Unmanned aerial vehicle path optimization method

Abstract

The invention discloses an unmanned aerial vehicle path optimization method, and belongs to the technical field of unmanned aerial vehicle path planning, and the method comprises the steps: S1, constructing an original data set based on the historical data of an unmanned aerial vehicle operation scene; s2, preprocessing the original data set based on a chaotic mapping strategy to obtain a training set and a verification set; s3, a sparrow search algorithm based on an improved individual updating mechanism is combined with the training set to construct and train a path optimization model; s4, inputting the verification set into the path optimization model for verification, and obtaining and outputting a verification result; s5, judging whether the path optimization model is qualified or not based on the verification result, if yes, outputting a qualified path optimization model, and if not, skipping to S1; s6, collecting the real-time data of the operation scene of the unmanned aerial vehicle as the input of the qualified path optimization model, and obtaining and outputting the optimal path of the unmanned aerial vehicle; according to the invention, the problem of poor accuracy in unmanned aerial vehicle path planning in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) path planning, and in particular to a UAV path optimization method. Background Art

[0002] With the development of drone technology, it is widely used in various industries to perform special operations. How to plan a safe and efficient drone path in a three-dimensional environment is the key to ensure that the drone successfully completes the task. At present, planning drone paths in a three-dimensional environment mainly relies on swarm intelligence optimization algorithms. However, the swarm intelligence optimization algorithms, especially the sparrow search algorithm, have some significant defects in practical applications: 1) The initialization of the population is random, resulting in a large number of inefficient individuals, which in turn increases the complexity of the algorithm and reduces the global search capability and accuracy of the algorithm; 2) As the number of iterations increases, the exponential decay component gradually approaches zero, resulting in a decrease in population diversity, making it difficult for the algorithm to converge to the optimal value; 3) There is a lack of information exchange between populations, and they rely entirely on their own search behavior to update their own positions and velocities, which causes the algorithm to fall into a local optimum and cannot obtain the result corresponding to the global optimum.

[0004] Chinese patent, publication number: CN 114485665 A, publication date: May 13, 2022, discloses a method for UAV trajectory planning based on a sparrow search algorithm, comprising the following steps: step (1): modeling the UAV trajectory planning to obtain a total cost function; step (2): optimizing the total cost function through a sparrow search algorithm; step (3): obtaining the optimal path for the UAV trajectory planning; and in the process of finding the optimal path for the UAV trajectory through the sparrow search algorithm, the invention suffers from the defects of high complexity and reduced individual search diversity of the sparrow search algorithm, which causes it to fall into a local optimum and makes it difficult to obtain the optimal path for the UAV. Summary of the invention

[0005] The purpose of the present invention is to solve the problem that in the prior art, due to the defects of high complexity and reduced individual search diversity of the sparrow search algorithm itself, the algorithm is easy to fall into the local optimum, resulting in poor accuracy of the drone path planning; a drone path optimization method is proposed, the original data set is preprocessed by a chaotic mapping strategy to obtain a uniformly distributed training set and a verification set, and the training set and the verification set are input into the sparrow search algorithm to make its population evenly distributed, avoid the aggregation of individuals in the population in the search space, reduce the inefficient individuals in the population, reduce the complexity of the algorithm, and improve the individual update mechanism of the sparrow search algorithm through a spiral search strategy and a population communication strategy, strengthen the flexibility of the individual and the communication between different individuals, enable the individual to explore more areas and paths, avoid the reduction of individual search diversity, enhance the global search ability of the algorithm, and effectively avoid falling into the local optimal solution during the operation of the algorithm; a path optimization model is constructed by the improved sparrow search algorithm and applied to the planning of the drone path to obtain a more accurate and reliable path planning solution.

[0006] In a first aspect, a technical solution provided in an embodiment of the present invention is a method for optimizing a drone path, comprising the following steps: S1, build the original data set based on the historical data of UAV operation scenes; S2. Preprocessing the original data set based on a chaotic mapping strategy to obtain a training set and a validation set; S3, constructing and training a path optimization model based on a sparrow search algorithm based on an improved individual update mechanism and combining the training set; S4, inputting the verification set into the path optimization model for verification, obtaining and outputting the verification result; S5, judging whether the path optimization model is qualified based on the verification result, if so, outputting the qualified path optimization model, if not, jumping to S1; S6. Collect real-time data of the UAV operation scene as input of the qualified path optimization model to obtain and output the optimal path of the UAV.

[0007] In this solution, an original data set is constructed based on historical data of the UAV operation scene, which effectively improves the efficiency of data processing of this application and significantly improves the accuracy and reliability of the subsequently constructed path optimization model; at the same time, the original data set is preprocessed based on the chaotic mapping strategy to obtain a training set and a verification set, so that the unevenly distributed data in the original data set are evenly distributed; secondly, the sparrow search algorithm with an improved individual update mechanism is combined with the training set to construct and train the path optimization model, which overcomes the defects of the sparrow search algorithm before the improvement and greatly improves the efficiency and accuracy of the path optimization model; and the verification set is used to verify the trained path optimization model to determine whether the path optimization model meets the requirements, ensuring that the trained path optimization model meets the needs of the actual operating environment, and can quickly and accurately plan the optimal flight path of the UAV, effectively preventing it from overfitting and underfitting; finally, the real-time data of the UAV operation scene is collected as the input of the qualified path optimization model for UAV path planning. At this time, the path optimization model not only meets the needs of the actual operating environment of the UAV, but also has extremely high efficiency and accuracy, and can accurately obtain and output the optimal path of the UAV in the shortest time.

[0008] Preferably, in S1, the specific process of constructing the original data set based on the historical data of the UAV operation scene is as follows: S11, establishing a three-dimensional terrain model based on the historical data of the UAV operation scene; S12, establishing a UAV path cost evaluation function based on the three-dimensional terrain model; S13, searching for a drone path based on the drone path cost evaluation function, and smoothing the searched drone path using a penalty function; S14. Integrate the smoothed UAV paths to obtain the original data set.

[0009] In this scheme, a three-dimensional terrain model is established based on the historical data of UAV operation scenes, which fully simulates the real UAV operation scenes and effectively improves the effectiveness and practicality of the three-dimensional terrain model; and a UAV path cost evaluation function is established based on the three-dimensional terrain model, through which the UAV flight path contained in the three-dimensional terrain model can be searched; secondly, after the UAV flight path is searched, the penalty function is used to smooth the searched UAV path, exclude unreasonable UAV paths, so that the accuracy of the final original data set is improved, and the efficiency and accuracy of the model are significantly improved.

[0010] Preferably, in S12, the specific process of establishing the UAV path cost evaluation function based on the three-dimensional terrain model is: S121, establishing a path length function and a flight height cost function based on the three-dimensional terrain model, and establishing a turning angle cost function based on the three-dimensional terrain model and the maximum turning angle range of the UAV; S122. Establish a UAV path cost evaluation function based on the path length function, the flight altitude cost function and the turning angle cost function.

[0011] In this scheme, when establishing the three-dimensional terrain model, not only the flight distance and flight altitude of the UAV are taken into consideration, but also the turning angle cost function is established in combination with the maximum turning angle range of the UAV, simulating the situation that the UAV may encounter obstacles during the actual flight process, so that the final optimal path of the UAV is more reasonable.

[0012] Preferably, in S2, the specific process of preprocessing the original data set based on the chaotic mapping strategy to obtain the training set and the validation set is: S21, taking the original data set as the input of the chaotic mapping strategy to obtain a mapping data set corresponding to the original data set; S22, integrating the original data set and the mapping data set into a total data set, and dividing the total data set into a training set and a test set.

[0013] In this scheme, the chaotic mapping strategy is used to process the original data set to obtain a mapping data set corresponding to the original data set, which makes the data authentic while exploring the possible real data, effectively improving the validity and comprehensiveness of the data.

[0014] Preferably, in S3, the specific process of constructing and training the path optimization model based on the sparrow search algorithm with the training set based on the improved individual update mechanism is as follows: S31, the individuals in the sparrow search algorithm with improved individual update mechanism include discoverers and joiners; Specifically, based on the discoverer update formula of the sparrow search algorithm combined with the spiral search strategy, the producer update formula is obtained, and based on the joiner update formula of the sparrow search algorithm combined with the population communication strategy, the beggar update formula is obtained; S32, constructing a path optimization model based on the producer update formula and the beggar update formula combined with the mirror reflection formula; S33, inputting the training set into the constructed path optimization model for training.

[0015] In this scheme, a spiral search strategy is adopted to update the producer update formula in the sparrow search algorithm, making the producer update process more flexible and effectively avoiding the sparrow search algorithm from falling into the local optimal situation; secondly, a population communication strategy is adopted to update the beggar update formula in the sparrow search algorithm, ensuring the convergence and population diversity of the sparrow search algorithm, effectively balancing local development and global development, thereby achieving the effect of improving the global search efficiency and accuracy.

[0016] Preferably, in S33, the specific process of inputting the training set into the constructed path optimization model for training is: S331, using the data in the training set as the sparrow individual component sparrow population; S332, calculating the fitness of each individual in the sparrow population; S333, selecting individuals with high fitness in the population as producers according to a preset proportion of producers in the population, and updating the producers according to a producer update formula; At the same time, a mirror operation is performed on the producer based on the mirror reflection formula to obtain the producer shadow, and the producer shadow is updated based on the update of the producer in combination with the mirror reflection formula; S334, selecting individuals in the population whose fitness is close to that of producers as beggars according to a preset proportion of beggars in the population, and updating the beggars according to a beggar update formula; At the same time, a mirror operation is performed on the beggar based on the mirror reflection formula to obtain the beggar's shadow, and the beggar's shadow is updated based on the update of the beggar and combined with the mirror reflection formula; S335, judging whether the sparrow population meets the preset stop updating condition, if so, stopping the updating of the sparrow population and completing the training of the path optimization model, if not, executing step S332.

[0017] In this scheme, the data in the training set is used as the sparrow individual component sparrow population. Based on the fact that the data in the training set is evenly distributed at this time, it can be obtained that the sparrow individuals are also evenly distributed, which makes the individuals in the sparrow population evenly distributed, effectively avoiding the emergence of a large number of invalid individuals, thereby reducing the complexity of the algorithm and greatly improving the efficiency and convergence speed of the algorithm; and while selecting producers and beggars for updating based on fitness as the standard, the mirror reflection formula is used to perform mirror operations on the producers and beggars to obtain their corresponding shadows, so that the exploration scope of the sparrow search algorithm is more comprehensive, effectively avoiding the sparrow search algorithm from falling into the local optimal situation, and significantly improving the convergence speed of the sparrow search algorithm; finally, it is judged whether the sparrow search algorithm meets the preset stop update condition, so that the path optimization model trained by the sparrow search algorithm meets the actual path exploration needs, ensuring the efficiency and accuracy of the path optimization model.

[0018] Preferably, in S333, the specific process of performing a mirror operation on the producer based on the mirror reflection formula to obtain the producer's shadow is: The position data corresponding to the producer is input into the mirror reflection formula for calculation to obtain the producer's shadow.

[0019] Preferably, in S334, the specific process of performing a mirror operation on the beggar based on the mirror reflection formula to obtain the beggar's shadow is: The position data corresponding to the beggar is input into the mirror reflection formula for calculation to obtain the beggar's shadow.

[0020] Preferably, in S335, the specific judgment process of judging whether the sparrow population meets the preset stop updating condition is: The current iteration number of the sparrow population is compared with the preset maximum iteration number. If the current iteration number is less than the maximum iteration number, it is determined that the sparrow population does not meet the preset stop update condition; if the current iteration number is greater than or equal to the maximum iteration number, it is determined that the sparrow population meets the preset stop update condition.

[0021] Preferably, in S5, the specific judgment process of judging whether the path optimization model is qualified based on the verification result is: The verification result is compared with the known UAV path. If the time consumed by the UAV flying along the path in the verification result is less than the time consumed by the UAV flying along the known UAV path, and the electric energy consumed by the UAV flying along the path in the verification result is less than the electric energy consumed by the UAV flying along the known UAV path, then the path optimization model is judged to be qualified; otherwise, the path optimization model is judged to be unqualified.

[0022] In this solution, the verification result is compared with the known UAV path to determine whether the path optimization model is qualified, which enhances the reliability of the path optimization model and promotes the iteration and improvement of the path optimization model.

[0023] The present invention has at least the following substantial effects: The present invention constructs an original data set based on historical data of UAV operation scenes, effectively improves the efficiency of data processing of the present application, and significantly improves the accuracy and reliability of the subsequently constructed path optimization model; at the same time, the original data set is preprocessed based on a chaotic mapping strategy to obtain a training set and a verification set, so that the unevenly distributed data in the original data set are evenly distributed, and the data in the training set is used as a sparrow individual component sparrow population to obtain a sparrow population with evenly distributed sparrow individuals, which effectively avoids the appearance of a large number of invalid individuals in the population, thereby reducing the complexity of the algorithm and greatly improving the efficiency and convergence speed of the algorithm; secondly, the sparrow search algorithm with an improved individual update mechanism is combined with the training set to construct and train the path optimization model, The defects of the sparrow search algorithm before the improvement are overcome, and the efficiency and accuracy of the path optimization model are greatly improved; and the validation set is used to verify the trained path optimization model to determine whether the path optimization model meets the requirements, ensuring that the trained path optimization model meets the needs of the actual operating environment, and can quickly and accurately plan the optimal flight path of the UAV, effectively preventing it from overfitting and underfitting; finally, real-time data of the UAV operation scene is collected as the input of the qualified path optimization model for UAV path planning. At this time, the path optimization model not only meets the needs of the actual operating environment of the UAV, but also has extremely high efficiency and accuracy, and can accurately obtain and output the optimal path of the UAV in the shortest time.

[0024] The above invention content is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0026] Figure 1 The present invention is a flow chart of a method for optimizing a UAV path. DETAILED DESCRIPTION

[0027] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0029] Embodiment 1: like Figure 1 As shown, this embodiment provides a method for optimizing a drone path, the method comprising the following steps: S1. Build the original dataset based on the historical data of UAV operation scenarios.

[0030] In one embodiment, in order to solve the problem of the original data set, this embodiment adopts a three-dimensional terrain model solution, which specifically includes the following steps: S11. Establish a three-dimensional terrain model based on historical data of UAV operation scenes; S12. Establishing a UAV path cost evaluation function based on the three-dimensional terrain model.

[0031] In one embodiment, in order to solve the problem of establishing a drone path cost evaluation function, this embodiment adopts a solution that comprehensively considers the drone flight conditions, specifically including the following steps: S121, establishing a path length function and a flight height cost function based on the three-dimensional terrain model, and establishing a turning angle cost function based on the three-dimensional terrain model and the maximum turning angle range of the UAV; S122. Establish a UAV path cost evaluation function based on the path length function, the flight altitude cost function and the turning angle cost function.

[0032] In this embodiment, the path length function is specifically as follows: Where: g(i) and g(i+1) are the coordinates of the i-th waypoint and the i+1-th waypoint, respectively, denoted as (xi ,y i ,z i ),(x i+1 ,y i+1 ,z i+1 );l i is the distance between the ith waypoint and the i+1th waypoint; n is the number of waypoints; L path is the path length function; By planning the path of the UAV using the above path length function, a shorter path can be obtained, which saves the fuel and time of the UAV and reduces the possibility of the UAV encountering unknown threats; The flight altitude cost function is as follows: Where z is the height of the obstacle in the 3D terrain model, and n is the number of obstacles; By using the above-mentioned flight altitude cost function to set the flight altitude of the drone, the terrain can be used to cover the position of the drone, which can help the drone avoid dangers such as unknown radars, and also avoid collisions between the drone and obstacles, the ground, mountains, etc. caused by flying too low. Since the maneuverability of the drone is restricted by the turning angle cost function, and during flight, the turning angle of the drone should be kept within the set maximum turning angle range, this embodiment establishes a turning angle cost function in combination with the maximum turning angle range of the drone, as follows: Among them, α i and α i+1 are the i-th and i+1-th road segment vectors, respectively, |α i | and |α i+1 | respectively α i and α i+1 The corresponding length, Φ is the maximum turning angle, δ is the current turning angle, J turn is the corner cost function; The above-mentioned drone path cost evaluation function based on the path length function, flight height cost function and turning angle cost function is as follows: J cost =w1L path +w2h height +w3J turn ; Among them, J cost is the total cost function, parameter w i (i=1,2,3…)The following conditions must be met:

[0033] S13, searching for a drone path based on the drone path cost evaluation function, and smoothing the searched drone path using a penalty function; S14. Integrate the smoothed UAV paths to obtain the original data set.

[0034] In this embodiment, the three-dimensional terrain model includes a baseline terrain model and an obstacle model. When the three-dimensional terrain model is being built, the flight area of ​​the UAV is set to a three-dimensional space of 100km×150km×3km, and a corresponding three-dimensional coordinate system is established. In addition, in order to make the flight environment of the UAV more realistic, an area that poses a threat to the UAV is added. The threat area is represented by a cylinder with a radius of R, and the cylinder is used to simulate the obstacle area, as follows: Where L k (x i ,y i ,z i ) is the kth obstacle area, (x k ,y k ) is the center coordinate of the kth obstacle, R is the area radius of the kth obstacle, and h is the height of the kth obstacle; In addition, since the path directly obtained by the UAV path cost evaluation function is too rough, it is easy to form a polygonal path that the UAV cannot fly, so the penalty function embedded in the objective function is combined with B-spline to smooth the path, as follows: Where, k∈{1,2}, is the penalty coefficient, is the penalty objective function, f(x) is the original fitness function, m represents the number of constraints, g i (x)(j=1,2...,m) is an inequality constraint, δ(g i (x)) is defined as follows:

[0035] S2. Preprocessing the original data set based on a chaotic mapping strategy to obtain a training set and a validation set.

[0036] In one embodiment, in order to pre-process the original data set, this embodiment adopts a chaotic mapping strategy, which specifically includes the following steps: S21, taking the original data set as the input of the chaotic mapping strategy to obtain a mapping data set corresponding to the original data set; S22, integrating the original data set and the mapping data set into a total data set, and dividing the total data set into a training set and a test set.

[0037] In this embodiment, a sine mapping strategy is specifically selected. By extending the sine mapping strategy from one dimension to three dimensions, the sine mapping strategy is transformed into a chaotic mapping strategy, as shown below: Where X k is a randomly generated solution, α is a parameter, X k+1 Represents the mapping solution; the core concept behind the sinusoidal mapping strategy is to systematically map candidate solutions in the search space, thereby facilitating a more unified and comprehensive search process.

[0038] S3. A sparrow search algorithm based on an improved individual update mechanism is combined with the training set to construct and train a path optimization model.

[0039] In one embodiment, in order to construct a path optimization model, this embodiment adopts a sparrow search algorithm solution, which specifically includes the following steps: S31, the individuals in the sparrow search algorithm with improved individual update mechanism include discoverers and joiners; Specifically, based on the discoverer update formula of the sparrow search algorithm combined with the spiral search strategy, the producer update formula is obtained, and based on the joiner update formula of the sparrow search algorithm combined with the population communication strategy, the beggar update formula is obtained; S32, constructing a path optimization model based on the producer update formula and the beggar update formula combined with the mirror reflection formula; S33, inputting the training set into the constructed path optimization model for training.

[0040] In one embodiment, in order to train the path optimization model, this embodiment adopts an improved sparrow search algorithm, which specifically includes the following steps: S331, using the data in the training set as the sparrow individual component sparrow population; S332, calculating the fitness of each individual in the sparrow population; S333, selecting individuals with high fitness in the population as producers according to a preset proportion of producers in the population, and updating the producers according to a producer update formula; At the same time, a mirror operation is performed on the producer based on the mirror reflection formula to obtain the producer shadow, and the producer shadow is updated based on the update of the producer in combination with the mirror reflection formula.

[0041] In one embodiment, the specific process of performing a mirror operation on the producer based on the mirror reflection formula to obtain the producer's shadow is: The position data corresponding to the producer is input into the mirror reflection formula for calculation to obtain the producer's shadow.

[0042] In this embodiment, the mirror reflection formula is as follows: Assuming B0 = λA0, where λ is a real number greater than 0, the original equation can be transformed into: b=(λ+1)x0-λa=(0.5λ+0.5)×(Z L +Z U )-λa; Among them, the value of b has the following possibilities: And the value of λ in the mirror reflection learning is calculated by the following formula: Combining mirror reflection learning and intelligent optimization algorithm, we can get the formula: x=(0.5λ+0.5)×(L j +U j )-λx j ,j=1,2,...D; In the above formula, α is the incident angle, β is the reflection angle, and x0 is the distance between point A0 and point B0 on ​​the horizontal axis and the origin O.

[0043] S334, selecting individuals in the population whose fitness is close to that of producers as beggars according to a preset proportion of beggars in the population, and updating the beggars according to a beggar update formula; At the same time, a mirror operation is performed on the beggar based on the mirror reflection formula to obtain the beggar's shadow, and the beggar's shadow is updated based on the update of the beggar and combined with the mirror reflection formula.

[0044] In one embodiment, the specific process of performing a mirror operation on the beggar based on the mirror reflection formula to obtain the beggar's shadow is: The position data corresponding to the beggar is input into the mirror reflection formula for calculation to obtain the beggar's shadow.

[0045] In this embodiment, the mirror reflection formula is as follows: Assuming B0 = λA0, where λ is a real number greater than 0, the original equation can be transformed into: b=(λ+1)x0-λa=(0.5λ+0.5)×(Z L +Z U )-λa; Among them, the value of b has the following possibilities: And the value of λ in the mirror reflection learning is calculated by the following formula: Combining mirror reflection learning and intelligent optimization algorithm, we can get the formula: x=(0.5λ+0.5)×(L j +U j )-λx j ,j=1,2,...D; In the above formula, α is the incident angle, β is the reflection angle, and x0 is the distance between point A0 and point B0 on ​​the horizontal axis and the origin O.

[0046] S335, judging whether the sparrow population meets the preset stop updating condition, if so, stopping the updating of the sparrow population and completing the training of the path optimization model, if not, executing step S332.

[0047] In one embodiment, the specific judgment process of judging whether the sparrow population satisfies the preset stop updating condition is as follows: comparing the current iteration number of the sparrow population with the preset maximum iteration number; if the current iteration number is less than the maximum iteration number, judging that the sparrow population does not satisfy the preset stop updating condition; if the current iteration number is greater than or equal to the maximum iteration number, judging that the sparrow population satisfies the preset stop updating condition.

[0048] In this embodiment, the discoverer update formula based on the sparrow search algorithm is combined with the spiral search strategy to obtain the producer update formula. The spiral search strategy is insensitive to the initial point, which also enhances the robustness and repeatability of the algorithm, as shown below: β=e bl *cos(2πb); Wherein, b is a random number ranging from 0 to 1, l changes dynamically with the number of iterations t, and the specific spiral change is controlled by the cosine function. The producer update formula is as follows: Among them, Q is a random number that obeys the normal distribution, R2 is the alarm value, and ST is the safety threshold; The beggar update formula is specifically as follows: C = a*(2*rand-1); Among them, iter max is the maximum number of iterations, t is the current number of iterations, and Z is a random value in the range of [-C, C] of the problem dimension; The preset maximum number of iterations is set to 500, and the sparrow population size is set to 30.

[0049] S4, inputting the verification set into the path optimization model for verification, obtaining and outputting the verification result; S5. Determine whether the path optimization model is qualified based on the verification result. If so, output the qualified path optimization model. If not, jump to S1.

[0050] In one embodiment, the specific judgment process of judging whether the path optimization model is qualified based on the verification result is: The verification result is compared with the known UAV path. If the time consumed by the UAV flying along the path in the verification result is less than the time consumed by the UAV flying along the known UAV path, and the electric energy consumed by the UAV flying along the path in the verification result is less than the electric energy consumed by the UAV flying along the known UAV path, then the path optimization model is judged to be qualified; otherwise, the path optimization model is judged to be unqualified.

[0051] S6. Collect real-time data of the UAV operation scene as input of the qualified path optimization model to obtain and output the optimal path of the UAV.

[0052] The following CEC2022 test set is used to compare and analyze this application with other swarm intelligence algorithms and other SSA variants to further verify the performance of this application. The CEC2022 test set contains 10 dimensions (10D) and 20 dimensions (20D), as well as unimodal functions (F1), basic functions (F2-F5), mixed functions (F6-F8), composite functions (F9-F12), and a total of 12 test functions.

[0053] Compared with other swarm intelligence algorithms: This section selects the sparrow search algorithm (SSA), particle swarm optimization algorithm (PSO), whale optimization algorithm (WOA) and artificial gorilla troop optimizer (GTO) for comparative analysis with this application (ISSA). Their algorithm parameters are set according to the literature, the population size is set to 30, and the maximum number of iterations is set to 500; at the same time, in order to reduce the contingency of the experiment, each algorithm is independently run 50 times on 12 benchmark functions, and the mean and standard deviation of the function are calculated. Among them, the black bold indicates the optimal value of different algorithms under the same function under the same dimension. The test results are as follows: Table 1. Mean and standard deviation (10D) of other swarm intelligence algorithms under the CEC 2022 test suite Table 2. Mean and standard deviation of other swarm intelligence algorithms under the CEC 2022 test suite (20D)

[0054] Through the analysis of Table 1, it can be seen that in the 10-dimensional space, the statistical results of the present application across 12 test functions are significantly better than the other four comparison algorithms; first, in terms of unimodal function (F1), the present application, particle swarm optimization algorithm, sparrow search algorithm and artificial gorilla army optimizer all obtained the best average value, but the present application showed better performance in terms of standard deviation. Although the best average value of the basic function (F2-F5) is not always obtained, the difference between individual results is very small, so the present application always shows an advantage in standard deviation; secondly, in terms of mixed functions (F6-F8), except for the large difference in the average value of the F6 function, the difference in results between other algorithms is very small, but it can still be seen that the optimization results of the present application are slightly better than those of other algorithms; finally, in terms of composite functions (F9-F12), the present application has always been better than other algorithms in terms of average value, and only the standard deviation of the composite function F10 is lower than that of the artificial gorilla army optimizer.

[0055] Through the analysis of Table 2, it can be seen that the statistical results of the present application corresponding to 12 groups of test functions in 20 dimensions are significantly better than those of the other four comparison algorithms; among the 12 test functions, except for the poor calculation result of the unimodal function (F1), the other functions all obtained good rankings; in addition, according to the results, the stability and robustness of the present application are significantly better than other algorithms, and are closer to the optimal solution of the theoretical value, which shows that the present application can fully and effectively explore the search space, while ensuring strong global optimization capabilities and local optimization capabilities.

[0056] Comparison with other SSA variants: This section selects three different improved sparrow search algorithms for comparative analysis with this application, including CLSSA, CSSA and CTSSA; in the process of comparative analysis, each algorithm adjusts parameters according to the settings in the reference, sets the sparrow population size to 30, sets the number of iterations to 500, and each algorithm runs each test function independently 50 times, and obtains the optimization results shown in the table: Table 3. Mean and standard deviation (10D) of other sparrow search algorithm variants under the CEC2022 test suite Table 4. Mean and standard deviation of other Sparrow Search algorithm variants under the CEC2022 test suite (20D)

[0057] It can be seen from Tables 3 and 4 that in the results of most test functions, the average values ​​and standard deviations of the present application on 12 functions are good, which reflects that the present application has higher convergence accuracy and robustness than CLSSA, CSSA and CTSSA; specifically, for the single-mode function (F1), the present application, CLSSA and CTSSA can all obtain the optimal value, but the standard deviation of the present application is relatively small; for the basic functions (F2-F5), only the present application can find the theoretical optimal solution of F3, and the standard deviation is small, followed by F2, F4 and F5, when the four algorithms are all unavailable. When the theoretical optimal value is obtained by the method, the present application shows good convergence among several algorithms; for the mixed functions (F6-F8), the results of the four algorithms are basically in the same order of magnitude, slightly better than other centralized variants; for the composite functions (F9-F10), the standard deviation calculated by the present application is slightly worse, but a better average value can be obtained; as shown in Table 4, the present application cannot obtain a smaller standard deviation in functions F4 and F10. In addition, under the other 10 test functions, the obtained average values ​​and standard deviations are optimal, which shows that the global search ability of the algorithm of the present application is superior.

[0058] So far, compared with the other three improved SSAs, this application proposes an efficient improvement strategy that is more in line with SSA from the perspectives of population initialization, individual iteration and mutation update; in the initial stage of the population, the above variables all adopt chaotic mapping; in the population iteration process, compared with the logarithmic spiral strategy and adaptive step strategy introduced by CLSSA, this application innovatively proposes a new improvement strategy, which makes full use of the information exchange between sparrow individuals and realizes the local development and global search capabilities of the equilibrium algorithm.

[0059] This embodiment has at least the following substantial effects: The present invention constructs an original data set based on historical data of UAV operation scenes, effectively improves the efficiency of data processing of the present application, and significantly improves the accuracy and reliability of the subsequently constructed path optimization model; at the same time, the original data set is preprocessed based on a chaotic mapping strategy to obtain a training set and a verification set, so that the unevenly distributed data in the original data set are evenly distributed, and the data in the training set is used as a sparrow individual component sparrow population to obtain a sparrow population with evenly distributed sparrow individuals, which effectively avoids the appearance of a large number of invalid individuals in the population, thereby reducing the complexity of the algorithm and greatly improving the efficiency and convergence speed of the algorithm; secondly, the sparrow search algorithm with an improved individual update mechanism is combined with the training set to construct and train the path optimization model, The defects of the sparrow search algorithm before the improvement are overcome, and the efficiency and accuracy of the path optimization model are greatly improved; and the validation set is used to verify the trained path optimization model to determine whether the path optimization model meets the requirements, ensuring that the trained path optimization model meets the needs of the actual operating environment, and can quickly and accurately plan the optimal flight path of the UAV, effectively preventing it from overfitting and underfitting; finally, real-time data of the UAV operation scene is collected as the input of the qualified path optimization model for UAV path planning. At this time, the path optimization model not only meets the needs of the actual operating environment of the UAV, but also has extremely high efficiency and accuracy, and can accurately obtain and output the optimal path of the UAV in the shortest time.

[0060] The above specific embodiments are preferred embodiments of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the present specific embodiments. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A method for optimizing a drone path, characterized in that: The method comprises the following steps: S1, build the original data set based on the historical data of UAV operation scenes; S2. Preprocessing the original data set based on a chaotic mapping strategy to obtain a training set and a validation set; S3, constructing and training a path optimization model based on a sparrow search algorithm based on an improved individual update mechanism and combining the training set; S4, inputting the verification set into the path optimization model for verification, obtaining and outputting the verification result; S5, judging whether the path optimization model is qualified based on the verification result, if so, outputting the qualified path optimization model, if not, jumping to S1; S6. Collect real-time data of the UAV operation scene as input of the qualified path optimization model to obtain and output the optimal path of the UAV.

2. A method for optimizing a drone path according to claim 1, characterized in that: In S1, the specific process of constructing the original data set based on the historical data of the UAV operation scene is as follows: S11. Establish a three-dimensional terrain model based on historical data of UAV operation scenes; S12, establishing a UAV path cost evaluation function based on the three-dimensional terrain model; S13, searching for a drone path based on the drone path cost evaluation function, and smoothing the searched drone path using a penalty function; S14. Integrate the smoothed UAV paths to obtain the original data set.

3. A method for optimizing a drone path according to claim 2, characterized in that: In S12, the specific process of establishing the UAV path cost evaluation function based on the three-dimensional terrain model is as follows: S121, establishing a path length function and a flight height cost function based on the three-dimensional terrain model, and establishing a turning angle cost function based on the three-dimensional terrain model and the maximum turning angle range of the UAV; S122. Establish a UAV path cost evaluation function based on the path length function, the flight altitude cost function and the turning angle cost function.

4. The method for optimizing a drone path according to claim 1, characterized in that: In S2, the original data set is preprocessed based on the chaotic mapping strategy to obtain the training set and the verification set. The specific process is: S21, using the original data set as the input of the chaotic mapping strategy to obtain a mapping data set corresponding to the original data set; S22, integrating the original data set and the mapped data set into a total data set, and dividing the total data set into a training set and a test set.

5. The method for optimizing a drone path according to claim 1, characterized in that: In S3, the specific process of constructing and training the path optimization model based on the sparrow search algorithm with the training set based on the improved individual update mechanism is as follows: S31, the individuals in the sparrow search algorithm with the improved individual update mechanism include discoverers and joiners; specifically, based on the discoverer update formula of the sparrow search algorithm combined with the spiral search strategy, the producer update formula is obtained, and based on the joiner update formula of the sparrow search algorithm combined with the population communication strategy, the beggar update formula is obtained; S32, constructing a path optimization model based on the producer update formula and the beggar update formula combined with the mirror reflection formula; S33, inputting the training set into the constructed path optimization model for training.

6. A method for optimizing a drone path according to claim 5, characterized in that: In S33, the specific process of inputting the training set into the constructed path optimization model for training is as follows: S331, using the data in the training set as the sparrow individual component sparrow population; S332, calculating the fitness of each individual in the sparrow population; S333, selecting individuals with high fitness in the population as producers according to a preset proportion of producers in the population, and updating the producers according to a producer update formula; At the same time, a mirror operation is performed on the producer based on the mirror reflection formula to obtain the producer shadow, and the producer shadow is updated based on the update of the producer in combination with the mirror reflection formula; S334, selecting individuals in the population whose fitness is close to that of producers as beggars according to a preset proportion of beggars in the population, and updating the beggars according to a beggar update formula; At the same time, a mirror operation is performed on the beggar based on the mirror reflection formula to obtain the beggar's shadow, and the beggar's shadow is updated based on the update of the beggar and combined with the mirror reflection formula; S335, judging whether the sparrow population meets the preset stop updating condition, if so, stopping the updating of the sparrow population and completing the training of the path optimization model, if not, executing step S332.

7. A method for optimizing a drone path according to claim 6, characterized in that: In S333, the specific process of performing a mirror operation on the producer based on the mirror reflection formula to obtain the producer's shadow is: The position data corresponding to the producer is input into the mirror reflection formula for calculation to obtain the producer's shadow.

8. The method for optimizing a drone path according to claim 6, characterized in that: In S334, the specific process of performing a mirror operation on the beggar based on the mirror reflection formula to obtain the beggar's shadow is: The position data corresponding to the beggar is input into the mirror reflection formula for calculation to obtain the beggar's shadow.

9. The method for optimizing a drone path according to claim 6, characterized in that: In S335, the specific judgment process of judging whether the sparrow population meets the preset stop updating condition is as follows: Compare the current iteration number of the sparrow population with the preset maximum iteration number, if the current iteration number is less than the maximum iteration number, it is determined that the sparrow population does not meet the preset stop update condition; If the current number of iterations is greater than or equal to the maximum number of iterations, it is determined that the sparrow population meets the preset stop updating condition.

10. The method for optimizing a drone path according to claim 1, characterized in that: In S5, the specific judgment process of judging whether the path optimization model is qualified based on the verification result is: The verification result is compared with the known UAV path. If the time consumed by the UAV flying along the path in the verification result is less than the time consumed by the UAV flying along the known UAV path, and the electric energy consumed by the UAV flying along the path in the verification result is less than the electric energy consumed by the UAV flying along the known UAV path, then the path optimization model is judged to be qualified; otherwise, the path optimization model is judged to be unqualified.

Images