Multi-intelligent aircraft cooperative trajectory planning method based on mixed local wolf pack optimization

Abstract

The invention discloses a multi-intelligent aircraft cooperative trajectory planning method based on hybrid local wolf pack optimization, and the method comprises the steps: carrying out the comprehensive evaluation of candidate path solutions of all sub-populations of a current population of a multi-intelligent aircraft, and selecting a plurality of individuals with the optimal fitness as current leader individuals; performing path disturbance on the plurality of leader individuals, and determining whether to replace one or more of the plurality of leader individuals by adopting a probability criterion according to the current fitness change and temperature parameters of the plurality of leader individuals so as to improve the global search capability; for common individuals in each sub-population, updating path node positions of the common individuals under the guidance of a plurality of current leader individuals, and performing boundary legality correction on updated path nodes to generate a new candidate path solution; and based on the new candidate path solutions, extracting path information of current optimal individuals of all sub-populations. According to the invention, the technical problem of inaccurate collaborative trajectory planning of multiple intelligent aircrafts is solved.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) navigation, and in particular to a multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization. Background Art

[0002] With the continuous development of drone technology, multi-UAV systems have been widely used in military reconnaissance, disaster search and rescue, environmental monitoring, logistics and distribution, and other fields. Compared with a single UAV, the coordinated operation of multiple UAVs can significantly improve mission execution efficiency, coverage, and system robustness, and has become a hot topic in intelligent flight control and path planning research. In such systems, collaborative trajectory planning technology plays a key role. Its core goal is to plan a reasonable trajectory for each UAV while meeting mission constraints, flight performance, communication requirements, and environmental safety, thereby achieving global coordination and conflict avoidance.

[0003] A large number of research results have been published on collaborative trajectory planning for multiple UAVs, primarily including graph search-based methods (such as A and D), model predictive control (MPC)-based methods, and intelligent optimization algorithm-based methods. Intelligent optimization algorithms have garnered widespread attention in recent years due to their strong adaptability, minimal parameter dependency, and ease of scalability to complex, high-dimensional scenarios. Typical algorithms, such as genetic algorithms, particle swarm optimization, ant colony algorithms, differential evolution algorithms, and gray wolf optimization algorithms, have all been applied to UAV path planning and collaborative decision-making. The gray wolf optimization algorithm (GWO), in particular, exhibits promising application prospects in the fields of multi-objective optimization and path planning due to its simple structure, strong convergence, and high computational efficiency.

[0004] However, existing optimization algorithms still face limitations in the face of dynamic changes, incomplete information, and complex environments in real-world applications. On the one hand, traditional algorithms often fall into local optima when dealing with high-dimensional coupled state spaces due to insufficient population diversity and a lack of global search strategies, resulting in degraded planning performance. On the other hand, the coordination mechanisms between multiple UAVs are still imperfect, often leading to path conflicts, task duplication, and wasted communication resources, which seriously affect the overall performance of the system. Furthermore, existing methods need to be improved in their ability to cope with sudden threats, avoid moving obstacles, and maintain system connectivity in dynamic environments. Therefore, designing a multi-UAV trajectory planning method with stronger global search capabilities, higher coordination efficiency, and improved environmental adaptability has become a key issue that urgently needs to be overcome in this field.

[0005] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0006] An embodiment of the present invention provides a method for collaborative trajectory planning of multiple intelligent aircraft based on hybrid local wolf pack optimization, so as to at least solve the technical problem of inaccurate collaborative trajectory planning of multiple intelligent aircraft.

[0007] According to one aspect of an embodiment of the present invention, a method for collaborative trajectory planning of multiple intelligent aircraft based on hybrid local wolf pack optimization is provided, comprising: comprehensively evaluating the candidate path solutions of each subpopulation of the current population of the multiple intelligent aircraft, and selecting multiple individuals with the best fitness as current leading individuals; performing path perturbations on the multiple leading individuals respectively, and adopting a probability criterion to decide whether to replace one or more of the multiple leading individuals based on the current fitness changes and temperature parameters of the multiple leading individuals, so as to improve the global search capability; for ordinary individuals in each subpopulation, updating their path node positions based on the current multiple leading individuals, and performing boundary legitimacy correction on the updated path nodes to generate new candidate path solutions; based on the new candidate path solutions, extracting the path information of the current optimal individuals of all subpopulations to plan the collaborative trajectory of the multiple intelligent aircraft.

[0008] According to another aspect of an embodiment of the present invention, a multi-intelligent aircraft collaborative trajectory planning device based on hybrid local wolf pack optimization is also provided, including: an evaluation module, configured to perform a comprehensive evaluation of the candidate path solutions of each subpopulation of the current population, and select multiple individuals with the best fitness as the current leading individuals; a perturbation module, configured to perform path perturbations on multiple leading individuals respectively, and use a probability criterion to decide whether to replace one or more of the multiple leading individuals based on the current fitness changes and temperature parameters of the multiple leading individuals, so as to improve the global search capability; an update module, configured to update the path node positions of ordinary individuals in each subpopulation according to the current multiple leading individuals, and perform boundary legitimacy correction on the updated path nodes to generate new candidate path solutions; a planning module, configured to extract the path information of the current optimal individuals of all subpopulations based on the new candidate path solutions, so as to plan the collaborative trajectory of multiple intelligent aircraft.

[0009] In an embodiment of the present invention, a comprehensive evaluation is performed on the candidate path solutions of each subpopulation of the current population of multi-intelligent aircraft, and multiple individuals with the best fitness are selected as the current leading individuals; the paths of the multiple leading individuals are perturbed, and a probability criterion is used to determine whether to replace one or more of the multiple leading individuals based on the current fitness changes and temperature parameters of the multiple leading individuals, so as to improve the global search capability; the path node positions of the ordinary individuals in each subpopulation are updated based on the guidance of the current multiple leading individuals, and the boundary legitimacy of the updated path nodes is corrected to generate a new candidate path solution; based on the new candidate path solution, the path information of the current best individuals of all subpopulations is extracted to plan the collaborative trajectory of the multi-intelligent aircraft. The above solution solves the technical problem of inaccurate collaborative trajectory planning of multi-intelligent aircraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0011] Figure 1 is a flowchart of an optional multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization according to an embodiment of the present invention;

[0012] Figure 2 is a flowchart of another optional multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization according to an embodiment of the present invention;

[0013] Figure 3 is an optional grey wolf algorithm hierarchical diagram according to an embodiment of the present invention;

[0014] Figure 4 is a schematic diagram of an optional hunting-driving position update of a gray wolf optimization algorithm according to an embodiment of the present invention;

[0015] Figure 5 is a schematic diagram of an optional path height cost according to an embodiment of the present invention;

[0016] Figure 6 is an optional obstacle collision area map according to an embodiment of the present invention;

[0017] Figure 7 is an optional convergence factor comparison diagram according to an embodiment of the present invention;

[0018] Figure 8 is a flowchart of another optional multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization according to an embodiment of the present invention;

[0019] Figure 9 is an optional autonomous obstacle avoidance three-dimensional rendering according to an embodiment of the present invention;

[0020] Figure 10 This is an optional three-dimensional bird's-eye view rendering of autonomous obstacle avoidance according to an embodiment of the present invention;

[0021] Figure 11 1 is a schematic structural diagram of an optional multi-intelligent aircraft collaborative trajectory planning device based on hybrid local wolf pack optimization according to an embodiment of the present invention;

[0022] Figure 12 A schematic structural diagram of an electronic device suitable for implementing the embodiments of the present disclosure is shown;

[0023] The parameters have the following meanings:

[0024] α, leader gray wolf; β, deputy leader gray wolf; δ, follower gray wolf; ω, ordinary gray wolf; D α , the distance vector from the common individual to α; D β , the distance vector from the common individual to β; D δ , the distance vector from the common individual to δ; H max , Maximum safe flight altitude; H min , minimum safe flight altitude; H ij , the actual flight height of the i-th UAV at trajectory node j; The coordinates of the kth track node of the i-th UAV along the j-th track plan; The coordinates of the k+1th track node of the i-th UAV along the j-th track plan; Coordinates and Coordinates of the point closest to the threat in the coordinate line track; R0, threat radius; Z, safety redundancy zone radius; E, escape zone radius; t, number of iterations; a(t), convergence factor. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] According to an embodiment of the present invention, a method embodiment of a multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] Figure 1 is a multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization according to an embodiment of the present invention, such as Figure 1 As shown, the method includes the following steps:

[0029] Step S102 : comprehensively evaluate the candidate path solutions of each subpopulation of the current population of multiple intelligent aircraft, and select multiple individuals with the best fitness as current leading individuals.

[0030] First, mission parameters are read, wherein the mission parameters include at least one of the following: the number of intelligent aerial vehicles, a flight speed range, a maximum flight distance, coordinates of the starting and ending points, coordinates of the threat area, boundaries of the no-fly zone, upper and lower altitude limits, and a safe distance; based on the mission parameters and the three-dimensional space modeling results, a path search space and constraint rules are determined; heuristic rules are combined with random strategies to generate a plurality of candidate path solutions that meet the constraint rules in the search space to form an initial population as the current population, and the current population is divided into a corresponding number of subpopulations according to the number of the multiple intelligent aerial vehicles.

[0031] Next, the flight cost of each candidate path solution is calculated, where the flight cost includes at least one of the following: flight energy consumption cost, flight altitude deviation cost, threat zone crossing risk, track curvature smoothness, arrival time error, and minimum distance conflict cost with other intelligent aircraft; the flight costs are weighted and combined to obtain a single evaluation index as the comprehensive fitness of the individual; the comprehensive fitness is sorted from best to worst, and the three individuals with the best, second best, and third best fitness in the current subpopulation are selected as the multiple leader individuals.

[0032] In some embodiments, the flight cost of each candidate path solution is calculated separately, including at least one of the following: estimating the flight energy consumption cost based on the total length of the path segment and the energy consumption coefficient; calculating the flight altitude cost based on the degree of deviation between the flight altitude of each path segment and the maximum and minimum flight altitudes; scoring based on whether the track point enters the threat area and the distance from the threat edge to obtain the threat avoidance cost; evaluating the path smoothness based on the angle or curvature change of adjacent track segments; calculating the time synchronization cost based on the error between the time when the intelligent aircraft arrives at the target point and the instruction time; and calculating the conflict cost by comparing the minimum distance between multiple intelligent aircraft at any time with the safety distance.

[0033] Step S104 , performing path perturbations on the plurality of leader individuals respectively, and using a probability criterion to decide whether to replace one or more of the plurality of leader individuals based on the current fitness changes and temperature parameters of the plurality of leader individuals, so as to improve the global search capability.

[0034] For example, disturbances are applied to the path node positions of multiple leading individuals respectively to generate multiple new candidate path solutions; the fitness of the new candidate path solutions is re-evaluated and compared with the fitness of the candidate path solutions to obtain a fitness difference; the acceptance probability is calculated based on the fitness difference and the temperature parameter, and based on the acceptance probability, it is determined whether to adopt the individual corresponding to the new candidate path solution as the new leading individual.

[0035] Step S106: for ordinary individuals in each subpopulation, the path node positions are updated according to the guidance of the current plurality of leader individuals, and boundary legitimacy correction is performed on the updated path nodes to generate new candidate path solutions.

[0036] For example, for ordinary individuals in each subpopulation, the position difference vector between them and the multiple leader individuals is calculated respectively; based on the nonlinear convergence factor of the current iteration and the position difference vector, the ordinary individuals are guided to move toward the multiple leader individuals to update their path node positions; boundary checking and legitimacy correction are performed on the updated path nodes so that the updated path nodes meet physical and task constraints.

[0037] Step S108: Based on the new candidate path solution, the path information of the current optimal individuals of all subpopulations is extracted to plan the collaborative trajectory of the multiple intelligent aircraft.

[0038] The embodiment of the present invention provides a multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization to solve the problems of poor planning path quality, slow search speed and easy falling into local optimality in the process of solving multi-UAV trajectory planning. It is a multi-UAV trajectory planning method with stronger global search capability, higher collaborative efficiency and better environmental adaptability, which effectively improves the speed and accuracy of multi-UAV optimization and can more accurately lock the global optimal solution during trajectory optimization.

[0039] The embodiment of the present invention addresses the problem that the path planning of multiple UAVs in a complex three-dimensional environment is prone to falling into local optimality and slow convergence, and also provides a multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization. The flow chart of this method is as follows Figure 2 As shown, the following steps are included:

[0040] Step S202: task constraint analysis and search space initialization.

[0041] First, all mission-related drone parameters are read from the input system, including maximum range, speed range, payload capacity, altitude limit, and mission constraints (such as minimum flight time and maximum turn angle). These parameters serve as inputs to the initialization of the Grey Wolf optimization algorithm, ensuring that the generated population meets the specific requirements of the mission. Next, the population is initialized based on the set population size, ensuring that a wider range of solution space is covered throughout the search process, thereby improving global search capabilities.

[0042] The Gray Wolf Optimization (GWO) algorithm mimics the social hierarchy and predation behavior patterns of gray wolf populations to optimize search. In the GWO algorithm, wolves are divided into four levels: α wolf, β wolf, δ wolf, and ω wolf. Among them, α is the wolf king, who is mainly responsible for various decisions of the group; β wolf assists α wolf and makes decisions; δ wolf obeys the instructions of α wolf and β wolf and can manage ω wolf. Gray wolves rely on this hierarchy to forage and hunt. Figure 3 shown.

[0043] The initial population is set not only by randomly generated individual positions but also by considering the specific constraints of the UAV mission. The size, distribution, and initialization of the population will affect the subsequent optimization efficiency and global search capabilities.

[0044] Step S204: Divide the heterogeneous sub-populations and establish a parallel evolution framework.

[0045] To further enhance the algorithm's global search capabilities, after initialization, the entire population P is divided into M subpopulations, each of which conducts its own independent search. Population diversity is crucial to avoid the convergence issues inherent in a single population. During this process, each subpopulation sets its own search range based on mission requirements (such as time, space, and fuel constraints). An information exchange mechanism between subpopulations will be implemented in subsequent steps to ensure that the best solutions from each subpopulation are globally updated.

[0046] After the population is divided, each subpopulation performs an independent gray wolf optimization search, where the movement of wolves in each subpopulation follows the update rules of equations (1)-(3):

[0047]

[0048] Gray wolf optimization algorithm hunting-driven position update diagram as shown below Figure 4 shown.

[0049] Step S206: Apply mapping based on the three-dimensional trajectory planning of multiple UAVs in a complex environment.

[0050] During the optimization process of each subpopulation, the fitness value of each individual needs to be calculated. The fitness function is used to evaluate the quality of each individual solution under the constraints (such as flight time, flight path, threat zone avoidance, etc.). Assume that the mission constraints include fuel consumption cost, flight altitude cost, threat cost, spatial coordination constraint, time coordination constraint, etc.

[0051] The path height cost is adapted as shown in the figure Figure 5 shown.

[0052] Fuel consumption or power consumption indirectly reflects the length of the UAV's flight range, and the consumption cost is proportional to the track distance. Assume that the track τ consists of n track segments The trajectory is composed of p1, the consumption cost coefficient, then the consumption cost f O-i As shown in formula (4):

[0053]

[0054] The flight altitude is too high or too low, and you need to pass f H-i Apply a penalty. Set the maximum flight altitude H max and the minimum flight altitude H min , the height of the kth track segment is H k , p2 and p3 are both height cost coefficients, height cost f H-i As shown in formula (5):

[0055]

[0056] The impact range of this threat source is modeled as a threat area, and drones are prohibited from entering this area. If a trajectory segment crosses the threat area, the mission is considered to have failed or the drone is destroyed. To this end, a corresponding threat cost function is constructed to quantify the risk. The collision cost is expressed as shown in Equation (6):

[0057]

[0058] Where p4 is the threat cost coefficient, M is the total number of UAV track points, C i is the neighborhood relationship between the UAV and the track point i. If the track point i is in the threat area, then C i =1; otherwise C i =0.

[0059] The spatial coordination constraint is also called the collision-free constraint, which requires that the minimum distance between drones should not be less than the minimum safe flight distance, that is:

[0060] d i-j ≥d s (7)

[0061] Where: d i-j is the minimum distance between the trajectory points of the i-th UAV and the j-th UAV during the flight;

[0062] d s The minimum safe flying distance between drones.

[0063] Assume that the command time for the UAV to reach the target destination is t c , through the command time constraint to ensure the UAV time coordination; according to the speed range and trajectory length of the UAV, the actual time range of the UAV to reach the target destination can be obtained but:

[0064]

[0065] Where: and are the maximum speed and minimum speed of the UAV respectively; L i is the trajectory length of the i-th UAV. Then the i-th UAV can complete the task within the specified time; otherwise, it cannot complete the task and the coordination of the formation is not achieved. It is necessary to calculate the time cost, which can be expressed by formula (9):

[0066]

[0067] For collision cost calculation, taking the i-th UAV as an example, we first calculate the distance between the current time of the UAV and the previous i-1 UAVs at each trajectory point. If the distance is less than the minimum safe flight distance, a collision is recorded. The collision counts of all trajectory points of the entire trajectory are counted in turn, and the total number of collisions of the UAV trajectory is obtained by summing them up. The calculation formula is shown in Equation (10):

[0068]

[0069] The objective function J is the weighted sum of all constraints, as shown in formula (11):

[0070]

[0071] Where: m is the number of drones; f O-i 、f H-i 、f T-i 、f m-i and f C-i It represents the fuel, altitude, threat, time and collision cost of the i-th UAV, and the obstacle collision area is as follows: Figure 6 As shown, w i After calculating the fitness of each agent, the solution with the smallest fitness value is selected as the α, β, and δ solutions. These three solutions will serve as the update benchmarks for the sub-population.

[0072] Step S208: adjusting the leader fitness temperature based on the simulated annealing criterion.

[0073] Since the core mechanism of MPGWO is still based on the standard GWO and lacks an effective escape mechanism, some MPGWO variants may still experience population "degeneration" or fall into local optimality problems during actual operation.

[0074] The SA algorithm is a global optimization algorithm inspired by the physical phenomena of metal quenching and searches for a solution close to the global optimal solution in the search space. To prevent the MPGWO algorithm from easily falling into a local optimal solution, the SA algorithm is introduced during the subpopulation iterative search process to calculate the probability P of each subpopulation α, β, and δ wolf being promoted to a suitable solution. The probability P expands the search space of these wolves and creates an alternative position update strategy that represents the diversity of capabilities. The formula for probability P is shown in Equation (12):

[0075]

[0076] T t+1 =η×T t (13)

[0077] To improve the algorithm's ability to handle constraints and prevent premature population convergence, this step updates and replaces the α, β, and δ solutions in the current subpopulation iteration process based on the solution obtained from the SA probabilistic update. By perturbing and probabilistically accepting the three optimal solutions α, β, and δ, revised performance indicators for the candidate solutions in each subpopulation are obtained. Then, the leader individuals of all subpopulations (i.e., the α, β, and δ wolves of each subpopulation) are comprehensively ranked according to their fitness, and the three top-ranked gray wolf individuals are selected as the new global optimal solution α, the second-best solution β, and the third-best solution δ, respectively. The resulting new α, β, and δ individuals then replace the original decision-making α, β, and δ individuals of each subpopulation, achieving dynamic updating of the subpopulation leaders. By replacing the optimal individual, the newly generated leader solution is guaranteed to meet the problem constraints and have a better evaluation, thereby improving the feasibility of the algorithm solution and the accuracy of the global optimal solution.

[0078] Step S210: performing adjustment based on the nonlinear convergence factor of the trigonometric function.

[0079] After completing step S208, the algorithm has obtained the latest set of leaders α, β, and δ, which have been perturbed and updated by the SA. To ensure that the subsequent iterative search of ordinary individuals can fully utilize the excellent global information carried by the new leader while avoiding the phenomenon of rapid decline in the early search radius and insufficient local development in the later stage caused by the traditional linear convergence factor, this step performs a nonlinear improvement on the core parameter a of the gray wolf model. The updated convergence factor a is shown in Equation (14).

[0080]

[0081] where t max To preset the maximum number of iterations, the initial value of a is usually 2. Figure 7 The following is an iterative comparison of the original convergence factor and the improved nonlinear convergence factor based on trigonometric functions. The improved trigonometric convergence factor a fluctuates slightly in the early stages but remains high for most subsequent iterations. Therefore, maintaining a high value for a long period of time helps enhance global search capabilities. At the end of the iteration, a's state reverses, improving local search capabilities. This adjustment achieves a more ideal balance between global and local search, resulting in higher accuracy. Together with the leader diversity generated in step S208, it promotes large-scale global exploration. Rapid convergence in the last 50% of iterations allows for precise integration with the subsequent local refinement in step S212. Through this dynamic adjustment of "high amplitude in the early stages, stable in the mid-stage, and rapid convergence in the late stages," ordinary ω-level individuals can achieve a more reasonable encirclement radius when performing position updates. This not only inherits the navigation advantages of the latest leader, but also provides ample margin for subsequent boundary projection and threat avoidance.

[0082] After completing the nonlinear convergence factor setting, the algorithm immediately proceeds to step S212, where the updated a(t) is used alongside the α, β, and δ traction vectors to drive the entire group of individuals to perform position iteration, boundary feasibility, and threat reconstruction. This lays the foundation for stable convergence and controllable accuracy for the subsequent dual-threshold convergence determination in step 8. This step thus establishes a parameter scheduling bridge between "leadership update" and "global position evolution," achieving an adaptive balance between the algorithm's temporal and spatial search scales.

[0083] Step S212: nonlinear grey wolf position update and feasible region correction.

[0084] In this step, after completing the update of the leader individual (α, β, δ), the position of each gray wolf individual in the population is calculated and updated using the improved gray wolf optimization position update formula. The formula is as follows:

[0085]

[0086] X'(t+1)=w α ×X1+w β ×X2+w δ ×X3 (18)

[0087] In the above dynamic weight position update formula, f(X α )、f(X β ) and f(X δ ) represent the fitness values ​​of the three “leader wolves” (α, β, δ) ranked 1st, 2nd, and 3rd in fitness in the current iteration respectively;

[0088] Using the fitness of these three leading wolves, the normalized weight w can be calculated α 、w β 、w δ The larger the weight, the better the path quality of the leader wolf and the stronger its guiding role on the population. In the formula, X1, X2, and X3 are the transient candidate position vectors obtained after the three leader wolves α, β, and δ pull the ordinary individuals in space; X'(t+1) is the three candidate vectors according to the weight w α 、w β 、w δ The final position of the individual in the next iteration is obtained after linear weighting. Through this weighted update with adaptive weights, individuals will be more inclined to the direction of the leader wolf with higher fitness, thereby accelerating convergence and improving the global optimality of the path solution.

[0089] This step then performs boundary checking and adjustment on each updated individual position: if a dimension's position exceeds a pre-defined allowable range, the value of that dimension is truncated to within a boundary threshold. This position update and boundary processing ensures that the new solutions for all gray wolf individuals adhere to the gray wolf optimization search strategy while satisfying the search space constraints defined by the problem, thus maintaining the validity of the solution.

[0090] Step S214: Generate the UAV dynamic obstacle avoidance flight and optimal solution.

[0091] In this step, the algorithm integrates and analyzes the global optimal solutions found during the iterative process to generate the final multi-UAV collaborative path planning solution. Specifically, the solution corresponding to the final α wolf (i.e., the global optimal solution) is extracted, which contains the optimal trajectory information for all UAVs. Based on this solution, a flight path for each UAV is then constructed, represented as a series of continuous three-dimensional trackpoints, so that each route meets the mission requirements and constraints (e.g., avoiding threat areas and meeting time synchronization requirements). By integrating these optimal trajectories, the algorithm produces the optimal path planning result for multi-UAV collaborative operations. This result fully demonstrates the optimization advantages of the algorithm: the trajectory of each UAV is optimally balanced at the global level, the overall flight cost (e.g., time and energy consumption) is reduced, and the safe and efficient completion of the collaborative mission is ensured. Finally, the multi-UAV path planning solution output in step S214 can be used for actual mission deployment, and the entire simulated annealing enhanced multi-population gray wolf optimization algorithm (SMPGWO) solution process for the multi-UAV path planning problem is completed.

[0092] The present invention proposes a collaborative trajectory planning method for multiple intelligent aircraft based on hybrid local wolf pack optimization. By introducing multiple swarm strategies, nonlinear convergence factors, simulated annealing algorithms, and dynamic weight mechanisms into the gray wolf optimization algorithm, the global search capability of the algorithm is enhanced and the emergence of local optimal solutions is avoided, thereby achieving efficient planning of three-dimensional trajectories for multiple UAVs. The main advantages of this invention are mainly reflected in the following aspects: first, the diversity of the population is improved through multiple swarm strategies, and the global exploration capability is enhanced; second, the introduction of nonlinear convergence factors effectively expands the search range and accelerates the convergence speed; third, the combination of simulated annealing algorithms and dynamic weight mechanisms enables the algorithm to escape from local optimality, ensuring higher trajectory optimization accuracy; finally, the superior performance of this method in multi-UAV trajectory planning is verified through simulation experiments, and both optimization speed and accuracy show a significant improvement.

[0093] The present application embodiment also provides another method for multi-intelligent aircraft collaborative trajectory planning based on hybrid local wolf pack optimization. The specific steps are as follows: Figure 8 Shown, including:

[0094] Step S802: task constraint analysis and search space initialization.

[0095] First, the algorithm reads the relevant parameters of each UAV in the mission, including the number of UAVs, starting and target positions, flight altitude range, trajectory turning radius limits, and three-dimensional environmental information (such as obstacle coordinates and dimensions, no-fly zones, and other constraints). Based on these parameters, the algorithm sets the initial conditions and constraint model required. For example, the path planning search space is defined as a three-dimensional space encompassing the flight areas of all UAVs. Constraint thresholds such as maximum flight distance and minimum safe distance are set to provide penalty factors when evaluating the quality of paths. Next, the algorithm population is initialized, generating several candidate multi-UAV path planning solutions as the initial individuals in the gray wolf population. Preferably, a combination of random initialization and heuristics is used to generate the initial population: Under the premise of satisfying basic flight constraints, a feasible path consisting of several discrete waypoints is randomly generated for each UAV. The set of all UAV paths is encoded into a high-dimensional position vector as a candidate solution. This generation of N individuals in the population ensures that the initial solutions are evenly distributed throughout the search space, resulting in a high degree of diversity.

[0096] The Gray Wolf Optimization (GWO) algorithm mimics the social hierarchy and hunting behavior of gray wolves to optimize search. In the GWO algorithm, wolves are divided into four hierarchies: α, β, δ, and ω. The α wolf is the leader, primarily responsible for all group decisions; the β wolf assists the α wolf and makes decisions; the δ wolf obeys the instructions of the α and β wolves and manages the ω wolf. Gray wolves rely on this hierarchical system to forage and hunt.

[0097] After initialization is completed, set the parameters of the gray wolf optimization algorithm (such as population size N, maximum number of iterations T max , coefficient initialization values, etc.) and simulated annealing algorithm parameters (initial temperature T0, cooling factor η, etc.) for subsequent iterative optimization.

[0098] Step S804: heterogeneous sub-population division and parallel evolution framework establishment.

[0099] The present invention employs a multi-population co-evolution strategy to improve the algorithm's search efficiency and adapt to optimization in complex, high-dimensional spaces. Step S804 divides the initial population into multiple subpopulations. Specifically, the population can be divided into M subpopulations based on the number of drones or the composition of their path codes, with each subpopulation focusing on optimizing a subset of variables within the overall path solution. Decision variables are divided by drone: the kth subpopulation contains the path portion of the i-th drone of all candidate solutions. Thus, each subpopulation optimizes the path of a single drone. Individuals in subpopulation k consist solely of the waypoint sequence of drone i; the path portions of other drones can be treated as if the environmental parameters were temporarily fixed. This problem decomposition divides the original high-dimensional search space into multiple subspaces of lower dimensions for parallel optimization. Simultaneously, each subpopulation is connected by a shared global evaluation function, allowing each to independently evolve its own component. The candidate solutions they generate, combined with the other currently optimal solutions, form a complete multi-drone path solution for fitness evaluation, ensuring that all subpopulations co-evolve toward the overall optimization goal.

[0100] Step S806: Apply mapping based on the three-dimensional trajectory planning of multiple UAVs in a complex environment.

[0101] During the optimization process of each subpopulation, the fitness value of each individual needs to be calculated. The fitness function is used to evaluate the quality of each individual's solution while satisfying constraints (such as flight time, flight path, threat zone avoidance, etc.). The improved fitness function calculation formula of the present invention is shown in Equation (11). After calculating the fitness of each agent, the solution with the smallest fitness value is selected as the α, β, and δ solutions. These three solutions will serve as the update benchmarks for the subpopulation.

[0102] Step S808: adjusting the leader fitness temperature based on the simulated annealing criterion.

[0103] To enhance the algorithm's ability to escape local optima, the present invention applies a simulated annealing algorithm to update the α, β, and δ solutions selected from each subpopulation during each iteration. Specifically, for each α, β, and δ individual in each subpopulation, a simulated annealing (SA) strategy is employed to attempt to generate a superior alternative solution. Starting with the current solution as the initial solution, a small perturbation is performed to generate a new candidate solution. The perturbation method can be selected based on the characteristics of the path encoding, adding small random offsets to the waypoint coordinates to generate a new path solution that differs slightly from the current solution. Accepting potentially inferior solutions as new solutions allows the algorithm to accept slightly inferior solutions with a higher probability in the early stages of the iteration, thereby increasing solution diversity. The linear cooling scheme provided by the present invention gradually reduces the temperature after each update attempt, ensuring that the algorithm tends to only accept superior solutions as it converges in the later stages of the iteration. Through this simulated annealing update process, the α, β, and δ solutions have a reasonable chance of escaping the original local optimum and migrating to a new path region. Even temporarily inferior solutions may be accepted, thus introducing new search directions for the population. In this step, the three leader wolves (α, β, δ) of each sub-population are subjected to SA perturbation update processing to maximize the search range of the population. In the specific implementation, it is also possible to choose to perform simulated annealing reinforcement only on the α wolf as needed, but optimizing multiple leaders simultaneously can provide more candidate search directions, thereby improving the performance of the algorithm.

[0104] After simulated annealing perturbation, each subpopulation will produce a set of candidate leader solutions that have been optimized by SA. Subsequently, the original α, β, and δ solutions are replaced and updated based on the penalized fitness values ​​of these candidate solutions: if the fitness of a candidate solution is better than the corresponding original solution after considering the constraint penalty, the candidate solution will replace the original α, β, and δ solutions. This step compares the evaluation value of the new solution obtained by simulated annealing in each subpopulation with the evaluation value of the original leader solution, where the "penalized" evaluation value indicates that the penalty for constraint violation is included. The α, β, and δ solutions updated in this step will serve as the leader leading the population search in the next iteration. In the multi-population gray wolf optimization framework, this mechanism of introducing new elites is equivalent to the survival of the fittest for the leadership of each subpopulation, so that the algorithm always maintains a set of high-quality guidance solutions, further improving the rigor of optimization and the speed of convergence.

[0105] Step S810: adjusting the nonlinear convergence factor based on the trigonometric function.

[0106] After completing step S808, the algorithm has obtained the latest set of leaders α, β, and δ, which have been perturbed and updated by the SA. To ensure that the subsequent iterative search of ordinary individuals can fully utilize the excellent global information carried by the new leader while avoiding the phenomenon of rapid decline in the early search radius and insufficient local development in the later stage caused by the traditional linear convergence factor, this step performs a nonlinear improvement on the core parameter a of the gray wolf model. The specific improvement method is shown in Equation (14).

[0107] After completing the nonlinear convergence factor setting, the algorithm immediately proceeds to step S812, where the updated a(t) and the α, β, and δ traction vectors are used to drive the entire group of individuals to perform position iteration, boundary feasibility, and threat reconstruction operations, laying the foundation for subsequent convergence stability and controllable accuracy. This step establishes a parameter scheduling bridge between "leadership update" and "global position evolution," achieving an adaptive balance between the algorithm's temporal and spatial search scales.

[0108] Step S812: nonlinear grey wolf position update and feasible region correction.

[0109] After completing the leader wolf update, the algorithm iteratively optimizes the positions of the remaining individuals in the subpopulation according to the position update rule of the gray wolf optimization (GWO). For each subpopulation, the current positions of the α, β, and δ wolves in the subpopulation are used to update the solution vectors of the other ω wolves (ordinary individuals) according to the gray wolf's strategy of surrounding prey. Specifically, the position update is performed using formulas (15)-(18). The position update formula shows that each individual will adjust itself according to the current positions of the three leader wolves, converging spatially to the positions of the α, β, and δ solutions, thereby simulating the process of the gray wolf group gradually approaching the prey. By introducing random factors in A and C, each update will retain a certain amount of random perturbation when moving towards the leader solution, which not only ensures convergence but also maintains population diversity to avoid premature maturity. The use of the improved nonlinear decreasing parameter a reflects the algorithm's gradual transition from exploration to exploitation: when a is large in the early stage, individuals have a larger movement stride to explore new areas; in the later stage, when a approaches 0, individuals fine-tune their positions to obtain a more optimal solution through fine search.

[0110] After the position update calculation is complete, the resulting new position vector needs to be bounded. Specifically, if any dimension of the wolf's position exceeds a predefined range (for example, a waypoint's coordinates exceed the flight area boundary, or the altitude component is outside the permitted altitude range), the value of that dimension is truncated to the boundary of the permitted range. This boundary adjustment ensures that each individual's new solution remains a legal path within the feasible domain of the problem. For example, if a waypoint is updated below ground level or above the maximum altitude, its altitude can be fixed to the ground level or maximum flight altitude, respectively. If the path coordinates exceed the map range, the coordinates are adjusted to the boundary. This boundary-adjusted position serves as the individual's position in the next iteration. After this step, the algorithm uses the GWO position update formula to ensure that each subpopulation individual converges toward the direction indicated by its leader wolf, while also performing necessary boundary processing to ensure the feasibility and safety of the solution. At this point, a new generation of candidate solutions for each subpopulation is formed, ready for the next iteration. Through the above position updates and boundary processing, it is ensured that the new solutions of all gray wolf individuals not only follow the gray wolf optimized search strategy, but also meet the search space constraints defined by the problem, thereby maintaining the validity of the solution.

[0111] Step S814: Dynamic obstacle avoidance flight of the UAV and generation of the optimal solution.

[0112] After the iterations are complete, the optimal individual from each subpopulation (i.e., the final α-wolf solution) represents the optimal planning solution for different components of the original problem. Finally, the α-solutions from all subpopulations are integrated to form a complete multi-UAV 3D path planning result.

[0113] Extract the drone trajectory plan corresponding to its α individuals from each subpopulation, and combine these trajectories to obtain the flight path set of multiple drones. Since the evaluation and evolution of each subpopulation are always guided by the global goal during the co-evolution process, the multi-drone path set obtained by the above integration can be regarded as the global optimal or suboptimal solution of the algorithm. Figure 9 As shown, the three-dimensional top view is as follows Figure 10 These paths, after repeated algorithmic optimization, have shown significant improvements over the initial paths in terms of path length, obstacle avoidance safety, and track smoothness. The total distance of each drone's flight path is close to the shortest, while avoiding obstacles as much as possible and meeting flight constraints such as turning radius, ensuring the feasibility and safety of the paths.

[0114] The Simulated Annealing-enhanced Multi-Population Grey Wolf Optimization (SMPGWO) method proposed in this paper aims to improve the optimization efficiency and accuracy of multi-UAV systems in complex three-dimensional trajectory planning, thereby avoiding local optimal solutions and accelerating the search process for the global optimal solution. This method effectively expands the capabilities of global and local search, enabling multi-UAV systems to complete more comprehensive tasks in more complex environments, meet the requirements of trajectory planning under multi-objective and multi-constraint conditions, and enable UAVs to achieve efficient collaborative operations in dynamic, highly complex mission environments.

[0115] This application also provides a multi-intelligent aircraft collaborative trajectory planning device based on hybrid local wolf pack optimization, such as Figure 11 As shown, it includes: an evaluation module 112, which is configured to perform a comprehensive evaluation on the candidate path solutions of each sub-population of the current population, and select multiple individuals with the best fitness as the current leading individuals; a perturbation module 114, which is configured to perform path perturbations on multiple leading individuals respectively, and adopt a probability criterion to decide whether to replace one or more of the multiple leading individuals according to the current fitness changes and temperature parameters of the multiple leading individuals, so as to improve the global search capability; an update module 116, which is configured to update the path node positions of ordinary individuals in each sub-population according to the current multiple leading individuals, and perform boundary legitimacy correction on the updated path nodes to generate new candidate path solutions; a planning module 118, which is configured to extract the path information of the current best individuals of all sub-populations based on the new candidate path solutions to plan the collaborative trajectory of multiple intelligent aircraft.

[0116] It should be noted that the multi-intelligent aircraft collaborative trajectory planning device based on hybrid local wolf pack optimization provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the multi-intelligent aircraft collaborative trajectory planning device based on hybrid local wolf pack optimization provided in the above embodiment and the multi-intelligent aircraft collaborative trajectory planning method embodiment based on hybrid local wolf pack optimization belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0117] Figure 12 Schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present disclosure is shown. Figure 12 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0118] like Figure 12 As shown, the electronic device includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage part 1008 into the random access memory (RAM) 1003. Various programs and data required for system operation are also stored in the RAM 1003. The CPU 1001, ROM 1002 and RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0119] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, and the like; an output section 1007 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 1008 including a hard disk; and a communication section 1009 including a network interface card such as a LAN card or a modem. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 1010 as needed, so that computer programs read therefrom can be installed into the storage section 1008 as needed.

[0120] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A multi-intelligent aircraft collaborative trajectory planning method based on hybrid local wolf pack optimization, characterized in that: include: Comprehensively evaluate the candidate path solutions of each sub-population of the current population of multi-intelligent aircraft and select multiple individuals with the best fitness as the current leader individuals; Performing path perturbations on the plurality of leader individuals respectively, and using a probability criterion to determine whether to replace one or more of the plurality of leader individuals based on current fitness changes and temperature parameters of the plurality of leader individuals, so as to improve global search capability; For ordinary individuals in each subpopulation, the path node positions are updated according to the guidance of the current multiple leader individuals, and the boundary legitimacy of the updated path nodes is corrected to generate new candidate path solutions; Based on the new candidate path solution, the path information of the current optimal individuals of all subpopulations is extracted to plan the collaborative trajectory of the multiple intelligent aircraft.

2. The method according to claim 1, characterized in that Before comprehensively evaluating the candidate path solutions of each subpopulation of the current population of multiple intelligent aircraft, the method further includes: Reading mission parameters, wherein the mission parameters include at least one of the following: the number of intelligent aircraft, flight speed range, maximum flight distance, start and end point coordinates, threat area coordinates, no-fly zone boundaries, upper and lower altitude limits, and safety distance; Determine a path search space and constraint rules based on the task parameters and the three-dimensional space modeling results; By utilizing heuristic rules in combination with random strategies, a plurality of candidate path solutions satisfying the constraint rules are generated in the search space, an initial population is formed as the current population, and the current population is divided into a corresponding number of sub-populations according to the number of the multi-intelligent aerial vehicles.

3. The method according to claim 1, characterized in that Comprehensively evaluate the candidate path solutions of each sub-population of the current population of multi-intelligent aircraft and select multiple individuals with the best fitness as the current leader individuals, including: Calculating a flight cost for each candidate path solution, the flight cost comprising at least one of the following: a flight energy consumption cost, a flight altitude deviation cost, a threat zone crossing risk, a track curvature smoothness, an arrival time error, and a minimum distance conflict cost with other intelligent aerial vehicles; The flight costs are weighted and combined to obtain a single evaluation index as the comprehensive fitness of the individual; The comprehensive fitness is sorted from best to worst, and the three individuals with the best, second best and third best fitness in the current sub-population are selected as the multiple leader individuals.

4. The method according to claim 3, characterized in that Calculate the flight cost for each candidate path solution, including at least one of the following: The flight energy cost is obtained by estimating based on the total length of the path segment and the energy consumption coefficient; Calculating the flight altitude cost based on the degree of deviation between the flight altitude of each path segment and the maximum and minimum flight altitudes; The threat avoidance cost is obtained by scoring the track point based on whether it enters the threat area and the distance to the threat edge. Evaluate path smoothness based on the angle or curvature change between adjacent track segments; The time synchronization cost is calculated based on the error between the time when the intelligent aircraft reaches the target point and the instruction time; The conflict cost is calculated by comparing the minimum distance between multiple intelligent aircraft at any time with the safety distance.

5. The method according to claim 1, wherein Performing path perturbations on the plurality of leader individuals respectively, and using a probability criterion to determine whether to replace one or more of the plurality of leader individuals based on current fitness changes and temperature parameters of the plurality of leader individuals, including: Applying disturbances to the path node positions of the plurality of leader individuals respectively to generate a plurality of new candidate path solutions; Re-evaluating the fitness of the new candidate path solution and comparing it with the fitness of the candidate path solution to obtain a fitness difference; An acceptance probability is calculated based on the fitness difference and the temperature parameter, and based on the acceptance probability, it is determined whether to adopt the individual corresponding to the new candidate path solution as a new leader individual.

6. The method according to claim 3, characterized in that For ordinary individuals in each subpopulation, their path node positions are updated according to the guidance of the current multiple leader individuals, and the boundary legitimacy of the updated path nodes is corrected to generate new candidate path solutions, including: For each common individual in the subpopulation, respectively calculating the position difference vector between the common individual and the plurality of leader individuals; Based on the nonlinear convergence factor of the current iteration and the position difference vector, guiding the ordinary individuals to move towards the plurality of leader individuals to update their path node positions; Bounds checking and legality correction are performed on the updated path nodes so that the updated path nodes meet physical and task constraints.

7. A multi-intelligent aircraft collaborative trajectory planning device based on hybrid local wolf pack optimization, characterized in that: include: An evaluation module is configured to comprehensively evaluate the candidate path solutions of each subpopulation of the current population and select multiple individuals with the best fitness as the current leader individuals; a perturbation module configured to perform path perturbations on each of the plurality of leader individuals and, based on current fitness changes and temperature parameters of the plurality of leader individuals, determine whether to replace one or more of the plurality of leader individuals using a probabilistic criterion to improve global search capability; An updating module is configured to update the path node positions of ordinary individuals in each subpopulation according to the guidance of the current plurality of leader individuals, and to perform boundary legitimacy correction on the updated path nodes to generate new candidate path solutions; The planning module is configured to extract the path information of the current optimal individuals of all sub-populations based on the new candidate path solution to plan the collaborative trajectory of multiple intelligent aircraft.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

9. A computer device, characterized in that: include: memory and processor, The memory stores a computer program; The processor is configured to execute a computer program stored in the memory, wherein the computer program enables the processor to execute the method according to any one of claims 1 to 6 when the computer program is executed.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images