A drone group cleaning operation coordination method and system

Abstract

The present invention relates to a method and system for coordinated cleaning operations of a drone group, wherein the method obtains and identifies the coordinates of the position to be cleaned, divides the cleaning operation space into multiple sub-areas, matches the coordinates of the drones and the position to be cleaned in the sub-areas, and first preliminarily generates a relatively excellent initial flight path set according to the matching results, and then optimizes the initial flight path set in each sub-area to obtain the optimal target flight path set. Compared with the prior art, the present invention can preliminarily generate a relatively excellent initial path and then optimize it, which can effectively improve the excellence of the final optimization result, and also greatly reduce the difficulty of obstacle avoidance in the path planning process, and improve the optimization efficiency.

Description

Technical Field

[0001] The present invention relates to the field of laser cleaning technology, and in particular to a method and system for coordinated cleaning operations of a swarm of unmanned aerial vehicles. Background Art

[0002] Solar photovoltaic panels are often installed in open-air environments, which are easily contaminated by dust and other dirt in the air, thus affecting the efficiency of photoelectric conversion. At the same time, since solar photovoltaic panels are mostly installed in environments that are not conducive to manual work, manual cleaning is not only cumbersome and dangerous, but may also cause damage to the solar photovoltaic panels, waste more water resources, and be inefficient.

[0003] Therefore, using drones to clean solar photovoltaic panels has become a more preferred method. However, the cleaning of solar photovoltaic panels requires the coordinated work of multiple drones. How to effectively coordinate the drone swarm for joint cleaning is an urgent problem to be solved. Summary of the invention

[0004] In view of this, it is necessary to provide a drone swarm cleaning operation coordination method to achieve the coordinated work of the drone swarm when performing cleaning operations.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a method for coordinated cleaning operations of a drone group, comprising:

[0007] Acquire an image of the target cleaning object and identify the coordinates of the position to be cleaned;

[0008] Dividing the cleaning operation space into a plurality of sub-areas, each of the sub-areas including a plurality of locations to be cleaned and a plurality of drones;

[0009] Matching the location to be cleaned in each of the sub-areas with the drone to obtain multiple matching results;

[0010] Generating an initial flight path set for each of the sub-areas according to the plurality of matching results;

[0011] The initial flight path set corresponding to each of the sub-areas is optimized respectively to obtain a target flight path set.

[0012] Furthermore, the location to be cleaned in each of the sub-areas is matched with the drone to obtain multiple matching results, including:

[0013] Randomly generate a plurality of chromosome individuals and establish an initial population, wherein the chromosome individuals include a plurality of gene fragments, and each of the gene fragments is used to characterize the matching relationship between the location to be cleaned and the drone;

[0014] Establish matching fitness model;

[0015] Evolving the initial population according to the matching fitness model to obtain a final population;

[0016] The matching result is obtained according to the last generation population.

[0017] Furthermore, evolving the initial population according to the matching fitness model to obtain a final population includes:

[0018] Acquire the initial population as the population to be evolved;

[0019] Crossover and mutation step: crossover and / or mutation of the chromosome individuals in the population to be evolved to obtain an intermediate population, wherein the intermediate population includes the chromosome individuals after crossover and / or mutation and the population to be evolved;

[0020] Selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value;

[0021] The offspring population is used as the population to be evolved, and the crossover mutation step and the selection step are repeated until a preset end condition is reached to obtain the final population.

[0022] Furthermore, the selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value, includes:

[0023] According to the matching fitness model, the individual matching fitness value of each chromosome in the intermediate population is calculated respectively;

[0024] Sort each chromosome in the intermediate population according to its corresponding individual matching fitness value;

[0025] A set number of chromosomes are selected according to the sorting result to constitute the offspring population.

[0026] Furthermore, the preset end condition includes:

[0027] The evolution coefficient is less than the set benchmark value;

[0028] The evolution coefficient is the difference between the smallest individual matching fitness value in the initial population and the smallest individual matching fitness value in the current offspring population.

[0029] Furthermore, generating an initial flight path set for each sub-area according to the plurality of matching results includes:

[0030] Establishing a path constraint condition, wherein the path constraint condition is used to constrain the paths of multiple drones within the same sub-area;

[0031] According to the path constraint conditions and the matching results, a plurality of flight paths are randomly generated for each UAV in the same sub-area, and the plurality of flight paths corresponding to the plurality of UAVs in the same sub-area constitute an initial flight path set.

[0032] Further, respectively optimizing the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set includes:

[0033] Establishing a search space according to the initial flight path set;

[0034] Establishing an optimization fitness model;

[0035] According to the search space and the optimization fitness model, the initial flight path set corresponding to each sub-area is optimized based on the SSA algorithm to obtain multiple target flight paths, the target flight paths correspond to the drones one-to-one, and the set of multiple target flight paths is the target flight path set.

[0036] Furthermore, the initial flight path set corresponding to each sub-area is optimized based on the SSA algorithm according to the search space and the optimization fitness model to obtain multiple target flight paths, including:

[0037] Select a plurality of the flight paths corresponding to the target UAV and map them to the search space to obtain a plurality of search individuals;

[0038] Optimizing the position of the search individual in the search space based on the SSA algorithm;

[0039] According to the optimized position of the search individual in the search space and the optimized fitness model, a flight path corresponding to the optimal search individual is selected as the target flight path corresponding to the target UAV.

[0040] Furthermore, the step of acquiring an image of the target cleaning object and identifying the coordinates of the position to be cleaned includes:

[0041] Acquire an image of a target cleaning object;

[0042] Based on the YOLOX network, the pixel coordinates of dirt in the image of the target cleaning object are extracted;

[0043] The coordinates of the position to be cleaned are obtained according to the pixel coordinates of the dirt.

[0044] In a second aspect, the present invention further provides a drone group cleaning operation collaborative system, comprising:

[0045] A coordinate acquisition unit, used to acquire an image of the target cleaning object and identify the coordinates of the position to be cleaned;

[0046] An area division unit, used to divide the cleaning operation space into a plurality of sub-areas, each of which includes a plurality of locations to be cleaned and a plurality of drones;

[0047] A matching unit, used for matching the location to be cleaned in each of the sub-areas with the drone to obtain a plurality of matching results;

[0048] A path initialization unit, used for generating an initial flight path set for each of the sub-areas according to the plurality of matching results;

[0049] The optimization unit is used to optimize the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set.

[0050] The present invention provides a method and system for the coordinated cleaning operation of a drone group, which obtains and identifies the coordinates of the position to be cleaned, divides the cleaning operation space into multiple sub-areas, matches the drones in the sub-areas with the coordinates of the position to be cleaned, and first preliminarily generates a relatively excellent initial flight path set according to the matching results, and then optimizes the initial flight path set in each sub-area respectively to obtain the optimal target flight path set. Compared with the prior art, the present invention divides the sub-areas, reasonably matches the drones and the coordinates of the position to be cleaned in each sub-area, and can preliminarily generate a relatively excellent initial path and then optimize it according to the matching results, which can effectively improve the excellence of the final optimization result. Secondly, the multiple sub-areas of the present invention are independent of each other, and the drones only move in their corresponding sub-areas, which greatly reduces the difficulty of obstacle avoidance in the path planning process. In addition, the matching, optimization and other operations for each sub-area in the present invention are independent of each other and do not affect each other, so that the method has excellent parallelism, and the operations of multiple sub-areas can be processed simultaneously, making full use of the computing performance of existing hardware equipment, and greatly improving efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A method flow chart of an embodiment of a collaborative method for cleaning operations by a swarm of drones provided by the present invention;

[0052] Figure 2 for Figure 1 The method flow chart of step S103 in FIG.

[0053] Figure 3 for Figure 1 The method flow chart of step S105 in FIG.

[0054] Figure 4 This is a system architecture diagram of an embodiment of a drone swarm cleaning operation collaborative system provided by the present invention. DETAILED DESCRIPTION

[0055] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0056] In the description of the present application, “plurality” means two or more than two, unless otherwise clearly and specifically defined.

[0057] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0058] The drone used in this embodiment includes a drone body and a cleaning device, wherein the cleaning device is connected to the drone body. The cleaning device uses a YAG laser. When the laser is emitted from the laser head, the pollutants on the surface of the object absorb the laser energy and either vaporize and volatilize, or expand instantly due to heat and overcome the adsorption force of the surface on the particles, so that the particles are separated from the surface of the object, thereby achieving the purpose of cleaning. For ease of understanding, this embodiment uses a solar photovoltaic panel as a specific target cleaning object for explanation. In practice, the target cleaning object can also be any other object.

[0059] The present invention divides the cleaning space into multiple sub-areas, and then reasonably matches the drones and the locations to be cleaned in each sub-area. According to the matching results, a preliminary optimized initial flight path can be obtained, and then these paths are optimized separately, so that the final result has a better level of excellence, and at the same time, the computing efficiency can be greatly improved.

[0060] The present invention provides a method and system for coordinated cleaning operations of a drone swarm, which are described below respectively.

[0061] Combination Figure 1 As shown, a specific embodiment of the present invention discloses a drone group cleaning operation coordination method, including:

[0062] S101, acquiring an image of a target cleaning object, and identifying the coordinates of a position to be cleaned;

[0063] S102, dividing the cleaning operation space into a plurality of sub-areas, each of the sub-areas including a plurality of locations to be cleaned and a plurality of drones;

[0064] S103, matching the location to be cleaned in each of the sub-areas with the drone to obtain a plurality of matching results;

[0065] S104, generating an initial flight path set for each of the sub-areas according to the plurality of matching results;

[0066] S105, respectively optimizing the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set.

[0067] The present invention provides a method and system for the coordinated cleaning operation of a drone group, which obtains and identifies the coordinates of the position to be cleaned, divides the cleaning operation space into multiple sub-areas, matches the drones in the sub-areas with the coordinates of the position to be cleaned, and first preliminarily generates a relatively excellent initial flight path set according to the matching results, and then optimizes the initial flight path set in each sub-area respectively to obtain the optimal target flight path set. Compared with the prior art, the present invention divides the sub-areas, reasonably matches the drones and the coordinates of the position to be cleaned in each sub-area, and can preliminarily generate a relatively excellent initial path and then optimize it according to the matching results, which can effectively improve the excellence of the final optimization result. Secondly, the multiple sub-areas of the present invention are independent of each other, and the drones only move in their corresponding sub-areas, which greatly reduces the difficulty of obstacle avoidance in the path planning process. In addition, the matching, optimization and other operations for each sub-area in the present invention are independent of each other and do not affect each other, so that the method has excellent parallelism, and the operations of multiple sub-areas can be processed simultaneously, making full use of the computing performance of existing hardware equipment, and greatly improving efficiency.

[0068] In a preferred embodiment, step S101 of the present invention, obtaining an image of the target cleaning object and identifying the coordinates of the position to be cleaned, specifically includes:

[0069] Acquire an image of a target cleaning object;

[0070] Based on the YOLOX network, the pixel coordinates of dirt in the image of the target cleaning object are extracted;

[0071] The coordinates of the position to be cleaned are obtained according to the pixel coordinates of the dirt.

[0072] Acquiring the image of the target cleaning object in the above process can be achieved through a drone or other camera equipment with a known position, and the mapping relationship between the real-world coordinate system and the pixel coordinate system can be established first according to the camera equipment. After the pixel coordinates of the dirt in the image are extracted by the YOLOX algorithm, the cleaning position coordinates in the real world can be obtained through the mapping relationship, and subsequent matching and other operations can be performed. However, it can be understood that the YOLOX algorithm recognizes images as a prior art and is not explained in detail in this application. For the object to be cleaned, that is, the dirt on the solar panel in this embodiment, because their sizes may be different, the center of the range that a drone can act on can be regarded as a coordinate in this embodiment. The dirt on the solar panel can be composed of multiple coordinates of the position to be cleaned, and multiple drones can be used to clean it together.

[0073] As a preferred embodiment, step S102 of this embodiment divides the cleaning operation space into multiple sub-areas, each of which includes multiple positions to be cleaned and multiple drones, and the cleaning operation space is divided by an average division method. Specifically, the target cleaning object in this embodiment is an array composed of multiple solar panels. For ease of understanding, the multiple solar panels in this embodiment are all located in the same plane, and the cleaning operation space is a plane above the solar panel and parallel to it. Multiple drones move in this plane, and the movement plane is evenly divided into multiple rectangular grids, each of which is a sub-area, and there are multiple positions to be cleaned on the solar panels in each sub-area.

[0074] Multiple drones can be pre-evenly hovered in each sub-area of ​​the cleaning operation space. After the sub-areas are divided, each drone can be manually controlled to enter a different sub-area based on the number of cleaning locations in each area, ultimately forming a correspondence between multiple sub-areas-drones-locations to be cleaned, and then performing subsequent matching, optimization and other operations.

[0075] The above process is a pre-processing process, which can be achieved by any existing means. After the above preparation work is completed, the locations to be cleaned in each area can be matched with the drones in the area. Specifically, Figure 2 As shown, in a preferred embodiment, step S103, matching the location to be cleaned in each of the sub-areas with the drone to obtain multiple matching results, specifically includes:

[0076] S201, randomly generating a plurality of chromosome individuals and establishing an initial population, wherein the chromosome individuals include a plurality of gene fragments, each of the gene fragments being used to characterize a matching relationship between the location to be cleaned and the drone;

[0077] S202, establishing a matching fitness model;

[0078] S203, evolving the initial population according to the matching fitness model to obtain a final population;

[0079] S204: Obtain the matching result according to the last generation population.

[0080] In a preferred embodiment, the encoding means of the chromosome in step S201 can be: when the number of positions to be cleaned in the sub-area is equal to the number of drones, the positions to be cleaned and the drones can be matched one by one. Each drone and the position to be cleaned can be numbered. For example, a chromosome can be generated in the form of an array. The value and content of each element in the array is a gene fragment, which constitutes the matching relationship between the position to be cleaned and the drone. For example, the array [3, 4, 1, 2, 5] means that drone No. 1 corresponds to position No. 3 to be cleaned, drone No. 2 corresponds to position No. 4 to be cleaned, etc. One array is a chromosome, and multiple arrays are randomly generated as the initial population.

[0081] Similarly, when the number of locations to be cleaned in a sub-area is less than the number of drones, an array can also be used for encoding. For example, the array [3, 4, 0, 1, 2, 5] indicates that drone No. 3 does not match any coordinates to be cleaned and is idle and unused in this cleaning.

[0082] When the number of positions to be cleaned in a sub-area is greater than the number of drones, the above method can also be used for one-to-one matching, retaining some of the positions to be cleaned. After the path is matched and optimized and cleaned, the path is matched and optimized again to clean the remaining positions to be cleaned. One-to-many matching can also be performed, and each drone can be matched with multiple positions to be cleaned. Subsequently, a path that passes through the matched multiple positions to be cleaned is planned to complete the cleaning at one time, or the cleaning is completed multiple times as in the above process. One-to-many matching can be encoded by a multidimensional array. For example, the array [(0, 0, 2, 4), (1, 0, 0, 0), (0, 3, 0, 0)] is used as a chromosome individual, indicating that drone No. 1 matches positions No. 2 and No. 4 to be cleaned, drone No. 2 matches position No. 1 to be cleaned, and so on.

[0083] During the encoding process, matching constraints can be set so that the encoded genes meet relevant requirements, such as constraining that the same number of the position to be cleaned will not be generated in the gene sequence generated by the above method (except for 0 representing mismatch). The above content is only a preferred encoding method. In practice, other existing gene encoding methods can also be used to characterize the matching relationship between the drone and the position to be cleaned.

[0084] In a preferred embodiment, the matching fitness model established in step S202 is as follows:

[0085]

[0086] In the formula, fitness x (y) represents the fitness value of the yth individual in the xth generation. The smaller the value, the better the individual's fitness, that is, the optimal matching relationship. d is the distance coefficient, which is artificially given and is a constant between 0 and 1. represents the distance between the UAV corresponding to the jth individual in generation x and its corresponding cleaning location, and m represents the total number of individuals, that is, the total number of chromosomes in the population.

[0087] After the initial population is generated, the initial population can be evolved, that is, step S203. In a preferred embodiment, step S203, evolving the initial population according to the matching fitness model to obtain the final population, specifically includes:

[0088] Acquire the initial population as the population to be evolved;

[0089] Crossover and mutation step: crossover and / or mutation of the chromosome individuals in the population to be evolved to obtain an intermediate population, wherein the intermediate population includes the chromosome individuals after crossover and / or mutation and the population to be evolved;

[0090] Selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value;

[0091] The offspring population is used as the population to be evolved, and the crossover mutation step and the selection step are repeated until a preset end condition is reached to obtain the final population.

[0092] As a preferred embodiment, in the crossover mutation step of this embodiment, the specific rule of crossover is: genes are randomly assigned between p chromosome individuals, each of which divides the gene sequence into two sequence segments, and then exchanged to form two new individuals.

[0093] The specific rules of mutation are: set the mutation probability to 1%, and the mutated chromosome individual will exchange the order of its two random gene fragments.

[0094] After the crossover mutation is completed, the selection step can be performed to eliminate individuals that do not meet the requirements, that is, individuals that do not meet the above matching constraints and individuals with low fitness. Eliminating individuals with low fitness is the above selection step. As a preferred embodiment, the selection step in this embodiment specifically includes:

[0095] According to the matching fitness model, the individual matching fitness value of each chromosome in the intermediate population is calculated respectively;

[0096] Sort each chromosome in the intermediate population according to its corresponding individual matching fitness value;

[0097] A set number of chromosomes are selected according to the sorting result to constitute the offspring population.

[0098] After the above process is completed, the initial population has completed an evolution. By repeating the above process until the preset end condition is reached, the final population can be obtained. In a preferred embodiment, the preset end condition is: the evolution coefficient is less than the set reference value, wherein the evolution coefficient is the difference between the smallest individual matching fitness value in the initial population and the smallest individual matching fitness value in the current offspring population.

[0099] After the last generation population is obtained, the individual with the lowest fitness value in the last generation population is selected, and the corresponding matching method is the desired optimal matching result, that is, step S204. The matching result obtained in the above process indicates that the distance relationship between the drone and the location to be cleaned has reached an optimal level, and the path generated according to the matching relationship has reached a certain optimization level.

[0100] After obtaining the matching results, step S104 may be executed to generate an initial flight path set for each sub-area according to the plurality of matching results. As a preferred embodiment, step S104 in this embodiment specifically includes:

[0101] Establishing a path constraint condition, wherein the path constraint condition is used to constrain the paths of multiple drones within the same sub-area;

[0102] According to the path constraint conditions and the matching results, a plurality of flight paths are randomly generated for each UAV in the same sub-area, and the plurality of flight paths corresponding to the plurality of UAVs in the same sub-area constitute an initial flight path set.

[0103] As a preferred embodiment, the path constraint condition in this embodiment can be expressed as:

[0104]

[0105] where π i represents the path of the ith UAV, X ijIndicates that the trajectories of drones i and j violate the same path constraint, which means that the paths intersect or collide, and n represents the number of drones in the sub-area. In this embodiment, the multiple flight paths in Shanghai need to meet the above constraints. Since the number of drones and cleaning positions in each sub-area in this embodiment is small, and the paths of multiple sub-areas can be calculated in parallel, even if a simple inspection method is used, such as directly traversing each flight path to check whether it meets the constraints, the constraints can be easily implemented to avoid path collisions.

[0106] As a preferred embodiment, the specific process of generating a flight path in this embodiment is as follows: For a pair of drones and objects to be cleaned, the position of the drone is the starting position, denoted by P i (x i ,y i ,θ i ), where (x i ,y i ) is the initial position, θ i is the initial direction, and the coordinate of the position to be cleaned is the final position, denoted as P f (x f ,y f ,θ f ), where (x f ,y f ) is the final position, θ f The final direction of the drone.

[0107] Generate a drone path r(t) between two points. The path generation method uses the fifth-order Pythagorean velocity vector path:

[0108]

[0109] Among them, b k =(x k ,y k ) represents the control point, t is a parameter, k = 0…5, and four of the control points are calculated by the first-order Hermite interpolation method, such as:

[0110] b0=(x i ,y i )

[0111] b5=(x f ,y f )

[0112]

[0113]

[0114] The four control points b0, b5, b1, and b4 are fixed points, and the remaining b2 and b3 are variable points. By changing these two control points, multiple initial flight path curves can be generated for each pair of drones and the location to be cleaned. In practice, other methods can also be used to generate multiple flight paths, such as using a polynomial function to represent a generated continuous path, and changing the coefficients in the polynomial function to obtain multiple flight paths.

[0115] After generating multiple flight paths for each drone in a sub-area, the initial flight path set of the sub-area can be obtained. Then, step S105 can be performed. Figure 3 As shown, as a preferred embodiment of the present invention, step S105 in the embodiment, respectively optimizing the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set, specifically includes:

[0116] S301, establishing a search space according to the initial flight path set;

[0117] S302, establishing an optimized fitness model;

[0118] S303. According to the search space and the optimization fitness model, based on the SSA algorithm, the initial flight path set corresponding to each sub-area is optimized respectively to obtain multiple target flight paths, wherein the target flight paths correspond to the UAVs one-to-one, and the set of multiple target flight paths is the target flight path set.

[0119] The SSA algorithm (Salp Swarm Algorithm) in the above process is inspired by the swarming behavior of salps in the ocean. Salps are transparent creatures that move in the water by pumping water through their bodies. When they hunt, they form salp chains to facilitate the rapid movement of the population to reach the location of the food source in the sea. The SSA algorithm uses the solution space of the problem as the ocean, that is, the search space, the multiple solutions to the problem as salps individuals, and the optimal solution to the problem as the food source to be found. By simulating the movement of salps to the food source, the solution of the problem converges to the optimal solution.

[0120] As a preferred embodiment, step S301 in this embodiment, based on the initial flight path set, establishes a search space, specifically: based on the specific expression form of the flight path in the initial flight path set, establishes a search space, takes the number of variables in the flight path model as the dimension of the search space, and takes the range of each variable as the boundary of the search space. For example, when a fifth-order polynomial is used to represent the flight path, there are six variable parameters in the fifth-order polynomial, including the terminal constant, so the established search space is a six-dimensional space, and the boundary of each dimension can be manually specified according to the specific situation.

[0121] As a preferred embodiment, the optimization fitness model established in step S302 in this embodiment specifically includes the comprehensive cost C generated during the flight of the drone. total :

[0122] C total =P1×C fuel +P2×C diver +P3XC gap

[0123] In the formula, P1, P2, and P3 are actual parameters, which are set manually according to specific circumstances. fuel is the fuel cost, C diver is the sharp turn cost, C gap is the gap cost, and the gap cost is 0 in the optimal solution.

[0124] Since the generated path can be represented by a series of track points, the sharp turn cost can be directly represented by the number of turns greater than the set angle in the flight point. Assuming that the speed of the drone is constant during flight, its fuel cost can be directly represented by the length of its flight path, as shown in the following formula:

[0125] C fuel =∑(D x，y，z )

[0126] Where D x,y,z Indicates the distance between two adjacent flight points. The specific calculation formula for the gap cost is:

[0127]

[0128] In the formula, (X end ,Y end ,Z end ) The endpoints of the path generated, (T X ,T Y ,T Z ) represents the target location. The comprehensive cost C calculated through the above process is total This is the optimization fitness value. The lower the value, the better the optimization fitness, and the path corresponding to the individual is optimal.

[0129] After the optimization fitness model is constructed, step S303 can be executed. As a preferred embodiment, step S303 in this embodiment, according to the search space and the optimization fitness model, optimizes the initial flight path set corresponding to each sub-area based on the SSA algorithm to obtain multiple target flight paths, specifically including:

[0130] Select a plurality of the flight paths corresponding to the target UAV and map them to the search space to obtain a plurality of search individuals;

[0131] Optimizing the position of the search individual in the search space based on the SSA algorithm;

[0132] According to the optimized position of the search individual in the search space and the optimized fitness model, a flight path corresponding to the optimal search individual is selected as the target flight path corresponding to the target UAV.

[0133] Specifically, in the above process, multiple flight paths corresponding to a UAV, that is, a target UAV, are used as multiple search individuals. According to different coefficients in the expression of its flight path, its coordinates in the search space are obtained, and then its position in the search space is optimized. The specific optimization process is as follows:

[0134] Step 1: Determine the location of each search individual.

[0135] Step 2: Calculate the optimization fitness of each search individual respectively, and sort the search individuals according to their optimization fitness.

[0136] Step 3: Select the individual with the best optimized fitness value as the food source, select the first several search individuals with better optimized fitness values ​​as the leader, and the remaining individuals as followers, following in order according to the optimized fitness ranking calculated in the above steps.

[0137] Step 4: The leader moves in the search space. The specific movement method is to move randomly according to one of the following formulas:

[0138]

[0139] In the formula, is the coordinate of the i-th leader after moving in the j-th dimension, where i is the sequence number sorted in step 2, and F j is the coordinate of the current food source in the jth dimension, ub j and lb j are the upper and lower bounds of the j-th dimension, c2 and c3 are random uniform numbers, and c1 is obtained by the following formula:

[0140]

[0141] Where l represents the current number of iterations, and L represents the set maximum number of iterations. The above formula is the movement of an individual in one dimension. The movement of an individual can be completed by calculating and combining the movements of all dimensions separately.

[0142] The above process means that the leader drone moves randomly one step in the search space combined with the current location of the food source to conduct a global search.

[0143] Step 5: The follower moves in the search space. The specific movement method is as follows:

[0144]

[0145] In the formula, is the coordinate of the i-th follower after moving in the j-th dimension, where i is the sequence number sorted in step 2. The meaning of this formula is that the follower moves in the direction of the search individual it follows, but the movement distance is half of the distance between the two. Unlike the leader, the follower mainly performs local search.

[0146] Step 6: Repeat steps 2 to 5 above until the set number of iterations is reached or the optimized fitness value of the individual meets the set requirements, then the iteration can be stopped. At this time, all search individuals converge to the location of the food source to the greatest extent, completing the optimization of the position of the search individual in the search space.

[0147] Afterwards, the optimized fitness value of each search individual can be calculated according to the optimized position of the search individual in the search space, and the search individual with the best optimized fitness value can be selected, and its corresponding flight path can be used as the target flight path of the target UAV.

[0148] After repeating the above process to optimize an optimal target flight path for each drone, the set of all target flight paths becomes the target flight path set, thus completing the collaborative process of the drone group cleaning operation.

[0149] In steps S103, S104, and S105 of the above embodiment, calculations for different sub-areas can be performed simultaneously, for example, using a multi-core processor for calculations, where each core acts as a calculation host to process the matching and path optimization process of a sub-area. This can maximize the excellent performance of today's advanced processors and greatly improve the optimization speed.

[0150] In order to better implement the drone group cleaning operation coordination method in the embodiment of the present invention, based on the drone group cleaning operation coordination method, correspondingly, please refer to Figure 4 , Figure 4 This is a schematic diagram of the structure of an embodiment of a drone swarm cleaning operation collaborative system provided by the present invention. A drone swarm cleaning operation collaborative method system 400 provided by an embodiment of the present invention includes:

[0151] A coordinate acquisition unit 410 is used to acquire an image of the target cleaning object and identify the coordinates of the position to be cleaned;

[0152] An area division unit 420 is used to divide the cleaning operation space into a plurality of sub-areas, each of which includes a plurality of locations to be cleaned and a plurality of drones;

[0153] A matching unit 430 is used to match the location to be cleaned in each of the sub-areas with the drone to obtain a plurality of matching results;

[0154] A path initialization unit 440, configured to generate an initial flight path set for each of the sub-areas according to the plurality of matching results;

[0155] The optimization unit 450 is used to optimize the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set.

[0156] It should be noted here that the corresponding drone swarm cleaning operation collaborative system 400 provided in the above embodiment can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above method embodiments, which will not be repeated here.

[0157] The present invention provides a method and system for the coordinated cleaning operation of a drone group, which obtains and identifies the coordinates of the position to be cleaned, divides the cleaning operation space into multiple sub-areas, matches the drones in the sub-areas with the coordinates of the position to be cleaned, and first preliminarily generates a relatively excellent initial flight path set according to the matching results, and then optimizes the initial flight path set in each sub-area respectively to obtain the optimal target flight path set. Compared with the prior art, the present invention divides the sub-areas, reasonably matches the drones and the coordinates of the position to be cleaned in each sub-area, and can preliminarily generate a relatively excellent initial path and then optimize it according to the matching results, which can effectively improve the excellence of the final optimization result. Secondly, the multiple sub-areas of the present invention are independent of each other, and the drones only move in their corresponding sub-areas, which greatly reduces the difficulty of obstacle avoidance in the path planning process. In addition, the matching, optimization and other operations for each sub-area in the present invention are independent of each other and do not affect each other, so that the method has excellent parallelism, and the operations of multiple sub-areas can be processed simultaneously, making full use of the computing performance of existing hardware equipment, and greatly improving efficiency.

[0158] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A drone group cleaning operation coordination method, characterized in that: include: Acquire an image of the target cleaning object and identify the coordinates of the position to be cleaned; Dividing the cleaning operation space into a plurality of sub-areas, each of the sub-areas including a plurality of locations to be cleaned and a plurality of drones; Matching the location to be cleaned in each of the sub-areas with the drone to obtain multiple matching results; Generating an initial flight path set for each of the sub-areas according to the plurality of matching results; Respectively optimizing the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set; The location to be cleaned in each of the sub-areas is matched with the drone to obtain multiple matching results, including: Randomly generate a plurality of chromosome individuals and establish an initial population, wherein the chromosome individuals include a plurality of gene fragments, and each of the gene fragments is used to characterize the matching relationship between the location to be cleaned and the drone; Establish matching fitness model; Evolving the initial population according to the matching fitness model to obtain a final population; Obtaining the matching result according to the last generation population; The step of evolving the initial population according to the matching fitness model to obtain a final population includes: Acquire the initial population as the population to be evolved; Crossover and mutation step: crossover and / or mutation of the chromosome individuals in the population to be evolved to obtain an intermediate population, wherein the intermediate population includes the chromosome individuals after crossover and / or mutation and the population to be evolved; Selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value; The offspring population is used as the population to be evolved, and the crossover mutation step and the selection step are repeated until a preset end condition is reached to obtain the final population; The selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value, comprises: According to the matching fitness model, the individual matching fitness value of each chromosome in the intermediate population is calculated respectively; Sort each chromosome in the intermediate population according to its corresponding individual matching fitness value; A set number of chromosomes are selected according to the sorting result to constitute the offspring population.

2. The drone group cleaning operation coordination method according to claim 1 is characterized in that: The preset end condition includes: The evolution coefficient is less than the set benchmark value; The evolution coefficient is the difference between the smallest individual matching fitness value in the initial population and the smallest individual matching fitness value in the current offspring population.

3. The drone group cleaning operation coordination method according to claim 1 is characterized in that: Generating an initial flight path set for each sub-area according to the plurality of matching results comprises: Establishing a path constraint condition, wherein the path constraint condition is used to constrain the paths of multiple drones within the same sub-area; According to the path constraint conditions and the matching results, a plurality of flight paths are randomly generated for each UAV in the same sub-area, and the plurality of flight paths corresponding to the plurality of UAVs in the same sub-area constitute an initial flight path set.

4. The drone group cleaning operation coordination method according to claim 3 is characterized in that: The step of respectively optimizing the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set includes: Establishing a search space according to the initial flight path set; Establishing an optimization fitness model; According to the search space and the optimization fitness model, the initial flight path set corresponding to each sub-area is optimized based on the SSA algorithm to obtain multiple target flight paths, the target flight paths correspond to the drones one-to-one, and the set of multiple target flight paths is the target flight path set.

5. The drone group cleaning operation coordination method according to claim 4 is characterized in that: The method of optimizing the initial flight path set corresponding to each sub-area based on the search space and the optimization fitness model based on the SSA algorithm to obtain multiple target flight paths includes: Select a plurality of the flight paths corresponding to the target UAV and map them to the search space to obtain a plurality of search individuals; Optimizing the position of the search individual in the search space based on the SSA algorithm; According to the optimized position of the search individual in the search space and the optimized fitness model, a flight path corresponding to the optimal search individual is selected as the target flight path corresponding to the target UAV.

6. The drone group cleaning operation coordination method according to claim 1 is characterized in that: The step of acquiring an image of the target cleaning object and identifying the coordinates of the position to be cleaned includes: Acquire an image of a target cleaning object; Based on the YOLOX network, the pixel coordinates of dirt in the image of the target cleaning object are extracted; The coordinates of the position to be cleaned are obtained according to the pixel coordinates of the dirt.

7. A drone group cleaning operation collaborative system, characterized in that: include: A coordinate acquisition unit, used to acquire an image of the target cleaning object and identify the coordinates of the position to be cleaned; An area division unit, used to divide the cleaning operation space into a plurality of sub-areas, each of which includes a plurality of locations to be cleaned and a plurality of drones; A matching unit, used for matching the location to be cleaned in each of the sub-areas with the drone to obtain a plurality of matching results; A path initialization unit, used for generating an initial flight path set for each of the sub-areas according to the plurality of matching results; an optimization unit, configured to optimize the initial flight path set corresponding to each of the sub-areas to obtain a target flight path set; The location to be cleaned in each of the sub-areas is matched with the drone to obtain multiple matching results, including: Randomly generate a plurality of chromosome individuals and establish an initial population, wherein the chromosome individuals include a plurality of gene fragments, and each of the gene fragments is used to characterize the matching relationship between the location to be cleaned and the drone; Establish matching fitness model; Evolving the initial population according to the matching fitness model to obtain a final population; Obtaining the matching result according to the last generation population; The step of evolving the initial population according to the matching fitness model to obtain a final population includes: Acquire the initial population as the population to be evolved; Crossover and mutation step: crossover and / or mutation of the chromosome individuals in the population to be evolved to obtain an intermediate population, wherein the intermediate population includes the chromosome individuals after crossover and / or mutation and the population to be evolved; Selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value; The offspring population is used as the population to be evolved, and the crossover mutation step and the selection step are repeated until a preset end condition is reached to obtain the final population; The selection step: calculating the individual matching fitness value of each chromosome in the intermediate population according to the matching fitness model, and obtaining the offspring population according to the individual matching fitness value, comprises: According to the matching fitness model, the individual matching fitness value of each chromosome in the intermediate population is calculated respectively; Sort each chromosome in the intermediate population according to its corresponding individual matching fitness value; A set number of chromosomes are selected according to the sorting result to constitute the offspring population.

Images