Multi-unmanned aerial vehicle cooperative coverage method based on weighted voronoi assignment

Abstract

The application provides a multi-unmanned aerial vehicle cooperative coverage method based on weighted Voronoi distribution, relates to the technical field of unmanned aerial vehicle data sensing and cooperative control, and the method comprises the following steps: constructing a weighted distance and a weighted Voronoi unit based on an unmanned aerial vehicle site weight, calculating a weighted centroid and a cost functional of the weighted Voronoi unit based on an environment weight, and generating adaptive spatial partition data. By establishing an information age dynamics model and generating time sequence constraint data, by constructing a distributed rolling time domain control optimization problem with the target of minimizing a weighted information age penalty function, and by taking the recursive dynamics of the information age state variable, the state dynamics of the unmanned aerial vehicle and the information age cutoff period as constraint conditions. By executing the first control input at the current sampling time and re-optimizing at the next sampling time, coverage and revisit of the task space are ensured under the premise of meeting the time sequence cutoff period constraint.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) data perception and collaborative control technology, and in particular to a multi-UAV collaborative coverage method based on weighted Voronoi assignment. Background Technology

[0002] In dynamic and uncertain environments, multi-UAV collaborative exploration faces the core technological challenge of achieving efficient, robust, and timely coverage. Existing technologies suffer from the following specific problems and their causes: First, there is a lack of time-series guarantees: Many existing methods focus on spatial coverage efficiency but fail to effectively guarantee timely revisits to high-priority areas. In time-sensitive tasks such as emergency search and rescue, outdated information may lead to missed detections; in continuous monitoring, delayed revisits may result in the loss of critical perception. This is because existing methods often neglect the explicit coupling between time-series indicators such as Information Age (AoI) and UAV control actions.

[0003] Second, they struggle to adapt to heterogeneous and dynamic environments: Traditional Voronoi-based coverage methods typically assume a static environment, isomorphic agents, and uniform spatial importance, which limits their applicability in real-world heterogeneous systems. While weighted Voronoi variants partially consider heterogeneity among UAVs, their modeling is still overly simplistic and fails to fully capture agent-specific capabilities or dynamic resource constraints.

[0004] Third, the limitations of centralized and decentralized approaches: On the one hand, centralized planning, while theoretically optimal, suffers from communication bottlenecks, high computational burden, and single points of failure, making it unsuitable for large-scale or time-critical tasks. On the other hand, decentralized approaches, while improving scalability and reducing communication requirements, often neglect timing constraints, uneven spatial workloads, and robustness to disturbances, which are crucial in time-sensitive and uncertain tasks.

[0005] Fourth, lack of formal guarantees for long-term performance: While model predictive control (MPC) strategies can enhance local robustness and short-term feasibility, most existing designs fail to provide explicit guarantees for long-term time-series coverage or revisit intervals, which are crucial for continuous monitoring tasks.

[0006] Therefore, the core challenge of existing technologies lies in designing a distributed, robust framework that can simultaneously ensure spatial efficiency (e.g., by adapting to UAV heterogeneity and environmental dynamics) and temporal feasibility (e.g., by guaranteeing bounded revisit intervals), while balancing heterogeneous workloads and maintaining robustness to disturbances, which has not yet been fundamentally solved in existing research. Summary of the Invention

[0007] This invention provides a multi-UAV cooperative coverage method based on weighted Voronoi assignment to address the shortcomings of existing technologies. By constructing a distributed control framework that couples environmental heterogeneity, UAV capability adaptability, and information age constraints, it achieves efficient, robust, and timely cooperative exploration and coverage by UAVs.

[0008] In a first aspect, the present invention provides a multi-UAV cooperative coverage method based on weighted Voronoi assignment, comprising: S102. Based on the heterogeneous capabilities of UAVs and the importance of environmental space, a weighted power Voronoi allocation model is constructed to generate adaptive spatial partitioning data. S104. Discretize the free space in the task space into multiple spatial units, establish information age state variables and their recursive dynamic models for each spatial unit, and generate time-series constraint data based on the information age cutoff date and priority weight. S106. Based on the adaptive spatial partitioning data, the temporal constraint data, the UAV state dynamics, and the local information exchanged with neighboring UAVs, a distributed rolling temporal domain control optimization problem is constructed to generate local control sequence data. The distributed rolling temporal domain control optimization problem aims to minimize the weighted information age penalty function in the prediction time domain and includes a temporal deadline constraint that the information age state variable does not exceed the information age deadline. S108. At the current sampling time, the first control input in the local control sequence data is used to drive the UAV to perform the coverage task, and S106 is re-executed at the next sampling time to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint.

[0009] Furthermore, based on the heterogeneous capabilities of UAVs and the spatial importance of the environment, a weighted power Voronoi allocation model is constructed to generate adaptive spatial partitioning data, including: Obtain the site weight data for each drone, whereby the site weight data represents the heterogeneous capabilities of the drone; Obtain environmental weight data for each spatial location point within the operating domain, wherein the environmental weight data represents the spatial importance or obstacle distribution of the spatial location point; Based on the site weight data, a weighted distance is defined between the UAV and the spatial location point. The weighted distance is the square of the Euclidean distance between the nominal position of the UAV and the spatial location point minus the site weight. Based on the environmental weight data, the weighted centroid and cost functional of each weighted Voronoi unit are calculated, so that the environmental weight does not participate in the calculation of the weighted distance, but only affects the update of the UAV's nominal position through the weighted centroid and cost functional. Based on the weighted distance, the free space in the mission space is divided into a set of weighted Voronoi units that correspond one-to-one with the UAV, generating the adaptive space partitioning data. The free space is the remaining area of ​​the mission space after deducting obstacles and their safety expansion regions. Each weighted Voronoi unit consists of all spatial location points in the free space that satisfy the minimum weighted distance condition.

[0010] Furthermore, the process of discretizing the free space in the task space into multiple spatial units, establishing information age state variables and their recursive dynamic models for each spatial unit, and generating time-series constraint data based on the information age cutoff date and priority weights includes: For each spatial location point within the weighted Voronoi unit, an information age state variable is defined, which is used to quantify the timeliness of the information at the spatial location point. Establish a discrete-time recursive equation for the information age state variable. The recursive equation resets the information age state variable when the UAV visits the corresponding spatial location point; otherwise, it increments the information age state variable by the discrete time step. Based on task requirements, information age cutoff periods are set for spatial location points of different priorities, and the time-series constraint data is generated. The time-series constraint data includes the information age cutoff periods and the corresponding priority weights. Among them, the information age state variable is according to A k ( t +1)=(1- x k ( t ))( A k ( t +1) recursion, x k ( t ) is the access indicator function, which is used when any UAV is at the sampling time. t Access the corresponding space unit g k hour x k ( t )=1, otherwise x k ( t =0; the timing constraint data also includes a revisit interval constraint determined by two adjacent access times, making the spatial unit... g k The difference between two consecutive access times does not exceed the corresponding information age cutoff period. t k .

[0011] Furthermore, the step of constructing a distributed rolling time-domain control optimization problem based on the adaptive spatial partitioning data and the temporal constraint data, and generating local control sequence data, includes: At each sampling moment, each UAV establishes a distributed rolling time-domain control optimization problem with the goal of minimizing the weighted information age penalty function in the prediction time domain. The weighted information age penalty function is the sum of the products of the priority weights and the convex function of the information age state variable. In the distributed rolling time-domain control optimization problem, the recursive dynamics of the information age state variable, the state dynamics of the UAV, control input constraints, state constraints, and the information age cutoff period are used as constraints, and a neighbor coordination term is introduced to maintain distributed coordination. The neighbor coordination term is a positive definite term and satisfies the decreasing property over time. It is used to constrain at least one of the following: power Voronoi boundary consistency, coverage area overlap, minimum safe distance, or communication connectivity of neighboring UAVs. Solving the distributed rolling time-domain control optimization problem yields the local control sequence data, which includes the control input at each time point in the prediction time domain. Furthermore, after the step of using the first control input in the local control sequence data to drive the UAV to perform the coverage task and re-execute S106 at the next sampling time, it further includes: In the distributed rolling time-domain control optimization problem, terminal state constraints, terminal invariant sets, terminal controllers, and terminal cost functions are configured. By shifting the feasible control sequence of the previous sampling time and continuing the terminal controller, candidate feasible solutions for the next sampling time are constructed, so that the distributed rolling time-domain control optimization problem of adjacent sampling times remains recursively feasible, generating stability guarantee data. The stability guarantee data includes recursive feasibility conclusions and input to state stability conclusions. By combining the adaptive spatial partitioning data, the temporal constraint data, the local control sequence data, and the stability guarantee data, a time-series perception-based distributed coverage control for heterogeneous UAV formations in a dynamic and uncertain environment is achieved.

[0012] Furthermore, the step of using the first control input in the local control sequence data to drive the UAV to perform a coverage task at the current sampling time, and re-executing S106 at the next sampling time to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint, includes: At the current sampling time, the control input corresponding to the current time in the local control sequence data is applied to the UAV, driving the UAV to perform a coverage task in the corresponding weighted Voronoi unit; At the next sampling moment, the current state information is reacquired and the distributed rolling time-domain control optimization problem is solved again to update the local control sequence data, thereby forming a closed-loop rolling time-domain control strategy, enabling the UAV to cover and revisit the task space while satisfying the timing deadline constraint.

[0013] Furthermore, the step of configuring terminal state constraints, terminal invariant sets, terminal controllers, and terminal cost functions in the distributed rolling time-domain control optimization problem, and constructing candidate feasible solutions for the next sampling time by shifting the feasible control sequence of the previous sampling time and continuing the terminal controller, ensures that the distributed rolling time-domain control optimization problem remains recursively feasible for adjacent sampling times, generating stability-guaranteed data, including: For each UAV, a terminal invariant set and a terminal controller are set, wherein the terminal invariant set is a compact set of states that remain unchanged under the action of the terminal controller; The terminal cost function is set as the control Lyapunov function to ensure that the terminal cost function monotonically decreases along the system trajectory within the terminal invariant set. Provided that the terminal invariant set satisfies the compact control invariance and the terminal cost function satisfies the control Lyapunov function descent condition, the feasible control sequence of the previous sampling time is shifted forward and connected to the terminal controller at the end of the prediction time domain to construct the candidate feasible control sequence of the next sampling time. This ensures that the distributed rolling time domain control optimization problem remains recursively feasible between adjacent sampling times and generates the stability guarantee data, which includes recursive feasibility conclusions and input to state stability conclusions.

[0014] Furthermore, the integration of the adaptive spatial partitioning data, the temporal constraint data, the local control sequence data, and the stability guarantee data to achieve time-series perception-based distributed coverage control of heterogeneous UAV formations in dynamic and uncertain environments includes: The drones determine their respective coverage sub-regions based on the adaptive spatial partitioning data; The time-series constraint data is used to ensure timely revisiting of high-priority regions. Specific flight control is executed based on the local control sequence data, and based on the stability guarantee data, the input-to-state stability and timing deadline constraints are satisfied when bounded disturbances exist, and the timing deadline constraints are satisfied with a probability not lower than a preset confidence level when sub-Gaussian disturbances exist.

[0015] Furthermore, the weighted power Voronoi assignment model is dynamically optimized through a Lloyd-type iterative update mechanism, which includes: Calculate the weighted centroid of the spatial location point within each weighted Voronoi cell. The weighted centroid is the weighted average position of all spatial location points within the weighted Voronoi cell, with the environmental weight as the weight. Update the nominal position of the drone to the weighted centroid; The iterative update mechanism is repeated until the nominal position of the UAV converges to the weighted centroid power Voronoi configuration. During the iteration of this mechanism, the cost functional remains monotonically non-increasing. The cost functional is the sum of the integrals of the squared Euclidean distances from all spatial points within the weighted Voronoi cells to the corresponding nominal UAV position, weighted by environmental weights. .

[0016] Furthermore, when the UAV is subjected to a sub-Gaussian random perturbation, the time-series constraint data is tightened through a probability cutoff guarantee mechanism, which includes: Obtain the variance proxy data of the sub-Gaussian random perturbation; Based on the preset confidence risk parameters d and probability guarantee time range T Calculate the safety margin data; the specific calculation method for the safety margin data is as follows: ,in, s k For the variance proxy s k The square root of ² is called the standard deviation, and the probability is guaranteed over a time range. T With the prediction time domain of the distributed rolling time-domain control optimization problem H Each setting is independent of the others; exist t k - β k Under the condition that >0, the information age cutoff period is tightened to a tightened cutoff period, where the tightened cutoff period is the original cutoff period. t k Subtract the aforementioned safety margin data β k ,Right now t k tight = t k - β k ; In the distributed rolling time-domain control optimization problem, the tightened deadline is used to replace the original deadline as a hard constraint. Based on the tail inequality and joint bound principle of sub-Gaussian random variables, it is guaranteed that within the time range of the probability guarantee...T Within this period, the information age status variable does not exceed the original cutoff period. t k The probability is not less than 1- d .

[0017] The multi-UAV cooperative coverage method based on weighted Voronoi allocation provided by this invention can produce the following beneficial technical effects compared with the prior art: This invention constructs weighted distance and weighted Voronoi units by introducing drone site weights, enabling more capable drones to cover larger or more complex sub-regions. By introducing environmental weights to calculate weighted centroids and cost functionals, hotspot areas, key monitoring areas, and areas with high travel costs receive higher influence weights in drone nominal position updates and trajectory planning. This adaptive spatial partitioning effectively addresses the limitations of traditional Voronoi methods that assume a static environment, isomorphic agents, and uniform spatial importance. By establishing an information age dynamics model and generating time-series constraint data, information timeliness indicators are explicitly introduced into the control framework, allowing the system to quantify the staleness of information at each spatial location and differentiate the importance of different areas through information age cutoff dates and priority weights. By constructing a distributed rolling time-domain control optimization problem with the objective of minimizing the weighted information age penalty function, and using the recursive dynamics of the information age state variable, the drone's state dynamics, and the information age cutoff date as constraints, each drone can independently solve local optimization problems while maintaining distributed coordination with neighboring drones through neighbor coordination terms. By executing the first control input at the current sampling time and re-optimizing it at the next sampling time, a closed-loop rolling time-domain control strategy is formed, enabling the UAV to continuously update control decisions based on real-time status, ensuring coverage and revisiting of the mission space while meeting the timing deadline constraints.

[0018] This invention can simultaneously ensure spatial efficiency, temporal feasibility, heterogeneous workload balancing, and robust coordination. It effectively solves the core problems in the prior art, such as difficulty in balancing spatial coverage and temporal assurance, insufficient heterogeneous adaptability, poor scalability of centralized control, and lack of formal guarantees for long-term performance. It is particularly suitable for time-sensitive and environmentally uncertain collaborative mission scenarios such as continuous surveillance and infrastructure monitoring. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating an optional multi-UAV cooperative coverage method based on weighted Voronoi allocation provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] It should be noted that, in the description of the embodiments of the present invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0023] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more.

[0024] Figure 1 This is a flowchart illustrating an optional multi-UAV cooperative coverage method based on weighted Voronoi allocation provided by the present invention.

[0025] This invention provides a multi-UAV cooperative coverage method based on weighted Voronoi allocation. By combining a weighted power Voronoi allocation mechanism with a time-aware distributed rolling temporal control strategy, it achieves efficient, timely, and robust cooperative coverage of heterogeneous multi-UAVs in dynamic and uncertain environments. The following detailed description of each step of the method is provided with reference to specific embodiments.

[0026] In step S102, a weighted power Voronoi allocation model is constructed based on the heterogeneous capabilities of the UAV and the importance of the environmental space to generate adaptive spatial partitioning data.

[0027] In this embodiment, site weight data for each drone is obtained. The site weight data represents the heterogeneous capabilities of the drone. For example, in a target search scenario, the deployed drone formation may include a large drone equipped with a high-resolution thermal imaging camera and a small drone equipped with only a regular optical camera. The former has stronger perception capabilities and longer endurance, and its site weight can be set to 2.0, while the latter's site weight can be set to 0.5. The value range of the site weight is usually between 0.5 and 2.0, and it is positively correlated with the perception range, which is between 12m and 18m.

[0028] Simultaneously, environmental weight data of each spatial location point within the operation domain is acquired. The environmental weight data represents the spatial importance of the spatial location point or the distribution of obstacles. For example, in a bounded two-dimensional region of 1000m×1000m, the environmental weight of the area where the target is suspected is higher, while the environmental weight of the area where the non-target is suspected is lower.

[0029] Based on the site weight data, a weighted distance is defined between the UAV and the spatial location point. This weighted distance is the squared Euclidean distance between the nominal position of the UAV and the spatial location point minus the site weight. , in, For drones i Spatial location point q The weighted distance between them p i For drones i The nominal position, w i For drones i Site authority, For free space Q f Any spatial location point within.

[0030] Optionally, multiple drones can form a connected undirected communication graph. G =( N , E Drones i The neighbor set is N i ={ j |( i , j )∈ E Each UAV exchanges at least one of the following with UAVs in the neighbor set at each sampling time: state information, nominal position information, power Voronoi boundary information, or local information age state variables, and constructs the distributed rolling time-domain control optimization problem based on the exchanged information.

[0031] Based on the weighted distance, the free space in the mission space is divided into a set of weighted Voronoi units corresponding one-to-one with the UAV, generating the adaptive spatial partitioning data. The free space is the remaining region of the mission space after deducting obstacles and their safety expansion regions. Each weighted Voronoi unit consists of all spatial location points within the free space that satisfy the minimum weighted distance condition. , in, For drones i The corresponding weighted (power) Voronoi unit (composed of all spatial locations in free space that satisfy the condition of minimum weighted distance); For all N The configuration consisting of the nominal positions of the drones; Free space, i.e., task space Q Deduct obstacle set O and the remaining area after its safe expansion region ( Q f = Q \ O This means that, all other things being equal, drones with higher site weights have larger weighted Voronoi units, thus covering a larger sub-area; environmental weights r ( q Instead of directly entering the weighted distance, the update of the UAV's nominal position is influenced by the weighted centroid calculation and cost functional in the following text, so that the high-importance region can obtain a greater influence weight in the nominal position iteration.

[0032] Optionally, based on the task space Q The set of obstacles in O Constructing free space with safe expansion distance Q f = Q \ O The power Voronoi unit partitioning, weighted centroid calculation, UAV nominal position update, and UAV flight trajectory are all confined to the free space. Q f Inside.

[0033] In step S104, the free space in the task space is discretized into multiple spatial units. For each spatial unit, an information age state variable and its recursive dynamic model are established, and time-series constraint data are generated based on the information age cutoff date and priority weight.

[0034] In this embodiment, for each spatial location point within the weighted Voronoi unit, an information age state variable is defined. This information age state variable is used to quantify the timeliness of information at that spatial location point. Representing spatial units At time t, the information age is determined by its value; a larger value indicates that the information is more outdated. A discrete-time recursive equation is established for the information age state variable. This recursive equation resets the information age state variable when the UAV visits the corresponding spatial location point; otherwise, it increments the information age state variable. A k ( t +1)=(1- x k ( t ))( A k ( t )+1); in, x k ( t )∈{0,1} is the access indicator function, when the drone is at time t Access Space Unit g k hour x k ( t )=1, at this time A k ( t +1)=0, the information age is reset to zero; when not accessed. x k ( t )=0, at this time A k ( t +1)= A k ( t +1, the information age increases by one.

[0035] Based on task requirements, information age cutoff periods are set for spatial location points of different priorities, generating the temporal constraint data. This temporal constraint data includes the information age cutoff periods and corresponding priority weights. For example, in target search, the information age cutoff period for hotspot areas (such as areas where the target may be located) is set to 30 seconds, with a higher priority weight; the information age cutoff period for corridor areas (such as main passageways) is set to 45 seconds, with a medium priority weight; and the information age cutoff period for ordinary areas (such as general streets) is set to 60 seconds, with a lower priority weight. The priority weights... l k Information age penalty function ( Ak The product of these components forms a weighted information age penalty function. l k · ( A k (), used to distinguish the importance of different regions in an optimization problem.

[0036] In step S106, a distributed rolling time-domain control optimization problem is constructed based on the adaptive spatial partitioning data and the temporal constraint data, generating local control sequence data. The distributed rolling time-domain control optimization problem aims to minimize the weighted information age penalty function and includes a temporal deadline constraint that the information age state variable does not exceed the information age deadline.

[0037] In this embodiment, each UAV, at each sampling time, establishes a distributed rolling time-domain control optimization problem with the objective of minimizing the weighted information age penalty function in the prediction time domain. The weighted information age penalty function can be the sum of the products of the priority weights and the convex function of the information age state variable, i.e. , in, For the prediction time domain length (prediction steps, e.g.) H =10 to 20 time steps); The total number of spatial units after discretization of free space (e.g., in a 30×30 grid). M =900); To predict the time step index in the time domain ( h =0,1,…, H 1); ( A k ) is a convex function, for example ( A k ) = A k (Linear function) or ( A k )= A k ² (quadratic function): When using a quadratic function, a stronger penalty is imposed on large information age values, prioritizing the reduction of timeliness loss in high information age regions.

[0038] In the distributed rolling time-domain control optimization problem, the recursive dynamics of the information age state variable, the state dynamics of the UAV, and the information age cutoff period are used as constraints, and a neighbor coordination term is introduced to maintain distributed coordination. The state dynamics of the UAV adopts a discrete-time kinematic model.x i ( t +1) = f i ( x i ( t ), u i ( t )) + d i ( t ),in x i ( t () represents the state variables, including position data and velocity data. u i ( t (This refers to) control inputs, including speed commands or acceleration commands. d i ( t () is a bounded perturbation or a sub-Gaussian perturbation; f i Discrete-time linear kinematics models can be used. Where A and B are system matrices, and the control input is constrained by the maximum speed and the maximum acceleration, for example, the maximum speed is 8 m / s and the maximum acceleration is 4 m / s².

[0039] The timeline deadline constraint is A k ( t + h )≤ t k Ensure in the prediction time domain H The information age of each spatial unit does not exceed its information age cutoff period. The neighbor coordination item... P t It is a positive definite function used to quantify at least one of the following factors among the current UAV and its neighboring UAVs: power Voronoi boundary consistency, coverage area overlap, minimum safe distance, or communication connectivity, and satisfies the descent property. P t+1 - P t ≤ - ξΠ t , x >0, thus maintaining boundary consistency, avoiding redundant coverage, avoiding coverage blind spots, and preventing collisions between adjacent drones while each drone independently solves the local optimization problem.

[0040] Solving the distributed rolling time-domain control optimization problem yields the local control sequence data, which includes the control inputs at each time point within the prediction time domain. ui (0), u i (1), ..., u i ( H -1).

[0041] In step S108, at the current sampling time, the first control input in the local control sequence data is used to drive the UAV to perform the coverage task, and S106 is re-executed at the next sampling time to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint.

[0042] In this embodiment, at the current sampling time, the control input corresponding to the current time in the local control sequence data is applied to the UAV, driving the UAV to perform a coverage task within the corresponding weighted Voronoi cell. For example, the UAV adjusts its flight trajectory according to the calculated acceleration control input to cover the spatial position points within its weighted Voronoi cell, while satisfying both velocity and acceleration constraints. At the next sampling time, the current state information (including current position, velocity, and information age state variables of each spatial cell) is reacquired, and the distributed rolling temporal domain control optimization problem is solved again, updating the local control sequence data to form a closed-loop rolling temporal domain control strategy. This enables the UAV to achieve coverage and revisiting of the task space while satisfying the temporal deadline constraints.

[0043] It should be noted that the sampling period Δ t It can be set to 1 second for prediction in the time domain. H Configured based on computing power and task requirements, for example H = 10 to 20 time steps. Through the rolling time-domain mechanism, the UAV can update the control input online according to the real-time status and dynamic environment, and adapt to disturbances in real time.

[0044] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment, which constructs a weighted power Voronoi allocation model based on the heterogeneous capabilities of UAVs and the spatial importance of the environment, and generates adaptive spatial partitioning data, includes: Obtain the site weight data for each drone, whereby the site weight data represents the heterogeneous capabilities of the drone; Obtain environmental weight data for each spatial location point within the operating domain, wherein the environmental weight data represents the spatial importance or obstacle distribution of the spatial location point; Based on the site weight data, a weighted distance between the UAV and the spatial location point is defined. The weighted distance is the square of the Euclidean distance between the nominal position of the UAV and the spatial location point minus the site weight. Based on the weighted distance, the free space in the mission space is divided into a set of weighted Voronoi units that correspond one-to-one with the UAV, generating the adaptive space partitioning data. The free space is the remaining area of ​​the mission space after deducting obstacles and their safety expansion regions. Each weighted Voronoi unit consists of all spatial location points in the free space that satisfy the minimum weighted distance condition.

[0045] In this embodiment, after using the first control input in the local control sequence data to drive the UAV to perform the coverage task and re-execute S106 at the next sampling time, the method further includes configuring terminal state constraints, terminal invariant sets, terminal controllers and terminal cost functions in the distributed rolling time domain control optimization problem to maintain recursive feasibility and generate stability-guaranteed data.

[0046] Specifically, a terminal invariant set is set for each drone. X f,i and terminal controller k i The terminal invariant set is the set of compact states that remain unchanged under the action of the terminal controller, that is, for all x ∈ X f,i ,have k i (x) ∈ U i and f i ( x , k i ( x )) ∈ X f,i This means that once the drone's state enters the terminal invariant set, it will remain within that set under the control of the terminal controller. (Setting the terminal cost function...) V f,i As a control Lyapunov function, it ensures that the terminal cost function monotonically decreases along the system trajectory within the terminal invariant set, i.e. , in, α f,i ∈ ∞ .

[0047] Based on the shifted construction of the feasible solution from the previous time step and the application of the terminal controller, the distributed rolling time-domain control optimization problem remains recursively feasible between adjacent time steps, generating the stability guarantee data. This stability guarantee data includes recursive feasibility conclusions and input-state stability conclusions. It is worth noting that the process of maintaining recursive feasibility can be as follows: at time step... t After obtaining the optimal solution, at time... t +1 moves the previous solution forward one step and applies the terminal controller at the terminal. k i Due to the properties of the terminal invariant set, the shifted solution satisfies all constraints, thus guaranteeing that at time t, the solution is invariant. t +1 has a feasible solution.

[0048] By combining the adaptive spatial partitioning data, the temporal constraint data, the local control sequence data, and the stability guarantee data, a time-series perception-based distributed coverage control for heterogeneous UAV formations in a dynamic and uncertain environment is achieved.

[0049] Optionally, the UAVs determine their respective coverage sub-regions based on the adaptive spatial partitioning data, meaning each UAV is responsible for covering spatial location points within its weighted Voronoi unit. Timely revisiting of high-priority areas is ensured based on the temporal constraint data; that is, through optimization of information age cutoff constraints and weighted information age penalty functions, UAVs prioritize accessing areas with higher information ages or higher priorities. Specific flight control is executed based on the local control sequence data, adjusting the flight trajectory according to the control input generated by the rolling temporal control strategy. Furthermore, based on the stability guarantee data, the system satisfies input-to-state stability and temporal cutoff constraints when bounded disturbances exist, and satisfies temporal cutoff constraints with a probability not lower than a preset confidence level when sub-Gaussian disturbances exist. Through the comprehensive application of the above data, a unified approach is achieved for spatial coverage efficiency, temporal guarantees, heterogeneous workload balancing, and robustness coordination.

[0050] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment constructs a weighted power Voronoi allocation model based on the heterogeneous capabilities of UAVs and the spatial importance of the environment, generating adaptive spatial partitioning data. This includes obtaining site weight data for each UAV, where the site weight data represents the heterogeneous capabilities of the UAV. For example, in a scenario deploying 6 to 15 heterogeneous UAVs, the site weight of each UAV is randomly sampled within the range of 0.5 to 2.0, and the site weight is positively correlated with the sensing range, which is between 12m and 18m.

[0051] It should be noted that drones with higher site weights have larger weighted Voronoi cells, enabling them to cover a larger area, reflecting their stronger perception and task execution capabilities. Environmental weight data for each spatial location point within the operational domain is acquired; this environmental weight data represents the spatial importance of the location point or the distribution of obstacles. For example, in a scenario where a bounded two-dimensional region of 1000m × 1000m is discretized into a 30 × 30 grid (900 cells in total), the environmental weights for hotspot areas are set to higher values, those for ordinary areas to lower values, and those for obstacle areas to a minimum value approaching zero.

[0052] Then, based on the site weight data, a weighted distance is defined between the UAV and the spatial location point. This weighted distance is the squared Euclidean distance between the nominal position of the UAV and the spatial location point minus the site weight. For example, before calculating the weighted distance, the task space coordinates can be normalized to the [0,1] interval to make the weighted distance and the site weight similar in magnitude; or a form such as... d i w ( q )=|| q - p i ||² / s i ²- w i The scale-normalized form, in which s i For drones i The sensing radius or overall capability scale. Example in scale-normalized form: UAV i Nominal position p i =(100, 200), Perception radius s i =18m, Site Weight w i =1.5, spatial location point q =(150, 250), then the weighted distance d i w ( q =(50²+50²) / 18²-1.5=13.93; Correspondingly, the less capable drones... s i Smaller or oh The smaller the value, the larger its weighted distance, and the smaller the corresponding Voronoi element.

[0053] Then, based on the weighted distance, the free space in the task space is divided into a set of weighted Voronoi units corresponding one-to-one with the UAVs, generating the adaptive spatial partitioning data. The free space is the remaining area of ​​the task space after deducting obstacles and their safety expansion regions. Each weighted Voronoi unit consists of all spatial location points in the free space that satisfy the minimum weighted distance condition. For example, for two UAVs A and B, UAV A has a site weight of 2.0 and a nominal location of (300, 400), while UAV B has a site weight of 0.8 and a nominal location of (350, 450). For the spatial location point q = (320, 420), the calculation is... d A w ( q ) = (320-300)² + (420-400)² - 2.0 = 400 + 400 - 2.0 = 798, d B w ( q ) = (320-350)² + (420-450)² - 0.8 = 900 + 900 - 0.8 = 1799.2, since d A w ( q )< d B w ( q The spatial location point belongs to the weighted Voronoi cell of UAV A.

[0054] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment discretizes the free space in the task space into multiple spatial units, establishes an information age state variable and its recursive dynamic model for each spatial unit, and generates time-series constraint data based on the information age cutoff date and priority weight. This includes defining an information age state variable for each spatial location point within the weighted Voronoi unit, whereby the information age state variable is used to quantify the timeliness of information at the spatial location point. For example, for a spatial unit... g k At the initial moment t When =0, the information age state variable A k (0) = 0, indicating that the information is up-to-date; if the spatial unit is in t =10 times was accessed by the drone, that is x k (10) = 1, then according to the recurrence relation we have Ak (11) = 0; if in t =11 was not visited at time 11, that is x k (11)=0, then A k (12)= A k (11)+1=1, and so on.

[0055] Establish a discrete-time recursive equation for the information age state variable. The recursive equation resets the information age state variable when the UAV visits the corresponding spatial location point, and increments the information age state variable otherwise.

[0056] Specifically, A k ( t +1) = (1 - x k ( t ))( A k ( t ) + 1), where x k (t) is the access indication function, when the drone is at time... t Access Space Unit g k hour x k ( t ) = 1, otherwise x k ( t = 0. For example, spatial unit g k exist t =0 to t =4 If none of the visits are made, then A k (0)=0, A k (1)=1, A k (2)=2, A k (3) = 3, A k (4) = 4, A k (5) = 5; in t If it is accessed at time 5, then A k (6) = 0; in t If the time interval from t=6 to t=9 is not visited, then A k (7)=1, Ak (8) = 2, A k (9) = 3, A k (10) = 4.

[0057] Based on task requirements, information age cutoff periods are set for spatial location points of different priorities, generating the temporal constraint data. This temporal constraint data includes the information age cutoff periods and corresponding priority weights. For example, in a target search scenario, information age cutoff periods and priority weights are set for three types of regions: information age cutoff periods for hotspot regions (high priority). t 1 = 30s, priority weight l 1 = 4.0; Information age cutoff for corridor area (medium priority) t 2 = 45s, priority weight l 2 = 2.0; Information age cutoff period for ordinary areas (low priority) t 3 = 60s, priority weight l 3 = 1.0. Here, the priority weight is used in the weighted information age penalty function, causing the optimization problem to prioritize reducing the information age of high-priority regions.

[0058] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi assignment provided in this embodiment constructs a distributed rolling temporal domain control optimization problem based on the adaptive spatial partitioning data and the temporal constraint data, generating local control sequence data, including the situation of each UAV at each sampling time. The goal is to minimize the weighted information age penalty function in the prediction time domain, establishing a distributed rolling temporal domain control optimization problem. The weighted information age penalty function is the sum of the products of the priority weights and the convex function of the information age state variable. For example, for the prediction time domain... H = 10, number of spatial units M = 900, the weighted information age penalty function is ,in ( A k ) = A k ².

[0059] Assume at time... t spatial unit g 1 (Hotspot Areas) l Information age state variable (1=4.0) A 1( t =20, spatial unit g 2 (Normal area, l Information age state variable (3=1.0) A2( t If ) = 35, then g The penalty contribution of 1 is 4.0 × 20² = 1600. g The penalty contribution of 2 is 1.0 × 35² = 1225. g The penalty of 1 contributes more, and optimization problems will prioritize planning against it when solving the problem. g Access to 1 reduces the information age of high-priority areas.

[0060] In the distributed rolling time-domain control optimization problem, the recursive dynamics of the information age state variable, the state dynamics of the UAV, and the information age cutoff period are used as constraints, and a neighbor coordination term is introduced to maintain distributed coordination. The recursive dynamics constraint of the information age state variable can be... A k ( t + h +1) = (1 - x k ( t + h ))( A k ( t + h ) + 1), to ensure that the evolution of information age conforms to actual access situation.

[0061] Optionally, the state dynamics constraints of the UAV are: x i ( t + h +1) = f i ( x i ( t + h ), u i ( t + h )) + d i ( t + h ), where the state variable xi includes position data and velocity data, and the control input u i Includes acceleration data, and satisfies u i ( t + h )∈ U i (Set of control constraints) and x i ( t + h )∈X i (Set of state constraints).

[0062] Information age cutoff period constraint is A k ( t + h )≤ t k This ensures that the information age of each spatial unit does not exceed its cutoff date within the prediction time domain. The neighbor coordination item... P t This is used to quantify the consistency of coverage boundaries, the degree of overlap in coverage areas, the minimum safe distance, or the degree of communication connectivity satisfaction between adjacent drones. For example, it can be set to... ,in N i For drones i The neighbor set consists of drones within a communication range of 180m. r safe The minimum safe distance allowed between adjacent drones. D bd ( i , j , t For drones i and j The change in power Voronoi boundary relative to the previous sampling time. α , β The coefficient is positive; the neighbor coordination term penalizes adjacent drones that are too close, severe fluctuations in coverage boundaries, and overlapping coverage areas, thereby avoiding duplicate coverage, coverage blind spots, and collisions between adjacent drones. Solving the distributed rolling time-domain control optimization problem yields the local control sequence data, which includes the control input at each moment in the predicted time domain, for example... u i (0) = (0.5, 0.3) m / s², u i (1) = (0.4, 0.2) m / s², ..., u i (9) = (0.1, 0.1)m / s², representing the acceleration control input at each time point in the prediction time domain.

[0063] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment uses the first control input in the local control sequence data to drive the UAV to perform a coverage task at the current sampling time, and re-executes S106 at the next sampling time to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint. This includes applying the control input corresponding to the current time in the local control sequence data to the UAV at the current sampling time, driving the UAV to perform the coverage task within the corresponding weighted Voronoi unit. For example, at time... t Solving the distributed rolling time-domain control optimization problem yields local control sequence data. u i (0), u i (1), ..., u i ( H -1)}, will be the first control input u i (0) = (0.5, 0.3) m / s² applied to the drone i The drone updates its speed and position based on this acceleration control input, with the speed updated as follows: v i (t+1) = v i (t) + u i (0)·Δ t Location updated to p i ( t +1)= p i ( t ) + v i ( t )·Δ t + 0.5· u i (0)·Δ t ², where Δ t = 1s. Assuming a drone i At any moment t If the velocity is (2.0, 1.5) m / s and the position is (100, 200) m, then the updated velocity is (2.5, 1.8) m / s and the position is (102.25, 201.65) m. The UAV continues to fly according to the updated state, covering the spatial position points within its weighted Voronoi cell.

[0064] At the next sampling time, the current state information is reacquired and the distributed rolling time-domain control optimization problem is solved again. The local control sequence data is updated to form a closed-loop rolling time-domain control strategy, enabling the UAV to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint. For example, at time... t +1, reacquire the current position (102.25, 201.65) m, current velocity (2.5, 1.8) m / s, and the current information age state variables of each spatial unit, and re-establish and solve the distributed rolling time-domain control optimization problem to obtain new local control sequence data { u i '(0), u i '(1), ..., u i '( H -1)}, will be the new first control input u i (0) is applied to the UAV, and this cycle repeats to form a closed-loop control. Through this rolling time-domain mechanism, the UAV can update the control input online according to the real-time status and dynamic environment (such as disturbances and changes in obstacles), ensuring continuous coverage and timely revisit of the mission space while meeting the time-deadline constraints.

[0065] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment configures the terminal invariant set, terminal controller, and terminal cost function to maintain the recursive feasibility of the distributed rolling temporal control optimization problem and generate stability-guaranteed data, including setting a terminal invariant set for each UAV. X f,i and terminal controller k i The terminal invariant set is a set of compact states that remain unchanged under the action of the terminal controller. For example, the terminal invariant set can be designed with a reference position as the center and a radius of... r The circular area, i.e. X f,i = { x i | || p i - p i ref || ≤ r , || v i || ≤ v max},in p i ref For dronesi The reference position (e.g., the weighted centroid of the current sub-region) is taken. r =5m, v max = 1.0 m / s.

[0066] Terminal controller k i It can be designed as a proportional controller. k i ( x i ) = - K p ( p i - p i ref ) - K d v i ,in K p and K d To control the gain. For all x ∈ X f,i ,have k i (x) ∈ U i (Satisfies control input constraints) and f i ( x , k i ( x )) ∈ X f,i (The state remains within the terminal invariant set). Define the terminal cost function. V f,i As a control Lyapunov function, it ensures that the terminal cost function monotonically decreases along the system trajectory within the terminal invariant set. For example, the terminal cost function can be designed as follows: V f,i ( x i ) =( p i - p i ref ) T · P p ·( p i - p i ref ) + vi T · P v · v i ,in P p and P v It is a positive definite matrix. The terminal cost function satisfies ,in, α f,i (|| x ||) = c ·|| x ||², c >0 ensures that the terminal cost function decreases monotonically along the system trajectory.

[0067] Based on the shift construction of the feasible solution of the previous time step and the application of the terminal controller, the distributed rolling time domain control optimization problem remains recursively feasible between adjacent time steps, generating the stability guarantee data, which includes recursive feasibility conclusions and input to state stability conclusions.

[0068] Specifically, at time t Obtain the optimal solution { u i (0) , u i (1) , ..., u i (H-1)}, the corresponding predicted state trajectory is { x i (0) , x i (1) , ..., x i (H)},in x i (H) ∈ X f,i At that moment t +1, construct candidate solutions as { u i (1) , u i (2) , ..., u i (H-1) , k i ( x i(H) That is, the previous solution is moved forward one step and the terminal controller is applied at the terminal. Because x i (H) ∈ X f,i Furthermore, the properties of terminal invariant sets can be applied. k i ( x i (H) The state remains unchanged afterward. X f,i Therefore, the candidate solution satisfies all constraints, making time... t +1 keeps it feasible, thus keeping the recursion feasible.

[0069] In an optional embodiment, the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment integrates the adaptive spatial partitioning data, the temporal constraint data, the local control sequence data, and the stability guarantee data to achieve time-aware distributed coverage control of heterogeneous UAV formations in a dynamic and uncertain environment. This includes the UAVs determining their respective coverage sub-regions based on the adaptive spatial partitioning data. For example, UAVs... i According to its weighted Voronoi unit V i w ( P The drone determines its coverage sub-region, which consists of all spatial locations that satisfy the condition of minimum weighted distance. i It is responsible for covering and revisiting spatial location points within the sub-region.

[0070] The time-series constraint data ensures timely revisiting of high-priority areas, such as drones. i When planning the flight trajectory, priority is given to spatial locations within its coverage sub-region whose information age is close to or exceeds the information age cutoff period, especially spatial locations in hotspot areas (information age cutoff period of 30 seconds), to ensure the information freshness of high-priority areas. Specific flight control is executed based on the local control sequence data, for example, by a drone. i The system adjusts its flight trajectory according to the control input sequence generated by the rolling time-domain control strategy, performs coverage tasks within the corresponding weighted Voronoi units, and avoids collisions with obstacles (such as collapsed buildings) and other UAVs. Based on the stability guarantee data, the system satisfies input-to-state stability and timing deadline constraints when bounded disturbances exist, and satisfies timing deadline constraints with a probability not lower than a preset confidence level when sub-Gaussian disturbances exist.

[0071] For example, when a UAV is subjected to bounded disturbances (position disturbance ±0.5m, velocity disturbance ±0.2m / s), due to the guarantee of input-state stability (ISS), the UAV's state will not diverge, and it can still maintain stable flight and meet the timing deadline constraints; when subjected to sub-Gaussian disturbances (position variance... s ² = 0.25, velocity variance s When 2=0.04), the time-series deadline constraint is met with a high probability (e.g., 99%) through the probability deadline guarantee mechanism.

[0072] By comprehensively applying the above data, this invention can realize time-series perception-based distributed coverage control of heterogeneous UAV formations in dynamic and uncertain environments, ensuring the unity of spatial coverage efficiency, timing guarantee, workload balancing, and robustness coordination.

[0073] In an optional embodiment, the weighted power Voronoi allocation model in the multi-UAV cooperative coverage method based on weighted Voronoi allocation provided in this embodiment is dynamically optimized through a Lloyd-type iterative update mechanism. The Lloyd-type iterative update mechanism includes calculating the weighted centroid of spatial location points within each weighted Voronoi unit. The weighted centroid is the weighted average position of all spatial location points within the weighted Voronoi unit with the environmental weight as the weight.

[0074] Specifically, for the weighted Voronoi unit of UAV i V i w ( P Its weighted centroid ,in, r ( q () represents a spatial location point q Environmental weights. For example, suppose a drone... i The weighted Voronoi cell contains spatial location points. q 1 = (100, 100) (Environmental weight) r ( q 1)=2.0), q 2 = (120, 110) (Environmental weight) r ( q 2)=1.5), q 3 = (110, 120) (Environmental weight) r ( q If 3) = 1.0), then the weighted centroid is... C i ( P= (2.0×(100,100) + 1.5×(120,110) + 1.0×(110,120)) / (2.0+1.5+1.0) = ((200+180+110) / 4.5, (200+165+120) / 4.5) = (108.9, 107.8).

[0075] Update the nominal position of the drone to the weighted centroid, that is... p i + = C i ( P The iterative update mechanism is repeated until the nominal position of the UAV converges to the weighted centroid power Voronoi configuration. During the iteration of the iterative update mechanism, the cost functional monotonically does not increase. The cost functional is the squared Euclidean distance from each spatial location point within a weighted Voronoi cell to the corresponding nominal position of the UAV multiplied by the environmental weight. r ( q The sum of weighted integrals, i.e. For example, in the first iteration, H ( P ) = 10000; After updating the nominal position, H ( P + ) = 8500; After the second iteration, H ( P ++ The cost functional remains monotonically constant until convergence, with the nominal position of the UAV converging to the weighted centroid power Voronoi configuration. p i * = C i ( P * The cost functional reaches a local minimum, and the system achieves optimal spatial coverage and workload balancing. This iterative update mechanism is distributed; each UAV only needs to exchange nominal position information with its neighboring UAVs to independently calculate its weighted centroid and update its nominal position, making it suitable for online adaptive adjustment of large-scale UAV formations.

[0076] In an optional embodiment, in the multi-UAV cooperative coverage method based on weighted Voronoi assignment provided in this embodiment, when the UAV is subjected to a sub-Gaussian random perturbation, the time-series constraint data is tightened through a probability cutoff guarantee mechanism. This probability cutoff guarantee mechanism includes obtaining variance surrogate data of the sub-Gaussian random perturbation. For example, in a post-disaster urban search and rescue scenario, the flight time of the UAV is uncertain due to random changes in wind speed. This uncertainty can be modeled as a sub-Gaussian random perturbation, and its variance surrogate data... s k ² = 0.25 (position) and s k ² = 0.04 (velocity).

[0077] Based on the preset confidence risk parameters d and probability guarantee time range T ( T With rolling time-domain prediction step size H (Set up independently of each other) Calculate safety margin data ,in s k For variance proxy s k The square root of ² is called the standard deviation. This is called the confidence level adjustment factor. For example, setting a confidence level of 1- d =0.99 (i.e.) d =0.01), probability guaranteed time range T =3600s (simulation time domain), then the confidence adjustment factor is ,Pick s k =0.5 (corresponding to positional variance proxy) s k If ²=0.25), then the safety margin data β k =0.5×5.06=2.53s.

[0078] exist t k - β k Under the condition that the age cutoff period is greater than 0, the cutoff period for the information age is tightened to a tightened cutoff period, which is the original cutoff period minus the safety margin data, i.e. t k tight = t k - β k ;like t k - β kIf the value is ≤0, the access frequency should be increased or the probability guarantee time range should be shortened. T To ensure safety margin β k Lower. For example, for hotspot areas, the age cutoff period for original information. t k =30s, tightening the deadline t k tight = 30 - 2.53 = 27.47s; For ordinary regions, the cutoff period for the age of the original information. t k =60s, tightening the deadline t k tight = 60 - 2.53 = 57.47s. In the distributed rolling time-domain control optimization problem, the tightened deadline is used to replace the original deadline as a hard constraint. That is, in the distributed rolling time-domain control optimization problem, the constraint condition becomes... A k ( t + h ) ≤ t k tight .

[0079] Based on the tail inequality and joint bound principle of sub-Gaussian random variables, it is guaranteed that within the time range of the probability guarantee... T Within this period, the information age status variable does not exceed the original cutoff period. t k The probability is not less than 1- d .

[0080] Specifically, for spatial units g k By the tail inequality Pr{ of the sub-Gaussian random variable e k ( j )≥ β}≤exp(- β ² / (2 s k ²)), for j ≤ T Taking the joint bound, we get Pr{max j≤T G j ( k )≥ D k + β} ≤ T ·exp(- β ² / (2 s k ²));Selectβ = β k Make the probability not exceed d Therefore, in the nominal revisit interval D k Not exceeding the tightening deadline t k tight Under the premise of probability 1- d For all t ∈[0, T All have A k ( t )≤ t k For example, when d =0.01、 T At 3600s, with a probability of no less than 99%, the spatial unit is guaranteed to remain within the 3600s simulation time domain. g k The information age state variable never exceeds the cutoff period of its original information age. t k .

[0081] The system 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.

[0082] 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.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some 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 the present invention.

Claims

1. A multi-UAV cooperative coverage method based on weighted Voronoi assignment, characterized in that, include: S102. Based on the heterogeneous capabilities of UAVs and the importance of environmental space, a weighted power Voronoi allocation model is constructed to generate adaptive spatial partitioning data. S104. Discretize the free space in the task space into multiple spatial units, establish information age state variables and their recursive dynamic models for each spatial unit, and generate time-series constraint data based on the information age cutoff date and priority weight. S106. Based on the adaptive spatial partitioning data, the temporal constraint data, the UAV state dynamics, and the local information exchanged with neighboring UAVs, a distributed rolling temporal domain control optimization problem is constructed to generate local control sequence data. The distributed rolling temporal domain control optimization problem aims to minimize the weighted information age penalty function in the prediction time domain and includes a temporal deadline constraint that the information age state variable does not exceed the information age deadline. S108. At the current sampling time, the first control input in the local control sequence data is used to drive the UAV to perform the coverage task, and S106 is re-executed at the next sampling time to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint.

2. The multi-UAV cooperative coverage method based on weighted Voronoi assignment according to claim 1, characterized in that, Based on the heterogeneous capabilities of UAVs and the importance of environmental space, a weighted power Voronoi assignment model is constructed to generate adaptive spatial partitioning data, including: Obtain the site weight data for each drone, whereby the site weight data represents the heterogeneous capabilities of the drone; Obtain environmental weight data for each spatial location point within the operating domain, wherein the environmental weight data represents the spatial importance or obstacle distribution of the spatial location point; Based on the site weight data, a weighted distance is defined between the UAV and the spatial location point. The weighted distance is the square of the Euclidean distance between the nominal position of the UAV and the spatial location point minus the site weight. Based on the environmental weight data, the weighted centroid and cost functional of each weighted Voronoi unit are calculated, so that the environmental weight does not participate in the calculation of the weighted distance, but only affects the update of the UAV's nominal position through the weighted centroid and cost functional. Based on the weighted distance, the free space in the mission space is divided into a set of weighted Voronoi units that correspond one-to-one with the UAV, generating the adaptive space partitioning data. The free space is the remaining area of ​​the mission space after deducting obstacles and their safety expansion regions. Each weighted Voronoi unit consists of all spatial location points in the free space that satisfy the minimum weighted distance condition.

3. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 1, characterized in that, The process of discretizing the free space in the task space into multiple spatial units, establishing information age state variables and their recursive dynamic models for each spatial unit, and generating time-series constraint data based on the information age cutoff date and priority weights includes: For each spatial location point within the weighted Voronoi unit, an information age state variable is defined, which is used to quantify the timeliness of the information at the spatial location point. Establish a discrete-time recursive equation for the information age state variable. The recursive equation resets the information age state variable when the UAV visits the corresponding spatial location point; otherwise, it increments the information age state variable by the discrete time step. Based on task requirements, information age cutoff periods are set for spatial location points of different priorities, and the time-series constraint data is generated. The time-series constraint data includes the information age cutoff periods and the corresponding priority weights. Among them, the information age state variable is according to A k ( t +1)=(1- χ k ( t ))( A k ( t +1) recursion, χ k ( t ) is the access indicator function, which is used when any UAV is at the sampling time. t Access the corresponding space unit g k hour χ k ( t )=1, otherwise χ k ( t =0; the timing constraint data also includes a revisit interval constraint determined by two adjacent access times, making the spatial unit... g k The difference between two consecutive access times does not exceed the corresponding information age cutoff period. τ k .

4. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 1, characterized in that, The process of constructing a distributed rolling time-domain control optimization problem based on the adaptive spatial partitioning data and the temporal constraint data, and generating local control sequence data, includes: At each sampling moment, each UAV establishes a distributed rolling time-domain control optimization problem with the goal of minimizing the weighted information age penalty function in the prediction time domain. The weighted information age penalty function is the sum of the products of the priority weights and the convex function of the information age state variable. In the distributed rolling time-domain control optimization problem, the recursive dynamics of the information age state variable, the state dynamics of the UAV, control input constraints, state constraints, and the information age cutoff period are used as constraints, and a neighbor coordination term is introduced to maintain distributed coordination. The neighbor coordination term is a positive definite term and satisfies the decreasing property over time. It is used to constrain at least one of the following: power Voronoi boundary consistency, coverage area overlap, minimum safe distance, or communication connectivity of neighboring UAVs. Solving the distributed rolling time-domain control optimization problem yields the local control sequence data, which includes the control input at each time point in the prediction time domain.

5. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 4, characterized in that, After the step of using the first control input in the local control sequence data to drive the UAV to perform the coverage task and re-execute S106 at the next sampling time, the method further includes: In the distributed rolling time-domain control optimization problem, terminal state constraints, terminal invariant sets, terminal controllers, and terminal cost functions are configured. By shifting the feasible control sequence of the previous sampling time and continuing the terminal controller, candidate feasible solutions for the next sampling time are constructed, so that the distributed rolling time-domain control optimization problem of adjacent sampling times remains recursively feasible, generating stability guarantee data. The stability guarantee data includes recursive feasibility conclusions and input to state stability conclusions. By combining the adaptive spatial partitioning data, the temporal constraint data, the local control sequence data, and the stability guarantee data, a time-series perception-based distributed coverage control for heterogeneous UAV formations in a dynamic and uncertain environment is achieved.

6. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 5, characterized in that, The step of using the first control input in the local control sequence data to drive the UAV to perform a coverage task at the current sampling time, and re-executing S106 at the next sampling time to achieve coverage and revisiting of the task space while satisfying the timing deadline constraint, includes: At the current sampling time, the control input corresponding to the current time in the local control sequence data is applied to the UAV, driving the UAV to perform a coverage task in the corresponding weighted Voronoi unit; At the next sampling moment, the current state information is reacquired and the distributed rolling time-domain control optimization problem is solved again to update the local control sequence data, thereby forming a closed-loop rolling time-domain control strategy, enabling the UAV to cover and revisit the task space while satisfying the timing deadline constraint.

7. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 5, characterized in that, The process involves configuring terminal state constraints, terminal invariant sets, terminal controllers, and terminal cost functions in the distributed rolling time-domain control optimization problem, and constructing candidate feasible solutions for the next sampling time by shifting the feasible control sequence of the previous sampling time and continuing the terminal controller, thereby ensuring the recursive feasibility of the distributed rolling time-domain control optimization problem between adjacent sampling times and generating stability-guaranteed data, including: For each UAV, a terminal invariant set and a terminal controller are set, wherein the terminal invariant set is a compact set of states that remain unchanged under the action of the terminal controller; The terminal cost function is set as the control Lyapunov function to ensure that the terminal cost function monotonically decreases along the system trajectory within the terminal invariant set. Provided that the terminal invariant set satisfies the compact control invariance and the terminal cost function satisfies the control Lyapunov function descent condition, the feasible control sequence of the previous sampling time is shifted forward and connected to the terminal controller at the end of the prediction time domain to construct the candidate feasible control sequence of the next sampling time. This ensures that the distributed rolling time domain control optimization problem remains recursively feasible between adjacent sampling times and generates the stability guarantee data, which includes recursive feasibility conclusions and input to state stability conclusions.

8. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 5, characterized in that, The method integrates the adaptive spatial partitioning data, the temporal constraint data, the local control sequence data, and the stability guarantee data to achieve time-aware distributed coverage control of heterogeneous UAV formations in dynamic and uncertain environments, including: The drones determine their respective coverage sub-regions based on the adaptive spatial partitioning data; The time-series constraint data is used to ensure timely revisiting of high-priority regions. Specific flight control is executed based on the local control sequence data, and based on the stability guarantee data, the input-to-state stability and timing deadline constraints are satisfied when bounded disturbances exist, and the timing deadline constraints are satisfied with a probability not lower than a preset confidence level when sub-Gaussian disturbances exist.

9. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 2, characterized in that, The weighted power Voronoi assignment model is dynamically optimized through a Lloyd-type iterative update mechanism, which includes: Calculate the weighted centroid of the spatial location point within each weighted Voronoi cell. The weighted centroid is the weighted average position of all spatial location points within the weighted Voronoi cell, with the environmental weight as the weight. Update the nominal position of the drone to the weighted centroid; The iterative update mechanism is repeated until the nominal position of the UAV converges to the weighted centroid power Voronoi configuration. During the iteration of the iterative update mechanism, the cost functional is monotonically non-increasing. The cost functional is the sum of the integrals of the squared Euclidean distances from all spatial position points within the weighted Voronoi cells to the corresponding nominal position of the UAV, weighted by the environmental weights.

10. The multi-UAV cooperative coverage method based on weighted Voronoi allocation according to claim 1, characterized in that, When the UAV is subjected to a sub-Gaussian random perturbation, the time-series constraint data is tightened through a probability cutoff guarantee mechanism, which includes: Obtain the variance proxy data of the sub-Gaussian random perturbation; Based on the preset confidence risk parameters δ and probability guarantee time range T Calculate the safety margin data; the specific calculation method for the safety margin data is as follows: ,in, σ k For the variance proxy σ k The square root of ² is called the standard deviation, and the probability is guaranteed over a time range. T With the prediction time domain of the distributed rolling time-domain control optimization problem H Each setting is independent of the others; exist τ k - β k Under the condition that >0, the information age cutoff period is tightened to a tightened cutoff period, where the tightened cutoff period is the original cutoff period. τ k Subtract the aforementioned safety margin data β k ,Right now τ k tight = τ k - β k ; In the distributed rolling time-domain control optimization problem, the tightened deadline is used to replace the original deadline as a hard constraint. Based on the tail inequality and joint bound principle of sub-Gaussian random variables, it is guaranteed that within the time range of the probability guarantee... T Within this period, the information age status variable does not exceed the original cutoff period. τ k The probability is not less than 1- δ .

Images