Multi-objective unmanned aerial vehicle path planning method based on improved melolontha algorithm
By improving the dung beetle algorithm and combining it with a multi-objective weighted function and an adaptive perturbation mechanism, the problem of coordinating the optimization of mission time, energy consumption and terrain threats in UAV trajectory planning is solved, thereby improving planning efficiency and trajectory feasibility and adapting to complex terrain and multi-threat source environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-02-09
- Publication Date
- 2026-05-29
Smart Images
Figure CN122108123A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of UAV flight path planning, specifically involving a multi-target UAV trajectory planning method based on an improved dung beetle algorithm. Background Technology
[0002] With the widespread application of UAV technology in fields such as terrain mapping, power line inspection, emergency rescue, and logistics transportation, multi-objective trajectory planning in complex terrains (such as mountains, urban building complexes, and areas with multiple threat sources) has become a core technological bottleneck. UAV trajectory planning must simultaneously meet three core objectives: shortest mission time, lowest flight energy consumption, and least terrain threat, while also taking into account trajectory smoothness, obstacle avoidance feasibility, and actual flight controllability. It is a typical multi-objective constrained optimization problem.
[0003] Currently, commonly used trajectory planning algorithms in the industry include traditional dung beetle optimization algorithm, genetic algorithm, particle swarm optimization algorithm, Osprey search algorithm, etc., but they have significant limitations in complex application scenarios: Insufficient multi-objective collaborative optimization capability: Existing algorithms mostly use single-objective optimization or simple weighting to handle multi-objective problems, which easily leads to the phenomenon of "paying attention to one thing while neglecting another" (such as focusing on shortening time while ignoring energy consumption, or increasing path length while reducing threats), making it difficult to achieve dynamic collaborative balance between task time, energy consumption, and terrain threats; The search strategy has shortcomings: the search mechanism of traditional dung beetle algorithm and other basic optimization algorithms is simple, and there is an imbalance between global exploration ability and local development ability. It either gets stuck in a local optimum due to insufficient global search range, or the feasibility of the planned trajectory is poor due to insufficient local development accuracy. The disturbance mechanism introduced by some algorithms is fixed and cannot be dynamically adjusted according to the iteration progress, resulting in insufficient adaptability. The aforementioned technical deficiencies result in existing trajectory planning methods having problems such as low planning efficiency, poor trajectory practicality, and limited adaptability in complex terrain applications, failing to meet the high-precision and high-reliability operational requirements of UAVs. Therefore, there is an urgent need for an optimized algorithm that takes into account multi-target collaboration, search efficiency, trajectory feasibility, and obstacle avoidance capabilities. Summary of the Invention
[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A multi-target UAV trajectory planning method based on an improved dung beetle algorithm includes the following steps: S1): Design objective function: Take mission time, flight energy consumption and terrain threat level as multi-objective optimization objects, and construct a weighted single objective function to achieve collaborative optimization; S2): Initialization phase: Adopting refraction-based reverse learning and elite selection strategies, combined with fast non-dominated sorting, to improve the diversity of the initial population and the ability to retain high-quality solutions; S3): Dung beetle stage: Integrating Pareto guidance and Osprey search ideas, adaptive search is achieved by dividing the search into accessible and accessible modes; S4): Dung beetle breeding stage: Introducing an adaptive t-distribution perturbation mechanism to dynamically balance global exploration and local development, generating new individuals and repairing; S5): Dung beetle foraging stage: Jacobi curve smoothing mechanism or interval disturbance mode is adopted to improve the smoothness and feasibility of the flight path; S6): Dung Beetle Stage: Integrating a Pareto-guided multi-objective collaborative sharing mechanism, updating and verifying individual positions; S7): Merge the offspring and parent populations, generate a new generation population through multi-objective screening, and update the external solution set; S8): After the iteration terminates, output the optimal trajectory.
[0005] Furthermore, step 1 specifically includes the following steps: S11): Design of the total flight path length function; S12): Flight energy consumption function design; S13): Terrain threat function design; S14): Design of multi-objective weighted functions.
[0006] Furthermore, step 2 specifically includes the following steps: S21): Reflection Backward Learning: Generates a reverse solution for each path individual to expand the search space; S22): Elite selection: merge the original population and the reverse population, and select the top M optimal individuals according to the multi-objective fitness function to form the initial population; S23): Fast non-dominated sorting: Calculates the rank and crowding distance of each individual in the initial population.
[0007] Furthermore, step 4 specifically includes the following steps: S41): Introducing an adaptive probability function: adjusting the global exploration probability according to the iteration progress; S42): When the random conditions are met, enter the global exploration mode, and introduce distribution perturbation with the global optimal individual as the center; S43): When the random conditions are not met, enter the local development mode and perform fine-grained perturbation based on the local superior position; S44): Path feasibility repair: Newly generated individuals need to undergo turning constraint function repair.
[0008] Furthermore, step 5 specifically includes the following steps: S51): If an individual does not enter the Jacobi curve search, an interval perturbation mechanism is used to perturb the trajectory within a local neighborhood. S52): If an individual enters the Jacobian curve mode, the Jacobian smooth path model is used to improve the curvature continuity and flightability of the UAV path, and enhance controllability to generate a smooth and stable new path.
[0009] Furthermore, step 6 specifically includes the following steps: S61): Model the relative gap between the current individual and the elite solution set as shareable information; S62): By transferring advantageous knowledge to disadvantaged individuals in the form of collaborative updates, the distribution quality of the Pareto frontier is further improved.
[0010] Furthermore, step 7 specifically includes the following steps: S71): Offspring merging: Merge the offspring individuals generated by the four behavioral subgroups into an offspring population, and merge it with the parent population to form a joint population; S72): Multi-objective screening: Perform multi-objective optimization screening on the joint population, and perform fast non-dominated sorting to calculate the rank and crowding distance of each individual; S73): Retain the first M individuals as the new generation population, and add the non-dominated solutions from the current population to the external solution set.
[0011] The beneficial effects of this invention are: 1. Excellent performance of multi-objective collaborative optimization, taking into account multiple core requirements: This invention takes mission time, flight energy consumption and terrain threat level as core optimization objectives. Through the construction of weighted single objective function and Pareto guidance mechanism, it achieves dynamic collaborative balance of the three objectives, which solves the problem of "single objective optimization and multi-objective imbalance" in existing algorithms. The planned trajectory can simultaneously meet the practical requirements of "shortest time, lowest energy consumption and least threat", and is suitable for diverse operation tasks in complex scenarios.
[0012] 2. Significantly improved initial population quality and convergence efficiency: By integrating refraction-based reverse learning and elite selection strategies with fast non-dominated sorting, the spatial coverage of the initial population is improved by more than 30%, the retention rate of high-quality solutions is increased, and the convergence time of the algorithm is effectively shortened; at the same time, the search blind zone caused by initial population aggregation is avoided, laying a good foundation for subsequent iterative optimization.
[0013] 3. Dynamic balance between global exploration and local development to avoid local optima: The adaptive t-distribution perturbation mechanism introduced in the dung beetle breeding stage dynamically adjusts the global exploration probability through iteration progress. Combined with the heavy-tailed characteristics of the t-distribution, it realizes the adaptive switching between "large-scale jump search + small-scale fine development", which effectively solves the problems of "premature convergence" and "local optimum trap" in traditional algorithms, and improves the probability of obtaining the global optimal solution by more than 40%.
[0014] 4. Strong obstacle avoidance capability in complex terrain and wide environmental adaptability: The rolling dung beetle stage integrates Pareto guidance and Osprey search concepts, achieving adaptive search in both obstacle-free and obstacle-covered modes; the multi-target collaborative sharing mechanism in the stealing dung beetle stage further enhances the path's ability to avoid complex obstacles. In complex terrains with multiple threat sources and high occlusion, the obstacle avoidance success rate reaches over 98%, adapting to various complex scenarios such as mountains, urban building clusters, and no-fly zones.
[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the original population of the present invention; Figure 2 This is a schematic diagram of the reverse population method of the present invention; Figure 3 This is a schematic diagram of the merged population of the present invention; Figure 4 This is the trajectory planning diagram for the present invention. Detailed Implementation
[0017] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention. Specific implementation examples: This paper proposes an improved dynamic bounded optimization (EDBO) algorithm for multi-target UAV trajectory planning in complex environments. The algorithm consists of five stages: initialization, dung beetle rolling, dung beetle breeding, dung beetle foraging, and dung beetle stealing. Each stage introduces innovative mechanisms to enhance global search and environmental adaptability.
[0019] Includes the following steps: S1): Design objective function: Take mission time, flight energy consumption and terrain threat level as multi-objective optimization objects, and construct a weighted single objective function to achieve collaborative optimization; S11): Design of the total flight path length function; This patent assumes that the individual drone is ∈ The path consists of n trajectory points, each trajectory point is composed of... = ( , , () represents a location point in three-dimensional space. The total path length is expressed as: , S12): Flight Energy Consumption Function Design; The energy consumption of a UAV during flight mainly consists of two parts: horizontal energy consumption and vertical energy consumption. Horizontal energy consumption is mainly affected by path length and is related to the UAV's flight distance, while vertical energy consumption is mainly affected by altitude and is related to the UAV's climb rate. The total energy consumption function is expressed as: ,in: ; It is the energy consumption coefficient for horizontal flight per unit distance; It is the energy consumption coefficient per unit height increase.
[0020] S13): Terrain Threat Function Design; This patent innovatively designs a three-dimensional terrain threat function formed by multi-source threat fusion, setting the UAV trajectory as... , where each trajectory point = ( , , () represents a location point in three-dimensional space. Simultaneously, various threat sources are normalized and their weights adjusted to obtain a terrain map. The threat function in the map is defined as: in Total number of threat sources; ( , , ): No. The source of the threat is at the trajectory point Threat intensity value at the location; : No. The weighting coefficients of the threat sources satisfy the following conditions: ; :path Total Threat Cost S14): Multi-objective weighted function design; This patent sets task time, energy consumption, and terrain threat level for multi-objective optimization: Task time Energy consumption Terrain threat level Weigh them into a single objective: .
[0021] S2): Initialization phase: Adopting refraction-based reverse learning and elite selection strategies, combined with fast non-dominated sorting, to improve the diversity of the initial population and the ability to retain high-quality solutions; To increase the diversity of the initial population, this patent introduces refraction-backward learning, which involves analyzing each individual along a path. ∈ Refraction inverse learning is performed to generate the corresponding inverse solution. Expanding the search space and increasing population diversity is beneficial for each individual's first... Dimensional search space boundary We use the following formula to generate the inverse solution:
[0022] This means that the inverse solutions for each dimension are generated by the symmetrical position of the current individual at the search boundary in that dimension. These inverse solutions improve the diversity of the initial population, making the correct solutions more widely distributed in the solution space. To verify the effectiveness of refraction inverse learning in three-dimensional multi-objective path planning, this paper conducts a visual comparative analysis of the distribution characteristics of the population in the three-dimensional search space before and after applying refraction inverse learning. The experiment is set with a population size of M=50, 5 trajectory points, and each individual containing 15 variables, forming a path representation in three-dimensional space. Figure 1 The image shows a three-dimensional scatter plot of the initial population. It can be observed that the individuals are distributed relatively randomly in the three-dimensional search space, with obvious clusters and gaps. Figure 2 This paper demonstrates the inverse population distribution map generated after introducing a refraction-based inverse learning mechanism. By applying the refraction symmetry formula to each individual in each dimension, this method generates new inverse solutions based on the original population, effectively expanding the search space boundary and improving the spatial coverage of the solutions. Figure 3 The diagram shows the three-dimensional distribution of the initial population A∪B obtained by merging the original population A and the reverse population B.
[0023] After merging the initial population and the population generated by refraction back learning, we perform elite selection on the merged population. For each individual, we calculate two objective values based on a multi-objective fitness function. and Then, these fitness values are used to measure the quality of individuals, and the top M individuals are selected to form a new population P. The core idea of the elite selection strategy is to retain the best-fitting individuals by merging the original population with the reverse population, ensuring that the population always contains the best solution. This strategy helps the algorithm maintain the quality of solutions and improves global search capabilities.
[0024] After generating the initial population P, Fast Non-Donminated Sorting is performed. This process calculates the rank and crowding distance for each individual. The rank is obtained by sorting each individual according to its non-dominance relationship in the target space. Rank=1 indicates that these individuals are not dominated by any other individuals in the target space and are the best individuals in the current population. The crowding distance measures the distribution density of an individual in the target space; the lower the crowding distance, the denser the solutions around that individual, and vice versa. The crowding distance guides the selection of individuals with good distribution.
[0025] For each individual Crowding in the target space The calculation formula is: , in and These respectively represent the individual's goal The objective values of two adjacent solutions on the , Let and be the maximum and minimum objective values in the objective space, respectively. Through this selection and sorting, both the final population P and the external solution set N can guarantee diversity and retain high-quality solutions.
[0026] S3): Dung beetle rolling stage: Based on the traditional dung beetle algorithm, this invention incorporates the Pareto solution set guidance mechanism and the search strategy ideas in the Osprey optimization algorithm (OOA) to construct a search model with multi-objective guidance and adaptive adjustment capabilities; Pareto index and Osprey search ideas in accessibility mode: In accessibility mode, this patent is inspired by the "long-distance dive" search behavior in Osprey optimization, and randomly selects two Pareto solutions from the current external Pareto solution set N. This facilitates information exchange and directional guidance between individuals.
[0027] Introducing an adaptive attenuation factor: , This patent achieves dynamic compression of the search range by mimicking the search intensity of random time decay in Osprey. This patent calculates the crowding difference of Pareto solutions. By integrating multi-objective information from individual solutions and excellent solutions, the search direction is updated, and the path is individual. The formula for updating the new position is: , This formula effectively improves the global exploration performance and solution set distribution quality of the population by remotely guiding and maintaining target diversity.
[0028] Historical trajectory deflection under obstacle mode: In obstacle-prone mode, this patent employs a historical trajectory deflection update mechanism to guide individual paths to adjust towards a safer area. Specifically, this involves individual paths... The positional difference at any given moment is used as the deflection direction, and displacement is updated by combining this with an angle function: , By preserving the inertial characteristics of individuals and introducing appropriate perturbations, the continuity and feasibility of paths can be improved while maintaining obstacle avoidance capabilities.
[0029] S4): Dung beetle breeding stage: Introducing an adaptive t-distribution perturbation mechanism to dynamically balance global exploration and local development, generating new individuals and repairing; This invention introduces an update strategy based on an adaptive t-distribution perturbation mechanism into traditional breeding dung beetles, thereby enhancing the dynamic balance control between global exploration capabilities and local exploitation capabilities during path search.
[0030] To achieve a dynamic balance between global and local reproduction strategies, an adaptive probability function is first constructed based on the current iteration number t:
[0031] in: These are empirical parameters used to adjust the speed and tendency of mode switching; This represents the maximum number of iterations. This indicates the probability that the current iteration will reproduce in global exploration mode.
[0032] As iterations proceed, the P-value gradually increases, driving the algorithm to transition from local exploitation to wide-area search, preventing premature convergence and getting trapped in local optima. This is while satisfying random conditions. At this time, enter global exploration mode and use the current best individual. Centered on the t-distribution perturbation, a new solution is generated:
[0033] in: Obeying the degree of freedom is of distributed; The distribution has a heavy-tailed characteristic, which can generate large perturbation amplitude within a certain range, enhancing the individual's ability to search and jump to regions far from the current local optimum.
[0034] This invention utilizes the current optimal individual The directional search guidance ensures that the search is biased towards a better solution space, which is especially suitable for individuals whose current search is stuck in a crowded or stagnant area, and effectively improves the algorithm's ability to escape local traps.
[0035] When the random value does not satisfy the global search condition When this happens, the system enters a local development mode, focusing on fine-grained development of the local space to improve the search density and efficiency of the population near excellent solutions. Fine-grained perturbations are performed using the current position as a baseline, combined with the upper and lower boundaries:
[0036] in: This indicates an individual's favorable position in the current or historical search process; These are the dynamic boundary estimates for the current search variables; This is the perturbation scaling factor, which controls the intensity and direction of the perturbation.
[0037] Regardless of whether the reproduction method is a global or local perturbation, the newly generated individuals All need to pass through the turning constraint function Perform path feasibility repair:
[0038] This repair operation ensures that the path solution meets the constraints, guaranteeing the physical feasibility and practical application value of the solution. S5): Dung beetle stage: Employing a Jacobi curve smoothing mechanism or interval perturbation mode to improve track smoothness and feasibility; the dung beetle primarily undertakes the important tasks of local search and detailed local exploration. This patent introduces two optional mechanisms: a Jacobi smoothing curve modeling mechanism and an adaptive interval perturbation mechanism. If an individual... Upon entering the Jacobi curve mode (with a probability of 0.5), its new position is updated using the following formula:
[0039] in: This represents the current globally optimal solution; It is a perturbation factor that follows a uniform distribution; This is a time-adaptive angle adjustment factor.
[0040] Fractional terms construct a Jacobi smooth path model, which improves indicators such as curvature continuity and flightability of UAV paths, and enhances controllability to generate smooth and stable new paths.
[0041] If an individual does not enter the Jacobi curve search, an interval perturbation mechanism is used to perturb the trajectory within a local neighborhood. The update formula is as follows:
[0042] in: This represents a perturbation guidance term constructed based on the boundary. The disturbance factor is either normally distributed or uniformly distributed.
[0043] An interval-perturbation local search mechanism is used for fine-grained adjustments, enhancing the ability to discover individual local optima. Simultaneously, it addresses the new locations generated for all foraging subgroups. Unify the execution of feasibility boundary correction operators:
[0044] in The path boundary constraint function ensures that the trajectory meets safety constraints and spatial limitations.
[0045] S6): Dung Beetle Stage: Integrating a Pareto-guided multi-objective collaborative sharing mechanism, updating and verifying individual positions; This patent proposes a Pareto-guided multi-objective collaborative sharing mechanism based on the traditional dung beetle model. It models the relative gap between the current individual and the elite solution set as "shareable information" and transfers advantageous knowledge to disadvantaged individuals through collaborative updates, further improving the distribution quality of the Pareto front. Each path individual... The new position is updated using the following formula:
[0046] in: This represents the current globally optimal solution; Represents an individual The local optimum to which it belongs; This represents a non-dominated solution randomly selected from the external Pareto solution set; S is a scale control parameter that controls the range of individual offset; This is a time-adaptive enhancement coefficient, which improves the convergence ability of the algorithm in the later stages.
[0047] This update mechanism achieves cross-solution set information sharing by integrating information from itself, local, and external high-performing individuals, thereby effectively guiding the direction of path evolution. Simultaneously, to avoid the search getting trapped in local optima, a time-adaptive mechanism is introduced to control the extent of the theft:
[0048] This adaptive cooperation factor has the ability to expand the search range and enhance exploration capabilities in the early stages, and to converge and shrink in the later stages to enhance local development intensity, giving the dung beetle a good search-development balance. Meanwhile, all stolen solutions also need to undergo unified boundary and obstacle verification. Ensure the path is feasible, free from obstacles and collisions, and meets the constraints. S7): Merge the offspring and parent populations, generate a new generation population through multi-objective screening, and update the external solution set; After updating the subpopulations for rolling, breeding, foraging, and stealing behaviors, all offspring individuals need to be merged and subjected to multi-objective screening to generate a new generation of population and update the external solution set. The specific steps are as follows: S71): Offspring merging: Merging the offspring individuals generated from the four behavioral subgroups into a single offspring population. and the parent population Merging to form a joint population :
[0049] This step enhances the diversity and optimization potential of the population by integrating a variety of search strategies, including global exploration (rolling ball behavior), local exploitation (reproductive behavior), path smoothing (foraging behavior along Jacobi curves), and multi-cooperative sharing (stealing behavior).
[0050] S72): Multi-objective screening: For joint population Calculate the rank and crowding distance for each individual.
[0051] S73): Prioritize individuals with lower rank; if ranks are the same, select individuals with larger dist to maintain diversity. (Final retention) Individuals as a new generation of population The current population Non-dominated solutions are added to the external solution set. .like Cut off by distance based on crowding level, retaining the most widely distributed Each solution set has multiple solutions, ensuring both diversity and coverage.
[0052] S8): After the iteration terminates, output the optimal trajectory.
[0053] To more intuitively demonstrate the multi-target UAV trajectory planning effect based on the improved dung beetle algorithm of this patent, see attached... Figure 4 A set of example results are given.
[0054] According to the iterative convergence curve, as the number of iterations increases from 0 to the maximum number of generations, the value of the dual objective function shows a trend of "rapid decline in the early stage and gradual flattening in the later stage"—the steep decline in the early stage proves that the rolling ball strategy with the hybrid Pareto guidance and Osprey idea and the adaptive t-distribution strategy has a strong global search capability; the flattening in the later stage indicates that the algorithm has converged stably and can continuously approach the Pareto optimal solution, verifying the convergence and reliability of multi-objective optimization.
[0055] Based on the 3D terrain track map and two corresponding 2D top-down views, the red track line traverses the entire path from the starting point (dot) to the ending point (triangle), clearly avoiding all no-fly zones. This illustration visually demonstrates the algorithm's obstacle avoidance capabilities and path feasibility in complex obstacle environments, proving the effectiveness and success of the planning technology in real-world scenarios.
[0056] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A multi-target UAV trajectory planning method based on an improved dung beetle algorithm, characterized in that: Includes the following steps: S1): Design objective function: Take mission time, flight energy consumption and terrain threat level as multi-objective optimization objects, and construct a weighted single objective function to achieve collaborative optimization; S2): Initialization phase: Adopting refraction-based reverse learning and elite selection strategies, combined with fast non-dominated sorting, to improve the diversity of the initial population and the ability to retain high-quality solutions; S3): Dung beetle stage: Integrating Pareto guidance and Osprey search ideas, adaptive search is achieved by dividing the search into accessible and accessible modes; S4): Dung beetle breeding stage: Introducing an adaptive t-distribution perturbation mechanism to dynamically balance global exploration and local development, generating new individuals and repairing; S5): Dung beetle foraging stage: Jacobi curve smoothing mechanism or interval disturbance mode is adopted to improve the smoothness and feasibility of the flight path; S6): Dung Beetle Stage: Integrating a Pareto-guided multi-objective collaborative sharing mechanism, updating and verifying individual positions; S7): Merge the offspring and parent populations, generate a new generation population through multi-objective screening, and update the external solution set; S8): After the iteration terminates, output the optimal trajectory.
2. The multi-target UAV trajectory planning method based on the improved dung beetle algorithm according to claim 1, characterized in that: Step 1 specifically includes the following steps: S11): Design of the total flight path length function; S12): Flight energy consumption function design; S13): Terrain threat function design; S14): Design of multi-objective weighted functions.
3. The multi-target UAV trajectory planning method based on the improved dung beetle algorithm according to claim 1, characterized in that: Step 2 specifically includes the following steps: S21): Reflection Backward Learning: Generates a reverse solution for each path individual to expand the search space; S22): Elite selection: merge the original population and the reverse population, and select the top M optimal individuals according to the multi-objective fitness function to form the initial population; S23): Fast non-dominated sorting: Calculates the rank and crowding distance of each individual in the initial population.
4. The multi-target UAV trajectory planning method based on the improved dung beetle algorithm according to claim 1, characterized in that: Step 4 specifically includes the following steps: S41): Introducing an adaptive probability function: adjusting the global exploration probability according to the iteration progress; S42): When the random conditions are met, enter the global exploration mode, and introduce distribution perturbation with the global optimal individual as the center; S43): When the random conditions are not met, enter the local development mode and perform fine-grained perturbation based on the local superior position; S44): Path feasibility repair: Newly generated individuals need to undergo turning constraint function repair.
5. The multi-target UAV trajectory planning method based on the improved dung beetle algorithm according to claim 1, characterized in that: Step 5 specifically includes the following steps: S51): If an individual does not enter the Jacobi curve search, an interval perturbation mechanism is used to perturb the trajectory within a local neighborhood. S52): If an individual enters the Jacobian curve mode, the Jacobian smooth path model is used to improve the curvature continuity and flightability of the UAV path, and enhance controllability to generate a smooth and stable new path.
6. The multi-target UAV trajectory planning method based on the improved dung beetle algorithm according to claim 1, characterized in that: Step 6 specifically includes the following steps: S61): Model the relative gap between the current individual and the elite solution set as shareable information; S62): By transferring advantageous knowledge to disadvantaged individuals in the form of collaborative updates, the distribution quality of the Pareto frontier is further improved.
7. The multi-target UAV trajectory planning method based on the improved dung beetle algorithm according to claim 1, characterized in that: Step 7 specifically includes the following steps: S61): Offspring merging: Merge the offspring individuals generated by the four behavioral subgroups into an offspring population, and merge it with the parent population to form a joint population; S62): Multi-objective screening: Perform multi-objective optimization screening on the joint population, and perform fast non-dominated sorting to calculate the rank and crowding distance of each individual; S63): Retain the first M individuals as the new generation population, and add the non-dominated solutions from the current population to the external solution set.