Unmanned aerial vehicle path optimization method based on improved fireworks algorithm

Abstract

The invention discloses an unmanned aerial vehicle path optimization method based on an improved fireworks algorithm, and belongs to the technical field of unmanned aerial vehicle path planning, and the method comprises the steps: S1, carrying out the preprocessing of unmanned aerial vehicle electric tower inspection historical data, and obtaining a training data set and a verification data set; s2, constructing a self-adaptive path planning model based on a fireworks algorithm improved by an optimized explosion strategy; s3, training the adaptive path planning model based on the training data set; s4, performing reliability verification on the trained adaptive path planning model based on the verification data set, if the reliability verification is qualified, executing S5, and if the reliability verification is not qualified, executing S3; s5, inputting the unmanned aerial vehicle electric tower inspection real-time data into the adaptive path planning model which is qualified in reliability verification for testing, and obtaining an unmanned aerial vehicle electric tower inspection optimal path; according to the invention, the problem of poor accuracy during 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 based on an improved fireworks algorithm. Background Art

[0003] With the development of drone technology, the existing technology no longer relies entirely on manual labor when inspecting power towers, but instead adopts drone-assisted methods to inspect power towers. The path planning of drones in power tower inspections mainly relies on classic traveling salesman problem-solving algorithms, including but not limited to heuristic algorithms, exact algorithms, and meta-heuristic algorithms. However, due to the complex tower inspection environment, the large number of inspection points, and the limited battery power of drones, the traveling salesman problem-solving algorithm has the following problems when planning drone tower inspection paths: 1) The traditional traveling salesman problem algorithm and its variant algorithms have the problems of high computational complexity and low efficiency when dealing with multiple constraints such as power constraints and charging requirements in drone path planning; 2) Many existing meta-heuristic algorithms are prone to falling into local optimality and are difficult to cover the entire search space, resulting in the final path being not the global optimal; 3) The dynamic planning of the charging station location is insufficient, making it difficult to achieve an effective combination of the charging station and the inspection path.

[0004] Chinese patent, publication number: CN 110836670 A, publication date: February 25, 2020, discloses a hybrid fireworks particle swarm collaborative method for solving UAV constrained route planning, including using two populations to search the optimal path in parallel and independently, one population uses an improved fireworks algorithm for search, and the other population uses a particle swarm optimization algorithm, wherein fireworks explode to find the approximate area of ​​the optimal solution in the entire search space and provide it to particles, guiding the direction of the subsequent search of the particles, and the particles perform detailed local searches in the area in the next iteration process, so that the two populations are combined to search and obtain the optimal solution to the planning problem; however, when planning the UAV path of this invention, the fireworks explosion range of the fireworks algorithm is difficult to cover the entire search space, causing it to fall into a local optimum, and ultimately making the obtained UAV path not the global optimal path. Summary of the invention

[0005] The purpose of the present invention is to address the problem that when planning the path of a UAV power tower inspection in the prior art, the explosion range of fireworks in the fireworks algorithm adopted cannot completely cover the search space, causing it to fall into a local optimum, resulting in poor accuracy in the path planning of the UAV; a UAV path optimization method based on an improved fireworks algorithm is proposed, and the fireworks algorithm is improved by optimizing the explosion strategy through a differential spark mechanism, a dual-channel selection mechanism and a reinitialization mechanism so that it can cover the entire search space, enhance the global search capability of the algorithm, and effectively avoid the algorithm from falling into a local optimal solution when planning the path of the UAV; and an adaptive path planning model is constructed through the improved fireworks algorithm to plan the UAV power tower inspection path, so as to obtain a more accurate and reliable UAV path planning scheme.

[0006] In a first aspect, a technical solution provided in an embodiment of the present invention is a method for optimizing a drone path based on an improved fireworks algorithm, comprising the following steps: S1. Preprocess the historical data of UAV tower inspection to obtain training data set and verification data set; S2, construct an adaptive path planning model based on the fireworks algorithm improved by optimizing the explosion strategy; S3, training the adaptive path planning model based on the training data set; S4, performing reliability verification on the trained adaptive path planning model based on the verification data set, if the reliability verification is qualified, executing S5, if the reliability verification is unqualified, executing S3; S5. Input the real-time data of the UAV tower inspection into the adaptive path planning model that has passed the reliability verification for testing to obtain the optimal path for the UAV.

[0007] In this solution, the historical data of drone tower inspections are preprocessed to obtain training data sets and verification data sets, which effectively improves the data quality, enables it to support a more effective data analysis process, and greatly reduces the time and resource consumption of data processing in this application; secondly, the fireworks algorithm is improved based on the optimized explosion strategy, and the improved fireworks algorithm is used to construct an adaptive path planning model for path planning, which overcomes the defects of the original fireworks algorithm itself, and at the same time enables drones to cover the entire area more efficiently when inspecting towers, effectively improving the efficiency of inspections; secondly, using the training data set to train the adaptive path planning model can optimize the performance of the adaptive path planning model, making it usable The optimal inspection path for the UAV can be comprehensively, accurately and quickly planned according to the actual operating environment of the UAV, and the verification data set can be used to verify whether the adaptive path planning model is qualified, and further determine whether the trained adaptive path planning model meets the needs of the actual operating environment, which greatly improves the accuracy of the adaptive path planning model and effectively prevents the adaptive path planning model from overfitting and underfitting. Finally, the real-time data of the UAV tower inspection is tested through the adaptive path planning model that meets the needs of the actual operating environment of the UAV, and the optimal path for the UAV tower inspection is obtained quickly and accurately, so that the UAV can complete the tower inspection task accurately in the shortest time according to the working standards.

[0008] Preferably, in S1, the specific process of preprocessing the historical data of the UAV power tower inspection to obtain the training data set and the verification data set is: Establish a three-dimensional coordinate system based on the historical data of the UAV power tower inspection, and convert the historical data of the UAV power tower inspection into a three-dimensional coordinate point representation; At the same time, the data converted into three-dimensional coordinate points are divided into a training data set and a validation data set.

[0009] In this solution, the historical data of drone tower inspections are marked in the form of three-dimensional coordinates, and a three-dimensional coordinate system is established to simulate the actual working environment of drone tower inspections. Not only is the obstacle area in the drone inspection environment shown in the form of data, but the inspection points and charging points can also be clearly identified, providing strong support for drone path planning.

[0010] Preferably, in S2, the specific process of constructing the adaptive path planning model based on the fireworks algorithm improved by optimizing the explosion strategy is: S21. Using the differential spark mechanism to improve the fireworks explosion operation of the fireworks algorithm, and using the dual-channel selection mechanism to improve the new fireworks selection operation of the fireworks algorithm, and using the reinitialization mechanism to improve the fireworks mutation operation of the fireworks algorithm; S22. Construct an adaptive path planning model based on the improved fireworks algorithm.

[0011] In this solution, in the fireworks explosion operation of the fireworks algorithm, a differential spark mechanism is used instead of the traditional Gaussian spark operator, which enhances the information exchange between multiple fireworks, promotes the global search capability of the fireworks algorithm, and effectively avoids the fireworks algorithm from falling into the local optimum; secondly, in order to avoid the problem of premature convergence of the fireworks algorithm, the present application adopts a dual-channel selection mechanism to improve the new fireworks selection operation of the fireworks algorithm, further enhances the global search capability of the fireworks algorithm, enables the fireworks algorithm to find the optimal solution faster, and effectively improves the accuracy and efficiency of the adaptive path planning model; finally, a reinitialization mechanism is used to improve the fireworks mutation operation of the fireworks algorithm, so that the evolution of individuals in the fireworks population is kept at the same pace as the evolution of the population, effectively saving computing resources, and effectively preventing the adaptive path planning model constructed based on the fireworks algorithm from being distorted, significantly improving the accuracy of the adaptive path planning model.

[0012] Preferably, in S21, the specific process of the fireworks explosion operation of the fireworks algorithm improved by using the differential spark mechanism is as follows: Where β is the differential scaling factor, is the differential vector, These are three fireworks chosen at random.

[0013] Preferably, in S21, the specific process of the new fireworks selection operation using the dual-channel selection mechanism to improve the fireworks algorithm is as follows: In the formula, is the differential fireworks selected from channel 2, x k is another part of the new firework selected from channel 1, x k+1 It's new fireworks.

[0014] Preferably, in S21, the specific process of using the reinitialization mechanism to improve the fireworks variation operation of the fireworks algorithm is as follows: In the formula, k max is the maximum number of iterations of the algorithm, k is the current number of iterations, is the current optimization increment, It is the expected optimization result of fireworks.

[0015] Preferably, in S3, the specific process of training the adaptive path planning model based on the training data set is: S31, generating a first generation of fireworks based on the adaptive path planning model in combination with the training data set, and calculating the fitness of the first generation of fireworks; S32, based on the battery capacity of the drone and the battery consumption rate per unit distance in the constraints of the adaptive path planning model, dividing the first-generation fireworks into at least flight path fireworks and charging path fireworks; S33, based on the fitness of the flight path fireworks, performing explosion operations and mutation operations on the flight path fireworks; at the same time, based on the fitness of the charging path fireworks, performing explosion operations and mutation operations on the charging path fireworks; S34, selecting new flight path fireworks based on the explosion operations and mutation operations of the flight path fireworks, and calculating the fitness of the new flight path fireworks; At the same time, selecting new charging path fireworks based on the explosion operation and mutation operation of the charging path fireworks, and calculating the fitness of the new charging path fireworks; S35, determining whether the training of the adaptive path planning model meets the stopping condition, if so, outputting the flight path fireworks with the minimum fitness and the charging path fireworks with the minimum fitness to obtain the optimal inspection path of the UAV, if not, executing S31.

[0016] In this solution, the adaptive path planning model is trained using a training data set, which can improve the performance of the adaptive path planning model so that it meets the needs of the actual operating environment of the drone and ensures that the optimal inspection path of the drone can be planned quickly and accurately even in a complex environment.

[0017] Preferably, in S35, the specific judgment process of judging whether the training of the adaptive path planning model satisfies the stopping condition is as follows: The current number of iterations of the adaptive path planning model is compared with the preset maximum number of iterations. If the current number of iterations is less than the maximum number of iterations, it is determined that the adaptive path planning model does not meet the stopping condition. If the current number of iterations is greater than or equal to the maximum number of iterations, it is determined that the adaptive path planning model meets the stopping condition.

[0018] In this solution, the preset maximum number of iterations is used as a standard to determine whether the adaptive path planning model meets the stopping condition, thereby ensuring the completion of the training of the adaptive path planning model, thereby ensuring the performance of the adaptive path planning model and enhancing the reliability of the adaptive path planning model.

[0019] Preferably, in S35, the specific process of outputting the flight path fireworks with the smallest fitness and the charging path fireworks with the smallest fitness to obtain the optimal inspection path of the drone is: The three-dimensional coordinates in the flight path fireworks and the charging path fireworks are sorted and connected according to the spatial relationship to obtain the optimal inspection path for the UAV.

[0020] In this scheme, the optimal inspection path of the drone not only takes into account the flight path of the drone, but also the charging path of the drone, so that the drone can complete the tower inspection task in one go, reducing the overall time cost and effectively improving the efficiency and time utilization of the drone inspection.

[0021] Preferably, in S4, the specific process of performing reliability verification on the trained adaptive path planning model based on the verification data set is: S41, counting the number of times the verification data set is input into the adaptive path planning model for testing and the optimal path of the drone corresponding to the verification data set is successfully obtained, and the total number of times the verification data set is input into the adaptive path planning model for testing; S42. Calculate the accuracy of the adaptive path planning model based on the number of successes and the total number of times. If the accuracy is greater than or equal to a preset threshold, determine that the adaptive path planning model reliability verification is qualified. If the accuracy is less than the preset threshold, determine that the adaptive path planning model reliability verification is unqualified.

[0022] In this solution, a verification data set is used to verify the reliability of the adaptive path planning model, which further ensures the performance of the adaptive path planning model, effectively guarantees the efficiency and accuracy of the adaptive path planning model, and greatly improves the reliability of the optimal path of the drone obtained in the end.

[0023] The present invention has at least the following substantial effects: In this application, the historical data of UAV power tower inspections are preprocessed to obtain training data sets and verification data sets, which effectively improves the data quality, enables it to support a more effective data analysis process, and greatly reduces the time and resource consumption of data processing in this application; secondly, the fireworks algorithm is improved based on the optimized explosion strategy, and the improved fireworks algorithm is used to construct an adaptive path planning model for path planning, which overcomes the defects of the original fireworks algorithm itself, and at the same time enables the UAV to cover the entire area more efficiently when inspecting power towers, effectively improving the inspection efficiency, and while constructing the adaptive path planning model, the battery life issues related to the UAV battery power are also considered, and a comprehensive model that can automatically schedule according to the UAV battery power is obtained; secondly, the training data set is used to train the automatic The adaptive path planning model can optimize the performance of the adaptive path planning model, so that it can comprehensively, accurately and quickly plan the optimal inspection path of the UAV according to the actual operating environment of the UAV, and use the verification data set to verify whether the adaptive path planning model is qualified, and further determine whether the trained adaptive path planning model meets the needs of the actual operating environment, which greatly improves the accuracy of the adaptive path planning model and effectively prevents the adaptive path planning model from overfitting and underfitting; finally, the real-time data of the UAV tower inspection is tested through the adaptive path planning model that meets the needs of the actual operating environment of the UAV, and the optimal path for the UAV tower inspection is obtained quickly and accurately, so that the UAV can accurately complete the tower inspection task in the shortest time according to the working standards.

[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 This is a flow chart of the UAV path optimization method based on the improved fireworks algorithm of the present invention; Figure 2 This is a distribution diagram of fireworks populations generated by differential sparks in the present invention. 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 based on an improved fireworks algorithm, and the method includes the following steps: S1. Preprocess the historical data of UAV tower inspection to obtain training data set and verification data set.

[0030] In one embodiment, in order to solve the problem of identifying the flight obstacles and charging points of the drone, this embodiment adopts a three-dimensional coordinate system solution, which specifically includes: Establish a three-dimensional coordinate system based on the historical data of the UAV power tower inspection, and convert the historical data of the UAV power tower inspection into a three-dimensional coordinate point representation; At the same time, the data converted into three-dimensional coordinate points are divided into a training data set and a validation data set.

[0031] In this embodiment, the drone tower inspection history data includes at least detailed tower data, inspection point data and charging point data; first, a three-dimensional graph of the tower is drawn in Matlab according to the detailed data of the tower, and important nodes and components such as insulators, connection points, support rods and lightning rods in the three-dimensional graph of the tower are marked as inspection points; in addition, the charging position set by the tower itself in the three-dimensional graph of the tower is marked as a charging point, and based on the three-dimensional graph of the tower and the marked inspection points and charging points, the three-dimensional coordinates of each inspection point and charging point are calculated, and the calculated three-dimensional coordinates are input into the three-dimensional graph of the tower to replace the marked inspection points and charging points; finally, the relevant data of the inspection points and charging points are recorded in a list, and the list is divided to obtain a training data set and a verification data set.

[0032] S2. Construct an adaptive path planning model based on the fireworks algorithm improved by optimizing the explosion strategy.

[0033] In one embodiment, in order to construct an adaptive path planning model, this embodiment adopts an improved fireworks algorithm solution, which specifically includes the following steps: S21. A differential spark mechanism is used to improve the fireworks explosion operation of the fireworks algorithm, and a dual-channel selection mechanism is used to improve the new fireworks selection operation of the fireworks algorithm. At the same time, a reinitialization mechanism is used to improve the fireworks mutation operation of the fireworks algorithm.

[0034] In one embodiment, in order to improve the fireworks explosion operation of the fireworks algorithm, this embodiment adopts a differential spark mechanism solution, which specifically includes: Where β is the differential scaling factor, is the differential vector, These are three fireworks chosen at random.

[0035] In this embodiment, the exploding fireworks of the fireworks algorithm are generated strictly according to the amplitude and radius of the explosion. They will gradually converge with the increase of the number of iterations and are very likely to fall into the local optimal situation. In the traditional fireworks algorithm, Gaussian mutation sparks are used to solve this problem, but Gaussian mutation sparks have the problem of uneven distribution, and Gaussian mutation sparks are only generated by randomly selected fireworks, resulting in a lack of information exchange between different fireworks individuals during the iteration process of the fireworks algorithm. Therefore, in this application, a differential spark mechanism is used instead of Gaussian mutation sparks to strengthen the information exchange between individual fireworks.

[0036] like Figure 2 As shown in the figure, a new fireworks population is generated using the differential spark mechanism. The red star represents the position of a single firework, the blue dot represents the position of the explosion firework, and the green X represents the position of the differential mutation firework. Figure 2 It can be clearly found that the positions of most differential mutation fireworks and fireworks maintain a considerable distance, which proves that the fireworks explosion operation of the fireworks algorithm improved by using the differential spark mechanism enhances the diversity of the fireworks population.

[0037] In one embodiment, in order to improve the new fireworks selection operation of the fireworks algorithm, this embodiment adopts a dual-channel selection mechanism solution, which specifically includes: In the formula, is the differential fireworks selected from channel 2, x k is another part of the new firework selected from channel 1, x k+1 It's new fireworks.

[0038] In this embodiment, when the traditional fireworks algorithm performs the selection operation of new fireworks, it adopts the elite selection strategy to select the guide fireworks and the explosion fireworks in one channel, and selects the better scheme as the new fireworks of the next generation. The elite selection strategy can speed up the convergence speed of the fireworks algorithm, but it is prone to the problems of premature convergence and lack of diversity. Therefore, the present application uses a dual-channel selection mechanism to select new fireworks, and the dual-channel selection mechanism formula and x k All are selected from the previous generation of fireworks, exploding fireworks and guiding fireworks.

[0039] In the process of selecting new fireworks by the dual-channel selection mechanism, both channels are used to select the next generation of new fireworks, and fireworks in the same channel compete with each other for the opportunity to become the next generation of fireworks. Specifically, fireworks, explosive fireworks and guide fireworks use the same channel, and differential fireworks use another channel. This increases the chance of retaining differential fireworks during the iteration process, thereby promoting the diversity of the fireworks population. In addition, the dual-channel selection mechanism can also fully retain the individual information of fireworks.

[0040] In one embodiment, in order to improve the fireworks mutation operation of the fireworks algorithm, this embodiment adopts a reinitialization mechanism solution, which specifically includes: In the formula, k max is the maximum number of iterations of the algorithm, k is the current number of iterations, is the current optimization increment, It is the expected optimization result of fireworks.

[0041] In this embodiment, the traditional fireworks algorithm adopts a new individual information interaction method based on competition when performing fireworks mutation operation. However, it is easy to distort when encountering complex optimization targets with large fitness changes, which leads to the final fitness of fireworks individuals being underestimated or overestimated. In addition, when the algorithm iteration process enters the middle and late stages, reinitialization will cause the individual to lag behind the evolution of the group, thereby causing a waste of computing resources, and the initialization of the individual may also lead to the appearance of a bad initial solution area, resulting in repeated reinitialization. Therefore, the present application adopts a reinitialization mechanism instead of the new individual information interaction method based on competition to solve the above problems, and in the initialization process of the reinitialization mechanism, the current optimization increment δ j k It is composed of the fitness difference of the individual in the past three iterations.

[0042] S22. Construct an adaptive path planning model based on the improved fireworks algorithm.

[0043] S3. Training the adaptive path planning model based on the training data set.

[0044] In one embodiment, in order to make the adaptive path planning model more suitable for the actual operation environment of the UAV, this embodiment adopts a training data set solution, which specifically includes the following steps: S31, generating a first generation of fireworks based on the adaptive path planning model in combination with the training data set, and calculating the fitness of the first generation of fireworks; S32, based on the battery capacity of the drone and the battery consumption rate per unit distance in the constraints of the adaptive path planning model, dividing the first-generation fireworks into at least flight path fireworks and charging path fireworks; S33, based on the fitness of the flight path fireworks, performing explosion operations and mutation operations on the flight path fireworks; at the same time, based on the fitness of the charging path fireworks, performing explosion operations and mutation operations on the charging path fireworks; S34, selecting new flight path fireworks based on the explosion operations and mutation operations of the flight path fireworks, and calculating the fitness of the new flight path fireworks; At the same time, selecting new charging path fireworks based on the explosion operation and mutation operation of the charging path fireworks, and calculating the fitness of the new charging path fireworks; S35, determining whether the training of the adaptive path planning model meets the stopping condition, if so, outputting the flight path fireworks with the minimum fitness and the charging path fireworks with the minimum fitness to obtain the optimal inspection path of the UAV, if not, executing S31.

[0045] In this embodiment, the drone battery capacity and the battery consumption rate per unit distance in the constraint conditions based on the adaptive path planning model divide the first generation fireworks into at least flight path fireworks and charging path fireworks, and also need to number the divided fireworks in sequence, and perform secondary numbering on the three-dimensional coordinates contained in the fireworks; At the same time, the adaptive path planning model needs to set the objective function and constraints before generating the first generation of fireworks. The objective function is as follows: The constraints are as follows: In the above formula, C represents the location of the checkpoint, M represents the location of the charging point, O represents the starting point, V represents the set of all points, and d gh is the distance between point g and point h, and g,h∈V; x gh is the decision variable; when the drone flies from point g to point h, x gh =1, in other cases, x gh=0; Q is the battery capacity of the drone, r is the battery consumption rate per unit distance; and is the residual power when the UAV arrives at and leaves point v, v∈V.

[0046] In one embodiment, in order to determine whether the adaptive path planning model has completed training, this embodiment adopts a maximum number of iterations scheme, which specifically includes: The current number of iterations of the adaptive path planning model is compared with the preset maximum number of iterations. If the current number of iterations is less than the maximum number of iterations, it is determined that the adaptive path planning model does not meet the stopping condition. If the current number of iterations is greater than or equal to the maximum number of iterations, it is determined that the adaptive path planning model meets the stopping condition.

[0047] In one embodiment, in order to obtain the optimal inspection path for the drone, this embodiment adopts a solution of combining flight path fireworks with charging path fireworks, specifically including: The three-dimensional coordinates in the flight path fireworks and the charging path fireworks are sorted and connected according to the spatial relationship to obtain the optimal inspection path for the UAV.

[0048] In this embodiment, the flight path fireworks and the charging path fireworks are first sorted in sequence according to the order before the division in step S32 to determine the sequence of each flight path and each charging path; secondly, the adjacent flight path fireworks and charging path fireworks are connected according to adjacent three-dimensional coordinate points. Specifically, if the flight path fireworks are in front of the charging path fireworks, the last three-dimensional coordinate point contained in the flight path fireworks is connected to the first three-dimensional coordinate point contained in the charging path fireworks to form a whole, and then the formed whole is converted into a complete route to obtain the optimal inspection path for the UAV. The optimal path for the UAV obtained in this way not only selects the shortest and safest flight path for the UAV, but also takes into account the battery power of the UAV and plans the charging path when the UAV is low on power, so that the UAV can solve its charging problem while performing the tower inspection task, effectively improving the UAV's endurance.

[0049] S4. Perform reliability verification on the trained adaptive path planning model based on the verification data set. If the reliability verification is qualified, execute S5; if the reliability verification is unqualified, execute S3.

[0050] In one embodiment, in order to verify the reliability of the adaptive path planning model, this embodiment adopts a verification data set solution, which specifically includes the following steps: S41, counting the number of times the verification data set is input into the adaptive path planning model for testing and the optimal path of the drone corresponding to the verification data set is successfully obtained, and the total number of times the verification data set is input into the adaptive path planning model for testing; S42. Calculate the accuracy of the adaptive path planning model based on the number of successes and the total number of times. If the accuracy is greater than or equal to a preset threshold, determine that the adaptive path planning model reliability verification is qualified. If the accuracy is less than the preset threshold, determine that the adaptive path planning model reliability verification is unqualified.

[0051] In this embodiment, the calculation formula of the accuracy of the adaptive path planning model is as follows: In the formula, p 准 is the accuracy of the adaptive path planning model, n is the number of successes, and N is the total number of times.

[0052] S5. Input the real-time data of the UAV tower inspection into the adaptive path planning model that has passed the reliability verification for testing to obtain the optimal path for the UAV.

[0053] The present application is compared and analyzed with other variants of the fireworks algorithm and other swarm intelligence algorithms through the CEC2013 test set to further verify the performance of the present application. The CEC2013 test set includes 28 benchmark functions, and the dimension of the variables can be changed.

[0054] Comparison with other fireworks algorithm variants: This section selects four classic fireworks algorithm variants - GFWA, dynFWA, COFFWA and MSCFWA ​​to compare with this application (IFWA-LSGO), and calculates the mean and standard deviation of the errors between the obtained solutions and the actual global optimal solution, as shown in the following table: Table 1. Mean and standard deviation of the CEC2013 test suite with FWA variants From the calculation results of the mean and standard deviation values ​​of the 28 benchmark functions in the table, it can be seen that the present application (IFWA-LSGO) leads the other 4 fireworks algorithm variants in 16 benchmark functions. At the same time, the optimization results of 17 benchmark functions of the present application (IFWA-LSGO) are better than the comparison documents in terms of standard deviation, which proves that the present application (IFWA-LSGO) has extremely strong stability; in the single-modal function test represented by function f2 and function f5, all algorithms can eventually approach the optimal solution, but in the early stage of iteration, the present application (IFWA-LSGO) can reduce the fitness value faster than other algorithms, and in the test process of multi-modal functions, the performance of the present application (IFWA-LSGO) is also more stable, which means that the present application (IFWA-LSGO) can not only maintain the diversity of the population and quickly reduce the fitness value in the early stage of iteration, but also outperform the comparison algorithm in the fitness value of the final iteration.

[0055] Compared with other swarm intelligence algorithms: This section uses four typical bee colony intelligence algorithms to compare with this application, including the invasive weed optimization algorithm, the firefly algorithm, the artificial bee colony algorithm, and the differential evolution algorithm. The specific comparison results are as follows: Table 2. Mean and standard deviation of five algorithms in the CEC2013 test suite (30D) Table 3. Mean and standard deviation (50D) of five algorithms in the CEC2013 test suite As shown in Table 2, the present application is in a leading position compared with the test results of other swarm intelligence algorithms in 30 dimensions (30D), especially in the test results of multimodal and composition test functions represented by functions f6-f28, the present application obtains better optimization results than the invasive weed optimization algorithm, firefly algorithm, artificial bee colony algorithm and differential evolution algorithm; As shown in Table 3, the dimension of the 28 benchmark functions is increased from 30 to 50 (50D) for further verification. According to the data in Table 3, it can be seen that the present application has achieved advantages in all 28 benchmark functions.

[0056] This embodiment has at least the following substantial effects: In this application, the historical data of UAV power tower inspections are preprocessed to obtain training data sets and verification data sets, which effectively improves the data quality, enables it to support a more effective data analysis process, and greatly reduces the time and resource consumption of data processing in this application; secondly, the fireworks algorithm is improved based on the optimized explosion strategy, and the improved fireworks algorithm is used to construct an adaptive path planning model for path planning, which overcomes the defects of the original fireworks algorithm itself, and at the same time enables the UAV to cover the entire area more efficiently when inspecting power towers, effectively improving the inspection efficiency, and while constructing the adaptive path planning model, the battery life issues related to the UAV battery power are also considered, and a comprehensive model that can automatically schedule according to the UAV battery power is obtained; secondly, the training data set is used to train the automatic The adaptive path planning model can optimize the performance of the adaptive path planning model, so that it can comprehensively, accurately and quickly plan the optimal inspection path of the drone according to the actual operating environment of the drone, and use the verification data set to verify whether the adaptive path planning model is qualified, and further determine whether the trained adaptive path planning model meets the needs of actual drone tower inspections, which greatly improves the accuracy of the adaptive path planning model and effectively prevents the adaptive path planning model from overfitting and underfitting; finally, the real-time data of drone tower inspections is tested through the adaptive path planning model that meets the actual needs of drone tower inspections, and the optimal inspection path of the drone is obtained quickly and accurately, so that the drone can accurately complete the tower inspection task in the shortest time according to the working standards.

[0057] 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. The UAV path optimization method based on the improved fireworks algorithm is characterized by: The method comprises the following steps: S1. Preprocess the historical data of UAV tower inspection to obtain training data set and verification data set; S2, construct an adaptive path planning model based on the fireworks algorithm improved by optimizing the explosion strategy; S3, training the adaptive path planning model based on the training data set; S4, performing reliability verification on the trained adaptive path planning model based on the verification data set, if the reliability verification is qualified, executing S5, if the reliability verification is unqualified, executing S3; S5. Input the real-time data of the UAV tower inspection into the adaptive path planning model that has passed the reliability verification for testing to obtain the optimal path for the UAV.

2. The method for optimizing the path of a UAV based on the improved fireworks algorithm according to claim 1, characterized in that: In S1, the specific process of preprocessing the historical data of UAV tower inspection to obtain the training data set and the verification data set is as follows: Establish a three-dimensional coordinate system based on the historical data of the UAV power tower inspection, and convert the historical data of the UAV power tower inspection into a three-dimensional coordinate point representation; At the same time, the data converted into three-dimensional coordinate points are divided into a training data set and a validation data set.

3. The method for optimizing the path of a UAV based on an improved fireworks algorithm according to claim 1, characterized in that: In S2, the specific process of constructing an adaptive path planning model based on the fireworks algorithm improved by optimizing the explosion strategy is as follows: S21. Using the differential spark mechanism to improve the fireworks explosion operation of the fireworks algorithm, and using the dual-channel selection mechanism to improve the new fireworks selection operation of the fireworks algorithm, and using the reinitialization mechanism to improve the fireworks mutation operation of the fireworks algorithm; S22. Construct an adaptive path planning model based on the improved fireworks algorithm.

4. The method for optimizing the path of a UAV based on the improved fireworks algorithm according to claim 3 is characterized in that: In S21, the specific process of the fireworks explosion operation of the fireworks algorithm improved by using the differential spark mechanism is as follows: Where β is the differential scaling factor, is the differential vector, These are three fireworks chosen at random.

5. The method for optimizing the path of a UAV based on an improved fireworks algorithm according to claim 3, characterized in that: In S21, the specific process of the new fireworks selection operation using the dual-channel selection mechanism to improve the fireworks algorithm is as follows: In the formula, is the differential fireworks selected from channel 2, x k is another part of the new firework selected from channel 1, x k+1 It's new fireworks.

6. The method for optimizing the path of a UAV based on the improved fireworks algorithm according to claim 3 is characterized in that: In S21, the specific process of using the reinitialization mechanism to improve the fireworks mutation operation of the fireworks algorithm is as follows: In the formula, k max is the maximum number of iterations of the algorithm, k is the current number of iterations, is the current optimization increment, It is the expected optimization result of fireworks.

7. The method for optimizing the path of a UAV based on an improved fireworks algorithm according to claim 3, characterized in that: In S3, the specific process of training the adaptive path planning model based on the training data set is: S31, generating a first generation of fireworks based on the adaptive path planning model in combination with the training data set, and calculating the fitness of the first generation of fireworks; S32, based on the battery capacity of the drone and the battery consumption rate per unit distance in the constraints of the adaptive path planning model, dividing the first-generation fireworks into at least flight path fireworks and charging path fireworks; S33, based on the fitness of the flight path fireworks, performing explosion operations and mutation operations on the flight path fireworks; at the same time, based on the fitness of the charging path fireworks, performing explosion operations and mutation operations on the charging path fireworks; S34, selecting new flight path fireworks based on the explosion operation and mutation operation of the flight path fireworks, and calculating the fitness of the new flight path fireworks; At the same time, selecting new charging path fireworks based on the explosion operation and mutation operation of the charging path fireworks, and calculating the fitness of the new charging path fireworks; S35, determining whether the training of the adaptive path planning model meets the stopping condition, if so, outputting the flight path fireworks with the minimum fitness and the charging path fireworks with the minimum fitness to obtain the optimal inspection path of the UAV, if not, executing S31.

8. The method for optimizing the path of a UAV based on the improved fireworks algorithm according to claim 7, characterized in that: In S35, the specific judgment process of judging whether the training of the adaptive path planning model meets the stopping condition is as follows: The current number of iterations of the adaptive path planning model is compared with the preset maximum number of iterations. If the current number of iterations is less than the maximum number of iterations, it is determined that the adaptive path planning model does not meet the stopping condition. If the current number of iterations is greater than or equal to the maximum number of iterations, it is determined that the adaptive path planning model meets the stopping condition.

9. The method for optimizing the path of a UAV based on the improved fireworks algorithm according to claim 7, characterized in that: In the above S35, the specific process of outputting the fireworks of the flight path with the minimum fitness and the fireworks of the charging path with the minimum fitness to obtain the optimal inspection path of the drone is as follows: The three-dimensional coordinates in the flight path fireworks and the charging path fireworks are sorted and connected according to the spatial relationship to obtain the optimal inspection path for the UAV.

10. The method for optimizing the path of a UAV based on an improved fireworks algorithm according to claim 1, characterized in that: In S4, the specific process of performing reliability verification on the trained adaptive path planning model based on the verification data set is as follows: S41, counting the number of times the verification data set is input into the adaptive path planning model for testing and the optimal path of the drone corresponding to the verification data set is successfully obtained, and the total number of times the verification data set is input into the adaptive path planning model for testing; S42. Calculate the accuracy of the adaptive path planning model based on the number of successes and the total number of times. If the accuracy is greater than or equal to a preset threshold, determine that the adaptive path planning model reliability verification is qualified. If the accuracy is less than the preset threshold, determine that the adaptive path planning model reliability verification is unqualified.

Images