Unmanned cluster collaborative task planning method for edge environment

Abstract

The invention provides an unmanned cluster collaborative task planning method for an edge environment, and belongs to the technical field of swarm intelligence, and the method comprises the steps: generating an initial population containing a plurality of individuals through employing a random strategy and a heuristic strategy based on an unmanned cluster composed of a plurality of unmanned platforms executing tasks in a geographic region, one individual represents one task planning scheme, and one task planning scheme comprises an allocation relationship between each region and each unmanned platform and a sequence of each unmanned platform accessing each region; and on the basis of propagation operation of adjacent region exchange, refraction operation of social learning and wave breaking operation of access order optimization, carrying out iterative updating on individuals in the initial population until the number of times of iterative updating reaches a preset maximum number of times of iterative updating, and obtaining an optimal task planning scheme. According to the method, the total completion time of the unmanned cluster for executing a plurality of area search tasks is minimized, and the capability of the unmanned cluster for executing complex tasks in an edge environment is improved.

Description

Technical Field

[0001] The present application relates to the field of swarm intelligence technology, and in particular to an unmanned cluster collaborative task planning method for edge environments. Background Art

[0002] In the context of the systematic development of networking, informatization, and intelligence, military tasks are showing a new trend of information-oriented, decision-making and control migrating to edge environments. The edge environment far away from the command center is highly dynamic and uncertain, resulting in a significant decline in the survivability and mission execution capabilities of high-value manned platforms in edge environments. Large-scale unmanned swarms distribute the perception, cognition, decision-making, and execution functions of manned platforms to different unmanned platforms, aggregate the capabilities of each platform through information exchange and functional interoperability between different platforms, form a system synergy, and then emerge with the ability to perform complex tasks in edge environments.

[0003] In recent years, unmanned swarms have been widely used in battlefields and disaster relief scenarios, performing tasks such as search, tracking, communication relay, and strike in a distributed and collaborative manner. The search task is the basis for unmanned swarms to make group decisions and collaborative control, and it is also the primary link for unmanned swarms to perform complex tasks. Generally speaking, search refers to the process of investigating the possible area of ​​the target in order to find the target you want to find. In the military field, autonomous unmanned swarms carrying differentiated sensors and detectors are used to search the entire target area and achieve high-precision positioning of the enemy; in the agricultural field, multiple unmanned platforms can cooperate with each other in farmland and forests to efficiently monitor the growth status of vegetation. As the edge environment becomes increasingly complex and diverse, the application of multiple unmanned swarms in various scenarios and search tasks achieved through complementary capabilities and functional collaboration will surely be a future development trend and research hotspot.

[0004] The unmanned swarm collaborative search problem can be described as multiple unmanned platforms carrying target detection equipment to conduct collaborative detection of areas where a given target exists or may exist in the target environment, which belongs to the category of unmanned swarm task planning. Collaboration is the process or ability to coordinate and accomplish a certain goal by planning the task allocation and flight trajectory between multiple unmanned platforms. Although unmanned swarms can significantly reduce the design difficulty of the system and improve the flexibility of the system, as the scale of the swarm continues to increase, the task planning problem of unmanned swarms has become more complicated. Especially when performing comprehensive detection tasks such as coverage search and target search and rescue, the coverage task path planning of the unmanned swarm must not only arrange a reasonable flight route for each unmanned platform in the swarm to scan the entire area, but also ensure that the goal of the search task and the safety of the unmanned platform can be met to the greatest extent. This problem has been proven to have NP-hard computational complexity, and it is difficult to obtain the optimal solution within a limited time. In addition, the coverage task path planning of the unmanned swarm must take into account many factors such as the edge environment constraints of the swarm, the capability constraints of the unmanned platform, and the target constraints of the task, and provide one or more effective paths for each unmanned platform within a reasonable time to ensure the efficient completion of the task.

[0005] Unmanned platform mission planning is one of the key basic technologies to achieve the autonomy and intelligence of unmanned swarm systems, and it is also an effective way to improve the implementation of complex target tasks by unmanned swarms in edge environments. Efficient unmanned platform mission planning methods can reduce the flight cost and task completion time of unmanned platforms and improve work efficiency. According to the different planning moments, unmanned platform path planning can be divided into three types: offline planning, online planning, and collaborative planning. If all the environmental information of the unmanned platform flight is known in advance, the path planning is performed before the unmanned platform takes off, which is offline planning; if the environmental information is unknown or partially unknown and the path planning occurs during the flight, then the planning is online planning; when there are too many or too complex tasks, the unmanned swarm adopts a cooperative approach to jointly cope with unforeseen environmental changes and achieve the overall goal, which is collaborative planning. The current mission planning methods rarely consider the impact of the flight capability of the unmanned platform and the geographical location of the region on path planning, and it is difficult to adapt to the heterogeneity of unmanned swarms and regional collections, and there are problems of unreasonable task allocation and path planning. Summary of the invention

[0006] In view of this, the embodiment of the present application provides an unmanned cluster collaborative task planning method for edge environments, which can generate a reasonable coverage area task planning plan for the unmanned cluster, determine the areas that each unmanned platform needs to search and the optimal order to access these areas, and minimize the total completion time of the unmanned cluster to perform multiple area search tasks, so as to improve the ability of the unmanned cluster to perform complex tasks in edge environments. Although the existing rule-based area allocation method or cluster center-based area set division method has a fast execution time, the single search path reduces the method's ability to find a better planning solution in a large-scale search space, the global search capability is weak, and the performance of the method also decreases rapidly with the increase in the number of areas.

[0007] The present application provides an unmanned cluster collaborative task planning method for edge environments, the method comprising:

[0008] Based on an unmanned cluster composed of multiple unmanned platforms performing tasks in a geographical area, a random strategy and a heuristic strategy are used to generate an initial population including multiple individuals, each individual represents a mission planning scheme, the geographical area includes multiple different areas, and a mission planning scheme includes an allocation relationship between each area and each unmanned platform and an order in which each unmanned platform visits each area;

[0009] Based on the propagation operation, the refraction operation and the breaking wave operation, the individuals in the initial population are iteratively updated until the number of iterative updates reaches a preset maximum number of iterations, so as to obtain the optimal task planning scheme.

[0010] According to a specific implementation of the embodiment of the present application, the method of using a random strategy and a heuristic strategy to generate an initial population including a plurality of individuals includes:

[0011] The random strategy and K-nearest neighbor heuristic strategy are used to generate an initial population containing multiple individuals.

[0012] According to a specific implementation of the embodiment of the present application, the use of a random strategy and a K-nearest neighbor heuristic strategy to generate an initial population comprising a plurality of individuals includes:

[0013] Assume that the geographic region R contains m regions, and the currently allocated region cluster is C, where C = {C1, C2, ..., C n}, C i Indicates the current unmanned platform U i The regional access set, the regional access number queue is Q i , Represents C i Unmanned Platform U i The number of the area visited in sequence, For unmanned platformi The first area currently visited, For unmanned platform i The last area visited;

[0014] When assigning an unmanned platform to one of the m areas, select the unmanned platform with the shortest current completion time and extract the last area currently visited by the unmanned platform

[0015] Traverse the remaining unvisited areas and find the distance The nearest K regions, where K is a preset integer value;

[0016] Randomly select an area from the K areas, add the selected area to the end of the area access set corresponding to the unmanned platform, and add the number of the selected area to the end of the area access number queue corresponding to the unmanned platform;

[0017] Continue to allocate unmanned platforms to areas that have not been allocated unmanned platforms until all areas within the geographic area R are allocated unmanned platforms, and generate a mission planning plan;

[0018] Based on the randomness of randomly selecting an area from the K areas, a plurality of task planning schemes are generated, and the initial population is formed based on the plurality of task planning schemes.

[0019] According to a specific implementation of the embodiment of the present application, the iterative updating of individuals in the initial population based on the propagation operation, the refraction operation and the breaking wave operation includes:

[0020] Based on the propagation operation of neighboring area exchange, each individual in the initial population is iteratively updated to obtain a first iterative update result;

[0021] For the individuals trapped in the local optimum in the first iterative update result, a refraction operation based on social learning is performed to obtain a second iterative update result;

[0022] The total task completion time is calculated for each individual in the second iterative update result, and for the individual whose total task completion time is less than the total task completion time of the current global optimal individual, a breaking wave operation based on access order optimization is performed to complete the iterative update of the individual.

[0023] According to a specific implementation of the embodiment of the present application, the propagation operation based on neighboring area exchange performs iterative updates on each individual in the initial population to obtain a first iterative update result, including:

[0024] Calculate the wavelength of each individual;

[0025] For each individual, determine the set of regions that the individual needs to perform local search on according to the wavelength corresponding to each individual;

[0026] The propagation operation of the adjacent region exchange is performed on the region set of the individual to complete the iterative update of the individual and obtain the first iterative update result.

[0027] According to a specific implementation of the embodiment of the present application, the calculation formula of the wavelength is:

[0028]

[0029] Among them, λ s is the wavelength of the current individual, h max is the maximum wave height, and h is the current individual wave height.

[0030] According to a specific implementation of the embodiment of the present application, performing a refraction operation based on social learning on the individuals trapped in the local optimum in the first iterative update result to obtain the second iterative update result includes:

[0031] Calculating the total task completion time for each individual in the first iterative update result, and arranging each individual in ascending order according to the total task completion time to form an individual sequence;

[0032] For the individual S trapped in the local optimum, an individual S' is randomly selected from the first several individuals in the individual sequence as a learning object, and the learning probability is calculated based on the learning object;

[0033] Generate a random number for each area in the individual S, set the area where the random number is less than the learning probability as the area to be refracted, replace the unmanned platform assigned to the area to be refracted, and arrange the access order of the area to be after the area closest to the area corresponding to the replaced unmanned platform;

[0034] When all the areas to be refracted are replaced, a second iterative update result is obtained.

[0035] According to a specific implementation of the embodiment of the present application, the calculation formula of the learning probability is:

[0036]

[0037] Among them, p learn is the learning probability, T S’ (U, R) is the total completion time of the task of individual S', T S (U, R) is the total completion time of the task performed by individual S.

[0038] According to a specific implementation of an embodiment of the present application, the step of performing a wave-breaking operation based on access order optimization on an individual whose total task completion time is less than the total task completion time of the current global optimal individual includes:

[0039] Among the individuals whose total task completion time is less than the total task completion time of the current global optimal individual, find the unmanned platform with the longest task execution time;

[0040] Two adjacent areas are randomly selected from the area access set corresponding to the unmanned platform with the longest mission execution time. If the exchange of the two adjacent areas can reduce the mission execution time of the unmanned platform, the two adjacent areas are exchanged to complete the individual's wave-breaking operation based on access order optimization.

[0041] According to a specific implementation of the embodiment of the present application, the individuals in the initial population are generated in a step-by-step manner.

[0042] Beneficial effects:

[0043] The unmanned cluster collaborative task planning method for edge environments in the embodiment of the present application, at the problem level, through in-depth mining of regional search task knowledge, makes a problem-specific design for the water wave optimization algorithm that solves general problems, mainly including the allocation relationship between each area and each unmanned platform and the order in which each unmanned platform visits each area. It can quickly and efficiently find a planning solution for the unmanned cluster collaborative search task in the edge environment. At the method level, by adopting a systematic approach to improve the water wave optimization algorithm, based on the propagation operation of neighboring area exchange, the refraction operation of social learning and the breaking wave operation of access order optimization, the ability of the water wave optimization algorithm to solve discrete combinatorial optimization problems is improved, and the application scope of the water wave optimization algorithm is broadened. In terms of scalability, by modifying the coding method of the problem and redesigning the operators of the water wave optimization algorithm, the proposed systematic method can be used to solve other task planning problems such as vehicle routing and logistics scheduling, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 The present invention is a flowchart of an unmanned cluster collaborative task planning method for edge environments according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0047] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0048] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0049] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. The illustrations only show the components related to the present application rather than being drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.

[0050] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0051] The present application embodiment provides an unmanned cluster collaborative task planning method for edge environments. The method is described in detail below. Figure 1 , the method comprising:

[0052] Step S101: Based on an unmanned cluster composed of multiple unmanned platforms performing tasks in a geographical area, a random strategy and a heuristic strategy are used to generate an initial population including multiple individuals, where each individual represents a task planning scheme. The geographical area includes multiple different areas, and a task planning scheme includes an allocation relationship between each area and each unmanned platform and an order in which each unmanned platform visits each area.

[0053] Step S102, based on the propagation operation of neighboring area exchange, the refraction operation of social learning and the breaking wave operation of visit order optimization, the individuals in the initial population are iteratively updated until the number of iterative updates reaches a preset maximum number of iterations, so as to obtain the optimal task planning solution.

[0054] In this embodiment, the unmanned cluster collaborative task planning method for edge environment is an evolutionary calculation method based on the water wave optimization algorithm (WWO). The problem-specific knowledge (the allocation relationship between each area and each unmanned platform and the order in which each unmanned platform visits each area) is used to redesign the generation strategy of the initial population of WWO, the propagation operator, the refraction operator, and the breaking operator. From the perspective of the problem and the method level, a solution suitable for unmanned cluster task planning is designed, and it has the ability to quickly find excellent planning solutions. The method of this embodiment can generate a reasonable coverage area task planning solution for the unmanned cluster, determine the areas that each unmanned platform needs to search and the optimal order to visit these areas, minimize the total completion time of the unmanned cluster to perform multiple area search tasks, and achieve the improvement of the ability of the unmanned cluster to perform complex tasks in the edge environment. The task planning method proposed in this embodiment is not only applicable to regional coverage task planning problems, but can also be extended to scheduling problems such as vehicle routing and logistics scheduling, and has a wide range of application scenarios.

[0055] In the specific implementation, the content of the task planning solution, i.e., the solution representation, is first determined according to the task planning problem. Specifically, the regional coverage task planning problem in the edge environment aims to determine the coverage area search tasks that each unmanned platform needs to perform and the paths to access these areas under the premise of ensuring the maximum flight limit of the unmanned platform. Suppose an unmanned cluster U consisting of n heterogeneous unmanned platforms = {U1, U2, ..., U n} In the m-block disconnected region R = {R1, R2, ..., R m} to perform coverage search tasks. Each unmanned platform can be controlled by U i = <V i , W i > indicates that Vi , W i Respectively represent the flight speed of the unmanned platform and the scanning radius of the sensor configured on the unmanned platform. j The scan area is represented by A j , Region R j and R k The Euclidean distance between j，k Indicates. TS i，j Indicates unmanned platform U i In region R j The time required to complete the scanning task, TF i，j , k Indicates unmanned platform U i From area R j To area R k Flight time. TS i，j and TF i，j，k They can be calculated by the following formulas:

[0056]

[0057] Assume that all unmanned platforms are initially located in the same initial area R0, and the unmanned platform needs to return to the initial area after completing the coverage task. Since the unmanned platform does not need to scan the initial area, the scanning area of ​​R0 is 0, that is, R0 = 0. 0，j (j∈[1,m]) represents the initial region to region R j Based on the above formal representation of unmanned platforms and geographical areas, the regional coverage task of unmanned swarm can be defined as: given n independent and heterogeneous unmanned platforms UAV = {U1, U2, ..., U n} and m geographically dispersed regions R = {R1, R2, ..., R m}, assign area coverage scanning tasks to each unmanned platform, and determine the optimal flight path for each unmanned platform. Under the premise of meeting the energy consumption constraints of the unmanned platform, minimize the completion time of the unmanned cluster to complete all target area coverage scanning tasks.

[0058] Using the three-dimensional matrix X = {x i，j，k |i∈[1,n],j,k∈「0,m]} represents the flight path of the unmanned platform, where x i，j，k It is a Boolean variable, representing the unmanned platform U i Whether from region R j Fly to area R k If the unmanned platform U i From R j Fly to Region R k , then x i，j，k =1, otherwise x i，j，k =0.

[0059] T(U i ) and T(U, R) represent the unmanned platform U i The time it takes to execute the task and the task completion time of the entire unmanned cluster, T(U i ) and T(U, R) are calculated as follows:

[0060]

[0061]

[0062] The regional coverage mission planning problem requires finding the optimal flight path matrix X for the unmanned swarm so that the mission completion time is minimized, that is,

[0063] minT(U, R).

[0064] When using evolutionary computing methods such as the water wave optimization algorithm to solve optimization problems, it is first necessary to determine an effective representation for the solution to the problem. The regional coverage mission planning problem requires determining the allocation mapping between the region and the unmanned platform and the path for the unmanned platform to access the region. Different region and unmanned platform allocation schemes and different region access orders constitute the complete solution space of the problem. Since the regional coverage mission planning problem contains two sub-problems, the method uses two different representations to characterize different sub-problems. A complete solution (mission planning scheme) can be represented as S, which contains the following two parts:

[0065] 1. In order to describe the allocation of regions to unmanned platforms, the first half of S can be represented by a vector Y of dimension m, where each element of vector Y is an integer between 1 and n. When the value of the jth element in vector Y is i, region R j By unmanned platform U i Perform the search task. The vector Y satisfies the non-redundant constraint, that is, each area can only be assigned to one unmanned platform.

[0066] 2. The second half of S is represented by another vector Z with dimension m. Each element of vector Z is an integer between 1 and m. Vector Z determines the order in which the unmanned platform visits the area. When the value of the jth element in vector Z is k, area R j It is an unmanned platform Y[j] The kth task visited. All areas visited by each unmanned platform form a full permutation.

[0067] In one embodiment, the use of a random strategy and a heuristic strategy to generate an initial population comprising a plurality of individuals includes:

[0068] The random strategy and K-nearest neighbor heuristic strategy are used to generate an initial population containing multiple individuals.

[0069] Furthermore, the individuals in the initial population are generated in a step-by-step manner.

[0070] In this embodiment, in order to balance the exploration and detection capabilities of the method, the initial population is iteratively generated using a K-nearest neighbor-based strategy based on the random strategy. Each individual is constructed in a generative manner, that is, the method selects an area each time, and then allocates an unmanned platform to this area until all areas are allocated.

[0071] In one embodiment, the method of using a random strategy and a K-nearest neighbor heuristic strategy to generate an initial population comprising a plurality of individuals specifically comprises the following steps:

[0072] Assume that the geographic region R contains m regions, and the currently allocated region cluster is C, where C = {C1, C2, ..., C n}, C i Indicates the current unmanned platform U i The regional access set, the regional access number queue is Q i , Represents C i Unmanned Platform U i The number of the area visited in sequence, For unmanned platform i The first area currently visited, For unmanned platform i The last area visited;

[0073] When assigning an unmanned platform to one of the m areas, select the unmanned platform with the shortest current completion time and extract the last area currently visited by the unmanned platform

[0074] Traverse the remaining unvisited areas and find the distance The nearest K regions, where K is a preset integer value;

[0075] Randomly select an area from the K areas, add the selected area to the end of the area access set corresponding to the unmanned platform, and add the number of the selected area to the end of the area access number queue corresponding to the unmanned platform;

[0076] Continue to allocate unmanned platforms to areas that have not been allocated unmanned platforms until all areas within the geographic area R are allocated unmanned platforms, and generate a mission planning plan;

[0077] Based on the randomness of randomly selecting an area from the K areas, a plurality of task planning schemes are generated, and the initial population is formed based on the plurality of task planning schemes.

[0078] During the specific implementation, each individual in the population represents a task planning solution for the problem. Since the quality of the initial population has a great influence on the quality of the final output solution, it is necessary to design a good initial population generation strategy, which is not only to generate a sufficiently high-quality initial population, but also to ensure that the individuals in the initial population have a good distribution. The randomly generated initial population has the best distribution and diversity, but lacks elite individuals and has weak local search capabilities; the initial population generated only by heuristic methods contains a large number of elite individuals, but has poor diversity and is easy to converge. Therefore, in order to generate an initial population that meets the requirements and balance the diversity and convergence of the initial population, the method of this embodiment combines random strategies and heuristic strategies such as K nearest neighbors to generate the initial population.

[0079] The individuals in the population are generated in a step-by-step manner. In the step-by-step manner, the method selects an area according to certain rules and allocates an unmanned platform to this area based on randomness and heuristic information. The area set contains m areas, so the method needs to perform m steps to generate a legal individual. In each step, the method selects the unmanned platform with the smallest current completion time, that is, And take out the last area currently visited by this unmanned platform, that is, R Qse The method then traverses the remaining unvisited area set to find the distance R Qse The nearest K regions are selected. If there are less than K regions left, the remaining regions are taken into consideration. To increase the distribution of individuals in the search space, the method randomly selects one from the K neighboring regions and adds it to C. s The method continues to iterate until all areas are assigned unmanned platforms and the visit order is determined. The method then continues the above process until an initial population that meets the predetermined number is generated.

[0080] In one embodiment, the iterative updating of individuals in the initial population based on the propagation operation, the refraction operation and the breaking wave operation includes:

[0081] Based on the propagation operation of neighboring area exchange, each individual in the initial population is iteratively updated to obtain a first iterative update result;

[0082] For the individuals trapped in the local optimum in the first iterative update result, a refraction operation based on social learning is performed to obtain a second iterative update result;

[0083] The total task completion time is calculated for each individual in the second iterative update result, and for the individual whose total task completion time is less than the total task completion time of the current global optimal individual, a breaking wave operation based on access order optimization is performed to complete the iterative update of the individual.

[0084] Specifically, as an evolutionary algorithm based on population iteration, WWO mainly updates and iterates the population through propagation, refraction and breaking wave operators. In the evolutionary process of each generation, individuals first perform propagation operations, and then selectively perform refraction and breaking wave operations according to the execution results. When an individual falls into a local optimum, the refraction operation will be performed to jump out of the local optimum by learning from elite individuals; if the performance of an individual exceeds the current optimal solution, the breaking wave operation will be performed around this individual to explore the neighborhood space. In order to better optimize the performance of WWO, the propagation operator based on neighboring area exchange, the refraction operator based on social learning, and the breaking wave operator based on visit order optimization are redesigned in combination with problem-specific knowledge.

[0085] Furthermore, the propagation operation based on neighboring region exchange performs iterative updates on each individual in the initial population to obtain a first iterative update result, including:

[0086] Calculate the wavelength of each individual;

[0087] For each individual, determine the set of regions that the individual needs to perform local search on according to the wavelength corresponding to each individual;

[0088] The propagation operation of the adjacent region exchange is performed on the region set of the individual to complete the iterative update of the individual and obtain the first iterative update result.

[0089] Specifically, the calculation formula of the wavelength is:

[0090]

[0091] Among them, λ s is the wavelength of the current individual, h max is the maximum wave height, and h is the current individual wave height.

[0092] In practice, the WWO propagation operator is executed based on the principle that the higher the fitness value of an individual, the smaller the change made to it, and the strength of the change is determined by the wavelength λ of the individual. s According to the calculation formula of wavelength, the larger the wave height, the more times the individual has not been updated, and the more changes it should receive. Since the regional search problem is a combinatorial optimization problem, the wavelength λ of the individual is first calculated here s , as the number of changes that the individual needs to perform, and then choose λ sPerform change operations on each area.

[0093] The propagation operator first calculates the probability λ of each region selection s / m, then traverse the region set and select λ s Each area will perform a local search operation based on the exchange of neighboring areas. For each selected area, first find the K neighboring areas of this area, randomly select an area assigned to another unmanned platform, and exchange the unmanned platform ownership and access areas of the two areas. After all areas have performed local search operations, update the wave height and determine whether the total time cost of the updated new individual is less than that of the original individual. If the new individual performs better, replace the original individual, otherwise it remains unchanged.

[0094] Furthermore, performing a refraction operation based on social learning on the individuals trapped in the local optimum in the first iterative update result to obtain a second iterative update result includes:

[0095] Calculating the total task completion time for each individual in the first iterative update result, and arranging each individual in ascending order according to the total task completion time to form an individual sequence;

[0096] For the individual S trapped in the local optimum, an individual S' is randomly selected from the first several individuals in the individual sequence as a learning object, and the learning probability is calculated based on the learning object;

[0097] Generate a random number for each area in the individual S, set the area where the random number is less than the learning probability as the area to be refracted, replace the unmanned platform assigned to the area to be refracted, and arrange the access order of the area to be after the area closest to the area corresponding to the replaced unmanned platform;

[0098] When all the areas to be refracted are replaced, a second iterative update result is obtained.

[0099] Specifically, the calculation formula of the learning probability is:

[0100]

[0101] Among them, p learn is the learning probability, T S’ (U, R) is the total completion time of the task of individual S', T S (U, R) is the total completion time of the task performed by individual S.

[0102] In specific implementation, the WWO refraction operator guides the individual to jump out of the local optimum when the individual falls into the local optimum, that is, when the wave height becomes zero. The traditional refraction operator allows the individual to learn from the global optimal individual to jump out of the local optimum, but all individuals evolve in the direction of the global optimal individual, which easily leads to a rapid reduction in the diversity of the population, which is particularly obvious in large-scale optimization problems. In order to slow down the convergence speed of the population, this embodiment proposes a refraction operation based on social learning. In social life, individuals will learn from different excellent individuals, not limited to the best individuals.

[0103] The refraction operator first arranges the individuals in the population in ascending order according to their total time cost, and the individuals with the shortest task completion time are placed at the front of the individual sequence. For an individual S to be refraction-operated, an individual S' is randomly selected from the individuals at the front of the sequence as the learning object, and the learning probability p is calculated according to the formula learn . Then, for each region in S, if the generated random number is less than the learning probability p learn , the unmanned platform assigned to the area is replaced, and its access order is arranged to be after the new unmanned platform is closest to this area. After all areas are allocated, the wave height is updated and the total time cost of the new individual is calculated.

[0104] Furthermore, the step of performing a wave-breaking operation based on access order optimization on an individual whose total task completion time is less than the total task completion time of the current global optimal individual includes:

[0105] Among the individuals whose total task completion time is less than the total task completion time of the current global optimal individual, find the unmanned platform with the longest task execution time;

[0106] Two adjacent areas are randomly selected from the area access set corresponding to the unmanned platform with the longest mission execution time. If the exchange of the two adjacent areas can reduce the mission execution time of the unmanned platform, the two adjacent areas are exchanged to complete the individual's wave-breaking operation based on access order optimization.

[0107] In the specific implementation, the WW0 wave-breaking operator performs local search near the global optimal individual, and improves the ability of the optimal individual to refine towards the optimal solution through a small search amplitude. The method uses local search based on access order optimization to implement the wave-breaking operation. Since the total time cost of the unmanned cluster is determined by the unmanned platform with the longest access time, the unmanned platform with the longest completion time is first determined, and two adjacent areas are randomly selected from the area access set of the unmanned platform. If exchanging these two areas can reduce the task completion time, then exchange them, otherwise remain unchanged.

[0108] The overall execution process of the method is described in detail below. The method of this embodiment adopts the water wave optimization algorithm framework, and combines the problem-specific knowledge to design a discretization method suitable for the problem to be solved, with an emphasis on improving the initial population generation strategy and the propagation, refraction, and wave-breaking operators of individual iterative updates. It mainly includes two stages: the initialization stage and the iterative update stage. The initialization stage generates an initial population that meets diversity and convergence. The iterative update stage continuously optimizes the individuals in the population. When the predetermined end condition is reached, the method ends and outputs the optimal planning solution found. The overall execution process of the method is as follows:

[0109] 1: Input. Input unmanned cluster UAV = {U1, U2, ..., U n}, region set R = {R1, R2, ..., R m}, maximum wave height h max , population size NP, number of neighbors K, maximum number of iterations G max .

[0110] 2: Initialization phase. Generate NP individuals in a step-by-step manner to form an initial population;

[0111] 2.1: Continue to select the unmanned platform U1 with the shortest completion time and the last visited area Find The K nearest neighbor regions of ;

[0112] 2.2: Randomly select one from the K nearest neighbor regions and append it to C s In the process, jump to 2.1 until all regions are allocated;

[0113] 2.3: If the current population size reaches NP, jump to 3 to execute, otherwise jump to 2.1 to iterate.

[0114] 3: Iteration phase. Continue to use propagation, refraction, and breaking wave operators to optimize individuals in the population until the end condition is reached;

[0115] 3.1: Calculate the wavelength of each individual;

[0116] 3.2: For each individual, calculate the set of areas where local search needs to be performed according to the wavelength and perform the propagation operation;

[0117] 3.3: If the wave height of the individual becomes zero, perform the refraction operation and optimize towards the direction of the elite individual through social learning;

[0118] 3.4: If the individual's current total completion time is less than the global optimum, perform the wave-breaking operation and perform local detection in the individual's neighborhood space;

[0119] 3.5: Determine whether the current number of iterations has reached the maximum number of iterations G max If reached, jump to 4, otherwise jump to 3.1 to iterate.

[0120] 4: Output. According to the optimal individual S found in the iterative optimization phase best , output S best The corresponding unmanned platform flight matrix X.

[0121] The embodiment of the present application aims at the problem characteristics of the regional search task in the edge environment, decomposes the problem into two sub-problems: unmanned platform area allocation and unmanned platform path planning (region access order), and establishes a coding method covering the two sub-problems; in order to generate an initial population with good distribution characteristics and excellent solutions, the method integrates heuristic strategies such as random strategies, K nearest neighbor algorithms, and minimum completion time to generate an initial population that meets the scale and distribution characteristics, and each individual in the initial population represents a planning scheme; in order to effectively solve the combinatorial optimization problem such as the regional search task, a systematic design method is adopted to design a propagation operator and a wave height update method; in order to avoid individuals in the population from being trapped in the local optimum for a long time, a refraction operator based on a social learning strategy is designed; in order to improve the detection ability of the method, a wave-breaking operator based on access order optimization is designed, which can effectively search the neighborhood space of the current optimal method. The method generally includes two stages: the initialization stage and the iterative update stage.

[0122] In the initialization phase, the method uses a combination of random strategies and heuristic strategies to generate an initial population of sufficient number. The initial population consists of a certain number of individuals. The randomly generated individuals make the population evenly distributed in the search space, and the method can fully search the solution space and has strong exploration ability. The heuristic strategy is used to generate individuals with excellent performance. The method performs local search around these individuals, which can accelerate convergence and has strong detection ability. In order to balance the exploration and detection capabilities of the method, this method uses a K-nearest neighbor-based strategy to iteratively generate the initial population. Each individual is constructed in a generative way, that is, the method selects an area each time, and then allocates an unmanned platform to this area until all areas are allocated.

[0123] In the iterative update phase, the method is updated according to the preset number of generations. In each generation, all individuals first perform propagation operations and update wave heights, and then selectively perform refraction operations or wave-breaking operations. If an individual has not been updated for a long time, that is, it is trapped in a local optimum, then the refraction operation is performed to jump out of the local optimum by learning from excellent individuals. In addition, if the current fitness value of the individual exceeds the global optimal individual, the wave-breaking operation is performed to perform a local search around this individual to improve the refinement ability of the method. In order to effectively solve the problem of unmanned cluster area search task planning in edge environments, the method makes targeted designs for propagation, refraction, and wave-breaking operators based on the traditional water wave optimization algorithm framework, and proposes a propagation operator based on neighboring area exchange, a refraction operator based on social learning, and a wave-breaking operator based on access order optimization. Individuals in the population perform propagation, refraction, and wave-breaking operations according to the rules until the preset number of generations is reached, and the planning scheme with the best fitness value is output.

[0124] Therefore, compared with the prior art, this application has three main advantages: 1. At the problem level, through in-depth mining of regional search task knowledge, a problem-specific design is made for the water wave optimization algorithm that solves general problems, which can quickly and efficiently find planning solutions for unmanned cluster collaborative search tasks in edge environments. 2. At the method level, by adopting a systematic approach to improve the water wave optimization algorithm, the ability of the water wave optimization algorithm to solve discrete combinatorial optimization problems is enhanced, and the scope of application of the water wave optimization algorithm is broadened. 3. In terms of scalability, by modifying the coding method of the problem and redesigning the operators of the water wave optimization algorithm, the proposed systematic method can be used to solve other task planning problems such as vehicle routing and logistics scheduling, and has good application prospects.

[0125] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for unmanned cluster collaborative task planning for edge environments, characterized in that: The method comprises: Based on an unmanned cluster composed of multiple unmanned platforms performing tasks in a geographical area, a random strategy and a heuristic strategy are used to generate an initial population including multiple individuals, each individual represents a mission planning scheme, the geographical area includes multiple different areas, and a mission planning scheme includes an allocation relationship between each area and each unmanned platform and an order in which each unmanned platform visits each area; Based on the propagation operation of neighboring area exchange, the refraction operation of social learning and the breaking wave operation of visit order optimization, the individuals in the initial population are iteratively updated until the number of iterative updates reaches a preset maximum number of iterations, thereby obtaining an optimal task planning scheme.

2. The unmanned cluster collaborative task planning method for edge environment according to claim 1 is characterized in that: The random strategy and the heuristic strategy are used to generate an initial population containing multiple individuals, including: The random strategy and K-nearest neighbor heuristic strategy are used to generate an initial population containing multiple individuals.

3. The unmanned cluster collaborative task planning method for edge environment according to claim 2 is characterized in that: The random strategy and the K-nearest neighbor heuristic strategy are used to generate an initial population containing multiple individuals, including: Assume that the geographic region R contains m regions, and the currently allocated region cluster is C, where C = {C1, C2, ..., C n }, C i Indicates the current unmanned platform U i The regional access set, the regional access number queue is Q i , Represents C i Unmanned Platform U i The number of the area visited in sequence, For unmanned platform i The first area currently visited, For unmanned platform i The last area visited; When assigning an unmanned platform to one of the m areas, select the unmanned platform with the shortest current completion time and extract the last area currently visited by the unmanned platform Traverse the remaining unvisited areas and find the distance The nearest K regions, where K is a preset integer value; Randomly select an area from the K areas, add the selected area to the end of the area access set corresponding to the unmanned platform, and add the number of the selected area to the end of the area access number queue corresponding to the unmanned platform; Continue to allocate unmanned platforms to areas that have not been allocated unmanned platforms until all areas within the geographic area R are allocated unmanned platforms, and generate a mission planning plan; Based on the randomness of randomly selecting an area from the K areas, a plurality of task planning schemes are generated, and the initial population is formed based on the plurality of task planning schemes.

4. The unmanned cluster collaborative task planning method for edge environment according to claim 3 is characterized in that: The individuals in the initial population are iteratively updated based on the propagation operation of neighboring area exchange, the refraction operation of social learning and the breaking wave operation of visit order optimization, including: Based on the propagation operation of neighboring area exchange, each individual in the initial population is iteratively updated to obtain a first iterative update result; For the individuals trapped in the local optimum in the first iterative update result, a refraction operation based on social learning is performed to obtain a second iterative update result; The total task completion time is calculated for each individual in the second iterative update result, and for the individual whose total task completion time is less than the total task completion time of the current global optimal individual, a breaking wave operation based on access order optimization is performed to complete the iterative update of the individual.

5. The unmanned cluster collaborative task planning method for edge environment according to claim 4 is characterized in that: The propagation operation based on neighboring region exchange is iteratively updated for each individual in the initial population to obtain a first iterative update result, including: Calculate the wavelength of each individual; For each individual, determine the set of regions that the individual needs to perform local search on according to the wavelength corresponding to each individual; The propagation operation of the adjacent region exchange is performed on the region set of the individual to complete the iterative update of the individual and obtain the first iterative update result.

6. The unmanned cluster collaborative task planning method for edge environment according to claim 5 is characterized in that: The wavelength is calculated as: Among them, λ s is the wavelength of the current individual, h max is the maximum wave height, and h is the current individual wave height.

7. The unmanned cluster collaborative task planning method for edge environment according to claim 4 is characterized in that: The step of performing a refraction operation based on social learning on the individuals trapped in the local optimum in the first iterative update result to obtain a second iterative update result includes: Calculating the total task completion time for each individual in the first iterative update result, and arranging each individual in ascending order according to the total task completion time to form an individual sequence; For the individual S trapped in the local optimum, an individual S' is randomly selected from the first several individuals in the individual sequence as a learning object, and the learning probability is calculated based on the learning object; Generate a random number for each area in the individual S, set the area where the random number is less than the learning probability as the area to be refracted, replace the unmanned platform assigned to the area to be refracted, and arrange the access order of the area to be after the area closest to the area corresponding to the replaced unmanned platform; When all the areas to be refracted are replaced, a second iterative update result is obtained.

8. The unmanned cluster collaborative task planning method for edge environment according to claim 7 is characterized in that: The calculation formula of the learning probability is: Among them, p learn is the learning probability, T S’ (U, R) is the total completion time of the task of individual S', T S (U, R) is the total completion time of the task performed by individual S.

9. The unmanned cluster collaborative task planning method for edge environment according to claim 4 is characterized in that: The step of performing a wave-breaking operation based on access order optimization on an individual whose total task completion time is less than the total task completion time of the current global optimal individual includes: Among the individuals whose total task completion time is less than the total task completion time of the current global optimal individual, find the unmanned platform with the longest task execution time; Two adjacent areas are randomly selected from the area access set corresponding to the unmanned platform with the longest mission execution time. If the exchange of the two adjacent areas can reduce the mission execution time of the unmanned platform, the two adjacent areas are exchanged to complete the individual's wave-breaking operation based on access order optimization.

10. The unmanned cluster collaborative task planning method for edge environment according to claim 1, characterized in that: The individuals in the initial population are generated in a step-by-step manner.

Images