the system uses a uniform sampling period as a benchmark and interpolates or performs nearest neighbor matching on the original data sequence of each UAV to generate aligned observations on the same sampling time sequence. For spatial location, linear interpolation can be used to estimate the aligned time position based on the position and velocity of the preceding and following times. For signal strength and packet loss rate, nearest neighbor scalars or linear interpolation can be used to obtain the estimated aligned time. Through this resampling and alignment operation, the system obtains a set of observation records that are completely aligned on the time axis. Each record contains a uniform sampling time number, UAV identifier, aligned spatial location coordinates, aligned signal strength value, and aligned packet loss rate.

[0068] After aligning the time axis, the system further transforms the physical dimensions of the signal strength. Since communication signal strength is typically expressed in logarithmic decibels and milliwatts (dB / mW), and subsequent interference strength calculations involve power ratios and multiplication, directly using logarithmic values ​​would lead to dimensional conflicts and unclear physical meaning of the calculation results. Therefore, the system converts the aligned signal strength values ​​from DB / mW to linear power values, with the conversion relationship satisfying the following formula: ,in, The linear power of the received signal of the i-th UAV at the n-th sampling time is expressed in milliwatts. This represents the corresponding received signal strength, measured in decibels and milliwatts. This exponential operation restores the signal strength from a logarithmic scale to a physical power scale, enabling subsequent power ratio calculations and weighted multiplication operations, thus ensuring the engineering feasibility and numerical stability of interference strength calculations.

[0069] Through the above processing, the system obtains a standardized observation dataset for each UAV at each aligned sampling moment, including spatial location coordinates, linear power values, and packet loss rate, providing a reliable data foundation for subsequent directional analysis.

[0070] S1.2: Deploy directional sampling units centered on the current position of each UAV to discretize the omnidirectional space into multiple directional sectors.

[0071] Based on the obtained standard observation dataset, the spatial position of each UAV at each sampling time is processed. Using this position as the origin, directional sampling units are established in the horizontal plane. The number M of directional sampling units can be set according to the angular resolution requirements of the actual scene, typically ranging from 8 to 36. A larger value for M allows for capturing more subtle directional changes, but the computational complexity increases accordingly. In practical applications, a trade-off can be made between the spatial scale of the substation busbar area and the distribution density of interference sources, for example, selecting 16 or 24 directions.

[0072] The angular interval between adjacent directional sampling units is determined by the total number of directions, satisfying the following formula: ,in, The angle interval between adjacent sampling units is expressed in radians; M is the number of sampling units in each direction, ranging from 8 to 36. This formula divides the omnidirectional space into M sector-shaped regions, providing a strictly consistent angular reference for subsequent calculations of gradient changes between adjacent directions.

[0073] The direction numbers increment sequentially in either clockwise or counterclockwise order, with a fixed reference datum for the zero direction. The zero direction can be set to the UAV's current heading or the geographic north direction. Using a fixed reference datum ensures that the direction numbers of the same UAV at different times, and different UAVs at the same time, have consistent physical meaning, avoiding direction misalignment caused by changes in the reference frame. In practice, if the UAV's heading is used as the reference, the heading angle needs to be recorded in real time, and the zero direction dynamically adjusted at each sampling moment. If the geographic north direction is used as the reference, the direction sampling unit is fixed relative to the absolute coordinate system, and each UAV only needs to project its direction based on its own position. Both methods can achieve standardized definition of the direction unit, which can be set according to the coordinate system of the actual system.

[0074] After the direction unit definition is completed, the system generates a direction sampling configuration table, which includes the total number of directions M and the angle interval. This table specifies the relative azimuth angle range corresponding to each direction number and the reference datum for the zero direction. It will serve as a unified geometric reference for all subsequent directional calculations, ensuring consistency in direction definitions throughout the entire methodology.

[0075] S1.3: Calculate the instantaneous interference intensity in each direction based on the linear power and packet loss rate in the standard observation dataset, and obtain a robust directional interference intensity sequence through exponential weighted fusion using a time window.

[0076] The system needs to determine a reference signal linear power value, which characterizes the baseline signal level under interference-free or low-interference conditions. The reference power can be calculated based on historical observation data, such as the moving quantile of the signal power received by the UAV during its stable flight period before entering the target area, or preset based on engineering calibration values. The reference power typically ranges from 0.001 mW to 0.1 mW, corresponding to typical reception levels in low signal strength scenarios. Furthermore, to avoid calculation divergence due to an excessively small denominator when the signal power is extremely low, the system introduces a power lower limit protection term, whose value range is... milliwatts to Milliwatts.

[0077] For the i-th UAV at the n-th sampling time, the system obtains its linear power value. and packet loss rate The link packet loss rate is defined as the proportion of lost data packets to the total number of data packets that should be received, with a theoretical range of [0,1]. In practical engineering, when communication is severely disrupted, the packet loss rate may approach 1. The system does not truncate the packet loss rate but directly uses the measured values ​​into the formula to ensure that the interference intensity truly reflects the degree of communication quality degradation. Based on the reference power, linear power, and packet loss rate, the system first calculates the instantaneous basic interference intensity in each direction at that moment. Since the direction sampling unit only defines the spatial orientation, and the current observation is a global measurement of the UAV's own position without distinguishing between different directions, the system treats the current linear power and packet loss rate as omnidirectionally consistent observations when calculating the interference intensity in each direction. That is, it assumes that within a small time scale, the instantaneous changes in interference are synchronous in all directions. Based on this, the system assigns the instantaneous basic interference intensity to each direction unit to obtain the instantaneous interference intensity in each direction. Subsequently, the spatial directional characteristics of the interference will be gradually revealed through multi-moment fusion and directional gradient calculation.

[0078] The instantaneous basic interference intensity, that is, the instantaneous interference intensity in each direction, is calculated according to the following formula:

[0079]

[0080] in, This represents the interference intensity of the i-th UAV at the nth sampling time in the direction number k. This value is dimensionless and its range is constrained to be between 0 and 1. Linear power; Reference linear power, in milliwatts; This is the lower power limit protection item, in milliwatts; This represents the link packet loss rate. The formula uses a min function to ensure the power ratio term does not exceed 1, thus ensuring the interference intensity is always non-negative, with its value constrained between 0 and 1. The physical meaning of this formula is: the lower the received power and the higher the packet loss rate, the stronger the interference intensity. Both the power ratio term and the packet loss penalty term are dimensionless, ensuring dimensional consistency in subtraction and multiplication operations.

[0081] Due to random fluctuations in communication links, the instantaneous interference intensity at a single moment may contain sudden spikes or occasional jumps. Directly using this data for subsequent gradient analysis and cone identification could lead to misjudgments. Therefore, the system employs an exponentially weighted fusion method with a finite time window to smooth the instantaneous interference intensity, suppressing short-term disturbances while preserving the continuous trend of interference changes. The fused directional interference intensity satisfies the following formula:

[0082]

[0083] in, Indicates the directional interference intensity after fusion; W is the fusion window length, ranging from 5 to 20. The fusion factor ranges from 0.6 to 0.9. The values ​​of these two parameters depend on the rate of change of interference in the actual scenario: the more frequent the interference changes, the smaller the window length can be to maintain response sensitivity; the more stable the interference background, the larger the window length can be to obtain a more stable fusion result. The summation sign represents a decay-weighted average of the instantaneous interference intensity over the past W time points. The exponential weighting method ensures that observations closer to the current time contribute more, and earlier data contribute less, thus smoothing noise while maintaining sensitivity to recent changes. This fusion process suppresses sudden spikes into robust signals that can be used for subsequent gradient determination, while preserving the changing trend of persistent interference.

[0084] Through the above calculations, the system obtains a fused directional interference intensity value for each UAV in each direction at each sampling time, forming a directional interference intensity sequence, which will serve as the basic input for subsequent directional gradient calculations and directional characterization.

[0085] S1.4: Calculate the interference gradient in adjacent directions and fuse the interference intensity and gradient to generate a directional characterization unit for cone recognition.

[0086] Based on the fusion interference intensity in adjacent directions, the directional interference gradient is calculated to characterize the drastic spatial variation of the interference. The interference intensity and the interference gradient are combined to form a directional characterization unit, providing a comprehensive judgment index for subsequent interference cone identification.

[0087] Based on the obtained fused directional interference intensity sequence, the interference change gradient between adjacent directions is calculated for each UAV at each sampling time. The directional interference gradient reflects the drastic change in interference intensity with direction. When there is a boundary where the interference intensity abruptly changes from a low-interference region to a high-interference region, the gradient value in that direction will increase significantly. The gradient calculation uses the square of the difference in interference intensity between adjacent directions to emphasize the contribution of the abrupt boundary, thus suppressing small fluctuations and highlighting large abrupt changes. To avoid numerical instability caused by an excessively small denominator when the number of directions is large, the system introduces an angle lower limit protection term. The calculation of the directional interference gradient satisfies the following formula:

[0088]

[0089] in, This represents the directional interference gradient intensity of the i-th UAV at the nth sampling time and in the direction k. This value is dimensionless. The directional interference intensity after fusion; This represents the fusion interference intensity in the adjacent preceding direction; The interval is the direction angle; This is the lower limit protection item for angles, in radians, with a value range of... arrive The use of the squared term reduces the contribution of the gradient value to small fluctuations and significantly amplifies its contribution to abrupt change boundaries, thereby more accurately identifying the spatial boundary location of disturbances.

[0090] After obtaining the directional disturbance gradient, the system needs to fuse the directional disturbance intensity information with the directional disturbance gradient information to form a comprehensive directional characterization index. This index reflects information in two dimensions: the magnitude of the disturbance intensity in the current direction and the steepness of the boundary between this direction and its adjacent directions. To balance the contributions of the two pieces of information, a weighted combination method is used for fusion, and the gradient term is compressed to the range of 0 to 1 to avoid the gradient value being too large and masking the influence of the intensity term.

[0091] The combined calculation of the directional interference characterization unit satisfies the following formula.

[0092]

[0093] in, It is a dimensionless unit for characterizing directional interference. and The combined weights range from 0.3 to 0.7, and the sum of the two is 1. The directional interference intensity after fusion; The gradient represents the directional interference gradient. A fractional term is used to compress the gradient strength to a range of 0 to 1. When the gradient is small, the fractional value is close to 0; when the gradient is large, the fractional value approaches 1, thus placing it on the same order of magnitude as the interference strength term, which is also within the range of 0 to 1. Through weighted combination, the system merges information from two dimensions—the strength of the interference and the steepness of the interference boundary—into a unified index. This index can be used for subsequent threshold determination and for similarity comparison during region expansion.

[0094] The spatial directional interference set, composed of directional interference characterization units from all UAVs at all sampling times and in all directional sectors, constitutes the spatial directional interference characterization model constructed in step S1. This model comprehensively and structurally depicts the directional distribution characteristics of interference in three-dimensional space, providing a data foundation for subsequent steps to identify interference cones.

[0095] In the technical solution of this disclosure, heterogeneous time-series observation data reported by swarm drones is processed by time axis alignment and physical dimension normalization, and omnidirectional space is discretized based on directional sampling units to calculate the interference intensity and its gradient in each direction, thereby constructing a spatial directional interference characterization model with direction resolution capability. This model can effectively capture the characteristics of interference changing with spatial direction, transforming the original communication signal strength and packet loss rate into a unified index that integrates interference intensity and boundary abrupt change information. This provides a reliable data foundation for subsequent accurate identification of interference regions with directional concentration and boundary abrupt change, avoiding the problem of traditional methods treating interference as a continuous uniform field and failing to identify directional abrupt changes.

[0096] S2: Based on the spatial directional interference characterization model, construct an interference cone identification model composed of interference cones generated by mapping the interference concentration region. The interference cone identification model includes the geometric parameters and risk level of the interference cones.

[0097] This step is used to construct an interference cone identification model. Based on the directional interference representation model obtained in S1, the directional interference representation is transformed into an interference cone structure with geometric parameters through adaptive threshold determination, time persistence confirmation, and adjacent directional continuity expansion. This provides a quantifiable risk area for subsequent task allocation and path planning. Specifically, this is achieved through the following sub-steps.

[0098] S2.1: Based on the directional disturbance gradient, potential disturbance mutation directions are identified by adaptive threshold determination.

[0099] Statistical analysis is performed on the interference gradients of all directions for each UAV at the same sampling time to generate a mutation judgment threshold that adapts to the environmental background. Based on this threshold, possible interference mutation directions are identified, providing a candidate set for subsequent persistence confirmation.

[0100] Extract the directional interference gradient value of each UAV in each direction at each sampling time from the spatial directional interference set obtained in step S1. For the i-th UAV at the nth sampling time, there are M directional gradient values ​​at that time, forming a gradient set. Since the electromagnetic background may vary significantly between different substation areas and at different flight altitudes, using a fixed threshold for abrupt change detection is difficult to adapt to complex and changing environments. Therefore, the system adopts an adaptive threshold generation method based on quantiles and dispersion, enabling the threshold to automatically adjust according to the current background interference level. The adaptive abrupt change threshold is calculated according to the following formula:

[0101]

[0102] in, Let be the adaptive mutation threshold for the i-th UAV at the n-th sampling time, which is dimensionless; The gradient values ​​in all directions at the current time. The quantile, with the quantile level p ranging from 70% to 90%, is used to characterize the background disturbance gradient level at the current moment; The scaling factor, ranging from 0.5 to 2, controls the leniency of the threshold relative to background fluctuations. The summation term calculates the sum of squared deviations of the gradients in each direction relative to the quantiles, divides by the degrees of freedom, and takes the square root to obtain the sample standard deviation, which characterizes the fluctuation amplitude of the gradient value at the current moment. The core idea of ​​this formula is to use the quantiles to represent the background level and the standard deviation to represent the fluctuation intensity, and the sum of the two yields a robust threshold that varies with the scene.

[0103] After obtaining the adaptive threshold, the system evaluates the degree of mutation for the gradient value in each direction. To transform the rigid threshold comparison into a continuous mutation score, facilitating subsequent persistence constraints and neighborhood expansion, the system employs a sigmoid function to map the difference between the gradient value and the threshold to a score between 0 and 1. The mutation score is calculated according to the following formula:

[0104]

[0105] in, Let k be the mutation score of the i-th UAV at the nth sampling time in direction k, with a value ranging from 0 to 1; The slope coefficient, ranging from 5 to 20, controls the steepness of the transition near the threshold. When the directional gradient is much greater than the threshold, the exponential term approaches 0, and the mutation score approaches 1; when the directional gradient is much less than the threshold, the exponential term approaches infinity, and the mutation score approaches 0; near the threshold, the score transitions smoothly. This processing transforms the binary determination of whether the threshold is exceeded into a continuous mutation score, providing richer quantitative information for subsequent temporal persistence confirmation and spatial continuity extension.

[0106] Through the above calculations, the system generates a mutation score for each UAV in each direction at each sampling time, forming a candidate set of interfering mutation directions. This set is used to further eliminate occasional spikes and confirm the true mutation directions with persistence.

[0107] S2.2: Perform continuous confirmation of the mutation direction over a time window to eliminate transient fluctuations and retain the true mutation direction.

[0108] The mutation scores in the same direction within a continuous time window are cumulatively evaluated to eliminate false triggers caused by instantaneous signal fluctuations, and only retain the true interference mutation directions that are persistent in the time dimension.

[0109] Mutation score sequence obtained based on S2.1 For each drone and each direction, a continuous time-dimensional analysis is performed independently. Since single-point spikes may be caused by communication interruptions or random interference, judging a sudden change solely based on a high score at a single moment would lead to frequent false triggers and invalid cone construction. Therefore, the system introduces a time-window cumulative scoring mechanism, requiring the sudden change signal to maintain a high level for a consecutive number of sampling moments before confirmation. The continuous confirmation score is calculated according to the following formula:

[0110]

[0111] in, The persistence confirmation score for the i-th UAV at the nth sampling time in direction k, with a value ranging from 0 to 1; The time window length, ranging from 2 to 6 sampling points; This is the time decay factor, ranging from 0.6 to 0.9. The formula uses an exponentially weighted average method to calculate the most recent... The system fuses the mutation scores from each time step, with scores closer to the current time step contributing more, thus smoothing noise while maintaining sensitivity to recent changes. A persistence confirmation threshold is set for the system. The value ranges from 0.6 to 0.8. Mutations with a persistence score below this threshold are considered occasional spikes and are removed. It should be noted that, due to... It is a normalized weighted average, with a theoretical value range between 0 and 1. A threshold of 0.6 to 0.8 is recommended. In practical applications, this can be understood as judging the result of a weighted sum rather than a weighted average. Those skilled in the art can flexibly set the corresponding threshold range according to the actual scoring calculation method used.

[0112] The system performs the aforementioned continuous determination for each direction. Only a mutation direction whose cumulative score exceeds the confirmation threshold within a continuous time window is determined as a confirmed mutation direction and recorded in the confirmed mutation direction set. Each element in this set contains the UAV identifier, sampling time, and confirmed direction number. This information will serve as seed points for subsequently constructing interference concentration regions.

[0113] S2.3: Construct a region of concentrated interference by continuously expanding in adjacent directions, with the confirmed mutation direction as the center.

[0114] Based on the obtained set of confirmed mutation directions, each confirmed mutation direction is used as the center direction to expand to the left and right. The expansion is based on the similarity of the interference representation of adjacent directions, and also requires that the adjacent directions themselves have high mutation score persistence to ensure that the expanded region is continuous and consistent.

[0115] Directional continuity is determined using a cost function to measure the differences between adjacent directions and the center direction in terms of both directional representation units and abrupt change scores. Directional representation unit Obtained from step S1.4, it integrates the directional disturbance intensity and the directional disturbance gradient; mutation score persistence Obtained from step S2.2, this reflects the stability of the direction being identified as a mutation direction. The continuity cost is calculated according to the following formula:

[0116]

[0117] in, The continuity cost of direction k relative to the center direction k0 is dimensionless; Let k be the directional characterization unit, with a value ranging from 0 to 1; The persistence confirmation score for direction k ranges from 0 to 1; and These are the directional characterization unit and the continuous confirmation score corresponding to the central direction k0, respectively; and This is an empirical scaling parameter used to normalize the fluctuation range of the two terms. Its value ranges from 0.1 to 0.5. It can be determined based on the standard deviation of directional characterization units and mutation scores in historical data, or preset based on calibration values ​​for typical scenarios. and This is a lower limit protection term, with a value range of 0.01 to 0.05, to avoid cost divergence caused by an excessively small denominator.

[0118] The system sets a similarity threshold. The value ranges from 0.1 to 0.3. For the adjacent direction to the left of the center direction, if its continuity cost relative to the center direction is less than or equal to... Furthermore, the continuous scoring of this direction itself. Not lower than the lower threshold (Values ​​range from 0.6 to 0.8), if the value is within the current set region, then that direction is included and expansion continues to the left; otherwise, expansion to the left stops. Similarly, expansion to the right continues. After expansion, the left boundary direction number of the set region is obtained. and right boundary direction numbering .

[0119] Based on the left and right boundaries, the expansion angle of this concentrated region can be calculated, which is the angular opening of the cone on the horizontal plane. The expansion angle is calculated according to the following formula:

[0120]

[0121] in, Let be the expansion angle of the r-th interference concentration area identified by the i-th UAV at time n, in radians; and These are the left and right boundary directions of the concentrated area, respectively; The directional angle interval is given in step S1.2. This formula converts the discrete range of directional numbers into a continuous range of angles, providing direct input for the subsequent description of the cone's geometric range.

[0122] The system iterates through all confirmed mutation directions, performs the aforementioned expansion operation on each mutation direction, and merges overlapping or adjacent concentrated regions. Finally, it obtains a set of interference concentrated regions from the current UAV's perspective at the current moment. Each region includes attributes such as the center direction, left and right boundaries, expansion angle, and corresponding directional representation units and persistence scores. Through the above expansion operation, each interference concentrated region is composed of spatially continuous directional sectors with a clearly defined continuous angular range.

[0123] S2.4: Map the concentrated interference region to a three-dimensional interference cone, estimate its geometric parameters and risk level, and output the set of interference structures.

[0124] Spatial mapping is performed on the obtained interference concentration regions. For each concentration region, its angular section defines the directional range of the cone on the horizontal plane, i.e., the extended angle. The axial direction of the cone can be taken as the spatial direction corresponding to the center of the region, or as the direction of the angle bisector of the left and right boundaries of the region. The vertex position of the cone can be taken as the current position coordinates of the UAV, but the perspective of a single UAV may be limited. To obtain a more robust cone description, the system can fuse the observation results of multiple UAVs at the same or close moments. For example, for the same spatial region, if multiple UAVs observe similar concentrated areas from different positions, the optimal estimated position of the cone vertex can be determined by geometric intersection or weighted averaging. For simplicity, this embodiment uses observation from a single UAV as an example; multi-UAV fusion can be achieved through subsequent data association algorithms.

[0125] The effective radius of the cone reflects the distance of the interference's influence along the axial direction. The system determines the radius boundary along the cone's axial direction based on the characteristic that interference intensity decays with distance. Specifically, the system selects a series of distance step points along the axial direction. For each distance point, it calculates the spatial coordinates of that point based on the UAV's current position and the axial direction, and estimates the expected interference intensity at that distance. Since the UAV cannot directly measure interference at distant points, the system can use an empirical model of interference intensity decaying with distance, or interpolate based on measured interference intensity at different distances along the same axial direction from historical flight data. When the interference intensity along the axial direction is below an available threshold... When the value is between 0.4 and 0.6, it is determined that the cone boundary has been reached. To avoid infinite extrapolation, the effective radius is limited to the range of 20 meters to 200 meters to match the typical operating distance of a substation.

[0126] For each identified cone, the system also needs to assess its risk level to distinguish between high-risk and low-risk areas in task allocation and path planning. The risk score comprehensively considers the cone's angular spread range, peak fusion interference intensity within the region, and effective radius. (Peak fusion interference intensity within the region...) The risk score can be obtained by taking the maximum value from the fused directional interference intensity at each direction and distance point within the interference concentration area, with a value ranging from 0 to 1. The risk score is calculated according to the following formula:

[0127]

[0128] in, The cone risk score is dimensionless, and its theoretical value range is greater than or equal to 0, and it usually does not exceed 4 in the actual parameter range. To expand the angle, the unit is radians; This is the angle smoothing constant, ranging from 0.1 to 0.3 radians, used to adjust the saturation rate of the angle item in the scoring; The value represents the peak fusion interference intensity within the region, ranging from 0 to 1. This is the interference smoothing constant, ranging from 0.2 to 0.4, used to adjust the saturation rate of the interference intensity term in the scoring; The effective radius is expressed in meters and ranges from 20 to 200. The radius scale constant, ranging from 20 to 100 meters, is used to adjust the contribution of the radius to the score. This formula, through the product of three factors, reflects the higher risk of a cone with a wider angle coverage, stronger peak interference, and greater influence distance. The logarithmic term ensures that the influence of the radius diminishes marginally, preventing excessive amplification of the score over long distances.

[0129] For the same cone observed by multiple drones, its global risk score This can be obtained by taking the maximum value or weighted average of the observation scores of each drone, to ensure that the scheduling system uses a uniform risk quantification index.

[0130] Through the above processing, the system outputs a set of interference structures, which includes attributes such as the vertex position, axial direction, expansion angle, effective radius, and risk score of each interference cone. This set constitutes the interference cone identification model constructed in this step. In subsequent steps, the scheduling system will directly call upon each interference cone in this model for task allocation and path planning.

[0131] In the technical solution of this disclosure embodiment, based on the directional interference characterization obtained in S1, through adaptive threshold determination, temporal dimension persistence confirmation, and adjacent direction continuity expansion, interference concentration regions with spatial direction concentration and boundary abrupt change characteristics are identified. These regions are then mapped as interference cones in three-dimensional space, and their vertex positions, axial directions, effective radii, and risk scores are estimated. This step transforms the abstract interference characterization into a geometric risk structure, quantifying the spatial range and danger level of the interference region. This provides clear avoidable targets and risk basis for subsequent task allocation and path planning, significantly improving the scheduling system's ability to perceive complex electromagnetic environments.

[0132] S3: Based on the interference cone, construct an anti-interference task allocation model, quantify the risk of crossing the interference cone into a task allocation cost, and generate a task sequence for the drone swarm after merging and organizing the drones.

[0133] This step is used to build an anti-interference task allocation model. Based on each interference cone in the interference cone identification model output by S2, the risk of the interference cone is explicitly incorporated into the cost calculation of the UAV-task pair. By solving the minimum total cost assignment scheme, the swarm can actively avoid high-risk cones in the task allocation stage, providing a task sequence that balances efficiency and safety for subsequent path planning. Figure 2 A flowchart illustrating the anti-interference task allocation and task sequence generation process provided in this embodiment is shown, such as... Figure 2As shown, this is achieved through the following sub-steps.

[0134] S3.1: Calculate the straight-line distance of each UAV-task pair, the total time it takes for the UAV to fly to the task point and perform the operation, the estimated energy consumption, and the feasibility markers to form a set of basic cost elements.

[0135] For each drone and each task to be assigned, calculate its straight-line distance, estimated arrival time, estimated energy consumption, and task feasibility marker to form a basic cost element table, providing basic data for subsequent risk correction and comprehensive cost calculation.

[0136] Obtain the current status information of all drones in the swarm, including the position coordinates, current speed, remaining battery percentage, and payload status of each drone. The position coordinates are provided by the positioning system and are typically in a three-dimensional Cartesian coordinate system, denoted as [reference needed]. , where i is the drone index. Current speed Used to estimate flight time, the value typically ranges from 3 to 15 meters per second, depending on the drone model and mission requirements. Remaining battery percentage. A small value between 0 and 1, reflecting the available energy margin of the drone. Payload status is used to determine whether the drone has the ability to perform specific types of tasks, such as whether it is equipped with a visible light camera or an infrared thermal imager.

[0137] At the same time, the system obtains a set of task points to be assigned, and each task point j contains spatial coordinates. Expected operation time And the task deadline. The estimated operation time refers to the time required for the drone to complete inspection or other tasks after arriving at the task location; its value typically ranges from 10 seconds to 600 seconds, depending on the task type. The task deadline is the time constraint that the task must begin or be completed before a certain point in time.

[0138] For each UAV i and each mission j, the system first calculates the straight-line distance, which will be used for subsequent flight time and energy consumption estimation. The straight-line distance is calculated according to the following formula:

[0139]

[0140] in, Let be the straight-line distance from the i-th UAV to the j-th mission point, in meters; The coordinates of the drone's current position; Here are the coordinates of the task point. This formula converts the three-dimensional spatial position difference into a geometric distance that can be directly used for time and energy consumption estimation.

[0141] Based on the straight-line distance and the drone's cruising speed, the system estimates the time required for the drone to fly from its current location to the mission point, and adds this to the mission duration to obtain the total time. The estimation of arrival time and total time satisfies the following formula:

[0142]

[0143] in, The total time taken for the i-th drone to perform the j-th task, in seconds; The drone's cruising speed is measured in meters per second and ranges from 3 to 15. This is the lower speed limit protection item, with the unit being meters per second and a value range of 0.5 to 2. It is used to prevent the denominator from diverging when the speed is zero. The duration of the task is expressed in seconds. This formula unifies flight time and task time under the same time dimension, providing a basis for subsequent time cost calculations.

[0144] Energy consumption estimation is another crucial foundational element of task allocation, requiring a comprehensive consideration of energy consumption during flight and hovering during operations. Flight energy consumption is directly proportional to flight distance, with energy consumption per unit distance... The value ranges from 0.01 to 0.05 watt-hours per meter, with the specific value depending on the aerodynamic efficiency and propulsion system characteristics of the UAV. Hovering energy consumption is determined by hovering power and hovering time, with hovering power... The value ranges from 100 to 600 watts, and the hovering time is the duration of the task. Energy consumption estimation satisfies the following formula:

[0145]

[0146] in, The estimated energy consumption for the i-th drone to perform the j-th task, in watt-hours; Energy consumption per unit distance, expressed in watt-hours per meter; Hovering power, measured in watts; The duration of the task is in seconds; 3600 is the conversion factor from seconds to hours to ensure consistency of units. This formula adds up the mobile energy consumption and the stationary energy consumption to obtain a total energy consumption estimate that can be compared with the drone's battery capacity.

[0147] The system also needs to generate mission feasibility markers to indicate whether the UAV-mission pair meets basic feasibility constraints. The determination of feasibility markers is based on two aspects: first, the total time spent... First, determine if the task deadline has been exceeded; if so, mark it as infeasible. Second, estimate energy consumption. Does it exceed the drone's remaining available battery power? Greater than The time marker is marked as infeasible, where The rated capacity of the battery is expressed in watt-hours, ranging from 100 to 1000. For infeasible combinations, the system marks them as unassignable and assigns them an extremely high cost in subsequent comprehensive cost calculations to avoid selection.

[0148] Based on the above calculations, the system generates a basic element record for each drone and each mission, including the straight-line distance. Estimated total time Expected energy consumption Including feasibility markers, these data constitute the basic element table, which will be called in subsequent sub-steps.

[0149] S3.2: Assess the risk of each UAV-mission pair traversing the interference cone along its path and calculate the quantified risk exposure.

[0150] Based on the obtained basic element table and the interference structure set output in step S2, a risk analysis is performed on each UAV i and each mission j. The interference structure set contains K interference cones, each cone having a vertex position, axial direction, and expansion angle. Effective radius and risk score Risk Score Calculated by step S2.4, it is dimensionless and usually takes a value greater than or equal to 0 and not exceeding 3.

[0151] The actual flight path of the drone from its current location to the mission point can be any curve, but during the mission allocation phase, to simplify calculations, the system approximates the flight path as a straight line. The core of risk analysis is to determine whether this straight path enters the influence area of ​​any interference cone. If it does, the crossing length is calculated and the cone risk is weighted.

[0152] For the r-th interference cone, first determine whether the straight line segment from UAV i to mission j might intersect it. This determination can be made using a geometric method: calculate the shortest distance from the straight line segment to the cone's axis, and perform a spatial inclusion test by combining the cone's expansion angle and effective radius. For simplicity, this embodiment uses an approximation method, which involves sampling several points at equal intervals along the straight path, determining whether each sampling point falls inside the cone, and accumulating the risk contribution corresponding to the point of entry.

[0153] For a straight path, the system divides it into Q equal segments. The recommended number of sampling points Q is 3 to 15. The locations of these sampling points are denoted as follows: , where m is the sampling point index. For each sampling point, calculate its normalized decision value relative to the r-th cone. This decision factor integrates constraints from both distance and angle boundaries. When When the value is less than 1, point p is considered to be located inside the cone or within its safe expansion boundary. The calculation of the point determination quantity inside the cone will be detailed in step S4, and its definition can be directly quoted here, i.e., it satisfies the following formula:

[0154]

[0155] in, Let p be the radial distance from point p to the vertex of the cone within the positive half-space along the cone's axis; The effective radius of the cone; For a safe expansion radius, the value should be between 3 and 15 meters. Let p be the angle between point p and the axis of the cone; For the cone's expansion angle; As a safety margin for the angle, the value ranges from 0.05 to 0.2 radians. This judgment quantity unifies the distance boundary and the angle boundary into a dimensionless quantity. The closer point p is to the center of the cone, the more... The smaller it is, the closer it is to 0; when point p is located on the boundary of the cone, Approaching 1; when point p is outside the cone, .

[0156] Iterate through all sampling points and all cones. For each sampling point, if... The risk contribution is then accumulated into the risk exposure of that path. The risk contribution uses an exponential decay method, meaning points closer to the center of the cone contribute more. Simultaneously, the risk score of the cone... As a weighting factor, the impact of the high-risk cone is made greater. The cumulative calculation of risk exposure satisfies the following formula:

[0157]

[0158] in, The cumulative risk exposure of the i-th UAV performing the j-th mission is dimensionless; K is the total number of interference cones. The risk score for the r-th interfering cone; The distance the UAV travels through to the mission path within the r-th cone is measured in meters. This value can be estimated by multiplying the proportion of sampling points falling into the cone by the total path length. Let be the straight-line distance between the i-th drone and the j-th mission; This is the lower limit protection item, in meters, with a value range of 1 to 10; This is the distance attenuation constant, in meters, ranging from 30 to 200, and is the exponent term. This design adjusts the weighting of risk contributions for missions at different distances in the overall cost calculation: for long-distance missions, the basic costs such as flight time and energy consumption are already significant, so the weight of risk items is appropriately reduced to avoid them excessively dominating the total cost; for short-distance missions, the weight of risk items is relatively increased to reflect the high sensitivity of mission failure if the mission enters the cone at short distances. This design allows the risk exposure to adaptively adjust its impact on the total cost based on the mission distance.

[0159] This formula integrates three factors—the length of traversal, the danger level of the cone, and the mission distance—into a dimensionless risk exposure. The traversal length percentage reflects the path's exposure level within the cone, the risk score reflects the cone's danger level, and the exponential decay means that for the same traversal length, close-range missions contribute a relatively larger risk, which aligns with the physical intuition that close-range missions are more likely to cause communication disruptions once they enter the cone.

[0160] Based on the above calculations, a risk exposure value is generated for each drone-mission pair. The system records crossing markers and cumulative risk scores to form a risk element table. This table, along with the basic element table, will be used to construct the subsequent comprehensive cost.

[0161] S3.3: Integrate the basic cost elements with the risk exposure to construct a comprehensive task cost.

[0162] Based on the basic element table obtained in S3.1 and the risk element table obtained in S3.2, the comprehensive task cost is calculated for each UAV i and each task j. The comprehensive cost requires considering the straight-line distance. Total time spent Expected energy consumption and risk exposure These four types of physical quantities are integrated into a single comparable dimensionless value. To achieve this, the system introduces a normalization reference parameter to compress each physical quantity to a similar order of magnitude.

[0163] Distance cost normalization adopts a distance normalization benchmark. The value ranges from 100 to 2000 meters and can be set according to the typical mission flight distance. Time cost normalization uses a time normalization benchmark. The value ranges from 60 to 1200 seconds, corresponding to the total time range of common tasks. Energy cost normalization needs to consider the drone's current remaining battery power, dividing the estimated energy consumption by the remaining available energy. The consumption percentage is obtained, and a denominator protection term is added to avoid division by zero. Risk cost normalization uses a saturation function to compress risk exposure to between 0 and 1, and a squared term is used to make the penalty for high-risk combinations grow faster. The calculation of the overall task cost satisfies the following formula:

[0164]

[0165] in, The overall task cost is dimensionless; , , , There are four weights, ranging from 0.1 to 0.6, with a total weight of 1. These weights can be adjusted according to the specific scenario and the relative importance of each factor. For example, in environments with severe interference, the risk weights can be increased appropriately. ; Let be the straight-line distance between the i-th drone and the j-th mission; As a distance normalization benchmark; Total time elapsed; As a time normalization benchmark; For projected energy consumption; This refers to the battery's rated capacity, measured in watt-hours. This represents the decimal value corresponding to the battery percentage of the i-th drone, ranging from 0.1 to 1; This is the energy consumption denominator protection item, with the unit being watt-hours and a value range of 1 to 10. It is used to prevent the cost from diverging when the energy consumption is extremely low due to an excessively small denominator. For risk exposure. The fractional square term makes when When the cost is small, the growth is slow. A larger time cost quickly approaches the weight value, thus imposing a stronger penalty on high-risk portfolios.

[0166] For drone-mission pairs marked as infeasible in step S3.1, such as combinations with timeouts or insufficient power, the system directly incorporates their overall cost. Set it to a maximum value that is much larger than the normal range, for example... This is to ensure that it will not be selected in subsequent assignment solving.

[0167] Through the above calculations, the system obtains a complete comprehensive cost matrix, where rows correspond to UAVs, columns correspond to tasks, and each element is... This matrix will serve as the input for the next step of the assigned solution.

[0168] S3.4: Solve for the minimum total cost assignment scheme under the task allocation constraints and generate the task allocation set.

[0169] Based on the comprehensive cost matrix, the algorithm solves for the allocation scheme that minimizes the total cost under the task allocation constraints, and outputs which drone each task is assigned to, along with the corresponding allocation confidence and a list of unassigned tasks.

[0170] First, let's define the basic constraints of task allocation: each task must be assigned to one and only one drone; each drone can undertake a maximum of a limited number of tasks within the same scheduling time window. These constraints are mathematically expressed by the following formula:

[0171]

[0172]

[0173] Where N represents the number of drones; M represents the number of missions; The assignment indicator has a value of 0 or 1, indicating whether task j is assigned to drone i; Let be the upper limit constant for the tasks of the i-th drone, with a value ranging from 1 to 3. This value can be determined based on a combination of factors including task complexity, drone payload capacity, and current battery level. For example, a value of 3 can be used when the battery level is above 80%, 2 when the battery level is between 50% and 80%, and 1 when the battery level is below 50%. The first constraint ensures that each task is uniquely assigned, and the second constraint ensures that the number of tasks undertaken by each drone does not exceed its capacity limit.

[0174] The total cost objective function is the sum of the combined costs of all selected UAV-mission pairs, and its mathematical expression satisfies the following formula:

[0175]

[0176] Where J is the total cost target value, which is dimensionless; N is the number of drones; and M is the number of missions. To assign the instruction quantity; To account for the overall task cost, the objective function transforms the allocation problem into an optimization problem of minimizing J under constraints.

[0177] This assignment problem can be solved using standard assignment algorithms, such as the Hungarian algorithm or its extensions, to handle cases where the maximum number of drone tasks is greater than one. The specific solution process is as follows: First, construct a cost matrix of size N rows and M columns, filling in maximum costs for unassignable items. Then, based on the relationship between the number of tasks and the number of drones, adopt an appropriate solution strategy. When the number of tasks M is less than or equal to the number of drones N multiplied by the maximum task limit, a capacity-constrained assignment solution can be directly performed. When the number of tasks is large, tasks can be sorted by their time urgency and solved in batches, with each batch containing 1-2 times the number of drones to ensure real-time performance.

[0178] After the solution is completed, the system obtains the allocation instruction. The value of is such that for each task j, there exists a unique i such that Then task j is assigned to drone i; if all If both values ​​are 0, the task is not assigned and is added to the unassigned task list. For each assigned drone-task pair, the system can calculate the cost based on the comprehensive cost. Calculate the confidence level of the sub-configuration, for example, take However, the confidence level is not a mandatory output and can be determined based on actual needs.

[0179] Through the above solution, the system obtains a task allocation set, which contains the correspondence between each task and a UAV. The system then organizes this set by UAV, generating a task sequence for each UAV. The task sequence is sorted according to task priority or expected execution order. Each task entry includes task point coordinates, expected operation duration, overall cost, risk exposure, and expected time and energy consumption. For unassigned tasks, the system adds them to the unassigned task list and can retry assigning them in subsequent scheduling cycles.

[0180] The generated task sequence is output to the path planning model in step S4. The path planning model will, for each UAV's task list, sequentially plan the flight path from the current position to the first task point and then to subsequent task points. Simultaneously, the risk exposure for each task is calculated. This can be used as a reference during path planning. If the risk exposure is too high, the path planning stage will more strictly detour around the relevant cones. If a feasible path cannot be found after path planning, a reallocation mechanism can be triggered, and the task can be reallocated back to step S3.

[0181] In the technical solution of this disclosure, the interference cone risk identified by S2 is explicitly incorporated into the task allocation decision process. By calculating the straight-line distance from each UAV to the task point, the estimated time, energy consumption, and the risk exposure of crossing the interference cone, a comprehensive task cost matrix containing multiple types of costs is constructed, and the assignment scheme with the minimum total cost is solved under allocation constraints. This step enables the swarm to actively avoid task combinations that may cross high-risk cones when allocating tasks, reducing the probability of UAVs entering areas with strong directional radiation from the source, improving the reliability and safety of task execution, while taking into account both task efficiency and energy consumption constraints.

[0182] S4: Based on the interference cone and the task sequence, construct a path planning model based on interference cone avoidance, and generate a flyable path for the UAVs assigned tasks in the task sequence that can effectively avoid the interference cone and satisfy kinematic constraints.

[0183] This step is used to construct a path planning model based on interference cone avoidance. Based on the task sequence output by S3 and each interference cone in the interference cone identification model output by S2, the interference cone is mapped to a high-cost occupied area and an avoidance path that satisfies curvature constraints is generated, providing the UAV with a flyable interference avoidance track. Figure 3A flowchart illustrating the interference cone avoidance path planning provided in an embodiment of this disclosure is shown, as follows: Figure 3 As shown, this is achieved through the following sub-steps.

[0184] S4.1: Transform the interference cone into a planarable high-cost occupied region and establish a decision function for spatial points falling into the cone.

[0185] Each interference cone identified in step S2 is transformed into a spatial constraint condition usable for path planning. A decision function for inaccessible and high-cost regions is established, and a safety expansion radius is added to the cone boundary to provide a spatial decision basis for subsequent path segment cost calculation.

[0186] Obtain the attribute parameters of each interference cone from the interference structure set output in step S2, including the vertex position. axial direction unit vector , Expanding angle Effective radius and risk score For each interference cone, the system establishes a function to determine whether a spatial point p falls within the cone's influence range. This determination considers two constraints: first, whether the radial distance from point p to the cone's apex is within the effective radius; and second, whether the angle of deviation of point p relative to the cone's axis is within the expansion angle. Simultaneously, to provide a safety margin in path planning and prevent the UAV from grazing across the cone boundary due to positioning errors or control deviations, the system introduces a safety expansion radius. and angle safety margin The recommended safe expansion radius is 3-15 meters, and the recommended angular safety margin is 0.05-0.2 radians. For each interfering cone, the system establishes a determination function to determine whether a spatial point p falls within the cone's influence range. The determination function... The calculation satisfies the following formula:

[0187]

[0188] in, Let p be the normalized decision quantity relative to the r-th cone, which is dimensionless; Let $p$ be the radial distance from point $p$ to the vertex of the cone within the positive half-space of the cone's axial direction, expressed in meters. It is calculated as follows: First, calculate the vector $\mathbf{p}$. In the axial direction If the projection length is negative, it means that point p is located behind the vertex of the cone and is not within the influence range of the cone. Pick To ensure the subsequent judgment quantity If the projection is positive, then That is, the length of the projection; The effective radius of the cone; For safe expansion radius; Let be the angle between point p and the axis of the cone, in radians, and its calculation method is as follows: ,like Then the included angle is 0; For the cone's expansion angle; For angle safety margin.

[0189] This formula unifies the distance and angle boundaries into dimensionless decision quantities. The distance term uses square normalization, and the angle term uses fourth-power normalization, making the penalty for angle deviation more sensitive and thus making it easier to avoid abrupt changes in the main lobe boundary. When point p is considered to be within the influence region defined by the cone and its safe expansion boundary; when At that time, point p is considered to be outside the safe zone.

[0190] The system uses the risk score of the cone. The cone is divided into two categories: impassable regions and high-cost regions. A high-risk threshold is set. The recommended value range is 1 to 2.5. If Then the cone is mapped to an impassable region, and any path segment entering this region will incur a very high cost; if If the cone is mapped to a high-cost region, a path segment entering this region will be penalized according to the degree of entry, but it is still allowed to pass through at a higher cost if necessary.

[0191] Through the above processing, the system establishes a set of planning constraint domains for all interference cones, which includes the spatial decision function, risk score, and region type label for each cone. This set will serve as the basis for subsequent path segment cost calculations.

[0192] S4.2: Discretize the flight path into path segments, calculate the length cost, interference penalty cost, and boundary proximity penalty to form the comprehensive cost of the path segment.

[0193] The flight path is discretized into several path segments. The length cost and interference penalty cost of each path segment are calculated, and a boundary proximity penalty is introduced to form a comprehensive cost of the path segment, which provides a basis for edge weights for subsequent global path search.

[0194] For each UAV, path planning is performed within its flight space based on its current position and target mission point. To simplify computation, the continuous space is typically discretized into a node grid or sampled nodes, with the connections between nodes forming candidate path segments. A node spacing of 2-8 meters is recommended to balance path planning precision with computational efficiency. The path search window radius can be set based on the distance to the mission point, with a recommended range of 100-500 meters. When the mission point is far away, segmented planning and rolling updates can be used.

[0195] For each candidate path segment k, the system first calculates its geometric length. The unit is meters, and the value ranges from 1.5 to 10 meters, depending on the node spacing and the connection method of the path segment. The path segment length will serve as the basis for the length cost.

[0196] To assess the interference risk of a path segment, the system uniformly selects a number of sampling points along the path segment. The recommended number of sampling points, Q, is 3 to 15, and the locations of these sampling points are denoted as […]. Where m is the sampling point index. For each sampling point, the system uses the decision function established in step S4.1 to calculate the normalized decision value of that sampling point relative to each cone. And based on cone risk score The interference contribution at this point is accumulated. The path segment interference penalty is calculated according to the following formula:

[0197]

[0198] in, The interference penalty for the k-th path segment is dimensionless; Q is the number of sampling points; K is the total number of cones. For the risk score of the r-th cone; This is the normalized decision value for the sampling points; This is a sample count protection term, ranging from 1 to 3, used to prevent the penalty value from becoming unstable due to an excessively small denominator when there are few sampling points. The core of this formula is: the deeper the sampling point is inside the cone, the better. The smaller the exponent, the smaller the exponent. The larger the value, the stronger the penalty; when the sampling point is located outside the cone... A value greater than or equal to 1, and an exponential term less than or equal to 0.367, contribute relatively little. The influence of multiple cones is summed and the average of the sampled points is taken to obtain the interference penalty for that path segment.

[0199] Boundary proximity penalty This is used to encourage paths to stay as far away from the cone boundary as possible, i.e., remain outside the safety buffer zone, avoiding accidental entry into high-risk areas due to control errors. The formula for calculating the boundary proximity penalty is defined as follows: That is, when the minimum decision value for a sampling point on a path segment is less than 1, the closer it is to 0, the greater the penalty; when the minimum decision value is greater than or equal to 1, the penalty is 0. The calculation of the comprehensive cost of a path segment satisfies the following formula:

[0200]

[0201] in, The total cost of the k-th path segment is dimensionless. , , The weights range from 0.1 to 0.7 and sum to 1. They can be adjusted according to the actual scenario to determine the importance of path length, interference avoidance, and boundary proximity. This represents the length of the path segment, in meters. The standard for length normalization is in meters, and the value ranges from 5 to 50. This is a length protection item, with the unit being meters, and a value range of 1-5; Punishment for interference; For a saturation function, Compressed to between 0 and 1, when When the value is small, this term is close to 0; when... When the value is large, this term approaches 1; Penalize the proximity to the boundary; As a normalization benchmark, its value ranges from 0.2 to 2; For protection purposes, the value ranges from 0.01 to 0.1. This formula unifies the path length cost, disturbance penalty cost, and boundary proximity penalty into a summable comprehensive cost, and uses a saturation function to prevent any single value from increasing infinitely, which could lead to planning instability.

[0202] Through the above calculations, the system generates a comprehensive cost for all candidate path segments, forming a candidate path segment cost table, which will serve as the basis for edge weights in subsequent global path search.

[0203] S4.3: The graph search method is used to find the minimum cost path from the starting point to the target point based on the comprehensive cost of the path segments, and an initial avoidance path is generated.

[0204] The flyable space is discretized into a node-based grid, with each node representing a possible location. The lines connecting adjacent nodes form path segments, and their combined cost is... Calculated from S4.2. For each UAV, the starting point is its current position, and the target point is the mission point. The goal of path planning is to find a sequence of nodes from the starting point to the target point that minimizes the sum of the total costs of all path segments. The total path cost is calculated according to the following formula:

[0205]

[0206] Where J is the total cost of a candidate path, which is dimensionless; This represents the number of path segments contained in the path. Let be the total cost of the k-th path segment. This formula breaks down the entire path into several segments and accumulates the costs; the path with the minimum total cost is the optimal avoidance path.

[0207] The system employs classic graph search algorithms, such as A* or Dijkstra's algorithm, for pathfinding. During the search, for each node, the algorithm maintains the cumulative cost from the starting point to that node and uses a heuristic function to estimate the remaining cost to accelerate the search. The heuristic function could be the time cost calculated by dividing the straight-line distance from the current node to the target point by the maximum expected speed; however, considering that the path segment cost already incorporates multiple factors, the heuristic function can be simplified to the straight-line distance multiplied by the average cost per unit distance. During the search, if a path segment causes a node to fall into an impassable region (i.e., there are sampling points on that path segment that cause the corresponding cone to...), the search is initiated accordingly. If the value is less than 1 and the cone is a non-passable region, then skip the edge and do not expand it.

[0208] To ensure real-time performance, the search window radius can be limited to a certain range, such as 100-500 meters. When the task points are far apart, segmented planning can be used, first planning up to the intermediate point, and then gradually updating.

[0209] After the search is completed, an initial avoidance path consisting of discrete path points is obtained. This path effectively avoids high-risk interference cones, but may contain broken lines or sharp turns, which do not meet the flight constraints of the UAV and require subsequent smoothing processing.

[0210] S4.4: Perform curvature constraints and smoothing on the initial path to generate an executable evasion path that satisfies the kinematic constraints of the UAV.

[0211] Geometric analysis is performed on the discrete point sequence of the initial path to calculate the heading angles and the rate of change of heading angles between adjacent path segments. For the k-th path segment, its heading angle... It can be calculated based on the coordinates of the start and end points of the segment. The heading angle difference between adjacent segments is... The curvature change test measure is used to assess the degree of path curvature, and its calculation satisfies the following formula:

[0212]

[0213] in, Let $\frac{k}{k}$ be the test value for the curvature change of the $k$ segment, in units of $\frac{k}{k}$. Let be the heading angle of the kth segment, in radians; The length of the k-th segment is given in meters. This is a length protection term, measured in meters, with a value ranging from 0.5 to 2, used to prevent the denominator from diverging when the segment length is too small. This test quantity approximately reflects the change in heading per unit length, i.e., curvature. UAVs typically have a minimum turning radius. The constraint, corresponding to the maximum allowable curvature is Therefore, it is necessary to ensure The minimum turning radius is recommended to be 5-30 meters, depending on the drone model.

[0214] The system starts with an initial path and adjusts the path points using optimization methods to ensure that the smoothed path satisfies curvature constraints while maintaining its ability to avoid interfering cones. The objective function of smoothing optimization is the weighted sum of the comprehensive cost of the path segment and the curvature penalty, and its calculation satisfies the following formula:

[0215]

[0216] in, To smoothly optimize the objective, it is dimensionless; For the length normalization constant, it is recommended to take 1 meter, so that Dimensionless; The total cost of the path segment; This is the curvature penalty coefficient, ranging from 1 to 20, used to balance risk avoidance and smooth flight. This is the curvature change check metric. The squared term severely penalizes excessive curvature, prompting the optimization algorithm to favor paths with smaller curvatures.

[0217] During optimization, the positions of path points are fine-tuned within permissible ranges, while avoiding the adjusted path segments from entering impassable regions. If the smoothed path locally enters the boundary of a high-cost region, the system can backtrack and increase the boundary proximity penalty weight. Then, re-execute sub-step S4.3 to generate a new path and perform smoothing again until the constraints are met or the maximum number of iterations is reached. It is recommended to take 3 to 8 iterations.

[0218] After smoothing, the system outputs a final set of paths, with each UAV corresponding to an executable path consisting of a smoothed sequence of path points and a corresponding heading sequence. This path not only effectively avoids interference cones but also satisfies the UAV's flight control constraints, enabling stable tracking by the flight control system.

[0219] In the technical solution of this embodiment, the interference cone is mapped as a planarable high-cost occupied region. A determination function for spatial points falling into the cone is established, the flight path is discretized, and the interference penalty cost and boundary proximity penalty of the path segments are calculated. A graph search algorithm is then used to generate a minimum-cost avoidance path from the starting point to the target point. Furthermore, curvature continuity constraints and smoothing processes ensure that the generated path meets the kinematic execution requirements of the UAV. This step enables the UAV to effectively avoid areas of strong directional interference during flight, ensuring stable communication links while maintaining path smoothness, thus improving the feasibility of flight control and path tracking accuracy.

[0220] S5: During the process of the UAV performing the mission according to the flightable path, a dynamic rescheduling mechanism based on interference mutation is constructed to monitor the changes in the interference environment in real time. When interference mutation is detected, the interference cone is updated, and the affected missions and UAVs are locally reassigned and replanned based on the updated interference cone and the mission sequence.

[0221] This step is used to construct a dynamic rescheduling mechanism based on interference mutations. During UAV mission execution, interference changes are monitored in real time. When a sudden interference mutation is detected, the interference structure is incrementally updated and the affected UAVs are identified. By performing local reallocation and path replanning for affected tasks, the swarm of UAVs gains the ability to adapt to dynamic electromagnetic environments. This is specifically achieved through the following sub-steps.

[0222] S5.1: Monitor the rate of change of interference in real time and make continuous judgments to trigger interference mutation events.

[0223] During the flight of the UAV along the planned path, directional interference characterization units are collected at fixed monitoring cycles to calculate the comprehensive interference intensity index and its rate of change. The system continuously identifies whether a sudden interference change occurs, providing a trigger signal for subsequent rescheduling.

[0224] The system continuously executes the calculation process of step S1 during the drone's flight, updating the directional interference characterization unit sequence of each drone at each sampling moment in real time. The monitoring period should be consistent with the sampling period in step S1, and it is recommended to use 0.05-0.2 seconds to ensure that rapid changes in interference can be captured.

[0225] For the i-th UAV at the n-th monitoring time, the directional characterization units in all directions at that time are first aggregated into a comprehensive interference intensity index. The purpose of aggregation is to compress multi-directional information into a single index, while highlighting the impact of abrupt changes in direction through appropriate nonlinear enhancement. The aggregation order q is used to control the amplification degree of large-value directions; the larger q is, the more prominent the contribution of large-value directions. The comprehensive interference intensity index is calculated according to the following formula:

[0226]

[0227] in, Let be the comprehensive interference intensity index of the i-th UAV at the n-th sampling time, which is dimensionless; is the directional characterization unit, with a value range of 0-1; M is the number of directions, with a value range of 8-36; q is the aggregation order constant, with a value range of 2-6, used to enhance the contribution of large values ​​and make the mutation direction more likely to raise the comprehensive index; As a protective term, its value ranges from 0.1 to 1 to avoid the denominator being too small. This formula uses aggregation in the form of the Lq norm to ensure that larger values ​​are within the range of 0.1-1. The value contributes more significantly to the overall index, thus avoiding the dilution of the characterization of mutation direction by the average effect.

[0228] After obtaining the comprehensive interference intensity index, the system calculates the interference change rate between adjacent time points. The change rate reflects the drastic change in interference intensity over time, and the squared term is used to emphasize abrupt jumps and suppress small-amplitude noise. The change rate is calculated according to the following formula:

[0229]

[0230] in, Let be the rate of change of interference for the i-th UAV at time n, expressed in seconds. and This represents the combined interference intensity at adjacent time points; The monitoring period is measured in seconds, with a range of 0.05-0.2. This is a time protection term, measured in seconds, with a value range of 0.005-0.02. The use of the square term ensures that the rate value responds to both positive and negative jumps, and large jumps are significantly amplified.

[0231] Since single-point rate jumps may be caused by transient noise, the system introduces a time window persistence determination, requiring the rate of change to remain at a high level for several consecutive monitoring periods to be considered a valid change. The persistence trigger score is calculated according to the following formula:

[0232]

[0233] in, The continuous trigger score for the i-th drone at time n is calculated based on the rate of change, in units of per second. The duration is the length of the continuous window, with a value ranging from 2 to 6. The decay factor ranges from 0.6 to 0.9. This formula uses an exponentially weighted average of recent change rates to remove the influence of single-point spikes, requiring that the abrupt change has temporal persistence.

[0234] System set trigger threshold The value ranges from 0.8 to 3.5 per second. At that moment, it is determined that the UAV has triggered an interference mutation, generating an interference mutation event and recording the trigger time, trigger confidence level, and corresponding directional range. The directional range can be extracted from the directional representation unit at that moment, for example, taking... For continuous directional intervals exceeding a certain threshold, or directly using the concentrated region identified in step S2, the system aggregates all UAV information that triggers a mutation, forming a mutation trigger marker set. Each marker includes the triggering UAV identifier, trigger time, and corresponding directional range. This interference mutation event will serve as the trigger signal for initiating incremental updates and rescheduling. The trigger marker set will serve as the basis for subsequent incremental updates and rescheduling.

[0235] This step allows the system to identify potential threats such as mobile interference sources or sudden intrusions from the main radiation direction in advance, preventing drones from being passively rescheduled only after entering a chain-break zone and giving time for proactive risk avoidance.

[0236] S5.2: Based on the directional data at the moment of triggering mutation, incrementally update the set of interference structures and generate or correct interference cones.

[0237] When a drone triggers a sudden change, the directional interference characterization unit at the trigger time recalculates the interference concentration area and generates a new interference cone or corrects the parameters of an existing cone. At the same time, it sets an expiration period for the cone to achieve automatic decay, thereby maintaining the real-time accuracy of the interference structure set.

[0238] After receiving the mutation trigger marker set output by S5.1, the system extracts the directional interference characterization unit sequence corresponding to the trigger time for each UAV that triggered the mutation. And extract the mutation score corresponding to the triggering time from step S2.1. Based on this data, the system re-executes the concentrated region construction process in step S2, including using an adaptive threshold to determine the direction of mutation, confirming persistence, and continuously expanding in adjacent directions to obtain the concentrated region of interference at the trigger time.

[0239] For each newly identified cluster region, the system maps it to a potential interference cone. The cone's apex can be the spatial location of the currently triggering UAV, or it can be optimized through fusion of multi-UAV observations. The cone's axial direction is taken as the center direction of the cluster region. The cone's expansion angle is the same as the expansion angle of the cluster region. The effective radius of the cone can be estimated based on the characteristic that the interference intensity decreases with distance, for example, by using the interference intensity in that direction at the trigger moment. With available thresholds By comparison, the maximum impact distance can be calculated. Cone risk score. The calculation can be recalculated according to the formula in step S2.4 based on the extension angle and peak interference intensity.

[0240] For newly generated cones, the system assigns an effective option weight to prevent them from occupying planning space for an extended period after the interference disappears in a short time. The decay of the effective option weight over time satisfies the following formula:

[0241]

[0242] in, Let be the effective weight of the r-th cone at time n, which is dimensionless; The sampling time number most recently confirmed for this cone; For monitoring cycle; The validity period time constant is in seconds and ranges from 1 to 10. This formula transforms the existence of the cone into a weight that decays naturally over time; if it is continuously observed in subsequent moments, it is continuously refreshed. This keeps the weight high; if it is not observed for a long time, the weight gradually decays to 0, and the cone is automatically removed from the set of interference structures.

[0243] For an existing cone, if the observation region that triggered the mutation matches it, then update the cone's parameters, including vertex position, spread angle, effective radius, and risk score, while also updating its most recent confirmation time. The validity period can be extended. Matching can be determined by the similarity of spatial location and directional range.

[0244] Through this step, the system achieves incremental updates to the set of interference structures, enabling the scheduling model to reflect changes in the electromagnetic environment in real time. The newly added or modified cones will be adopted in subsequent rescheduling processes.

[0245] S5.3: Based on the updated interference cone structure, identify affected drones and form a rescheduling trigger set.

[0246] Based on the updated set of interference structures, a forward-looking risk assessment is performed on the drones currently performing tasks, identifying those drones that are about to enter new or upgraded high-risk cones in the remaining paths, and collecting the tasks affected by them to form a rescheduling trigger set.

[0247] The system acquires the updated set of interference structures, which includes newly added or modified interference cones and their effective weights. For each UAV performing a mission, the system acquires its current sequence of remaining path segments. The remaining path segments consist of the unflyed portions of the final path output in step S4, and include a series of discrete points or path segments moving forward from the current position along the path.

[0248] Set look-ahead distance It is recommended to use a range of 20-100 meters to define the spatial scope of the risk assessment. For each UAV, the system starts from the current position and extends forward along the remaining path to all path segments within the look-ahead distance, performing cone intersection verification on the sampling points on these path segments. The verification method uses the decision quantity defined in S4.1. For each sampling point, calculate its decision value relative to each updated cone. If there exists a sampling point such that... And the effective weight of the corresponding cone Above a threshold (e.g., 0.5), and the cone risk score is high (e.g., If the result is positive, it is determined that the drone will enter a high-risk area and be affected by this mutation.

[0249] Simultaneously, the system checks the drone's assigned but not yet executed tasks. For each task, it assesses whether a straight path or planned path from the current location to the task point might cross a newly emerged high-risk cone. If the probability of crossing is high, the task is added to the list of affected tasks.

[0250] For each drone identified as affected, the system records its drone identifier, current location, list of affected tasks, and index of the path segments that need to be rolled back. This information constitutes a rescheduling trigger set, which will be used in the next sub-step to perform reallocation and replanning.

[0251] S5.4: Perform local reallocation and path replanning on the affected tasks, and merge the output to get the final set of scheduling paths.

[0252] Local redistribution is performed on the affected tasks within the trigger set. Only the task costs of the affected drones and their neighboring supported drones are recalculated and reassigned. Then, new avoidance paths are planned for the redistributed drones and merged with the paths of the unaffected drones to form the final scheduling path set.

[0253] First, the scope of the reassignment is determined. To reduce computational jitter, a local reassignment strategy is adopted, including only the affected UAV and its neighboring supported UAVs in the reassignment candidate set. Neighborhood range It is recommended to select drones within a 100-500 meter range as potential support candidates. For affected missions, if the originally assigned drone is unable to continue, it should be reassigned from these candidate drones.

[0254] Re-execute the anti-interference task assignment model from step S3, but only for UAVs and affected tasks within a local scope. The cost matrix only includes combinations of candidate UAVs and affected tasks, and the constraints still apply. After solving the assignment that minimizes the total cost, a new assignment result for the affected tasks is obtained.

[0255] For the reassigned UAV, the system needs to replan the avoidance path from the current location to the new task point. If the UAV has remaining path segments that have not been executed, but the new task point has changed or the environmental interference structure has been updated, the original remaining path is discarded, and the path planning model in step S4 is executed again to generate a new path from the current location to the new task point that avoids the updated interference cone. The new path is also smoothed for curvature to meet the executability requirements.

[0256] For unaffected drones, their original paths remain unchanged and they continue to operate.

[0257] Finally, the system merges the original paths of unaffected drones with the new paths of affected drones to form the final set of scheduling paths for the entire swarm of drones. Each drone in this set has a defined, flyable path, and the swarm as a whole is adaptive to dynamically changing interference environments.

[0258] In the technical solution of this disclosure, changes in the interference environment are continuously monitored during the actual flight of the UAV. By calculating the comprehensive interference intensity index and its rate of change and making continuous determinations, interference mutation events are detected in real time. Once a mutation is triggered, the interference structure set is incrementally updated based on the directional interference characterization at the trigger time. A forward-looking risk assessment is performed on the affected UAVs, and local reallocation and path replanning are executed for the affected tasks. This step enables the swarm to have a rapid adaptive capability to the dynamic electromagnetic environment, allowing for timely adjustment of task allocation and flight paths when new interference cones appear or existing interference intensifies. This avoids communication interruptions or task failures caused by interference mutations, significantly improving the robustness and environmental adaptability of swarm scheduling.

[0259] According to embodiments of this disclosure, an electronic device is also provided, which may include a processor, a communications interface, a memory, and a communication bus, wherein the processor, the communications interface, and the memory communicate with each other via the communication bus. The processor can invoke logical instructions stored in the memory to execute the methods provided in the above embodiments.

[0260] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0261] On the other hand, this disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the methods provided in the above embodiments.

[0262] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0263] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0264] It should be understood that the above embodiments are only used to illustrate the technical solutions of this disclosure, and not to limit them; although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure.

Claims

1. A method for scheduling unmanned aerial vehicle (UAV) swarms with anti-interference capabilities, characterized in that, The method includes: S1: Based on the time-series observation data of UAV swarms, construct a spatial directional interference characterization model that includes the directional interference intensity and directional interference gradient on each directional sector; S2: Based on the spatial directional interference characterization model, construct an interference cone identification model composed of interference cones generated by mapping the interference concentration region. The interference cone identification model includes the geometric parameters and risk level of the interference cones; S2 includes: Based on the directional interference gradient, potential interference mutation directions are identified through adaptive threshold determination. The potential interference mutation directions are continuously confirmed over a time window, and transient jitters are eliminated while confirmed mutation directions are retained; Centered on the confirmed mutation direction, an interference concentration region with a continuous angular range is constructed by continuously expanding in adjacent directions; The interference concentration area is mapped as an interference cone in three-dimensional space, and its geometric parameters and risk level are estimated. The generated interference cone and its geometric parameters and risk level constitute the interference cone identification model. S3: Based on the interference cone, construct an anti-interference task allocation model, quantify the risk of crossing the interference cone into a task allocation cost, and generate a task sequence for the drone swarm after merging and organizing the drones. S4: Based on the interference cone and the task sequence, construct a path planning model based on interference cone avoidance, and generate a flyable path for the UAVs assigned tasks in the task sequence that can effectively avoid the interference cone and satisfy kinematic constraints. S5: During the process of the UAV performing the mission according to the flightable path, a dynamic rescheduling mechanism based on interference mutation is constructed to monitor the changes in the interference environment in real time. When interference mutation is detected, the interference cone is updated, and the affected missions and UAVs are locally reassigned and replanned based on the updated interference cone and the mission sequence.

2. The anti-interference UAV swarm scheduling method according to claim 1, characterized in that, S1 includes: The time-series observation data of the UAV swarm are processed by time axis alignment and dimension normalization to generate a standard observation dataset that is consistent in time and space. Centered on the current location of the drone, the omnidirectional space is discretized into multiple directional sectors; Based on the standard observation dataset, the directional interference intensity in each directional sector is calculated, and a robust directional interference intensity sequence is formed by fusion through a time window. Calculate the directional interference gradient based on the directional interference intensity of adjacent directional sectors; The directional interference intensity and the directional interference gradient on the same directional sector are fused to generate a directional interference characterization unit; The spatial directional interference characterization model is composed of directional interference characterization units of the UAV swarm at all sampling times and in all directional sectors.

3. The anti-interference drone swarm scheduling method according to claim 1, characterized in that, S3 includes: Calculate the basic cost elements between each UAV and the task to be assigned. The basic cost elements include the straight-line distance between the UAV and the task pair, the total time it takes for the UAV to fly to the task point and perform the operation, the estimated energy consumption, and the feasibility marker. Assess the risk exposure of each UAV-mission pair's flight path crossing the interference cone; The basic cost elements are integrated with the risk exposure to construct a comprehensive task cost; Under the task allocation constraint, the solution is performed with the goal of minimizing the comprehensive task cost, generating a task allocation set, and then the task allocation set is merged and organized into a task sequence according to the UAV. The task sequence includes the task point coordinates, expected operation time and comprehensive task cost of each task.

4. The anti-interference UAV swarm scheduling method according to claim 3, characterized in that, S4 includes: The interference cone is transformed into a high-cost occupied area in path planning, and a determination function is established to determine whether a spatial point falls into the cone's influence range. The flight path is discretized into path segments, and the interference penalty cost for each path segment is calculated according to the determination function. The interference penalty cost is fused with the path segment length cost and boundary proximity penalty to construct the comprehensive cost of each path segment; Based on the comprehensive cost of the path segment, an initial avoidance path from the starting point to the target point is generated using a graph search method; The initial evasion path is subjected to curvature constraints and smoothing to generate an executable evasion path that satisfies the kinematic constraints of the UAV.

5. The anti-interference UAV swarm scheduling method according to claim 4, characterized in that, The decision function considers both the radial distance of the spatial point relative to the vertex of the cone and the angle relative to the axis of the cone, and introduces a safety expansion boundary.

6. The anti-interference UAV swarm scheduling method according to claim 1, characterized in that, S5 includes: During the flight of the UAV, the comprehensive interference intensity index and its rate of change are calculated based on the directional interference characterization unit collected in real time, and interference mutation events are triggered by continuous determination. Based on the directional data at the moment of triggering mutation, the set of interference cones is incrementally updated to generate new interference cones or correct the parameters of existing interference cones; Based on the updated interference cone, the affected UAVs that are about to enter the high-risk area in the remaining flight path are identified, and their associated affected tasks are collected to form a rescheduling trigger set; Local reassignment is performed on the affected tasks within the rescheduling trigger set, and new evasion paths are planned for the reassigned UAVs.

7. An anti-interference drone swarm scheduling system, applied to the method according to any one of claims 1-6, characterized in that, The system includes: The spatial directional interference characterization model construction module is used to construct a spatial directional interference characterization model based on the time-series observation data of UAV swarms, which includes the directional interference intensity and directional interference gradient in each directional sector. An interference cone identification model construction module is used to construct an interference cone identification model based on the spatial directional interference characterization model, which is composed of interference cones generated by mapping interference concentration regions. The interference cone identification model includes the geometric parameters and risk levels of the interference cones. Specifically, it includes: identifying potential interference mutation directions based on the directional interference gradient using an adaptive threshold; continuously confirming the potential interference mutation directions over a time window, eliminating instantaneous jitter, and retaining the confirmed mutation directions; constructing an interference concentration region with a continuous angular range by continuously expanding adjacent directions with the confirmed mutation direction as the center; mapping the interference concentration region to interference cones in three-dimensional space, and estimating their geometric parameters and risk levels; the generated interference cones and their geometric parameters and risk levels constitute the interference cone identification model. The anti-interference task allocation and task sequence generation module is used to construct an anti-interference task allocation model based on the interference cone, quantify the risk of crossing the interference cone into a task allocation cost, and generate a task sequence for the drone swarm after being merged and organized by drones. The interference cone avoidance path planning module is used to construct a path planning model based on interference cone avoidance based on the interference cone and the task sequence, and generate a flyable path that can effectively avoid the interference cone and meet the kinematic constraints for the UAVs that have been assigned tasks in the task sequence. The dynamic rescheduling module based on interference mutation is used to construct a dynamic rescheduling mechanism based on interference mutation during the process of UAVs executing tasks according to the flightable path. It monitors changes in the interference environment in real time, updates the interference cone when interference mutation is detected, and performs local reallocation and path replanning for the affected tasks and UAVs based on the updated interference cone and the task sequence.

8. An electronic device, characterized in that, The electronic device includes a memory and at least one processor, the memory storing a computer program, and the processor executing the computer program to implement the anti-interference drone swarm scheduling method according to any one of claims 1-6.

9. A computer storage medium, characterized in that, It stores a computer program, which, when executed, implements the anti-interference drone swarm scheduling method according to any one of claims 1-6.

Images