A path planning method based on multi-collaborative particle swarm optimization algorithm

Abstract

This invention relates to the field of path planning technology and provides a path planning method based on a multi-cooperative particle swarm optimization algorithm. This method constructs a multi-cooperative particle swarm structure with a main population and auxiliary populations evolving in tandem, and introduces a dual-center guidance mechanism to achieve a dynamic balance between global exploration and local fine-tuning during UAV path search. Furthermore, a closed-loop adaptive parameter control mechanism is constructed, enabling key particle swarm parameters to adaptively adjust with changes in the search state, avoiding premature convergence and improving search stability. Simultaneously, through a diversity-driven population information exchange control method, the ability of the particle swarm to escape local optimal path solutions is enhanced, achieving stable and high-quality planning of UAV flight paths. This invention effectively improves the autonomous path planning performance and engineering applicability of UAVs in complex 3D environments.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and in particular to a path planning method based on a multi-cooperative particle swarm optimization algorithm. Background Technology

[0002] As the application of artificial intelligence technology in the field of UAV autonomous navigation continues to deepen, existing UAV path planning methods are gradually evolving from traditional graph search and sampling methods to intelligent algorithms based on particle swarm optimization. These methods have certain advantages in continuous space search and multi-objective optimization. However, in complex three-dimensional spatial environments and under multiple constraints, existing intelligent path planning technologies generally use fixed or empirically set algorithm parameters, making it difficult to perceive changes in the evolutionary state of the search population in real time. This can easily lead to problems such as rapid decay of search diversity, premature convergence, or getting trapped in local optima. At the same time, existing multi-swarm or cooperative intelligent optimization methods mostly rely on static information exchange rules and lack a closed-loop control mechanism based on a unified population state, making it difficult to achieve a dynamic balance between global exploration and local development. In addition, although some path planning methods based on deep learning or reinforcement learning have improved the level of decision-making intelligence, they still face problems such as high training costs, limited stability, and limited generalization ability in highly constrained and strongly non-convex three-dimensional path solution spaces. Therefore, existing intelligent UAV path planning technologies still need further improvement in terms of search stability, path quality, and robustness in complex three-dimensional environments. Summary of the Invention

[0003] To address the problems of existing intelligent UAV path planning methods in complex 3D environments, such as decreased search diversity, reliance on experience for parameter adjustment, static population cooperation mechanisms, and difficulty in balancing global exploration and local fine-grained search, this invention proposes a path planning method based on a multi-cooperative particle swarm optimization algorithm. The core of this method lies in constructing a multi-cooperative particle swarm structure where a main population and auxiliary populations co-evolve, and introducing a dual-center guidance mechanism to collaboratively constrain the global search behavior and local development behavior of the particle swarm, thereby maintaining stable search capabilities in high-dimensional, strongly constrained path solution spaces. Furthermore, this invention uses population diversity during particle swarm evolution as a unified state observation to construct a path planning method based on the "population diversity" of the particle swarm. The closed-loop control mechanism consisting of "state observation, adaptive parameter adjustment, and search behavior feedback" enables the inertia weight and learning factor to adaptively adjust with changes in the search state, avoiding premature convergence or ineffective divergence of the particle swarm. Simultaneously, through a diversity-driven population information exchange and control method, dynamic synergy is achieved between the main population's local fine-grained search capability and the auxiliary population's global exploration capability, effectively enhancing the particle swarm's ability to escape local optimal path solutions. This invention does not rely on large-scale offline training or scenario priors, and can achieve stable and high-quality planning of UAV flight paths under complex three-dimensional spatial constraints, significantly improving the robustness and adaptability of intelligent path planning algorithms in practical UAV application scenarios.

[0004] This invention provides a path planning method based on a multi-cooperative particle swarm optimization algorithm. This method is applied to autonomous path planning scenarios for unmanned aerial vehicles (UAVs) in complex three-dimensional spatial environments, and specifically includes the following steps:

[0005] Step S1: Path Planning Problem Modeling: Construct a path planning search space in the UAV 3D path planning scenario. The path planning search space describes the feasible flight path range of the UAV from the takeoff point to the target point. Obtain environmental element information related to path planning, including the planning start point, planning end point, no-fly zone boundary, threat area parameters, and energy consumption constraint information during UAV flight. Based on the environmental element information, construct a path cost function and feasibility constraints. The path cost function includes path length cost, threat risk cost, track curvature cost, flight energy consumption cost, and constraint violation penalty term. Minimize the path cost function as the UAV path planning objective. Define the UAV flight path representation corresponding to a particle in any iteration state as a candidate path. Use particle position encoding to parameterize the candidate path. Encode each candidate path as a particle position vector. The particle position vector consists of several waypoints, control points, and spline curve parameters, forming a path solution space for UAV path search.

[0006] Step S2: Multi-cooperative Particle Swarm Initialization: Based on the path solution space, an improved dual-center multi-cooperative particle swarm optimization algorithm is constructed to initialize the multi-cooperative particle swarm structure used for UAV path planning. The improved dual-center multi-cooperative particle swarm optimization algorithm is constructed as follows: based on PSO, and addressing the search requirements of high-dimensional, strongly constrained, and non-convex path solution spaces in UAV path planning, a multi-cooperative particle swarm structure with co-evolution of the main population and auxiliary population is constructed. During the iterative evolution of the particle swarm, a multi-path search trajectory modeling mechanism based on velocity decomposition is introduced to generate multiple candidate motion path forms within a single iteration. Furthermore, the search coverage of the particle swarm in the path solution space is expanded; simultaneously, a radial kernel-weighted dispersion calculation mechanism and an adaptive perturbation enhancement mechanism based on search convergence stagnation judgment are introduced to dynamically adjust the dispersion state and search activity level of the particle swarm; further, under the multi-cooperative particle swarm structure, a dual-center reference structure composed of a group virtual center and an individual virtual center is constructed to coordinately optimize the global guidance and individual guidance process of the particle swarm, thereby achieving an adaptive balance between global exploration capability and local development capability, forming a particle swarm optimization framework oriented towards multi-cooperative evolution and dual-center guidance, and constructing an improved dual-center multi-cooperative particle swarm optimization algorithm;

[0007] Step S3: Population Diversity Observation: In each iteration, based on the multi-cooperative particle swarm structure and the dual-center evolution mechanism of the improved dual-center multi-cooperative particle swarm optimization algorithm, the particle position distribution characteristics, particle velocity distribution characteristics and their evolutionary relationships are jointly characterized to construct a population diversity observation mechanism oriented towards search behavior and evolutionary state; on the basis of the population diversity observation mechanism, the current particle state of the main population and auxiliary population is observed to obtain the population diversity index.

[0008] Step S4: Closed-loop adaptive parameter control: Based on the population diversity index, a closed-loop adaptive parameter control mechanism is constructed through a preset monotonic mapping function. The population diversity index is used as a unified observation input to generate closed-loop parameter control results for controlling the particle swarm search behavior. The closed-loop parameter control results include inertial weights, cognitive learning factors, and population guidance factors in the UAV path planning process.

[0009] Step S5: Multi-cooperative population information exchange regulation: Based on the closed-loop parameter regulation results, using population diversity indicators as cooperative observation input, a closed-loop regulation mechanism for information exchange between the main population and auxiliary populations is constructed, and population cooperation regulation results are adaptively generated to constrain the cooperative evolution process of the main population and auxiliary populations.

[0010] Step S6: Particle Swarm Position and Velocity Update: Under the joint constraints of closed-loop parameter control results and population cooperation control results, particle velocity and particle position are updated for the main population and auxiliary population respectively; new UAV candidate paths are generated based on the updated particle positions, and the new UAV candidate paths are evaluated in combination with path cost function and feasibility constraints; accordingly, the individual historical best path of the particles and the globally best path shared by the main population and auxiliary population are updated.

[0011] Step S7: Termination Judgment and Path Output: During the particle swarm iteration process, the iteration process is terminated based on the stability of the globally optimal UAV path solution in consecutive iterations. When the termination condition is met, the UAV flight path corresponding to the current globally optimal particle is output as the final planned path, and the final planned path is output as the UAV 3D flight path result.

[0012] Furthermore, the process of observing the current particle states of the main population and auxiliary populations and calculating the population diversity index specifically includes the following steps:

[0013] Step S31: Obtain the particle position distribution characteristics and particle velocity distribution characteristics of each particle in the main population and auxiliary population, and jointly model the particle position distribution characteristics and particle velocity distribution characteristics to characterize the overall dispersion and search dynamics of the particle swarm in the path solution space.

[0014] Step S32: Based on step S31, the dispersion state and evolution characteristics of the particle swarm in the path solution space are structurally measured based on the population diversity observation mechanism, and the population diversity index is calculated.

[0015] Furthermore, step S32 specifically includes the following steps:

[0016] Step S321: Based on the particle position distribution features and particle velocity distribution features obtained in step S31, and according to the multi-path search trajectory modeling mechanism based on velocity decomposition, the motion direction structure implicit in the particle velocity distribution features is decomposed and analyzed into three velocity components corresponding to the inertial guidance direction, the individual cognitive guidance direction, and the group collaborative guidance direction; the three velocity components are combined in different arrangements to construct various particle motion path forms, and multiple intermediate candidate positions are generated based on different path combinations; the intermediate candidate positions and the current particle position are included in the particle position distribution feature set.

[0017] Step S322: The current position state and the historical best state of each particle in the particle position distribution feature set are weighted and summarized to construct a group virtual center to characterize the overall evolution trend of the group and an individual virtual center to characterize the optimal aggregation trend of individuals; and the global best particle position in the current iteration is used as the core reference center, and the group virtual center and the individual virtual center are used as auxiliary reference structures to be introduced into the measurement process of the group diversity index to form a reference center system for diversity calculation.

[0018] Step S323: Using a radial kernel-weighted dispersion calculation mechanism, the spatial deviation of the current position and intermediate candidate positions of the particles relative to the reference center system is calculated by kernel function weighting to construct the basic dispersion metric results;

[0019] Step S324: Perform adaptive enhancement processing on the basic dispersion measure results to obtain enhanced dispersion measure results to reflect potential search capability degradation; and generate a population diversity index based on the enhanced dispersion measure results.

[0020] Furthermore, in step S6, when updating the particle velocity and position of the main population, the main population adopts a velocity update strategy that enhances local development. Based on the inertia and learning terms of the particle swarm velocity update model, the velocity update strategy introduces a second-order momentum term and an oscillation enhancement term to suppress high-frequency oscillations during the particle velocity update process and improve the local convergence stability and fine search capability of the particle in the neighborhood of the current good solution. When updating the particle velocity and position of the auxiliary population, the auxiliary population adopts a perturbation update strategy that enhances global exploration. The perturbation update strategy introduces a joint perturbation term formed by Levy flight perturbation and chaotic perturbation during the particle position process to achieve a large-scale jump search of the particle in the path solution space and broaden the global search coverage.

[0021] By adopting the above solution, the beneficial effects achieved by the present invention are as follows:

[0022] First, this invention constructs a multi-cooperative particle swarm structure in which the main population and auxiliary populations evolve in tandem, and introduces a dual-center guidance mechanism. This enables effective coordination between global exploration and local fine-grained search during UAV path planning, improving the coverage and search stability of UAVs in complex three-dimensional environments with high-dimensional and strongly constrained path solution spaces. It also solves the problem that existing intelligent path planning methods are prone to having a single search direction and unstable path solution quality under conditions of dense obstacles and complex constraints in no-fly zones. Thus, it enhances the autonomous planning capability and path feasibility assurance level of UAVs in complex mission scenarios.

[0023] Second, this invention uses the diversity of the population during the particle swarm evolution process as a unified state observation to construct a closed-loop adaptive parameter control mechanism. This enables the inertial weight and learning factor to be adaptively adjusted according to the changes in the UAV path search stage, realizing dynamic adjustment of the particle swarm search behavior. This improves the search efficiency and convergence reliability in the path planning process and solves the technical problem that traditional particle swarm algorithms rely on fixed parameters and have difficulty in balancing early exploration and late convergence. As a result, it enhances the robustness and consistency of UAV path planning results under different environmental conditions and task requirements.

[0024] Third, by introducing a population information exchange and regulation mechanism driven by population diversity, this invention enables the local development capability of the main population and the global exploration capability of the auxiliary population to dynamically cooperate during the path planning process. This enhances the ability of the particle swarm to escape local optimal path solutions and solves the problems of static and insufficient adaptability of population cooperation rules in existing multi-swarm intelligent optimization methods. Thus, without relying on large-scale offline training and scenario priors, it significantly enhances the practicality, stability, and engineering deployability of UAV 3D path planning methods in complex real-world application scenarios. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a path planning method based on a multi-cooperative particle swarm optimization algorithm proposed in this invention.

[0026] Figure 2 This is a schematic diagram of the dual-center reference structure proposed in Example 3. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] Example 1, according to Figure 1 This invention provides a path planning method based on a multi-cooperative particle swarm optimization algorithm. This method is applied to autonomous path planning scenarios for unmanned aerial vehicles (UAVs) in complex three-dimensional spatial environments, and specifically includes the following steps:

[0029] Step S1: Path Planning Problem Modeling: Construct a path planning search space in the UAV 3D path planning scenario. The path planning search space describes the feasible flight path range of the UAV from the takeoff point to the target point. Obtain environmental element information related to path planning, including the planning start point, planning end point, no-fly zone boundary, threat area parameters, and energy consumption constraint information during UAV flight. Based on the environmental element information, construct a path cost function and feasibility constraints. The path cost function includes path length cost, threat risk cost, track curvature cost, flight energy consumption cost, and constraint violation penalty term. Minimize the path cost function as the UAV path planning objective. Define the UAV flight path representation corresponding to a particle in any iteration state as a candidate path. Use particle position encoding to parameterize the candidate path. Encode each candidate path as a particle position vector. The particle position vector consists of several waypoints, control points, and spline curve parameters, forming a path solution space for UAV path search.

[0030] Step S2: Multi-cooperative Particle Swarm Optimization Initialization: Based on the path solution space, an improved dual-center multi-cooperative particle swarm optimization algorithm is constructed. This algorithm initializes the multi-cooperative particle swarm structure used for UAV path planning. The multi-cooperative particle swarm structure includes a main population and auxiliary populations. The main population performs local fine-grained search within the neighborhood of the current best UAV path solution, while the auxiliary population performs global exploration in the path solution space. The particle position vector, particle velocity vector, individual historical best path, and global best path are initialized for each particle in both the main and auxiliary populations, enabling them to enter a cooperative iterative state while sharing the global best path. The improved dual-center multi-cooperative particle swarm optimization algorithm is constructed based on PSO, addressing the search requirements of high-dimensional, strongly constrained, and non-convex path solution spaces in UAV path planning. This paper constructs a multi-cooperative particle swarm optimization (PSO) structure in which the main population and auxiliary populations co-evolve. During the iterative evolution of the PSO, a multi-path search trajectory modeling mechanism based on velocity decomposition is introduced to generate multiple candidate motion path forms within a single iteration, thereby expanding the search coverage of the PSO in the path solution space. Simultaneously, a radial kernel-weighted dispersion calculation mechanism and an adaptive perturbation enhancement mechanism based on search convergence stagnation discrimination are introduced to dynamically adjust the dispersion state and search activity level of the PSO. Furthermore, under the multi-cooperative PSO structure, a dual-center reference structure composed of a group virtual center and an individual virtual center is constructed to coordinately optimize the global guidance and individual guidance processes of the PSO, thereby achieving an adaptive balance between global exploration capability and local development capability, forming a PSO optimization framework oriented towards multi-cooperative evolution and dual-center guidance, and constructing an improved dual-center multi-cooperative PSO optimization algorithm.

[0031] Step S3: Population Diversity Observation: In each iteration, based on the multi-cooperative particle swarm structure and the dual-center evolution mechanism of the improved dual-center multi-cooperative particle swarm optimization algorithm, the particle position distribution characteristics, particle velocity distribution characteristics, and their evolutionary relationships are jointly characterized to construct a population diversity observation mechanism oriented towards search behavior and evolutionary state. On the basis of the population diversity observation mechanism, the current particle state of the main population and auxiliary population is observed to obtain the population diversity index. The population diversity index is then normalized to map it to a preset bounded interval to eliminate the influence of path solution spatial scale differences on the diversity observation results. The population diversity index is used to reflect the global exploration intensity, local aggregation degree, and potential convergence trend of the particle swarm in the current iteration during the UAV path planning process based on the improved dual-center multi-cooperative particle swarm optimization algorithm. It also serves as a unified state observation for the subsequent closed-loop adaptive parameter control mechanism and the cooperative evolution scheduling mechanism of the main population and auxiliary population.

[0032] Step S4: Closed-Loop Adaptive Parameter Control: Based on the population diversity index, a closed-loop adaptive parameter control mechanism is constructed through a preset monotonic mapping function. The population diversity index is used as a unified observation input to generate closed-loop parameter control results for controlling the particle swarm search behavior. These results include inertial weights, cognitive learning factors, and swarm guidance factors in the UAV path planning process. Through the monotonic mapping function, when the population diversity index is high, the inertial weights are increased and the cognitive learning factor is strengthened to ensure the UAV path search maintains strong global exploration capabilities. When the population diversity index is low, the inertial weights are decreased and the swarm guidance factor is strengthened to accelerate the aggregation of the particle swarm towards the current optimal UAV path solution and enhance local development capabilities. Therefore, based on the exploration level and convergence state of the particle swarm at different iteration stages, an adaptive combination of inertial weights, cognitive learning factors, and swarm guidance factors matching the current search state is generated, forming the closed-loop parameter control results to constrain subsequent particle velocity and position update processes.

[0033] The closed-loop adaptive parameter control mechanism satisfies the following monotonic mapping relationship:

[0034] ;

[0035] ;

[0036] ;

[0037] in, This indicates the first step in the drone path planning process. The inertial weight of the next iteration is used to adjust the strength of the particle's inheritance of historical velocity during the velocity update process, thereby affecting the overall exploration stride of the UAV path search. This represents the preset minimum boundary value (lower limit) of the inertia weight. This represents the preset maximum boundary value (upper limit) of the inertia weight. Indicated by group diversity indicators The input is a monotonic mapping function, which is used to map the dispersion of the current UAV path candidate solution in the path solution space to the parameter adjustment intensity, thereby constructing a closed-loop control relationship consisting of "search state observation - parameter adjustment - search behavior change"; Indicates the first The cognitive learning factor in the next iteration is used to control the approach strength of the particle toward its individual historical best position, which reflects the degree to which the particle depends on its own historical flight path experience. This represents the preset lower limit of the cognitive learning factor. This indicates the preset upper limit of the cognitive learning factor; Indicates the first The group guidance factor in the next iteration is used to control the approach strength of particles toward the group's globally optimal UAV path, which reflects the degree of cooperative following of particles toward the group's optimal flight path during the path search process. This represents the preset lower limit of the group guiding factor. Indicates the preset upper limit of the group guiding factor;

[0038] Step S5: Multi-cooperative population information exchange regulation: Based on the closed-loop parameter regulation results, using population diversity indicators as cooperative observation inputs, a closed-loop regulation mechanism for information exchange between the main population and auxiliary populations is constructed. This adaptively generates population cooperation regulation results to constrain the cooperative evolution process between the main population and auxiliary populations. The population cooperation regulation results represent the information exchange cycle between the main population and auxiliary populations. This information exchange cycle is dynamically adjusted based on the population diversity indicators and satisfies the following diversity-driven monotonic mapping relationship:

[0039] ;

[0040] in, Indicates the first The information exchange cycle (closed-loop control quantity) between the primary population and the secondary population corresponding to each iteration. This represents the preset maximum threshold (upper limit) for the information exchange cycle. The preset minimum threshold representing the information exchange cycle; The indicator parameter (positive number) is used to adjust the information exchange cycle when the population diversity index decreases and shows a convergence trend, so as to increase the information exchange frequency between the main population and the auxiliary population.

[0041] When the preset information exchange triggering conditions are met, elite particle exchange and particle reset operations are performed between the main population and the auxiliary population according to the information exchange cycle. Specifically, this includes: merging the particle sets in the main population and the auxiliary population, selecting a preset proportion of elite particles according to fitness to form the parent particle pool, and cross-injecting the elite particles into the main population and the auxiliary population; at the same time, performing local reset based on the global optimal particle neighborhood on a preset proportion of inferior particles to introduce new search directions into the current search process; thereby, through the synergistic interaction between the local development capability of the main population and the global exploration capability of the auxiliary population, the ability of the particle swarm to escape local optimal solutions is dynamically enhanced, forming a population cooperation regulation result used to constrain the subsequent particle swarm evolution process;

[0042] Step S6: Particle Swarm Position and Velocity Update: Under the joint constraints of closed-loop parameter control results and population cooperation control results, particle velocity and particle position are updated for both the main population and the auxiliary population. New UAV candidate paths are generated based on the updated particle positions, and the new UAV candidate paths are evaluated in conjunction with the path cost function and feasibility constraints. Accordingly, the individual historical best paths of particles and the globally best paths shared by the main population and the auxiliary population are updated, so that the UAV path planning process continues to evolve under closed-loop control.

[0043] Step S7: Termination Judgment and Path Output: During the particle swarm iteration process, the iteration process is terminated based on the stability of the globally optimal UAV path solution in consecutive iterations. When the termination condition is met, the UAV flight path corresponding to the current globally optimal particle is output as the final planned path, and the final planned path is output as the UAV's 3D flight path result. The UAV's 3D flight path is output in the form of waypoint sequence, spline curve control parameters, and executable flight trajectory commands to guide the UAV to complete the autonomous flight mission from the planned starting point to the planned ending point during actual flight. Feasibility verification and engineering constraint consistency checks are further performed on the UAV's 3D flight path. The engineering constraints include maximum climb rate constraint, minimum turning radius constraint, minimum obstacle clearance safety distance constraint, and energy consumption-speed constraint. When the final planned path meets the above engineering constraint conditions, it is output as the UAV's executable flight path to the UAV flight control system, mission planning module, and ground control station. This achieves safe avoidance of obstacle areas, no-fly zones, and threat areas in complex 3D environments, while taking into account the comprehensive optimization goals of flight path length, energy consumption efficiency, and flight smoothness, thus completing the UAV 3D path planning process based on the multi-cooperative particle swarm optimization algorithm.

[0044] Example 2, based on Example 1, involves observing the current particle states of the main population and auxiliary populations to calculate the population diversity index. The specific steps include:

[0045] Step S31: Obtain the particle position distribution characteristics and particle velocity distribution characteristics of each particle in the main population and auxiliary population, and jointly model the particle position distribution characteristics and particle velocity distribution characteristics to characterize the overall dispersion and search dynamics of the particle swarm in the path solution space.

[0046] Step S32: Based on step S31, the dispersion state and evolution characteristics of the particle swarm in the path solution space are structurally measured based on the population diversity observation mechanism, and the population diversity index is calculated.

[0047] Example 3, according to Figure 2 This embodiment is based on Embodiment 2. In this embodiment, step S32 specifically includes the following steps:

[0048] Step S321: Based on the particle position distribution features and particle velocity distribution features obtained in step S31, and according to the multi-path search trajectory modeling mechanism based on velocity decomposition, the motion direction structure implicit in the particle velocity distribution features is decomposed and analyzed into three velocity components corresponding to the inertial guidance direction, the individual cognitive guidance direction, and the group collaborative guidance direction. The three velocity components are combined in different arrangements to construct multiple particle motion path forms, and multiple intermediate candidate positions are generated based on different path combinations. The intermediate candidate positions and the current particle position are included in the particle position distribution feature set to enhance the ability to characterize the search coverage and path expansion behavior of the particle swarm in the path solution space, thereby avoiding the underestimation of the dispersion of the swarm based solely on a single path update method.

[0049] Step S322: To reduce the bias of a single reference center on the observation results of population diversity, a weight allocation mechanism based on particle fitness is introduced according to the dual-center reference structure. The current position state of each particle in the particle position distribution feature set and the historical best state of each individual are weighted and summarized to construct a population virtual center to characterize the overall evolution trend of the population and an individual virtual center to characterize the optimal aggregation trend of individuals. The global best particle position in the current iteration is used as the core reference center, and the population virtual center and individual virtual center are used as auxiliary reference structures to be introduced into the measurement process of population diversity index, so as to form a reference center system for diversity calculation.

[0050] In this embodiment, the overall structure of the particle swarm diversity modeling mechanism based on the dual-center reference structure, the correlation between each reference center, and their distribution in the path solution space are as follows: Figure 2 As shown; Figure 2 This is a schematic diagram of a dual-center reference structure; where the horizontal axis represents the path solution space dimension 1, and the vertical axis represents the path solution space dimension 2; blue dots represent the current position of the particle (the current set of candidate paths), orange crosses represent the individual's historical best position (each particle's own best path memory); star (green) represents the globally optimal particle (core reference center); rhombus (red) represents the group's virtual center, representing the overall evolution trend of the group; square (purple) represents the individual's virtual center, representing the individual's optimal clustering trend;

[0051] Step S323: Using a radial kernel-weighted dispersion calculation mechanism, the spatial deviation of the current particle position and intermediate candidate positions relative to the reference center system is calculated using a kernel function weighting. Specifically, the global optimal particle position in the current iteration is used as the core reference center, and combined with the auxiliary reference structure formed by the group virtual center and the individual virtual center, the spatial deviation of the particle relative to the core reference center is measured by kernel weighting to construct a basic dispersion metric result for characterizing the spatial expansion state and aggregation trend of the particle swarm.

[0052] The formula for calculating the basic dispersion measure result is as follows:

[0053] , ;

[0054] in, This represents the iteration index of the particle swarm optimization algorithm in the UAV path planning process. This represents the particle number index, used to distinguish the candidate flight paths of the drone corresponding to different particles; This represents the number of particles participating in the basic dispersion measure calculation, which is the total number of particles currently participating in the statistics in the main population and the auxiliary population. Indicates the first The particle in the first In the UAV 3D path planning scenario, the particle position vector at the next iteration corresponds to the parameterized representation of a candidate flight path of the UAV. It consists of several waypoint 3D coordinates, control point parameters or spline curve parameters, and is used to describe the flight trajectory shape of the UAV in space. Indicates the first The globally optimal particle position vector at the next iteration serves as the reference center for the radial kernel weighted metric. Indicates the first The particle and the globally optimal particle are in the... The distance metric at each iteration is used to reflect the degree of deviation of the candidate path from the optimal path in terms of overall track shape, waypoint distribution, or curvature characteristics. The square of the distance is used to construct the exponential term of the radial kernel function; Represents an exponential function; Indicates the first The kernel scale parameter (bandwidth) at the next iteration. The median operator represents the median of the set of distances from all particles to the globally optimal particle. Indicates the first The results of the basic dispersion measure in the next iteration are used as the initial quantity for calculating the population diversity index;

[0055] Step S324: To avoid distortion of diversity assessment due to slowed dispersion changes during the search stagnation phase, an adaptive perturbation enhancement mechanism based on search convergence stagnation criteria is used to adaptively enhance the basic dispersion metric results, resulting in enhanced dispersion metric results that reflect potential search capability degradation. Based on the enhanced dispersion metric results, a population diversity index is generated. It is used to quantitatively describe the overall dispersion, search activity level, and potential convergence state of the particle swarm in the current iteration round;

[0056] To prevent the particle swarm from getting trapped in local optima, a stagnation criterion and a mutation probability incrementing mechanism are introduced, for example, for the th... Definition of a particle:

[0057] ;

[0058] in, This indicates the search convergence stagnation indicator (also known as the fitness improvement ratio indicator). Indicates the first The particle in the first The fitness function value of the corresponding candidate path in the next iteration round. Indicates the first The particle in the first The fitness function value of the corresponding candidate path in the next iteration round;

[0059] when After several consecutive rounds (5 times), the threshold condition is met. When the random number is less than the current mutation probability, mutation is performed.

[0060] ;

[0061] in, Indicates the first The particle position vector in the current iteration round The Dimensional components; This represents the variation mixing coefficient, used to control the weight ratio of "retaining the original position component" and "introducing the reference optimal component"; Indicates the first The individual historical best position vector of each particle The Dimensional components.

[0062] Example 4, based on Example 3, in step S6, when updating the particle velocity and position of the main population, the main population adopts a velocity update strategy that enhances local development. The velocity update strategy introduces a second-order momentum term and an oscillation enhancement term on the basis of the inertia term and learning term of the particle swarm velocity update model, so as to suppress high-frequency oscillations in the particle velocity update process and improve the local convergence stability and fine search capability of the particle in the neighborhood of the current good solution. When updating the particle velocity and position of the auxiliary population, the auxiliary population adopts a perturbation update strategy that enhances global exploration. The perturbation update strategy introduces a joint perturbation term formed by Levy flight perturbation and chaotic perturbation in the particle position process, so as to realize the large-scale jump search of the particle in the path solution space and broaden the global search coverage.

[0063] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A path planning method based on a multi-cooperative particle swarm optimization algorithm, characterized in that, The method includes the following steps: Step S1: Construct the path planning search space and obtain environmental element information; based on the environmental element information, construct the path cost function and feasibility constraints to form the path solution space; Step S2: Based on the path solution space, construct an improved dual-center multi-cooperative particle swarm optimization algorithm, and initialize the multi-cooperative particle swarm structure using the improved dual-center multi-cooperative particle swarm optimization algorithm; Step S3: In each iteration, based on the multi-cooperative particle swarm structure and the dual-center evolution mechanism of the improved dual-center multi-cooperative particle swarm optimization algorithm, a population diversity observation mechanism is constructed, population diversity is observed, and population diversity index is calculated. Step S4: Construct a closed-loop adaptive parameter control mechanism using a monotonic mapping function, taking population diversity indicators as a unified observation input, and generating closed-loop parameter control results; Step S5: Based on the closed-loop parameter regulation results, using population diversity indicators as collaborative observation input, construct a closed-loop regulation mechanism for information exchange between the main population and auxiliary populations, and generate population collaborative regulation results; Step S6: Under the joint constraints of the closed-loop parameter control results and the population cooperation control results, the particle velocity and particle position are updated respectively to generate new UAV candidate paths, and the paths are evaluated and the globally optimal path shared in the multi-cooperative particle swarm structure is updated. Step S7: Terminate the test based on the stability of the global optimal path and output the UAV's 3D flight path result.

2. The path planning method based on multi-cooperative particle swarm optimization algorithm according to claim 1, characterized in that: The path planning search space describes the feasible flight path range of the UAV from the takeoff point to the target point; environmental element information includes the planning start point, planning end point, no-fly zone boundary, threat area parameters, and energy consumption constraint information during the UAV flight; the path cost function includes path length cost, threat risk cost, trajectory curvature cost, flight energy consumption cost, and constraint violation penalty.

3. The path planning method based on multi-cooperative particle swarm optimization algorithm according to claim 1, characterized in that: Multi-cooperative particle swarm architecture includes a multi-cooperative population system consisting of a main population and auxiliary populations.

4. The path planning method based on multi-cooperative particle swarm optimization algorithm according to claim 1, characterized in that: The improved dual-center multi-cooperative particle swarm optimization algorithm is constructed as follows: based on PSO, a multi-cooperative particle swarm structure in which the main population and auxiliary population co-evolve is constructed; a multi-path search trajectory modeling mechanism based on velocity decomposition is introduced to expand the search coverage; a radial kernel weighted dispersion calculation mechanism and an adaptive perturbation enhancement mechanism based on search convergence stagnation discrimination are introduced to dynamically adjust the dispersion state and search activity level of the particle swarm. In a multi-cooperative particle swarm optimization (MCS) architecture, a dual-center reference structure is constructed to collaboratively optimize the global and individual guidance processes of the particle swarm, resulting in an improved dual-center MCS optimization algorithm.

5. The path planning method based on multi-cooperative particle swarm optimization algorithm according to claim 1, characterized in that: The process of conducting population diversity observations and calculating population diversity indices includes the following steps: Step S31: Obtain the particle position distribution characteristics and particle velocity distribution characteristics of each particle in the main population and auxiliary population, and jointly model the particle position distribution characteristics and particle velocity distribution characteristics to characterize the overall dispersion degree and search dynamic characteristics of the particle swarm in the path solution space. Step S32: Based on step S31, the dispersion state and evolution characteristics of the particle swarm in the path solution space are structurally measured based on the population diversity observation mechanism, and the population diversity index is calculated.

6. The path planning method based on multi-cooperative particle swarm optimization algorithm according to claim 5, characterized in that: Step S32 specifically includes the following steps: Step S321: Based on the particle position distribution features and particle velocity distribution features, and according to the multi-path search trajectory modeling mechanism based on velocity decomposition, the motion direction structure implicit in the particle velocity distribution features is decomposed and analyzed into three velocity components corresponding to the inertial guidance direction, the individual cognitive guidance direction, and the group collaborative guidance direction; the three velocity components are combined in different arrangements to construct various particle motion path forms, and intermediate candidate positions are generated based on different path combinations; the intermediate candidate positions and the current particle position are included in the particle position distribution feature set. Step S322: Introduce a weight allocation mechanism based on particle fitness to weight and summarize the current position state and the individual's historical best state of each particle in the particle position distribution feature set, and construct the group virtual center and the individual virtual center; and take the global best particle position in the current iteration as the core reference center, and use the group virtual center and the individual virtual center as auxiliary reference structures to introduce them into the measurement process of the group diversity index, forming a reference center system. Step S323: Using a radial kernel-weighted dispersion calculation mechanism, the spatial deviation of the current position and intermediate candidate positions of the particles relative to the reference center system is calculated by kernel function weighting to construct the basic dispersion metric results; Step S324: Perform adaptive enhancement processing on the basic dispersion measure results through an adaptive perturbation enhancement mechanism based on search convergence stagnation judgment to obtain the enhanced dispersion measure results; and generate a population diversity index based on the enhanced dispersion measure results.

7. The path planning method based on multi-cooperative particle swarm optimization algorithm according to claim 1, characterized in that: The closed-loop parameter regulation results include inertia weight, cognitive learning factor, and group guidance factor.

Images