Multi-uav flight path generation method and generation system thereof

Abstract

The application relates to the technical field of unmanned aerial vehicle flight path optimization, in particular to a multi-unmanned aerial vehicle flight path generation method and a generation system thereof, which comprises the following steps: step 1: taking a base station as an origin, establishing a three-dimensional space coordinate system, connecting all unmanned aerial vehicles with the base station signal, and transmitting state information to the base station; and based on the state information transmitted by each unmanned aerial vehicle, the base station renders the positions of the unmanned aerial vehicles and environmental obstacle information in the three-dimensional space coordinate system; the multi-unmanned aerial vehicle flight path generation method and system provided by the application effectively overcome the defects that the existing single particle swarm algorithm is easy to fall into a local optimal solution, and significantly improve the overall performance and robustness of the multi-unmanned aerial vehicle cooperative path planning.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) flight path optimization technology, and more specifically, to a method and system for generating flight paths for multiple UAVs. Background Technology

[0002] The content in this section provides only background information related to this application and may not constitute prior art.

[0003] With the rapid development of drone technology, single drones are gradually showing limitations in complex, large-scale mission scenarios, such as limited coverage, low mission execution efficiency, and insufficient fault tolerance. In contrast, multi-drone systems, through collaborative work, can effectively overcome these bottlenecks and accomplish tasks that single drones cannot. For example, in scenarios such as wide-area search and rescue, large-scale agricultural and forestry plant protection monitoring, urban three-dimensional security patrols, rapid disaster site assessment, and cluster logistics delivery, multiple drones can execute tasks in parallel, cooperate, cover a wider area, provide mutual information support, and possess stronger system robustness.

[0004] Currently, particle swarm optimization (PSO) is generally used to optimize the path planning of multiple UAVs. This method treats each possible UAV flight path (or its key parameters) as a "particle," and the entire particle swarm flies in a solution space defined by task constraints (such as obstacle avoidance, collision avoidance, endurance, and time windows) and objective functions (such as minimizing total flight time, minimizing total path length, and maximizing coverage efficiency). The particles dynamically adjust their "flight" direction and speed based on their individual and collective historical best positions. Through iterative search, the PSO selects the relatively optimal (or suboptimal) cooperative flight scheme that satisfies the constraints from a vast array of possible path combinations.

[0005] The core of the Particle Swarm Optimization (PSO) algorithm lies in constructing an objective function that comprehensively considers various performance indicators (such as path length, time, safety, and energy consumption), and relying on the algorithm to dynamically iterate and find the minimum or maximum value of this function. Essentially, PSO is a heuristic random search algorithm, and its search process is highly dependent on the initial population settings and parameter adjustments. When solving multi-UAV path optimization problems with complex constraints and multi-modal characteristics (multiple local optima), the algorithm is very prone to prematurely converging to a local optimum, failing to find the global optimum. This means that the generated path solution may only be a relatively good, but not the best, among many path solutions, limiting further performance improvements in multi-UAV systems. Especially in applications with extremely high requirements for path efficiency and collaborative accuracy, the accuracy of PSO is not high. Summary of the Invention

[0006] The summary section of this application is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this application propose a method and system for generating flight paths for multiple unmanned aerial vehicles (UAVs) to address the technical problems mentioned in the background section above.

[0008] As a first aspect of this application, some embodiments of this application provide a method for generating flight paths for multiple unmanned aerial vehicles (UAVs), comprising the following steps: Step 1: Establish a three-dimensional spatial coordinate system with the base station as the origin. All drones are connected to the base station and transmit status information to the base station. Based on the status information transmitted by each drone, the base station renders the location of each drone and environmental obstacle information in the three-dimensional spatial coordinate system. Step 2: The base station loads the flight mission for each drone and generates K alternative paths for each drone based on the flight mission. Step 3: Each drone randomly selects one candidate path from K candidate paths as a particle. Multiple particles are randomly generated, and the genetic algorithm is used to iterate and select the best population. Step 4: Construct the first objective function based on the flight resource constraints of the UAV, including UAV energy consumption constraints, UAV operation constraints, and UAV collision constraints; Step 5: Construct a second objective function based on the base station's communication resource allocation. The communication resource allocation includes communication task constraints and sensing task constraints. Step 6: Iterate the optimal population using the ant colony algorithm with the first objective function and the second objective algorithm respectively to generate the Pareto solution set of the optimal path for each UAV.

[0009] The multi-UAV flight path generation method and system provided by this invention effectively overcomes the shortcomings of existing single particle swarm optimization algorithms, which are prone to getting trapped in local optima, by introducing a dual-objective collaborative optimization framework and a hybrid intelligent algorithm strategy. This significantly improves the overall performance and robustness of multi-UAV collaborative path planning. First, this application utilizes a genetic algorithm to conduct a global exploration in a vast space of candidate path combinations, selecting a potentially "optimal population." Then, it innovatively decouples the complex path optimization problem into two key dimensions—the UAV's flight resource constraints (energy consumption, operation, collision) and the base station's communication resource constraints (communication, sensing), and constructs independent objective functions for each. Finally, using an ant colony algorithm, which excels at handling multi-objective optimization and discrete path selection, a Pareto optimal solution set search is performed on the high-quality population obtained from the genetic algorithm for both objectives. This hybrid strategy of "global initial screening using genetic algorithm + dual-objective modeling + Pareto optimization using ant colony algorithm" not only significantly reduces the risk of the algorithm getting stuck in local optima and ensures the global superiority of the path scheme, but also takes into account the collaborative needs of individual UAV flight capability limitations and system-level communication resource allocation. Ultimately, it generates a set of Pareto-optimal flight path schemes that achieve the best balance in terms of flight efficiency, safety, energy consumption, and communication reliability, significantly improving the mission execution efficiency and collaborative reliability of multi-UAV systems in complex environments.

[0010] Furthermore, step 3 includes the following steps: Step 31: Each drone randomly selects one alternative path from the K alternative paths, as a particle. Randomly generate J particles to obtain the initial population IND, and set the fitness function fit; ; ; Where j represents the index of the particle, and J represents the number of particles in the initial population. Let i represent the j-th particle, I represent the total number of drones, and i represent the index of the drone. Let $k$ represent the alternative path randomly selected by particle $j for drone $i$, where $k \in {1, 2, ..., K}. Step 32: Calculate the fitness function value of each particle, and perform J rounds of roulette selection based on the fitness function value of each particle to obtain the parent population; Step 33: Pair up the parent populations in pairs, and assign each particle in the paired parent population a probability. The process involves polling, performing a crossover operation on selected particles, and a mutation operation on unselected particles.

[0011] Step 34: Repeat steps 32 to 33 to iterate until the maximum number of iterations is reached, and take the population from the last iteration as the optimal population.

[0012] Furthermore, the fitness function is `fit`. ; ; ; Where j represents the index of the particle, and J represents the total number of particles. This represents the fitness function value of the j-th particle. For particles The total cost calculated in the first objective function, For hyperparameters greater than 0, For a specific multi-machine collaborative mode, A set of path features, Represents particles Multi-machine collaborative mode The degree of correlation; Indicates the current iteration number; This indicates the multi-machine cooperative mode from the beginning of evolution to the cGA generation. The weighted total number of times an individual adopts and successfully passes on the technology; Indicates the inheritance decay factor. ; This is the current generation increment, among which For relative fitness; The importance weight of multi-machine collaborative mode 'a'.

[0013] Furthermore, step 32 includes the following steps: Step 321: Calculate the fitness function values ​​for all particles. Calculate the probability P of each particle being selected. j ; ; j represents the index of the particle, and J represents the total number of particles. This represents the fitness function value of the j-th particle; Step 322: The probability P of selecting all particles j Particles are mapped to a circle according to their size. The proportion of the area occupied within the circle is equal to the probability P. j ; Step 323: Randomly generate a point within the circle, select the particle that maps to the position of that point as a parent particle, and repeat this process for J points to select J parent particles, thus obtaining the parent population.

[0014] Furthermore, ; in, This is the maximum number of iterations for the genetic algorithm. This indicates the current iteration number.

[0015] Furthermore, for the crossover operation, the two pairs of paired particles x j and x j-1 Randomly select intersection point i0, and move particle x j The first i0 elements and particle x j-1 Swap the first i0 elements in the middle; For mutation operations, in particle x j Randomly select a mutation point i1, and then randomly select a new alternative path for the drone corresponding to mutation point i1 as the replacement.

[0016] Furthermore, step 6 includes the following steps: Step 61: Initialize the optimization parameters of the ant colony optimization algorithm, including the maximum number of iterations. Total number of ants Pheromones as heuristic factors Expected heuristic factors pheromone intensity value and pheromone evaporation factor ; Step 62: Using the optimal population obtained in Step 3, reconstruct the space of each particle in the population into a multi-stage supergraph G; Step 63: Based on the supergraph G, construct the pheromone matrix T and the heuristic matrix H; Initialize the pheromone matrix T: Initialize the pheromone concentration on each edge of the collaborative decision graph. For the same smaller positive constant ; ; , , These represent the weights of the first synergistic factor, the second synergistic factor, and the third synergistic factor, respectively. This indicates that the supernode has been reached. The path combination scheme determined at that time, This indicates the risk of a collision between any two drones in this path combination. When representing regional coverage, the redundant coverage area between paths. This indicates an estimate of the degree of competition among drones for communication spectrum and sensing resources; Step 64: Each ant crawls through the supergraph G based on the pheromone matrix T and the heuristic matrix H to obtain a path solution; Step 65: Calculate the first objective function value and the second objective function value of the path solution respectively; Step 66: Let Found for all ants of the current generation There are path solutions, each path solution Corresponding to a target vector For any two path solutions and Define the dominance relationship: Dominate If and only if, Denotes the first objective function. Describe the second objective function: (1) ,Right now Not inferior to in all objectives ; (2) Existence or That is, it is better to be stricter on at least one objective; Step 67: Sort all path solutions hierarchically according to Pareto non-dominance relations: Front1, Front2, ... Front FG Where Front1 is the optimal first non-dominated layer, Front2 is the second non-dominated layer dominated by Front1, and FG represents the total number of non-dominated layers; Step 68: Update the global pheromone based on the ants' movement results; ; Where e and d represent the row and column indices in the pheromone matrix and heuristic matrix, respectively. It is an indicator function; it is 1 when ant h passes through edge (e,d), and 0 otherwise. Let (e,d) be the information enhancement factor. Indicates the pheromone evaporation factor. This represents the pheromone level at (e,d) in the w-th iteration. Let represent the pheromone at (e, d) in the (w+1)th iteration, and h represent the index of the ant. Let represent the path solution traversed by the h-th ant; ; ; ; ; ; Let Q be the information enhancement factor for edge (e,d), and let Q represent the pheromone intensity value. Indicates the overall weight. Represents the path solution The cost of travel , These represent the weights of the first objective function and the weights of the second objective function, respectively. Denotes the first objective function. Describes the second objective function. Represents the path solution Total congestion Represents the path solution The non-dominated layer corresponding to the sorting, The rank indicates the sequence number of the non-dominated layer; W(rank) represents the hierarchical weight function. This represents the normalized congestion degree of each path solution. Indicates the weighting factor; Step 69: Repeat steps 64-68 to continuously update the Pareto solution set. During the iteration loop, determine whether the current iteration count has reached the preset maximum iteration count. If the path is not reached, return to step 64, and all ants begin a new path construction based on the updated pheromone matrix; if the path has been reached, perform a non-dominated sort on all V solutions of the final generation, and select Front1 as the output Pareto optimal solution set.

[0017] Furthermore, the first objective function : ; ; The intermediate objective function is... It is a negative constraint function. and These are the first constraint weight and the second constraint weight, respectively. It is an energy consumption item. It is a flight risk item; ; ; I represents the number of drones, and i represents the drone index. Let represent the energy consumption of the i-th drone. The energy consumption of a drone includes flight energy consumption and communication energy consumption. This is a reference value for the maximum energy consumption allowed for the entire drone swarm to complete its path mission. This represents the maximum battery capacity of a single drone. ; ; ; I represents the number of drones, i represents the drone index, N represents the number of waypoints, and n represents the waypoint index. This is the relative flight risk prediction for the i-th drone. This represents the maximum predicted relative flight risk for drone I. From the nth waypoint to the nth waypoint Flight risk prediction for each waypoint Represents environmental risk weighting factors. From the nth waypoint to the nth waypoint Environmental risk prediction for each waypoint Represents the high-altitude risk weighting factor. From the nth waypoint to the nth waypoint High-altitude risk prediction for each waypoint This represents the maximum risk value of a single drone on a path with n waypoints; Negative constraint function for: ; in, It is a corner constraint term; ; ; in, A negative penalty value. It is the maximum turning angle of the drone. For at the route node The steering angle, For the turning angle constraint at route node n, Number of waypoints For waypoint indexes, Let i represent the number of drones, and i represent the index of the drone. These are motion region constraints; ; ; Regional constraint information at route node n, It is a negative penalty value; These are obstacle avoidance constraints: ; ; Where v≠i, I is the total number of UAVs; n is the waypoint index, and N is the number of waypoints. It is the minimum safe distance for obstacle avoidance. This is the obstacle avoidance information for the i-th UAV at the n-th waypoint; It is the nth waypoint on the path of the i-th UAV and the nth waypoint on the path of the v-th UAV. The minimum Cartesian distance between waypoints This represents the waypoint index of the v-th UAV, and PS represents the conversion factor.

[0018] Furthermore, the second objective function ; ; in, For the first The utility function of communication sensing for unmanned aerial vehicles (UAVs). For the first The weighting factor of the drone utility function, where i is the drone index. Number of drones; The calculation is as follows: ; ; ; in, Resources allocated to the UAV's i-communication mission Resources allocated to the drone's sensing mission For the utility function of communication quality and effectiveness, The utility function for the quality and effectiveness of the sensing task; and They are , The signal-to-noise ratio.

[0019] The beneficial effects of this application are as follows: This invention jointly models the multi-UAV path planning problem and the resource allocation problem, employs Pareto optimization theory, and uses a non-dominated sorting algorithm to evaluate the path solutions in multiple dimensions. It can directly output a set of Pareto optimal solutions in a single run. Each solution in this set represents the optimal trade-off among multiple conflicting objectives such as flight energy consumption, collision risk, and mission timeliness, providing decision-makers with the possibility of flexible selection based on real-time mission requirements and overcoming the limitations of traditional single-objective optimization.

[0020] This invention introduces a congestion calculation mechanism to assign higher weights to unique solutions in sparse regions during path selection and pheromone updates, effectively preventing the algorithm from converging to local regions of the Pareto front. The final output solution set is evenly distributed on the front, covering various extreme trade-offs from "lowest energy consumption" to "lowest risk," meeting the differentiated needs of different task scenarios and ensuring the diversity of the solution set.

[0021] The first objective function of this invention incorporates multiple risks, including energy risk and flight risk, establishing a more accurate and intelligent multi-UAV path planning model and improving the efficiency of finding the optimal path. The design of this function ensures that the optimization algorithm can systematically search for the optimal path with the lowest energy consumption and least risk, while satisfying flight safety and mission rules, providing an accurate and reliable evaluation benchmark for the entire multi-objective optimization process.

[0022] The second objective function of this invention is a joint optimization objective design of communication and sensing utility. Compared with the traditional resource allocation method (independent allocation of communication and sensing), this invention can dynamically adjust resources such as power and bandwidth according to real-time channel status and task requirements, effectively improving the system's spectrum efficiency and energy efficiency, and further enhancing the system's scalability and robustness.

[0023] This invention innovatively designs a dynamic fitness function and an adaptive crossover probability mechanism based on traditional genetic algorithms. The dynamic fitness function accelerates the inheritance and convergence of superior genes, while the adaptive crossover probability is automatically adjusted according to the number of generations. In the early stages, a high crossover probability is maintained, which is beneficial for global exploration, while in the later stages, the crossover probability decreases, focusing on local development, effectively balancing the algorithm's exploration and development capabilities.

[0024] This invention employs an improved ant colony algorithm to achieve efficient collaborative path search. The pheromone update mechanism of the classic ant colony algorithm is extended from single-objective to multi-objective scenarios, and the pheromone increment formula is redesigned. By introducing a hierarchical weight factor, the Pareto front solution (first non-dominated layer) receives the highest pheromone enhancement weight, while the weight of the suboptimal layer gradually decreases. This retains the positive feedback search advantage of the ant colony algorithm while guiding the entire ant colony towards the Pareto front, achieving a multi-objective extension of the positive feedback mechanism. The improved pheromone update rule integrates convergence and diversity considerations. On the one hand, hierarchical weights ensure faster pheromone accumulation for high-quality solution paths; on the other hand, a crowding adjustment factor rewards unique solutions, avoiding premature convergence. This mechanism allows the algorithm to explore a broader trade-off space while maintaining a relatively fast convergence speed, obtaining a higher-quality, more evenly distributed Pareto solution set. For multi-UAV sequential decision-making scenarios, this invention designs an incremental collaborative index calculation method. When planning the current drone path, it is only necessary to calculate the collision risk, coverage overlap and resource competition with the already planned drones, which reduces the computational complexity and greatly improves the real-time performance of online planning.

[0025] This invention innovatively proposes a quantitative model for resource competition. By calculating the product of the number of overlapping resource blocks between any two UAVs and their spatial proximity, it simultaneously captures the impact of hard spectrum resource conflicts and electromagnetic wave propagation attenuation on cooperative efficiency. This model enables the algorithm to proactively avoid paths that would lead to dense clustering of UAVs in physical space and the use of the same frequency band, significantly reducing inter-UAV interference.

[0026] This invention reconstructs multi-UAV path planning into a multi-stage collaborative decision-making process, constructing a complete collaborative evaluation system that includes collision risk, coverage overlap, and resource competition, all integrated into a heuristic matrix. This design allows the ant colony algorithm to consider not only the merits of the path itself at each decision step, but also the overall synergy between the path and existing alternatives, thereby generating a globally optimal cluster path solution.

[0027] This invention systematically addresses the shortcomings of existing UAV swarm path planning methods in terms of collaboration, real-time performance, and flexibility through multi-objective Pareto optimization, improved genetic ant colony algorithm, resource contention modeling, and hierarchical architecture design, providing a reference technical solution for efficient collaborative task execution of multiple UAVs in complex environments. Attached Figure Description

[0028] Figure 1 This is a flowchart of a method for generating flight paths for multiple unmanned aerial vehicles (UAVs). Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments. The same reference numerals in the accompanying drawings represent the same components. It should be noted that the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the described embodiments of this application without creative effort are within the scope of protection of this application.

[0030] Compared to the embodiments shown in the accompanying drawings, feasible embodiments within the scope of this application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or components with different connections, etc. Furthermore, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0031] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” and similar terms used in this specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “upper” and “lower” are used only to indicate relative positional relationships, and these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0032] refer to Figure 1 Example 1: A method for generating flight paths for multiple unmanned aerial vehicles (UAVs), comprising the following steps: Step 1: Establish a three-dimensional spatial coordinate system with the base station as the origin. All drones are connected to the base station and transmit status information to the base station. Based on the status information transmitted by each drone, the base station renders the location of each drone and environmental obstacle information in the three-dimensional spatial coordinate system. Step 2: The base station loads the flight mission of each drone and generates K alternative paths for each drone based on the flight mission of each drone.

[0033] The alternative paths can be generated using Yen's algorithm to find the top K shortest paths, random path generation based on Rapid Random Tree Exploration (RRT), and by adjusting path cost weights (such as distance, risk, terrain complexity, etc.) to produce diverse paths. These alternative path sets will serve as input for online collaborative planning, allowing subsequent multi-objective optimization algorithms to select the best path. All the path generation algorithms used above are well-known techniques in the field of robot path planning.

[0034] Step 3: Each drone randomly selects one alternative path from the K alternative paths as a particle. Multiple particles are randomly generated, and the genetic algorithm is used to iterate and select the best population.

[0035] Step 3 includes the following steps: Step 31: Each drone randomly selects one alternative path from the K alternative paths, as a particle. Randomly generate J particles to obtain the initial population IND, and set the fitness function fit; ; ; Where j represents the index of the particle, and J represents the number of particles in the initial population. Let i represent the j-th particle, I represent the total number of drones, and i represent the index of the drone. Let $k$ represent the alternative path randomly selected by particle $j for drone $i$, where $k \in {1, 2, ..., K}. The fitness function is `fit`. ; ; ; in, This represents the fitness function value of the j-th particle. For particles The total cost calculated in the first objective function (proposed in step 4), For hyperparameters greater than 0, For a specific multi-machine collaborative mode, A set of path features, Represents particles Multi-machine collaborative mode The degree of correlation; Indicates the current iteration number; This indicates the multi-machine cooperative mode from the beginning of evolution to the cGA generation. The weighted total number of times the technology has been adopted and successfully passed on by outstanding individuals; Indicates the inheritance decay factor. ; This is the current generation increment, among which For relative fitness; Multi-machine collaborative mode Importance weights.

[0036] Step 32: Calculate the fitness function value of each particle, and perform J rounds of roulette selection based on the fitness function value of each particle to obtain the parent population.

[0037] Step 32 includes the following steps: Step 321: Calculate the fitness function values ​​for all particles. Calculate the probability P of each particle being selected. j ; ; Step 322: The probability P of selecting all particles j Particles are mapped to a circle according to their size. The proportion of the area occupied within the circle is equal to the probability P. j ; Step 323: Randomly generate a point within the circle, select the particle that maps to the position of that point as a parent particle, and repeat this process for J points to select J parent particles, thus obtaining the parent population.

[0038] Step 33: Pair up the parent populations in pairs, and assign each particle in the paired parent population a probability. The process involves polling, performing a crossover operation on selected particles, and a mutation operation on unselected particles.

[0039] ; in, This is the maximum number of iterations for the genetic algorithm. Indicates the current iteration number; For the crossover operation, the two pairs of particles x j and x j-1 Randomly select intersection point i0, and move particle x j The first i0 elements and particle x j-1 Swap the first i0 elements in the middle; For mutation operations, in particle x j Randomly select a mutation point i1, and then randomly select a new alternative path for the drone corresponding to mutation point i1 as the replacement.

[0040] Step 34: Repeat steps 32 to 33 to iterate until the maximum number of iterations is reached, and take the population from the last iteration as the optimal population.

[0041] Step 4: Construct the first objective function based on the flight resource constraints of the UAV, including UAV energy consumption constraints, UAV operation constraints, and UAV collision constraints; First objective function : ; ; The intermediate objective function is... It is a negative constraint function. and These are the first constraint weight and the second constraint weight, respectively. It is an energy consumption item. It is a flight risk item; ; ; I represents the number of drones, and i represents the drone index. Let represent the energy consumption of the i-th drone. The energy consumption of a drone includes flight energy consumption and communication energy consumption. This is a reference value for the maximum energy consumption allowed for the entire drone swarm to complete its path mission. This represents the maximum battery capacity of a single drone. ; ; ; I represents the number of drones, i represents the drone index, N represents the number of waypoints, and n represents the waypoint index. This is the relative flight risk prediction for the i-th drone. This represents the maximum predicted relative flight risk for drone I. From the nth waypoint to the nth waypoint Flight risk prediction for each waypoint Represents environmental risk weighting factors. From the nth waypoint to the nth waypoint Environmental risk prediction for each waypoint Represents the high-altitude risk weighting factor. From the nth waypoint to the nth waypoint High-altitude risk prediction for each waypoint This represents the maximum risk value of a single drone on a path with n waypoints; The environmental risk weighting factor Weighting factor for high-altitude risk To meet the preset adjustable parameters based on the task scenario. .

[0042] ; in, From the nth waypoint to the nth waypoint Environmental risk prediction for each waypoint For the nth waypoint to the nth waypoint The set of sampling points for the flight segment between waypoints. Sampling points Terrain risk value, Sampling points Threat risk value (calculated based on distance to the nearest threat source). Sampling points The meteorological risk value is given by l, where l represents the index of the sampling point and L represents the total number of sampling points. , , These are terrain weight, threat weight, and meteorological weight, respectively. The sampling point is the point between two waypoints; ; ; in, Sampling points The slope of the terrain, Sampling points Terrain risk value, This is the maximum permissible slope for drones. , Indicates sampling point Partial derivatives of coordinates, This represents the arctangent function.

[0043] like : ; like : ; like : .

[0044] in, Sampling points Threat risk value, This indicates the sampling point of the threat source t. distance, The minimum threat radius (absolute danger radius). The maximum threat radius (effective influence radius) is t, and the threat source t represents the pre-imported locations of each object. ; ; ; in, Sampling points Meteorological risk value, Sampling points wind speed, Sampling points Wind speed risk value, Visibility risk value, The maximum wind speed at which the drone can safely perform its mission. Sampling points Visibility, The minimum visibility required for safe drone flight was determined using wind speed information published by the meteorological bureau. ; ; in, From the nth waypoint to the nth waypoint High-altitude risk prediction for each waypoint For the nth waypoint to the nth waypoint The average altitude of the flight segment between waypoints Let n be the altitude of the nth waypoint. For the first The elevation of each waypoint This refers to the maximum permissible flight altitude for drones. ; This represents the maximum risk value of a single drone on a path with N waypoints, assuming that each flight segment simultaneously bears the maximum high-altitude risk (related to altitude ZH) and the maximum environmental risk, where ZH is the altitude. Negative constraint function for: ; in, It is a corner constraint term; ; ; in, A negative penalty value. It is the maximum turning angle of the drone. For at the route node The steering angle, For the turning angle constraint at route node n, Number of waypoints For waypoint indexes, Let i represent the number of drones, and i represent the index of the drone. These are motion region constraints; ; ; Regional constraint information at route node n, It is a negative penalty value; These are obstacle avoidance constraints: ; ; Where v≠i, I is the total number of UAVs; n is the waypoint index, and N is the number of waypoints. It is the minimum safe distance for obstacle avoidance. This is the obstacle avoidance information for the i-th UAV at the n-th waypoint; It is the nth waypoint on the path of the i-th UAV and the nth waypoint on the path of the v-th UAV. The minimum Cartesian distance between waypoints This represents the waypoint index of the v-th UAV, and PS represents the conversion factor; Step 5: Construct the second objective function based on base station communication resource allocation Communication resource allocation includes communication task constraints and sensing task constraints; ; in, For the first The utility function of communication sensing for unmanned aerial vehicles (UAVs). For the first Weighting factors for the utility function of unmanned aerial vehicles (UAVs); The calculation is as follows: ; ; ; in, Resources (including bandwidth and power) allocated to the drone's communication mission. Resources (including bandwidth and power) allocated to the drone's sensing mission. For the utility function of communication quality and effectiveness, For the utility function of the quality and effectiveness of the sensing task, and They are , The signal-to-noise ratio, where i is the drone index. This refers to the number of drones.

[0045] The objective is to maximize the second objective function, and the constraints are given: ; ; ; ; ; Where i is the drone index. For the number of drones, The downlink transmit power allocated to the i-th UAV, The maximum total transmit power allowed by the system. To be assigned to the The uplink receiving power of the drone. The maximum total received power allowed by the system. To be assigned to the The bandwidth of the drone The maximum total bandwidth allowed by the system. For the first The current amount of sensing data of the drone. For the first The minimum amount of sensing data that a drone needs to complete.

[0046] In ISAC resource allocation, the joint optimization of communication and sensing is considered. The goal is to maximize the joint utility of communication and sensing while meeting the minimum data rate (or total data volume) requirements of the sensing task. That is, to maximize the second objective function through the formula.

[0047] Step 6: Iterate the optimal population using the ant colony algorithm with the first objective function and the second objective algorithm respectively to generate the Pareto solution set of the optimal path for each UAV.

[0048] Step 6 includes the following steps: Step 61: Initialize the optimization parameters of the ant colony optimization algorithm, including the maximum number of iterations. Total number of ants Pheromones as heuristic factors Expected heuristic factors pheromone intensity value and pheromone evaporation factor ; Step 62: Using the optimal population obtained in Step 3, reconstruct the space of each particle in the population into a multi-stage supergraph G.

[0049] Specifically, from the optimal population (particle swarm) after step 3 iteration, each particle selects a super node SN for each drone i's alternative path. q Two adjacent super nodes are designated as super edges SE. q SE q The starting point is SN q SE q The endpoint is SN q+1All super edges are constructed into a super graph G; each graph node represents a "decision state" (i.e., a set of unselected path candidates for each UAV), and each edge represents a "decision action" (i.e., selecting a path for one or more UAVs).

[0050] Step 63: Based on the supergraph G, construct the pheromone matrix T and the heuristic matrix H; Initialize the pheromone matrix T: Initialize the pheromone concentration on each edge of the collaborative decision graph. For the same smaller positive constant ; Initialize the heuristic matrix H: Calculate the collaborative heuristic value corresponding to each edge in the collaborative decision graph. This value is determined by the collaborative heuristic function. Calculations ensure that decisions are made in a way that promotes overall collaboration. Corresponding to the state The immediate attractiveness of choosing action d, that is, the (partial) solution that will be formed after executing the decision action corresponding to that edge. The higher the value, the greater the potential for immediate synergy in that decision.

[0051] ; , , These represent the weights of the first synergistic factor, the second synergistic factor, and the third synergistic factor, respectively. This indicates that the supernode has been reached. The path combination scheme determined at that time, This indicates the risk of a collision between any two drones in this path combination. When representing regional coverage, the redundant coverage area between paths. This indicates an estimate of the degree of competition among drones for communication spectrum and sensing resources; ; ; ; in, This represents the risk of a collision between any two drones in this path combination, where i and v represent the indices of the drones, respectively. Indicates the time point of drone i Location, This indicates that the drone v is at a certain point in time. The position, where g represents the index of the time point. G0 represents the step size at a given time point, and G0 represents the total number of time points. To determine the time point based on the drone and v The instantaneous collision risk intensity is calculated from the predicted location. This indicates that drone i and drone v are at time points. Euclidean distance, This represents the minimum safe distance / safety threshold parameter. A function is selected for the maximum value, used for comparison. and And choose the larger of the two as the final value. This represents the formula for calculating Euclidean distance. ; When representing regional coverage, the redundant coverage area between paths. This represents the total area effectively covered. This represents the number of overlapping grid cells.

[0052] The calculation process is as follows: Discretize the continuous task region into a set of grids U with equal areas; For reaching the supernode The path combination determined by time For each UAV i in the dataset (including the path determined by the node for the current UAV), a subset of the grid that it can cover during the mission is mapped based on its flight path and the onboard sensor model. , .

[0053] Calculate the union of all single-machine coverage sets, and denote its cardinality (number of grids) as . This represents the total area effectively covered. For each covered grid cell... Count the number of drones covering it (count(u)). Then, the sum of the additional coverage counts for all grids. That is, the number of overlapping coverage grids. It quantifies the redundant consumption of sensing resources, where u represents the grid point index.

[0054] ; ; RB v and RB i Let V and i represent the sets of resource blocks required by drone v and drone i, respectively. To define the intensity of resource conflict between any two drones i and v, This indicates the number of overlapping resource blocks occupied by drone i and drone v; The estimated average distance between the two machines. This is the interference-sensitive distance threshold. The smaller the value, the lower the disturbance to the existing resource pool caused by introducing this path, and the better the synergy.

[0055] RB u ⊆{rb1,rb2, ..., rb Z Z represents the total system bandwidth B. total The total number of resource blocks divided into orthogonal groups, rb z This represents the z-th resource block; Resource blocks are based on the amount of UAV i-communication data D i Based on the required perception accuracy, a set of communication and perception resource blocks is allocated to all planned UAVs i according to a preset fixed strategy (such as the proportional fairness algorithm); Based on the expected communication load and perception task volume of the UAV on the candidate path, estimate the required resource block set RB. v .

[0056] Step 64: Each ant crawls through the supergraph G to obtain a path solution based on the pheromone matrix T and the heuristic matrix H.

[0057] Specifically: Ant starting point: Each ant starts from the initial decision state (all drones have not selected a path, only a candidate set exists); Multi-round decision-making: Ants make decisions in multiple rounds. In each round, the ant is in a certain decision state; the ant's decision state is determined by the state transition probability. Choosing an edge (i.e., a decision action) has a probability determined by the pheromone concentration on that edge. and heuristic values Joint decision: ; Where e and d represent the row and column indices of the pheromone matrix and heuristic matrix, respectively, A represents the pheromone heuristic factor, B represents the expected heuristic factor, and r represents the index of the edge chosen by the ant when it is in a certain decision state; Complete the solution: The ant performs the selected action (e.g., choosing path A for drone 2), and the state is updated accordingly. This process is repeated until, after several rounds of decision-making, the ant reaches the termination state—the paths of all drones have been uniquely determined. At this point, the ant has completed a complete cooperative path planning solution.

[0058] Step 65: Calculate the first objective function value and the second objective function value of the path solution respectively.

[0059] When all After all the ants have completed their own plans, we obtain a set containing... A population of schemes is established. For each scheme in the population, the first objective function defined in step 4 and the second objective function defined in step 5 are called for calculation. The value of the first objective function comprehensively reflects the scheme's energy consumption, risk, and degree of constraint violation; the smaller its value, the better the scheme. The second objective function reflects the maximization of the joint utility of all UAV communication and perception; the larger its value, the better the scheme.

[0060] Step 66: Let Found for all ants of the current generation There are path solutions, each path solution Corresponding to a target vector For any two path solutions and Define the dominance relationship: Dominate If and only if, Denotes the first objective function. Describe the second objective function: (1) ,Right now Not inferior to in all objectives ; (2) Existence or That is, it is better to be stricter on at least one objective.

[0061] Step 67: Sort all path solutions hierarchically according to Pareto non-dominated relations to obtain layers of Pareto fronts (Front1, Front2, ...), where Front1 is the optimal first non-dominated layer (Pareto optimal solution), Front2 is the second-best layer dominated by Front1, and so on. In step 67, a “crowding distance” is calculated for each solution within a non-dominated layer to characterize its density with neighboring solutions in the target space.

[0062] Specifically, for each path solution, the solutions are sorted according to the first objective function value and the second objective function value, respectively. Boundary solutions located at both ends of the objective are assigned infinitely large crowding distances to ensure their preservation; the crowding distance of intermediate solutions is accumulated by adding the normalized intervals of their adjacent solutions in each objective direction. A larger crowding distance indicates a sparser region where the solution is located, and thus a higher degree of uniqueness. During pheromone updates or path selection, the crowding distance serves as an important weighting factor, enabling the algorithm to uniformly explore the entire tradeoff space while approximating the Pareto front. The implementation is as follows: For each non-dominated layer Front, calculate the total congestion degree for each solution. ; The calculation method is as follows: first, sort the solutions of the non-dominated layer according to the function value of the first objective function, and then set the crowding degree of the boundary solutions: ; For the non-dominated layer Front o intermediate solution : ; in, This represents the number of solutions in the o-th non-dominated layer. and These are the maximum and minimum values ​​of the first objective function in the non-dominated layer. This represents the crowding level of the first objective function.

[0063] For the second objective function, repeat The steps to obtain the results are as follows: ,Will and Add them together to calculate the total congestion for each solution. That is, add them together... The first objective function in the equation is replaced with the second objective function.

[0064] The greater the overall congestion, the further away the solution is from its neighbors in all objective dimensions, and the stronger its "uniqueness" and "irreplaceability".

[0065] Step 68: Update the global pheromone based on the ants' movement results; ; Where e and d represent the row and column indices in the pheromone matrix and heuristic matrix, respectively. It is an indicator function; it is 1 when ant h passes through edge (e,d), and 0 otherwise. Let (e,d) be the information enhancement factor. Indicates the pheromone evaporation factor. This represents the pheromone level at (e,d) in the w-th iteration. Let (e, d) represent the pheromone at (e, d) in the (w+1)th iteration, and h represent the index of the ant. Let represent the path solution traversed by the h-th ant; ; ; ; ; ; Let Q be the information enhancement factor for edge (e,d), and let Q represent the pheromone intensity value. Indicates the overall weight. Represents the path solution The cost of travel , These represent the weights of the first objective function and the weights of the second objective function, respectively. Denotes the first objective function. Describes the second objective function. Represents the path solution Total congestion Represents the path solution The non-dominated layer corresponding to the sorting, The rank indicates the sequence number of the non-dominated layer; W(rank) represents the hierarchical weight function. This represents the normalized congestion degree of each path solution. Indicates the weighting factor; Step 69: Repeat steps 64-68 to continuously update the Pareto solution set. During the iteration loop, determine whether the current iteration count has reached the preset maximum iteration count. If the goal is not achieved, return to step 64, where all ants begin a new path construction based on the updated pheromone matrix. If the goal is achieved, perform a non-dominated sort on all V solutions of the final generation, and select Front1 as the output Pareto optimal solution set. These solutions constitute the Pareto optimal front, representing the set of best trade-offs achievable among multiple conflicting objectives.

[0066] Example 2: Multi-UAV flight path generation system, which uses the multi-UAV flight path generation method described in Example 1 to allocate tasks for multiple UAVs.

[0067] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for generating flight paths for multiple unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Step 1: Establish a three-dimensional spatial coordinate system with the base station as the origin. All drones are connected to the base station and transmit status information to the base station. Based on the status information transmitted by each drone, the base station renders the location of each drone and environmental obstacle information in the three-dimensional spatial coordinate system. Step 2: The base station loads the flight mission for each drone and generates K alternative paths for each drone based on the flight mission. Step 3: Each drone randomly selects one candidate path from K candidate paths as a particle. Multiple particles are randomly generated, and the genetic algorithm is used to iterate and select the best population. Step 4: Construct the first objective function based on the flight resource constraints of the UAV, including UAV energy consumption constraints, UAV operation constraints, and UAV collision constraints; Step 5: Construct a second objective function based on the base station's communication resource allocation. The communication resource allocation includes communication task constraints and sensing task constraints. Step 6: Iterate the optimal population using the ant colony algorithm with the first objective function and the second objective function respectively to generate the Pareto solution set of the optimal path for each UAV; Step 6 includes the following steps: Step 61: Initialize the optimization parameters of the ant colony optimization algorithm, including the maximum number of iterations. Total number of ants Pheromones as heuristic factors Expected heuristic factors pheromone intensity value and pheromone evaporation factor ; Step 62: Using the optimal population obtained in Step 3, reconstruct the space of each particle in the population into a multi-stage supergraph G; Step 63: Based on the supergraph G, construct the pheromone matrix T and the heuristic matrix H; Initialize the pheromone matrix T: Initialize the pheromone concentration on each edge of the collaborative decision graph. For the same smaller positive constant ; ； , , These represent the weights of the first synergistic factor, the second synergistic factor, and the third synergistic factor, respectively. This indicates that the supernode has been reached. The path combination scheme determined at that time, This indicates the risk of a collision between any two drones in this path combination. When representing regional coverage, the redundant coverage area between paths. This indicates an estimate of the degree of competition among drones for communication spectrum and sensing resources; Step 64: Each ant crawls through the supergraph G based on the pheromone matrix T and the heuristic matrix H to obtain a path solution; Step 65: Calculate the first objective function value and the second objective function value of the path solution respectively; Step 66: Let Found for all ants of the current generation There are path solutions, each path solution Corresponding to a target vector For any two path solutions and Define the dominance relationship: Dominate If and only if, Denotes the first objective function. Describe the second objective function: (1) ,Right now Not inferior to in all objectives ; (2) Existence or That is, it is better to be stricter on at least one objective; Step 67: Sort all path solutions hierarchically according to Pareto non-dominance relations: Front1, Front2, ... Front FG Where Front1 is the optimal first non-dominated layer, Front2 is the second non-dominated layer dominated by Front1, and FG represents the total number of non-dominated layers; Step 68: Update the global pheromone based on the ants' movement results; ; Where e and d represent the row and column indices in the pheromone matrix and heuristic matrix, respectively. It is an indicator function; it is 1 when ant h passes through edge (e,d), and 0 otherwise. Let (e,d) be the information enhancement factor. Indicates the pheromone evaporation factor. This represents the pheromone level at (e,d) in the w-th iteration. Let represent the pheromone at (e, d) in the (w+1)th iteration, and h represent the index of the ant. Let represent the path solution traversed by the h-th ant; ； ； ； ； ； Let Q be the information enhancement factor for edge (e,d), and let Q represent the pheromone intensity value. Indicates the overall weight. Represents the path solution The cost of travel , These represent the weights of the first objective function and the weights of the second objective function, respectively. Denotes the first objective function. Describes the second objective function. Represents the path solution Total congestion Represents the path solution The non-dominated layer corresponding to the sorting, The rank indicates the sequence number of the non-dominated layer; W(rank) represents the hierarchical weight function. This represents the normalized congestion degree of each path solution. Indicates the weighting factor; Step 69: Repeat steps 64-68 to continuously update the Pareto solution set, iterate, and determine whether the current iteration count has reached the preset maximum iteration count. If the path is not reached, return to step 64, and all ants begin a new path construction based on the updated pheromone matrix; if the path is reached, perform a non-dominated sort on all V solutions of the final generation, and select Front1 as the output Pareto optimal solution set. 2.The method of claim 1, wherein, Step 3 includes the following steps: Step 31: each UAV randomly selects 1 candidate path from K candidate paths as a particle ; randomly generate J particles to obtain an initial population IND, and set a fitness function fit; ； ； wherein j represents the index of the particle, J represents the number of particles in the initial population, represents the jth particle, I represents the total number of UAVs, i represents the index of the UAV, represents the candidate path randomly selected by the particle j for the UAV i, k e {1, 2,... K}; Step 32: Calculate the fitness function value of each particle, and perform J rounds of roulette selection based on the fitness function value of each particle to obtain the parent population; Step 33: Pair the parent population two by two, and pair each particle in the paired parent population with a probability of 0.5 Polling, polling selected particles to perform crossover operation, and polling unselected particles to perform mutation operation; Step 34: Repeat steps 32 to 33 to iterate until the maximum number of iterations is reached, and take the population from the last iteration as the optimal population. 3.The method of claim 2, wherein, The fitness function is `fit`. ； ； ; Where j represents the index of the particle, and J represents the total number of particles. This represents the fitness function value of the j-th particle. For particles The total cost calculated in the first objective function, For hyperparameters greater than 0, For a specific multi-machine collaborative mode, A set of path features, Represents particles Multi-machine collaborative mode The degree of correlation; Indicates the current iteration number; This indicates the multi-machine cooperative mode from the beginning of evolution to the cGA generation. The weighted total number of times an individual adopts and successfully passes on the technology; Indicates the inheritance decay factor. ; This is the current generation increment, among which For relative fitness; The importance weight of multi-machine collaborative mode 'a'. 4.The method of claim 3, wherein, Step 32 includes the following steps: Step 321 : Calculate the fitness function value of all particles , calculate the probability P of each particle being selected j ; ； j denotes the index of the particle, J denotes the total number of particles, Fjdenotes the fitness function value of the jth particle; Step 322: Calculate the probability P that all particles are selected j Particles are mapped into a circle according to their size The area proportion in the circle equals the probability P j ; Step 323: Randomly generate a point within the circle, select the particle that maps to the position of that point as a parent particle, and repeat this process for J points to select J parent particles, thus obtaining the parent population.

5. The multi-UAV flight path generation method according to claim 2, characterized in that, ； wherein, is the maximum number of iterations of the genetic algorithm, denotes the current iteration number. 6.The method of claim 2, wherein, For the crossover operation, the two pairs of particles x j and x j-1 Randomly select intersection point i0, and move particle x j The first i0 elements and particle x j-1 Swap the first i0 elements in the middle; For mutation operations, in particle x j Randomly select a mutation point i1, and then randomly select a new alternative path for the drone corresponding to mutation point i1 as the replacement.

7. The multi-UAV flight path generation method according to claim 3, characterized in that, first objective function : ； ； is an intermediate objective function, is a negative constraint function, and are a first constraint weight and a second constraint weight, respectively, is an energy expenditure term, is a flight risk term; ； ； I represents the number of drones, and i represents the drone index. Let represent the energy consumption of the i-th drone. The energy consumption of a drone includes flight energy consumption and communication energy consumption. This is a reference value for the maximum energy consumption allowed for the entire drone swarm to complete its path mission. This represents the maximum battery capacity of a single drone. ； ； ； I represents the number of drones, i represents the drone index, N represents the number of waypoints, and n represents the waypoint index. This is the relative flight risk prediction for the i-th drone. This represents the maximum predicted relative flight risk for drone I. From the nth waypoint to the nth waypoint Flight risk prediction for each waypoint Represents environmental risk weighting factors. From the nth waypoint to the nth waypoint Environmental risk prediction for each waypoint Represents the high-altitude risk weighting factor. From the nth waypoint to the nth waypoint High-altitude risk prediction for each waypoint This represents the maximum risk value of a single drone on a path with n waypoints; Negative constraint function for: ； in, It is a corner constraint term; ； ； in, A negative penalty value. It is the maximum turning angle of the drone. For at the route node The steering angle, For the turning angle constraint at route node n, Number of waypoints For waypoint indexes, Let i represent the number of drones, and i represent the index of the drone. These are motion region constraints; ； ； Regional constraint information at route node n, It is a negative penalty value; These are obstacle avoidance constraints: ； ； Where v≠i, I is the total number of UAVs, and n is the waypoint index. Number of waypoints It is the minimum safe distance for obstacle avoidance. This is the obstacle avoidance information for the i-th UAV at the n-th waypoint; It is the nth waypoint on the path of the i-th UAV and the nth waypoint on the path of the v-th UAV. The minimum Cartesian distance between waypoints This represents the waypoint index of the v-th UAV, and PS represents the conversion factor.

8. The method for generating flight paths for multiple unmanned aerial vehicles according to claim 7, characterized in that, Second objective function ; ； in, For the first The utility function of communication sensing for unmanned aerial vehicles (UAVs). For the first The weighting factor of the drone utility function, where i is the drone index. Number of drones; The calculation is as follows: ； ； ； in, Resources allocated to the UAV's i-communication mission Resources allocated to the drone's sensing mission For the utility function of communication quality and effectiveness, The utility function for the quality and effectiveness of the sensing task; and They are , The signal-to-noise ratio.

9. A multi-UAV flight path generation system, characterized in that, The flight paths of each UAV are generated using the multi-UAV flight path generation method described in any one of claims 1 to 8.

Images