Path planning method and device, medium and equipment

Abstract

The invention discloses a path planning method and device, a medium and equipment, and relates to the technical field of path planning. Comprising the steps of determining a starting point, an ending point and a to-be-planned path area; constructing a path planning cost function; all paths between the starting point and the ending point are used as a population of a moss growth optimization algorithm, each path between the starting point and the ending point is used as an individual of the moss growth optimization algorithm, the moss growth optimization algorithm is initialized, and the maximum number of iterations is determined; operating the moss growth optimization algorithm, and optimizing the moss growth optimization algorithm by using the comprehensive learning strategy formula in the operation process; performing iterative optimization on the path planning cost function by using the optimized moss growth optimization algorithm until the number of iterations meets the maximum number of iterations; and outputting an optimal solution of the path planning cost function, and applying the optimal solution as a target path to unmanned aerial vehicle path planning. According to the invention, the shortest path can be found, and the path planning speed is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of path planning, and in particular relates to a path planning method, device, medium and equipment. BACKGROUND

[0002] In the context of the increasing application of unmanned aerial vehicles (UAVs) today, when they perform flight tasks in complex mountainous terrain and areas with obstacles, they face many challenges. Unmanned aerial vehicles need to quickly and safely plan the optimal path in a dynamically changing environment to avoid obstacles, respond to unexpected threats, and efficiently complete tasks. However, traditional path planning methods often have many shortcomings when faced with such complex and variable scenarios, such as slow convergence speed, easy to fall into local optimal solution, etc., making it difficult to meet the high requirements of path planning performance and safety in actual applications.

[0003] With the rapid development of unmanned aerial vehicle technology and its wide application in various fields, dynamic unmanned aerial vehicle path planning has gradually become an important research direction that attracts much attention. The core of this problem is how to design an optimal or near-optimal path for an unmanned aerial vehicle from the starting point to the end point in an environment full of uncertainty and dynamic changes, while meeting various complex constraint conditions. The following are several key analysis points for dynamic unmanned aerial vehicle path planning in mountainous terrain: 1. Complexity of the environment; 2. Diversity of tasks; 3. Cost constraints; 4. Adaptability. First, in mountainous terrain, the ups and downs of the terrain greatly increase the difficulty of path planning, and new obstacles may appear at any time during flight, which puts higher requirements on the flexibility and adaptability of path planning. Second, the unmanned aerial vehicle may receive new task instructions during flight, which requires it to quickly adjust the flight target and direction. Third, factors such as flight distance, flight height, and turning angle of the unmanned aerial vehicle will affect its flight performance and energy consumption, so path planning needs to balance and optimize among these cost factors. In addition, real-time performance is also a key factor. In a dynamic environment, rapid response is crucial, which requires the path planning algorithm to have strong performance and be able to find a better path in as few iterations as possible. Finally, the algorithm needs to be able to adapt to various environmental changes and uncertainties to ensure that the unmanned aerial vehicle can safely and efficiently complete tasks under complex and variable conditions.

[0004] In recent years, researchers have conducted extensive research on unmanned aerial vehicles, which can better solve the unmanned aerial vehicle path planning problem, but as the complexity of the unmanned aerial vehicle flight environment continues to deepen, the performance of the algorithm is also constantly challenged.

[0005] In the prior art, the moss optimization algorithm (MGO) is a swarm intelligence optimization algorithm, but it is easy to fall into local optimum and has a slow convergence speed, which limits its performance to some extent.

[0006] Therefore, the prior art has the problem that in a dynamic unmanned aerial vehicle path planning model, due to slow convergence speed, low convergence accuracy and large path cost of the algorithm, the best path of the unmanned aerial vehicle has the problem of slow convergence speed in determination, which further leads to low planning efficiency of the best path. SUMMARY

[0007] Therefore, it is necessary to provide a path planning method, device, medium and equipment in view of the above technical problems.

[0008] The present specification adopts the following technical solutions: The present specification provides a path planning method, comprising: determining a starting point, an ending point and a path planning region to be planned; constructing a path planning cost function; initializing a moss growth optimization algorithm by taking all paths between the starting point and the ending point as a population of the moss growth optimization algorithm and taking each path between the starting point and the ending point as an individual of the moss growth optimization algorithm, and determining a maximum number of iterations; running the moss growth optimization algorithm and optimizing the moss growth optimization algorithm by using a comprehensive learning strategy formula during the running process; iteratively optimizing the path planning cost function by using the optimized moss growth optimization algorithm until the number of iterations meets the maximum number of iterations; outputting an optimal solution of the path planning cost function, taking the optimal solution as a target path and applying it to unmanned aerial vehicle path planning.

[0009] Optionally, the step of comprehensive learning strategy comprises: generating a random number greater than 0 and less than 1; if the random number is greater than or equal to the learning probability, setting the d-th dimension value of the best position of each individual in the population as the d-th dimension value of the learning sample vector; if the random number is less than the learning probability, selecting a first individual and a second individual from the population, when the fitness value of the best position of the first individual is less than the fitness value of the best position of the second individual, setting the d-th dimension value of the first individual as the d-th dimension value of the learning sample vector, when the fitness value of the best position of the first individual is greater than or equal to the fitness value of the best position of the second individual, setting the d-th dimension value of the second individual as the d-th dimension value of the learning sample vector; wherein the first individual and the second individual are different from other individuals, and the first individual and the second individual are different; When the value of all dimensions in the learning sample vector is determined by the d-th dimension value of the best position of each individual in the population, a target individual is selected from the population, and a target dimension is selected from the dimensions of the learning sample vector, and the value of the best position of the target individual in the target dimension is set to the dimension value of the learning sample vector in the target dimension; wherein the target individual is different from other individuals; The sample vector is used to guide the individual to update; The learning probability The calculation formula is: Wherein, is equal to , is equal to 0.5, represents the d-th individual, represents the total number of individuals.

[0010] Optionally, the running process of the moss growth optimization algorithm is: Determine the reference value, compare each individual in the moss growth optimization algorithm with the reference value, determine at least two subsets, and take the subset with the most number of individuals in the at least two subsets as the population to be analyzed; According to the positional relationship of the individuals in the population to be analyzed, the wind direction of the moss growth optimization algorithm is calculated; According to the wind direction, the spore diffusion operation is performed on the individuals in the population to be analyzed to generate a first individual; The individuals that meet the preset conditions are searched and bred according to the double propagation strategy to generate a second individual; The first individual and the second individual are compared with the individuals in the population to be analyzed respectively, and the individual with a better path planning cost function is retained to update the population to be analyzed; When the maximum number of iterations is reached, stop running.

[0011] Optionally, the blowing direction of the wind direction is from the region to the reference value individual; The calculation formula of the wind direction is: Wherein, represents the wind vector, represents the number of individuals in the population to be analyzed, represents the differential of , represents the spore of the d-th moss individual.

[0012] Optionally, according to the wind direction, the spore diffusion operation is performed on the individuals in the population to be analyzed, and the calculation formula used is: ​​​ wherein, represents the spore transmission process of the first moss individual, newly generated moss individual, represents the spore of the first moss individual, , represents a random number, represents a threshold parameter, when, represents the spore transmission distance under stable wind conditions, , represents a fixed parameter, represents a random vector uniformly distributed in the interval , and has the same dimension as , represents the wind strength, , represents the current number of function evaluations, and represents the maximum number of iterations, represents the transmission distance under turbulent wind conditions, , , represents the ratio of the number of individuals in the population to be analyzed to the number of individuals in the total population .

[0013] Optionally, the individuals that meet the preset conditions are searched and bred according to the double transmission strategy, and the calculation formula used is: wherein, represents the newly generated individual at the index , and is the th element of , is randomly selected but is subject to the constraint of the dimension of the individual, represents the individual with the best performance at present, and represents the th component of , represents the th element of the wind vector , and the random variable is used to determine the breeding strategy,

[0014] Optionally, the method further comprises: The balanced experiment test and the diversity experiment test are performed on the optimized moss growth optimization algorithm to evaluate the optimization performance of the optimized moss growth optimization algorithm.

[0015] The present specification provides a path planning device, comprising: A data acquisition module is specifically configured to determine a starting point, an ending point and a path planning region to be planned. A function construction module is specifically configured to construct a path planning cost function. An algorithm optimization module is specifically configured to take all paths between the starting point and the ending point as a population of the moss growth optimization algorithm, take each path between the starting point and the ending point as an individual of the moss growth optimization algorithm, initialize the moss growth optimization algorithm and determine a maximum iteration number, run the moss growth optimization algorithm, optimize the moss growth optimization algorithm by using a comprehensive learning strategy formula during the running process, and iteratively optimize the path planning cost function by using the optimized moss growth optimization algorithm until the iteration number meets the maximum iteration number. A path output module is specifically configured to output an optimal solution of the path planning cost function, take the optimal solution as a target path and apply the target path to unmanned aerial vehicle path planning.

[0016] The present specification provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the path planning method.

[0017] The present specification provides a computer device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the path planning method when executing the program.

[0018] The above at least one technical solution adopted by the present specification can achieve the following beneficial effects: In the path planning method provided by the present specification, firstly, a comprehensive learning strategy is proposed to balance the abilities of global search and local search and accelerate the convergence speed of the algorithm; secondly, the comprehensive learning strategy is added to the moss optimization algorithm to improve the ability of the algorithm to jump out of a local optimum; and finally, the application of the algorithm in dynamic unmanned aerial vehicle path planning is explored, which significantly improves the path planning performance of the unmanned aerial vehicle in a dynamic and complex environment, enhances the safety and reliability of the task execution of the unmanned aerial vehicle and has practical significance.

[0019] The present specification improves the convergence speed of the moss optimization algorithm and the ability of the moss optimization algorithm to jump out of a local optimum. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0021] Figure 1 A path planning method flow diagram is provided in the specification; Figure 2 The experimental results of the balance analysis of CLMGO and MGO; Figure 3 The results of the diversity analysis of CLMGO and MGO; Figure 4 The function experimental results; Figure 5 The three-dimensional view and top view of cases 1-8; Figure 6 A path planning device schematic diagram is provided in the specification; Figure 7 A computer device schematic diagram for implementing the path planning method is provided in the specification. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the specification clearer, the technical scheme of the present application will be described in detail below with reference to the specific embodiments of the specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the specification, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0023] The technical scheme provided by each embodiment of the present application will be described in detail below with reference to the drawings.

[0024] Figure 1 A path planning method flow diagram is provided in the specification, specifically including the following steps: S101: Determine the starting point, the ending point and the path planning region to be planned.

[0025] S102: Construct a path planning cost function.

[0026] Exemplarily, the path planning cost function can include a path length cost function, a flight height cost function, an obstacle avoidance cost function and a path smoothing cost function.

[0027] The path length cost function is: The path length cost function is: ,​​ ) and ( , , ) are the coordinates of waypoints and respectively.

[0028] Each path is considered as a candidate solution, and each waypoint on the path is denoted by = ( , , ). The total path length is calculated by accumulating the Euclidean distance between consecutive waypoints.

[0029] Second, to ensure the safety of UAV flight, the flight height needs to be limited. The minimum and maximum flight heights are set as and respectively. The corresponding flight height cost function is calculated as follows:

[0030] where the flight height is given by , which is relative to the ground. The variable is used to maintain the average height, and the values beyond the range will be penalized. C is a constant to penalize the paths that exceed the height limit. To ensure safe navigation, the generated path must be able to effectively guide the UAV to bypass spatial obstacles. To ensure safe navigation, the generated path must effectively guide the UAV to bypass spatial obstacles. Each obstacle, or threat region, is modeled as a cylindrical region characterized by its projection center and a given threat radius Rk.

[0031] The obstacle avoidance cost function is: where is the distance from the th threat center to the th path segment. The diameter is determined by the size of the UAV. is the dangerous distance, which is related to the strength of the GPS signal. is a constant to penalize the routes that enter the dangerous region.

[0032] Finally, to consider the turning performance of the UAV, the turning angle is the angle between the projections of two road segments on the horizontal Oxy plane. Let k represent the unit vector along the z-axis, the projection vectors can be calculated as:

[0033] The climbing angle ψij represents the inclination angle between the path segment and its projection on the horizontal plane The turning angle between adjacent vector projections can be calculated as: The turning angle between adjacent vector projections can be calculated as: The turning angle between adjacent vector projections can be calculated as: The path smoothness cost function can be calculated as:

[0034] wherein, and denote the penalty coefficients related to the steering angle and the climbing angle, respectively.

[0035] The path planning cost function can be calculated as: , , and c4 are the weight proportions of the path length cost function, the flight height cost function, the obstacle avoidance cost function and the path smoothness cost function, respectively.

[0036] The server mentioned in the specification can be a server arranged in a service platform, or a device such as a desktop computer, a notebook computer, etc. capable of executing the scheme of the specification. For the convenience of description, the following will be described taking the server as the execution subject.

[0037] S103: initializing the moss growth optimization algorithm by taking all paths between the starting point and the ending point as the population of the moss growth optimization algorithm and taking each path between the starting point and the ending point as an individual of the moss growth optimization algorithm, and determining the maximum number of iterations.

[0038] S104: running the moss growth optimization algorithm and optimizing the moss growth optimization algorithm by using the comprehensive learning strategy formula during the running process.

[0039] In the embodiment, the steps of the comprehensive learning strategy include: generating a random number greater than 0 and less than 1; if the random number is greater than or equal to the learning probability, setting the d-th dimension value of the best position of each individual in the population as the d-th dimension value of the learning sample vector; If the random number is less than the learning probability, the first individual and the second individual are selected from the population, when the fitness value of the best position of the first individual is less than the fitness value of the best position of the second individual, the d-th dimension value of the first individual is set as the d-th dimension value of the learning sample vector, when the fitness value of the best position of the first individual is greater than, or equal to, the fitness value of the best position of the second individual, the d-th dimension value of the second individual is set as the d-th dimension value of the learning sample vector; wherein the first individual and the second individual are different from other individuals, and the first individual and the second individual are different; When the values of all dimensions in the learning sample vector are determined by the d-th dimension values of the best positions of each individual in the population, the target individual is selected from the population and the target dimension is selected from the dimensions of the learning sample vector, and the value of the target dimension of the best position of the target individual is set as the dimension value of the target dimension of the learning sample vector; wherein the target individual is different from other individuals; The sample vector is used to guide the individual to update; The learning probability The calculation formula is as follows: Wherein, is equal to , is equal to 0.5, represents the d-th individual, represents the total number of individuals.

[0040] The running process of the moss growth optimization algorithm is as follows: A reference value is determined, each individual in the moss growth optimization algorithm is compared with the reference value, at least two subsets are determined, and the subset with the most number of individuals in the at least two subsets is taken as the population to be analyzed; According to the position relationship of the individuals in the population to be analyzed, the wind direction of the moss growth optimization algorithm is calculated; According to the wind direction, the spore diffusion operation is performed on the individuals in the population to be analyzed to generate the first individual; The individuals meeting the preset condition are searched and bred according to the double propagation strategy to generate the second individual; The first individual and the second individual are compared with the individuals in the population to be analyzed respectively, and the individual with a better path planning cost function is retained to update the population to be analyzed; When the maximum number of iterations is reached, the running is stopped.

[0041] The blowing direction of the wind direction is from the region to the reference value individual; The calculation formula of the wind direction is as follows: Wherein, ​represents the wind vector, represents the number of individuals in the population to be analyzed, represents the derivative of represents the spores of the th moss individual.

[0042] According to the wind direction, the spore diffusion operation is performed on the individuals in the population to be analyzed, and the calculation formula used is: wherein, represents the newly generated moss individual through the spore propagation process of the th moss individual, represents the spores of the th moss individual, , represents a random number, represents a threshold parameter, when , it represents the spore propagation under stable wind conditions, represents the spore propagation distance under stable wind conditions, , represents a fixed parameter, represents a random vector uniformly distributed in the interval (0, 1), which has the same dimension as , represents the wind strength, , represents the current number of function evaluations, and represents the maximum number of iterations, represents the propagation distance under turbulent wind conditions, , , represents the proportion of the number of individuals in the population to be analyzed to the number of individuals in the total population .

[0043] The individuals that meet the preset conditions are searched and reproduced according to the double propagation strategy, and the calculation formula used is: wherein, represents the newly generated individual at the index , and is the th element of , is randomly selected but is constrained by the dimension of the individual, represents the individual with the best current performance, and represents the th component of , represents the wind vector represents a fixed threshold value.

[0044] Exemplarily, let represent the best individual in the population . In order to distinguish individuals, the value of the jth dimension of each individual is taken as a reference (i.e., a reference value). By comparing the jth dimension component of each individual in the population with the jth dimension component of , two subsets can be formed: wherein represents the value of the jth dimension of the ith individual, and represents the total number of dimensions. Then the subset containing more elements is selected as follows:

[0045] wherein, is a function for determining the total number of moss individuals in a particular set. In the case of a set obtained after multiple segmentation steps, the calculation process is as shown below:

[0046] wherein, the variable represents the number of segmentation times, which is defined as in this embodiment, and its minimum value is limited to 1. The symbol represents a floor function, which returns the largest integer less than or equal to the input value.

[0047] corresponds to the jth value randomly selected from the range , and satisfies .

[0048] In this embodiment, the wind always blows from the area to the optimal individual . The wind direction is further determined according to the calculation formula of the wind direction.

[0049] The variable num represents the number of individuals contained in the set , and the calculation formula is: wherein, refers to a set of distance values, which is used to measure the distance between each individual in the set and the optimal individual . ​​​​​

[0050] The average distance between the elite individuals and is calculated, which helps to smooth the convergence trajectory and thus improves the optimization performance of the multi-objective optimization (MGO) algorithm.

[0051] In the exploration phase of the multi-objective optimization (MGO) algorithm, the search process simulates the process of spore propagation. When the wind is strong, spores will be scattered in an unpredictable way, usually resulting in long-distance propagation. In contrast, in stable wind flow, spores can be blown further; while in unstable or turbulent wind conditions, their propagation range is more limited. Most spores are released during periods of stable wind, while only a few spores are scattered in more chaotic wind patterns. As the wind weakens, spores gradually settle near the original moss source. In this embodiment, the position of each spore is considered as a potential new candidate solution. And by performing spore diffusion operations on the individuals in the population under analysis according to the wind direction, the calculation formula used simulates how spores propagate through the wind.

[0052] In the development phase of the multi-objective optimization (MGO) algorithm, a double propagation strategy is adopted, which simulates both sexual and vegetative reproduction processes. This mechanism generates new moss individuals near their parents, thereby enhancing the local search capability. When the randomly generated value satisfies the condition , the double propagation strategy is activated.

[0053] In the context of sexual reproduction, genetic-level recombination occurs, enabling offspring to inherit genetic characteristics from the current individual and the best-performing individual. In contrast, vegetative reproduction involves generating new solutions from existing moss fragments, which are also affected by wind-driven movement (similar to spore propagation). Compared to spore-based reproduction, the double propagation strategy promotes reproduction within a narrower spatial range, but this limitation helps to converge more quickly to the optimal environmental conditions. Unlike multi-agent meta-heuristic algorithms, this approach increases the frequency of modifying only one dimension of the individual solution, thereby enhancing the fine-grained local search capability.

[0054] Individuals that satisfy the preset condition are searched and reproduced according to the double propagation strategy, and in this embodiment, a fixed threshold of 0.5 is set Control behavior: when the sexual reproduction mode of the double propagation mechanism is applied; otherwise, the vegetative reproduction variant is used using an alternative formula, as shown below: where, is a random vector with values uniformly distributed in the interval (0, 1), with the same dimension as . .

[0055] For example, the comprehensive learning strategy determines a learning sample vector to guide the information transfer during the search process. (in From 1 to , is the total number of populations), this method generates a learning sample vector , which has the same dimension as the problem search space .

[0056] Specifically, for the learning sample vector Every dimension of (in From 1 to ): First, according to an individual Related learning probability Make a judgment. It is Individual learning sample vector, is an individual No. The value of the dimension, yes All individuals in the dimension The optimal solution of represents the total number of individuals, Represents the total dimension, , represents two randomly selected individuals, Represents the learning probability, which is calculated as follows:

[0057] in, , is equal to 0.5, represents the i-th individual, The total number of individuals.

[0058] In generating learning sample vectors In the process, for the current processing dimension , perform the following steps: (1) Generate a random number: Generate a random number between 0 and 1.

[0059] (2) Probability judgment: If the random number is greater than or equal to the learning probability , then the best position of individual i itself ( ) is assigned the d-th dimension value to the learning sample vector The dth dimension of .

[0060] If the random number is less than the learning probability , then do the following: a. Randomly select two different individuals: Randomly select two individuals from the population that are different from individual i and different from each other, and mark them as a and b respectively.

[0061] b. Compare fitness and select: Compare the best position of individual a ( ) and the best position of individual b ( )’s fitness value.

[0062] *If the fitness value of the best position of individual a is less than the fitness value of the best position of individual b, the d-th dimension value of the best position of individual a is assigned to the learning sample vector The dth dimension of .

[0063] * Otherwise (i.e. the fitness value of the best position of individual b is less than or equal to the fitness value of the best position of individual a), the d-th dimension value of the best position of individual b is assigned to the d-th dimension of the learning sample vector fi, i.e. .

[0064] (3) After completing the above steps, if for all dimensions d from 1 to D, the learning sample vector The value of each dimension d of points to the individual with the same dimension (i.e. For all d, an additional adjustment is required to introduce diversity. The specific adjustments are as follows: Randomly select a different individual: Randomly select an individual from the population that is different from individual i, labeled j. Randomly select a dimension: Randomly select a learning sample vector A dimension of . Replace the optimal solution: the selected individual The best position ( ) Dimension values ​​are assigned to learning sample vectors No. Dimension, that is .

[0065] (4) Through the above process, a learning sample vector fi is generated for each individual i, and each dimension value in the vector Points to the best position of an individual in the population (it may be itself or another individual) in the dth dimension. This learning sample vector will be used to guide the next update of individual i.

[0066] S105: iteratively optimize the path planning cost function using the optimized moss growth optimization algorithm until the number of iterations meets the maximum number of iterations.

[0067] S106: output the optimal solution of the path planning cost function, and apply the optimal solution as the target path to the unmanned aerial vehicle path planning.

[0068] In this embodiment, the method further comprises: The optimized moss growth optimization algorithm is tested for balance and diversity to evaluate the optimization performance of the optimized moss growth optimization algorithm.

[0069] By way of example, a comprehensive learning strategy is added to the MGO algorithm to obtain a CLMGO algorithm, which is tested on a balance experiment, and the experimental results are as shown in Figure 2 , Figure 2 Balance analysis experimental results of CLMGO and MGO. Figure 2 Balance analysis of 12 functions in the CEC 2017 test set is shown, including F1, F4, F7, F9, F11, F13, F14, F15, F16, F19, F28 and F30. Each subgraph visualizes three key curves: the red line corresponds to the exploration stage, the blue line represents the development stage, and the green line shows the increase and decrease trend. In the early stage of iteration, CLMGO usually maintains a high level of exploration search, while as the algorithm converges, the development search gradually dominates. The upward trend of the green line indicates that the global exploration search exceeds the local development search, while the downward trend indicates the opposite. By removing the implicit mechanism, CLMGO demonstrates better development performance than the original MGO on the 12 functions tested. The results show that CLMGO can more effectively balance the trade-off between exploration and development than MGO, thereby enhancing the global optimization performance.

[0070] Figure 3 Diversity analysis results of CLMGO and MGO. As shown in Figure 3 , Figure 3The diversity analysis results of the algorithm are shown. It depicts the average distance between agents during the iteration process of some benchmark functions (specifically F1, F4, F7, F9, F11, F13-F16, F19, F28 and F30). In these graphs, the x-axis represents the number of iterations, while the y-axis represents the level of diversity maintained in the population. At the beginning of the optimization process, both CLMGO and MGO exhibit high diversity due to random initialization. As iterations proceed, diversity decreases, indicating that the population is converging. Notably, in most functions, such as F13, F15 and F19, CLMGO's diversity decreases faster than MGO. This indicates that CLMGO has a stronger exploitation search capability. Overall, the diversity results verify that CLMGO not only maintains sufficient exploration search in the early stage, but also achieves effective exploitation search in the later stage, demonstrating superior balance and convergence performance.

[0071] Optionally, the optimized moss growth optimization algorithm can also be subjected to function experiments. As shown in Figure 4 , Figure 4 are the function experiment results. As can be seen from Figure 4 , CLMGO performs well in function test tasks, significantly outperforming MGO and obtaining the optimal solution; as shown in Figure 4 , CLMGO converges more easily and obtains the minimum value of the function. The comprehensive learning strategy enables individuals to obtain information from multiple high-quality solutions, rather than relying solely on a single historical optimal solution, thereby significantly enhancing global exploration ability in the early stage of optimization and reducing the risk of falling into local optima. In addition, the comprehensive learning strategy can adaptively adjust the search strategy - expanding the exploration range in the initial stage and promoting faster convergence in the later iterations, thereby improving the overall accuracy of the obtained solution. Furthermore, the adaptive learning mechanism of the comprehensive learning strategy helps to improve the stability of the solution, effectively alleviating the severe fluctuations in the convergence process. On the other hand, the original cryptobiosis mechanism in MGO, although intended to avoid local stagnation by preserving and reviving certain individuals, often introduces redundant calculations and repeated information, thereby slowing down the optimization process. Removing this mechanism in CLMGO simplifies the algorithm structure and eliminates unnecessary overhead, thereby improving overall computational efficiency.

[0072] In summary, the excellent performance of CLMGO on the CEC 2017 benchmark test set verifies the effectiveness of combining the comprehensive learning strategy with the removal of the cryptobiosis mechanism. This synergy significantly improves convergence speed, global search ability and solution stability. The proposed method not only improves optimization efficiency, but also provides more reliable solutions for complex real-world problems (such as unmanned aerial vehicle path planning), demonstrating its application potential in practical engineering scenarios.

[0073] Optionally, asFigure 5 Figures 6A-6H show three-dimensional views and top views of Case 1-8. Figure 5 To evaluate the effectiveness and advantages of the proposed CLMGO algorithm for UAV path planning in complex three-dimensional environments, CLMGO was compared with several state-of-the-art metaheuristic algorithms, including MGO, RIME, RUN, DBO, OBLGWO, CGSCA, and WDE. All experiments were conducted in a three-dimensional terrain simulation environment with increasing complexity and threat density. Eight different test scenarios (Cases 1-8) were designed to simulate realistic and challenging conditions, each containing different configurations of threat objects, distribution patterns, and terrain elevations.

[0074] To evaluate the effectiveness and advantages of the proposed CLMGO algorithm for UAV path planning in complex three-dimensional environments, CLMGO was compared with several state-of-the-art metaheuristic algorithms, including MGO, RIME, RUN, DBO, OBLGWO, CGSCA, and WDE. All experiments were conducted in a three-dimensional terrain simulation environment with increasing complexity and threat density. Eight different test scenarios (Cases 1-8) were designed to simulate realistic and challenging conditions, each containing different configurations of threat objects, distribution patterns, and terrain elevations.

[0075] The results show that CLMGO consistently produces smoother, safer trajectories with better obstacle avoidance capabilities and global path planning abilities. CLMGO also exhibits competitive convergence behavior compared to other methods. Figure 5 Three-dimensional views and top views of UAV path planning results are presented, covering different terrain configurations and obstacle distributions (Cases 1-8). The paths generated by various optimization algorithms are visualized, with the CLMGO algorithm (red solid line) consistently exhibiting superior smoothness, obstacle avoidance, and path efficiency. Compared to other algorithms, the paths generated by CLMGO are more optimized, effectively balancing flight safety and efficiency. The algorithm successfully navigates around obstacles while maintaining a stable altitude profile, minimizing unnecessary detours, and reducing total flight time. Figure 5 The top views in Figures 6A-6H further validate the effectiveness of CLMGO, as it consistently produces the most direct and efficient paths while avoiding obstacles with minimal deviation. In contrast, other methods such as WDE (red dashed line) and MGO (green dotted line) often exhibit longer and more tortuous trajectories. The results confirm that CLMGO provides the best trade-off between path length and obstacle avoidance, making it a robust solution for complex three-dimensional UAV navigation tasks. These findings highlight the adaptability and reliability of CLMGO in real-world UAV applications, ensuring safe and efficient autonomous flight.

[0076] The beneficial effects of the present application are: The present application proposes a model for realizing path planning using moss growth optimization algorithm improved by comprehensive learning strategy, which is based on comprehensive learning strategy and moss optimization algorithm, introduces the iterative behavior of comprehensive learning into the MGO optimization algorithm to obtain an improved arithmetic optimization algorithm (CLMGO) for selecting the optimal solution and obtaining higher quality parameters.

[0077] The CLMGO proposed in the present application exhibits strong search ability and superior global convergence in path optimization tasks, can realize more efficient unmanned aerial vehicle trajectory planning, and thus enhances the safety and reliability of task execution. The present application verifies the effectiveness of CLMGO in unmanned aerial vehicle path planning, and provides a solid experimental basis for further improvement of intelligent path optimization algorithms.

[0078] The path planning method provided by one or more embodiments of the present application is based on the same idea, and the present application also provides a corresponding path planning device, as shown in Figure 6 .

[0079] Figure 6 A path planning device provided by the present application is shown in the schematic diagram, which comprises: The data acquisition module is specifically configured to determine the starting point, the ending point and the path planning region to be planned. The function construction module is specifically configured to construct a path planning cost function. The algorithm optimization module is specifically configured to take all paths between the starting point and the ending point as the population of the moss growth optimization algorithm, take each path between the starting point and the ending point as an individual of the moss growth optimization algorithm, initialize the moss growth optimization algorithm and determine the maximum number of iterations, run the moss growth optimization algorithm, use a comprehensive learning strategy formula during the running process to optimize the moss growth optimization algorithm, and use the optimized moss growth optimization algorithm to iteratively optimize the path planning cost function until the number of iterations meets the maximum number of iterations. The path output module is specifically configured to output the optimal solution of the path planning cost function, take the optimal solution as the target path, and apply it to unmanned aerial vehicle path planning.

[0080] The specific limitations of the path planning device can be referred to the limitations of the path planning method described above, and will not be repeated here. The various modules in the above path planning device can be realized by software, hardware and their combinations in whole or in part. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to call and execute the operations corresponding to the above modules by the processor.

[0081] The specification also provides a computer readable storage medium storing a computer program, the computer program being used to execute the above Figure 1 The path planning method is provided.

[0082] The specification also provides a computer readable storage medium storing a computer program, the computer program being used to execute the above Figure 7 The structure diagram of the computer device is shown in the specification, and the computer device includes a processor, an internal bus, a network interface, a memory and a non-volatile memory. Figure 7 The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs to implement the above Figure 1 The path planning method is provided.

[0083] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory. The volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0084] The technical features of the above embodiments can be combined in any way. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the specification.

Claims

1. A path planning method, characterized in that: include: Determine the starting point, end point and path area to be planned; Construct path planning cost function; All paths between the starting point and the ending point are used as a population of the moss growth optimization algorithm, each path between the starting point and the ending point is used as an individual of the moss growth optimization algorithm, the moss growth optimization algorithm is initialized, and a maximum number of iterations is determined; Running the moss growth optimization algorithm, and optimizing the moss growth optimization algorithm using a comprehensive learning strategy during the running process; Iteratively optimizing the path planning cost function using the optimized moss growth optimization algorithm until the number of iterations meets the maximum number of iterations; The optimal solution of the path planning cost function is output, and the optimal solution is used as the target path for drone path planning.

2. The path planning method according to claim 1, wherein: The steps of the comprehensive learning strategy include: Generate a random number, where the random number is greater than 0 and less than 1; If the random number is greater than or equal to the learning probability, the d-th dimension value of the best position of each individual in the population is set as the d-th dimension value of the learning sample vector; If the random number is less than the learning probability, a first individual and a second individual are selected from the population; when the fitness value of the best position of the first individual is less than the fitness value of the best position of the second individual, the d-th dimension value of the first individual is set as the d-th dimension value of the learning sample vector; when the fitness value of the best position of the first individual is greater than or equal to the fitness value of the best position of the second individual, the d-th dimension value of the second individual is set as the d-th dimension value of the learning sample vector; wherein the first individual and the second individual are different from other individuals, and the first individual and the second individual are different; When the values ​​of all dimensions in the learning sample vector are determined by the d-th dimension value of the best position of each individual in the population, a target individual is selected from the population and a target dimension is selected from the dimensions of the learning sample vector, and the value of the best position of the target individual in the target dimension is set as the dimension value of the learning sample vector in the target dimension; wherein the target individual is different from other individuals; Using the sample vector to guide the individual to update; The learning probability The calculation formula is: in, equal , is equal to 0.5, Indicates the Individuals, Represents the total number of individuals.

3. The path planning method according to claim 1, wherein: The operation process of the moss growth optimization algorithm is as follows: Determining a benchmark value, comparing each individual in the moss growth optimization algorithm with the benchmark value, determining at least two subsets, and selecting the subset with the largest number of individuals among the at least two subsets as the population to be analyzed; Calculating the wind direction of the moss growth optimization algorithm according to the positional relationship of the individuals in the population to be analyzed; performing a spore spreading operation on individuals in the population to be analyzed according to the wind direction to generate a first individual; Search and reproduce the individuals that meet the preset conditions according to the double propagation strategy to generate the second individual; Comparing the first individual and the second individual with individuals in the population to be analyzed respectively, retaining individuals that are more optimal for the path planning cost function, and updating the population to be analyzed; When the maximum number of iterations is reached, the execution stops.

4. The path planning method according to claim 3, wherein: The blowing direction of the wind is from the area to the reference value individual; The calculation formula for the wind direction is: in, represents the wind vector, represents the number of individuals in the population to be analyzed, express The differential of Indicates the Spores from moss individuals.

5. The path planning method according to claim 3, wherein: According to the wind direction, the spore diffusion operation is performed on the individuals in the population to be analyzed, and the calculation formula used is: in, Indicates that through spores of moss individuals The newly generated moss individuals during the propagation process, Indicates the spores of moss individuals , represents a random number, represents the threshold parameter, when When , it means that the spores are spread under stable wind conditions, represents the spore spread distance under steady wind conditions, , represents fixed parameters, Indicates a A random vector uniformly distributed in same, Indicates strong wind. , represents the current function evaluation number, and represents the maximum number of iterations, represents the propagation distance under turbulent wind conditions, , , Indicates the population to be analyzed The number of individuals and the total population The proportion of the number of individuals.

6. The path planning method according to claim 3, wherein: The individuals that meet the preset conditions are searched and reproduced according to the double propagation strategy, and the calculation formula used is: in, Indicates that the index The newly generated individual yes No. elements, are randomly selected, but are constrained by individual dimensionality restrictions, represents the best performing individual at the moment, and express No. A quantity, Represents the wind vector The jth element of the random variable To determine reproductive strategy, Indicates a fixed threshold.

7. The path planning method according to claim 1, wherein: The method further comprises: The optimized moss growth optimization algorithm will be subjected to a balance experiment test and a diversity experiment test to evaluate the optimization performance of the optimized moss growth optimization algorithm.

8. A path planning device, characterized in that: include: The data acquisition module is specifically used to determine the starting point, the end point and the area of ​​the path to be planned; Function construction module, specifically used to construct the path planning cost function; The algorithm optimization module is specifically configured to use all paths between the starting point and the ending point as a population of the moss growth optimization algorithm and each path between the starting point and the ending point as an individual of the moss growth optimization algorithm, initialize the moss growth optimization algorithm, and determine a maximum number of iterations; run the moss growth optimization algorithm, and optimize the moss growth optimization algorithm using a comprehensive learning strategy formula during the running process; and iteratively optimize the path planning cost function using the optimized moss growth optimization algorithm until the number of iterations meets the maximum number of iterations; The path output module is specifically used to output the optimal solution of the path planning cost function, and use the optimal solution as the target path for drone path planning.

9. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

Images