Intelligent resource combining and scheduling method oriented to hierarchical organization relationship

Abstract

The invention discloses a hierarchical organization relationship-oriented intelligent resource combination and scheduling method. The method comprises the following steps of: 1, quantitatively describing a combat management entity, a combat execution entity and a combat target entity; 2, constructing a target clustering model based on a combat target entity; 3, generating a target clustering result by using a target clustering method according to the target clustering model; 4, constructing a resource combination and scheduling model based on the combat management entity, the combat execution entity and the combat target entity according to a target clustering result; and 5, solving the model in the step 4 through a resource combination and scheduling model solving algorithm, and generating a resource combination and scheduling result. According to the method, target clustering modeling and optimization problem modeling are adopted to construct a resource combination and scheduling problem oriented to a hierarchical organization relationship, an intelligent optimization method is adopted to solve the model, and an efficient coding and decoding mechanism is introduced into a corresponding algorithm design, so that the algorithm solving efficiency is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of aerial collaboration technology, and in particular to a resource intelligent combination and scheduling method oriented to hierarchical organizational relationships. Background Art

[0002] Combining and scheduling combat resources facing hierarchical organizational relationships is the key to ensuring efficient allocation and reasonable scheduling of combat resources, and is also an important basis for achieving collaborative and agile air combat. In the existing technology, the hierarchical-hierarchical method is mainly used to conduct corresponding research on the intelligent combination and scheduling of air combat resources. For example, Li Xiangmin et al. proposed a three-layer structure of formation, group and platform in the paper "A Review of Key Technologies for Collaborative Combat Command and Control of Formations with / Combat Execution Entities". The formation layer takes combat missions and battlefield situations as inputs, and mainly completes the formation of tactical groups and mission allocation of combat management entities-combat execution entities; the group layer takes tactical grouping plans and task information as inputs, and mainly completes the time-series task allocation of combat execution entities; the platform layer takes the task allocation plan as input, and carries out detailed track planning and load application planning, etc. In the paper "Hierarchical Method of Task Assignment for Multiple Cooperating UAV Teams", Hu Xiaoxuan et al. divided the problem of task assignment for multiple combat execution entities into three sub-problems in order to reduce the scale of problem solving: task clustering, task cluster assignment, and intra-cluster task assignment. Each combat execution entity formed a de facto collaborative formation to perform tasks within each cluster. Finally, this hierarchical task assignment method was compared with the non-hierarchical task assignment method, showing the advantages of this method in terms of timeliness of assignment. In the paper "Design Method of Organizational Structure for MAVs and UAVs Heterogeneous Team with Adjustable Autonomy", Chen Jun et al. established multiple constraints including payload resources, decision resources, network effectiveness, network survivability, task execution trust, interaction delay, decision information processing capability, etc., and constructed a mathematical model with the optimization goal of optimizing the weighted sum of the above factors. The design results of the collaborative relationship, supervisory control relationship, and decision authorization relationship in the combat organization were obtained through optimization solution.

[0003] A Chinese patent with the patent number CN112734278B discloses a method for dynamically scheduling command and control organizational resources for time-sensitive target strikes. The method takes maximizing the quality of task completion and minimizing the time required for task completion as the optimization goals, constructs a dynamic scheduling model for command and control organizational resources under multi-objective conditions, and uses an improved quantum particle swarm algorithm to solve the model, thereby improving the ability to strike time-sensitive targets.

[0004] In the Chinese patent with patent number CN113392521B, a method and device for constructing a resource grouping model for joint sea and air combat tasks are disclosed. A hierarchical behavior tree network is established based on the decomposed tasks, and resources are decomposed to construct a resource multi-objective optimization model, and single-objective optimization solutions are performed in order of priority, thereby improving the ability to make rapid tactical decisions under complex, uncertain and highly dynamic conditions.

[0005] In a Chinese patent with patent number CN111274524B, a resource scheduling method under sudden alarm events is disclosed. According to the threat level of the sudden alarm event, defense weapons are allocated to the sudden alarm events to be allocated after judgment. It has the advantages of simple method and rapid scheduling.

[0006] However, most of the above existing resource combination and scheduling methods have certain application limitations, with numerous constraints and large solution scales. The optimization problem-solving algorithms are difficult to handle complex constraints and have low solution efficiency, making it difficult to generate resource combination and scheduling solutions with high solution quality. Summary of the invention

[0007] In view of the above-mentioned problems, the present invention proposes a resource intelligent combination and scheduling method for hierarchical organizational relationships. The resource combination and scheduling problem for hierarchical organizational relationships is constructed by adopting target clustering modeling and optimization problem modeling, and an intelligent optimization method is used to solve the model. An efficient encoding and decoding mechanism is introduced in the corresponding algorithm design to improve the efficiency of the algorithm solution.

[0008] The present invention provides a resource intelligent combination and scheduling method oriented to hierarchical organizational relationships, comprising the following steps:

[0009] Step 1: Define the combat management entity, combat execution entity, and combat target entity;

[0010] Step 2: Construct a target clustering model based on combat target entities;

[0011] Step 3: Based on the target clustering model, the target clustering result is calculated using the greedy clustering algorithm that selects the initial cluster center;

[0012] Step 4: Based on the target clustering results, construct a resource combination and scheduling model based on combat management entities, combat execution entities and combat target entities;

[0013] Step 5: Based on multi-objective constraint processing, discrete artificial bee colony algorithm is used to solve the resource combination and scheduling model;

[0014] Step 6: Combine and schedule the current resources according to the obtained resource combination and scheduling results.

[0015] Furthermore, the specific operation steps of step 1 are:

[0016] Step 1.1: Define the combat management entity set consisting of |M| combat management entities as M = {M1, M2, …, M |M|},for Each combat management entity M in the set i The basic attribute is the charge load threshold CCM i ;

[0017] Step 1.2: Define the set of combat execution entities consisting of |U| combat execution entities as U = {U1, U2, …, U |U|},for Each combat execution entity U in the set j The basic attribute is the resource capability vector Among them, |R| is the number of resource types, and the superscript u represents the combat execution entity; if but Represents the combat execution entity U j The number of combat resources of the nth category, if It means U j Not having the nth type of combat resources;

[0018] Step 1.3: Define the combat target set consisting of |TAR| combat target entities as for Each combat target entity TAR in the collection k The basic attribute is the resource demand vector and the position vector

[0019]

[0020] Furthermore, the specific steps of establishing the target clustering planning model in step 2 include:

[0021] Step 2.1: Set the number of targets in each target cluster to a constant or and and will contain the target number or The number of target clusters is defined as A1 and A2 respectively, and the values ​​of A1 and A2 are calculated by the following formula:

[0022]

[0023] Among them, |TAR| represents the number of combat targets, and |TC| represents the number of target clusters;

[0024] Step 2.2: Construct the target clustering model:

[0025]

[0026] in, d(TAR k ,TC m ) normalized value; d(TAR k ,TC m ) represents the combat target entity TAR k and the mth target cluster TC m The target position distance between is the decision variable for whether the kth target will be classified into the mth target cluster; is the number of targets in each target cluster.

[0027] Furthermore, the specific operation steps of step 3 are:

[0028] Step 3.1: Calculate the target position distance between each combat target entity in the combat target entity set according to the target position distance calculation formula, generate the distance matrix D0, and find the maximum value d in D0 max , and d max The corresponding targets TAR1 and TAR2 are used as the initial cluster centers of different target clusters;

[0029] The target position distance calculation formula is:

[0030]

[0031] Among them, d(TAR k ,TAR k' ) represents two combat target entities TAR k and TAR k' The target position distance between They represent the abscissa of the k-th target, the abscissa of the k'-th target, the ordinate of the k-th target, and the ordinate of the k'-th target respectively;

[0032] Step 3.2: Determine whether |TC|=2 holds. If so, go to step 3.5; otherwise, calculate the target position distances between all remaining targets and TAR1 and TAR2 to form matrix D1, and perform non-dominated sorting on all 2D distance vectors in D1, select the target corresponding to the distance vector with the largest sorting value, and if there is only one target, use that target as the cluster center TAR3 of the third target cluster; otherwise, execute step 3.3;

[0033] Step 3.3: Calculate the crowding distance for all 2D distance vectors with the largest ranking value, and select the target corresponding to the maximum crowding distance as the cluster center TAR3 of the third target cluster;

[0034] Step 3.4: Determine whether the number of initial cluster centers has reached the preset |TC|. If so, execute step 3.5; if not, return to step 3.2 to continue calculating all remaining targets and all existing cluster centers TAR1, TAR2, ..., TAR m The distance values ​​form the matrix D m-1 , until the number of initial cluster centers reaches |TC|, and then the corresponding non-dominated sorting and crowding distance calculation are performed;

[0035] Step 3.5: Calculate the distances of all remaining targets to the centers of |TC| clusters respectively. For each cluster, select a target closest to the cluster to be merged. When different clusters need to merge the same target, the target is preferentially merged into the cluster with the closer distance, and the other cluster is merged into the target with the second closest distance.

[0036] Step 3.6: Update the cluster center coordinates using the following formula and determine the number of targets included: Whether the number of target clusters reaches A1, or contains the target number Whether the number of target clusters reaches A2, if so, the target clusters that reach the target number limit will no longer be merged into the new target;

[0037] The cluster center coordinate update formula is:

[0038]

[0039] Among them, a * is the updated cluster center coordinate, N is the number of targets, a n is the nth target coordinate;

[0040] Step 3.7: Determine whether all targets have been incorporated into the cluster. If so, terminate. Otherwise, execute step 3.5.

[0041] Furthermore, the resource combination and scheduling model constructed in step 4 is:

[0042]

[0043] and

[0044]

[0045] Among them, η0 is the minimum redundancy of the command and control capability of the combat management entity in each target cluster; η1 is the minimum redundancy of various resource capabilities in each target cluster; The decision variable indicating whether the i-th combat management entity executes the m-th target cluster task; The decision variable indicating whether the jth combat execution entity executes the mth target cluster task.

[0046] Furthermore, the specific steps of step 5 include:

[0047] Step 5.1: First, randomly generate the initial honey source within the feasible interval to obtain the initial platform-target cluster matching matrix Among them, the platform refers to the combat management entity or combat execution entity. The matrix X P-TC The number of rows is the number of nectar sources N fs , the number of columns is the sum of the number of combat execution entities and combat management entities |U|+|M|, the first |U| columns represent the combat execution entity-target cluster matching relationship, and the |U|+1~|U|+|M| columns represent the combat management entity-target cluster matching relationship;

[0048] Next, the matrix X is calculated using the following honey source location generation formula: P-TC The element x in row i and column j ij :

[0049]

[0050] in, and Represents the matrix X P-TC The upper and lower bounds of the j-th column variable, rand represents a random number in the interval (0,1);

[0051] Finally, set the capacity N of the external archive ψ used to store the space for saving the previous optimal solutions rf N rf =N fs , and save the solutions corresponding to all initial nectar sources in ψ;

[0052] Step 5.2: The employed bees look for a better nectar source location near the initial nectar source and use the following nectar source update formula to update the nectar source location:

[0053]

[0054] Among them, x ij ' is the element of the newly generated nectar source, To randomly select the element corresponding to the nectar source from ψ, x i'j is the corresponding position element of other nectar sources, RAND is the update threshold, which is generally taken as 0.5; if x ij 'Beyond , then take x ij ' is the nearest boundary value, that is:

[0055]

[0056] The fitness value of the new nectar source is calculated. If the new nectar source dominates the old nectar source, it is replaced; if the new nectar source is dominated, the old nectar source is kept unchanged; if they are equivalent, a random choice is made to replace or keep it unchanged;

[0057] Step 5.3: After all employed bees have completed their search, they exchange information about the location of the nectar source with the observation bees. The observation bees use the tournament operator to select the best nectar source. fs Select two nectar sources from the nectar sources, and the corresponding two fitness values ​​are recorded as F1 = (f1', f2') and F2 = (f1", f2"). If F1 dominates F2, it satisfies At this time, the first nectar source is selected and updated according to the nectar source update formula; otherwise, the second nectar source is selected and updated according to the nectar source update formula;

[0058] Step 5.4: When the preset iter_limit iterations are reached, if the solution quality of the nectar source has not improved, the employed bees corresponding to the nectar source will become scout bees, abandon the original nectar source, and generate a new nectar source location according to the nectar source location generation formula.

[0059] Furthermore, we calculate the matrix X P-TC The steps include:

[0060] Step 5.1.1: Decode the nectar source. Each nectar source is a vector with 1 row and n columns, and the individual position value range is [1,|TC|+1). After decoding, X P-TC Mapped to a 3D matrix

[0061]

[0062] Step 5.1.2: For W P-TC All non-zero elements in the row vector formed by the first row and the first to |U| columns of the first 2D matrix are set to 0 from left to right. After each setting, check whether the following equation is true. If not, reset the elements that were set to 0 to 1:

[0063]

[0064] in, and They are the resource capacity satisfaction of the nth resource before and after each non-zero element is set to 0;

[0065] Next, for W P-TC The above operations are performed on the |TC|-1 row vectors consisting of the 2nd to |TC| rows and the 1st to |U| columns of the first 2D matrix in ;

[0066] Step 5.1.3: Determine whether the resource capacity of all resource classes meets the requirements. If so, execute step 5.1.5; if not, P-TC All non-1 elements in the row vector formed by the first row and the first to |U| columns of the first 2D matrix are set to 1 from left to right. After each setting, it is determined whether the following equation is true. If so, the set elements are reset to 0:

[0067]

[0068] Among them, |R'| is the number of resource types whose resource capacity does not reach the demand threshold. and They are the capacity satisfaction of the n'th resource before and after each non-1 element is set to 1;

[0069] Next, determine whether to stop the setting operation according to the following formula. If it is satisfied, stop. Otherwise, continue to set it to 1:

[0070]

[0071] Finally, for W P-TC The above operations are performed on the |TC|-1 row vectors consisting of the 2nd to |TC| rows and the 1st to |U| columns of the first 2D matrix in ;

[0072] Step 5.1.4: For W P-TC Repeat step 5.1.2 for the first 2D matrix in W P-TC 2nd to Nth fs Execute steps 5.1.2 to 5.1.4 for each 2D matrix;

[0073] Step 5.1.5: For the matrix W P-TC The |TC| row vectors consisting of the 1st to |TC|th rows and the |U|+1st to |U|+|M|th columns of each 2D matrix in execute steps 5.1.2 to 5.1.4. When executing step 5.13, the command capability satisfaction of the target cluster is calculated by the following formula, and it is judged whether the command capability satisfaction meets the requirements, and finally the feasible solution matrix Z is obtained. P-TC ;

[0074] The target cluster command capability satisfaction degree represents the degree to which all UAV command requirements matching each target cluster are met. m The calculation formula of the target cluster command capability satisfaction is:

[0075]

[0076] Among them, CCS m The target cluster is charged with capability satisfaction;

[0077] Step 5.1.6: Let X P-TC =Z P-TC , we can get the combat management entity-target cluster matching relationship matrix X M-TC And the combat execution entity-target cluster matching relationship matrix X U-TC .

[0078] Therefore, the present invention adopts the above-mentioned resource intelligent combination and scheduling method oriented to hierarchical organizational relationships, which has the following beneficial effects:

[0079] First, from the perspective of improving the robustness of the formation plan, the present invention establishes a target clustering model that minimizes the target distance and a resource combination and scheduling model that minimizes the redundant cost of (command and control, resource) capabilities, and respectively adopts a greedy clustering algorithm for selecting the initial cluster center and a discrete artificial bee colony algorithm for multi-objective constraint processing to solve the problem; the defects of the existing resource combination and scheduling methods are overcome, and the original problem is divided into two sub-problems of target clustering and resource scheduling from a hierarchical perspective, which effectively reduces the scale of the solution, is feasible and robust;

[0080] Second, since the construction method of the present invention adopts a hierarchical modeling mechanism, target clustering is first performed, and then resource combination and scheduling are performed, a single large-scale problem is decomposed into multiple small-scale problems, which reduces the complexity of problem solving and effectively realizes problem transformation;

[0081] Third, since the present invention adopts a greedy clustering algorithm that selects the initial cluster center to solve the target clustering model during the model solving process, it effectively avoids the problem of poor clustering effect caused by random selection of the initial cluster center of the k-best algorithm;

[0082] Fourthly, the present invention adopts a multi-objective constraint processing discrete artificial bee colony algorithm to solve the resource combination and scheduling model, which solves various problems such as low encoding and decoding efficiency and difficult processing of constraints in existing resource combination and scheduling problems.

[0083] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1It is the overall flow chart of resource combination and scheduling in the present invention;

[0085] Figure 2 It is a flow chart of the greedy clustering algorithm for selecting the initial cluster center in the present invention;

[0086] Figure 3 It is the relationship diagram between the decision variable space and the target space in the present invention;

[0087] Figure 4 is a decoding rule diagram in the present invention;

[0088] Figure 5 It is a comparison chart of clustering measurement values ​​of various algorithms in the present invention;

[0089] Figure 6 It is the target clustering result diagram under the preferred initial cluster center greedy clustering algorithm of the present invention;

[0090] Figure 7 It is a distribution diagram of non-inferior feasible solutions for each iteration number under the greedy clustering algorithm for selecting the initial cluster center in the present invention;

[0091] Figure 8 It is a comparison chart of coverage index in the present invention;

[0092] Fig. 9 It is a comparison chart of uniformity index in the present invention;

[0093] Fig.10 It is a comparison chart of the breadth index in the present invention. DETAILED DESCRIPTION

[0094] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art make various changes or modifications to the present invention, and these equivalent forms also fall within the scope limited by the claims attached to the application.

[0095] The present invention proposes a resource intelligent combination and scheduling method oriented to hierarchical organizational relationships, which comprises the following steps:

[0096] Step 1: Quantitatively describe the combat management entity, combat execution entity, and combat target entity;

[0097] Step 1.1: The combat management entity is used to perform combat management and, when necessary, to coordinate tactics with the combat execution entity to jointly complete combat tasks; the combat management entity set is denoted as M = {M1, M2, …, M |M|}, where |M| is the number of combat management entities; Each management entity Mi The basic attribute is the charge load threshold CCM i ;

[0098] Step 1.2: Under the management of the combat management entity, the combat execution entity performs various combat tasks at the front line of the battlefield and makes limited autonomous decisions when conditions permit. The set of combat execution entities is denoted as U = {U1,U2,…,U |U|}, where |U| is the number of combat execution entities; Each combat execution entity U j The basic attribute is the resource capability vector Where, the superscript u represents the combat execution entity (drone), and the superscript u indicates that all quantities are related to the combat execution entity; |R| is the resource type; if but Represents the combat execution entity U j The number of combat resources of the nth category, if It means U j Not having the nth type of combat resources;

[0099] Step 1.3: The combat target entity is the direct combat target; the combat target entity set is recorded as Among them, |TAR| is the number of combat targets; Each combat target entity TAR k The basic properties are: resource demand vector and location

[0100] Step 2: Construct a target clustering model based on combat target entities;

[0101] Step 2.1: Calculate the target position distance using formula (1):

[0102]

[0103] Among them, d(TAR k ,TAR k' ) represents two combat target entities TAR k and TAR k' The target position distance between them is used to measure the difference between the positions of two combat target entities. k and TAR k' When the position difference is not large, the two targets should belong to the same target cluster. By clustering targets, the targets with close geographical distances can be grouped into one category, which can effectively reduce the navigation time of the combat execution entity to execute the two targets sequentially. They represent the abscissa of the k-th target, the abscissa of the k'-th target, the ordinate of the k-th target, and the ordinate of the k'-th target respectively;

[0104] Step 2.2: Based on the target location distance, a target clustering model is established to ensure that the target load of each group is relatively balanced;

[0105] Specifically, the specific operation steps of step 2.2 to establish the target clustering planning model include:

[0106] Step 2.2.1: To ensure that the target load of each group is relatively balanced, set the number of targets in each target cluster to a constant. or in,

[0107] and will contain the target number or The number of target clusters is defined as A1 and A2 respectively, and the values ​​of A1 and A2 are calculated by the following formula:

[0108] A1+A2=|TC| (2)

[0109]

[0110] Among them, TC represents the target cluster;

[0111] According to formulas (2) and (3), we can get

[0112] Step 2.2.2: Establish the following target clustering planning model:

[0113]

[0114] in, d(TAR k ,TC m )’s normalized value; TC m is the mth target cluster; is the decision variable for whether the kth target will be included in the mth target cluster; the first constraint condition indicates that any target can only belong to one target cluster; the second constraint condition indicates that all targets must belong to a certain target cluster; the third constraint condition ensures that the target load of each combat formation is relatively balanced; the fourth constraint condition indicates The value is 0 or 1.

[0115] Step 3: Calculate the target clustering result using the greedy clustering algorithm that selects the initial cluster center;

[0116] like Figure 2 As shown, the present invention proposes a greedy clustering algorithm for selecting the initial cluster center, which includes the following steps:

[0117] Step 3.1: Calculate the target position distances of all combat target entities in the set TAR according to formula (1), generate the distance matrix D0, and select the maximum value d in D0 max , and d max The corresponding targets TAR1 and TAR2 are used as the initial cluster centers of different target clusters;

[0118] Step 3.2: Determine whether |TC|=2 holds. If so, go to step 3.5; otherwise, calculate the target position distances between all remaining targets and the initial cluster centers TAR1 and TAR2, respectively, to form a matrix D1 of size (|TAR|-2)×2, and perform non-dominated sorting on the (|TAR|-2)×2 2-dimensional distance vectors in the matrix D1, and select the target corresponding to the distance vector with the largest sorting value. If there is only one corresponding target, then the target entity is used as the cluster center TAR3 of the third target cluster; if there are multiple targets, execute step 3.3;

[0119] Step 3.3: Calculate the crowding distance for all 2D distance vectors with the largest ranking value, and select the target corresponding to the maximum crowding distance as the cluster center TAR3 of the third target cluster;

[0120] Step 3.4: Determine whether the number of initial cluster centers has reached the preset |TC| value (such as |TC| = 5). If so, execute step 3.5; if not, return to step 3.2 and continue until the number of initial cluster centers reaches |TC|. The difference is that when returning to step 3.2, all remaining targets and all existing cluster centers TAR1, TAR2, ..., TAR m The distance value forms a (|TAR|-m)×m matrix D m-1 , and then perform the corresponding non-dominated sorting and crowding distance calculation;

[0121] Step 3.5: Calculate the distances of all remaining targets to the centers of |TC| clusters respectively. Each cluster selects a remaining target closest to it for incorporation. To avoid conflicts, when different clusters need to incorporate the same target, the target is preferentially incorporated into the cluster that is closer to it, and the other cluster is incorporated into the target that is the next closest.

[0122] Step 3.6: Update the cluster center coordinates using the following formula and determine the number of targets included: Whether the number of target clusters reaches A1, or contains the target number Whether the number of target clusters reaches A2, if so, the clusters that have reached the target number limit will no longer be merged into the new target;

[0123] The cluster center coordinate update formula is:

[0124]

[0125] Among them, a * is the updated cluster center coordinate, N is the number of targets, a n is the nth target coordinate;

[0126] Step 3.7: Determine whether all targets have been clustered. If so, the algorithm stops. Otherwise, execute step 3.5.

[0127] Step 4: Construct a resource combination and scheduling model based on combat management entities, combat execution entities, and combat target entities, which is divided into a three-layer hierarchy of formations, groups, and platforms;

[0128] In order to ensure the command and control requirements and resource requirements for mission execution, a certain redundancy mechanism needs to be adopted, while at the same time, a reasonable amount of redundancy needs to be controlled to maximize the efficiency of capability utilization, that is, to meet the following requirements:

[0129] (1) Average redundancy cost of command and control capability at the marshaling level Minimum, ensuring efficient configuration of control capabilities:

[0130]

[0131] in, The decision variable indicating whether the i-th combat management entity executes the m-th target cluster task; The decision variable indicating whether the jth combat execution entity executes the mth target cluster task;

[0132] (2) Average Redundancy Cost of Resource Capacity at the Marshalling Layer Minimum, ensuring efficient resource capacity configuration:

[0133]

[0134] Based on this, the resource combination and scheduling model established is:

[0135]

[0136] Among them, η0 is the minimum redundancy of the combat management entity command and control capability in each target cluster; η1 is the minimum redundancy of each resource capability in each target cluster; the first constraint condition indicates that the combat management entity command and control capability in each target cluster must meet the minimum redundancy; the second constraint condition indicates that each resource capability in each target cluster must meet the minimum redundancy; the third constraint condition indicates The value is 0 or 1; the fourth constraint indicates The value is 0 or 1.

[0137] Step 5: Based on multi-objective constraint processing, discrete artificial bee colony algorithm is used to solve the resource combination and scheduling model;

[0138] Specifically, the multi-objective constraint processing discrete artificial bee colony algorithm includes the following steps:

[0139] Step 5.1: In the initialization phase, the initial honey sources are randomly generated within the feasible interval, and the initial platform-target cluster matching matrix is ​​randomly generated Among them, the platform refers to the combat management entity (manned aircraft) or the combat execution entity (unmanned aircraft), and the number of rows in the matrix is ​​the number of honey sources N fs ; The number of columns of the matrix is ​​the sum of the number of combat execution entities and combat management entities |U|+|M|, the first |U| columns represent the combat execution entity-target cluster matching relationship, and the |U|+1~|U|+|M| columns represent the combat management entity-target cluster matching relationship;

[0140] Calculate the matrix X by formula (8): P-TC The element x in row i and column j ij :

[0141]

[0142] in, and Represents the matrix X P-TC The upper and lower bounds of the j-th column variable, rand represents a random number in the interval (0,1);

[0143] Set the capacity of the external archive used to store the optimal solutions to N rf =N fs , after the initial nectar source is formed, all the solutions corresponding to the initial nectar sources are saved in the external archive ψ;

[0144] Step 5.2: In the hired bee stage, after the initial nectar source location is generated by formula (8), the hired bees look for a better nectar source location near the initial nectar source and update the nectar source location using the update formula shown in formula (9):

[0145]

[0146] Among them, x ij ' is the element of the newly generated nectar source, To randomly select the element corresponding to the nectar source from ψ, x i'j is the corresponding position element of other nectar sources, RAND is the update threshold, which is generally taken as 0.5; if x ij 'Beyond , then take x ij ' is the nearest boundary value, that is:

[0147]

[0148] The fitness value of the new nectar source is calculated. If the new nectar source dominates the old nectar source, it is replaced; if the new nectar source is dominated, the old nectar source is kept unchanged; if they are equivalent, a random choice is made to replace or keep it unchanged;

[0149] Step 5.3: Observer bee stage: After all employed bees have completed their search, they exchange the location information of the nectar source with the observer bees. The observer bees use the tournament algorithm with replacement to select the tournament operator to perform the selection operation and obtain the better nectar source. fs Select two nectar sources from the nectar sources, and the corresponding two fitness values ​​are recorded as F1 = (f1', f2') and F2 = (f1", f2"). If F1 dominates F2, it satisfies:

[0150]

[0151] Then select the first nectar source and update it according to formula (9); otherwise, select the second nectar source and update it according to formula (9); where, represents Pareto, not inferior to;

[0152] Step 5.4: Scouting bee stage. When the preset iter_limit iterations are reached, if the solution quality of the nectar source has not improved, the employed bees corresponding to the nectar source will become scout bees, abandon the original nectar source, and generate a new nectar source location according to formula (8).

[0153] After each nectar source location is mapped and feasible, it corresponds to a solution in the target space. The specific steps are:

[0154] Step 5.1.1: Multi-objective constraint processing The discrete artificial bee colony algorithm uses real number coding and specific decoding rules for decoding. Each honey source is a vector of 1 row and n columns. In order to ensure that each target cluster can be matched with a platform, the individual position value range is [1,|TC|+1);

[0155] Taking 6 combat execution entities, 3 combat management entities, and 5 target clusters as an example, the first 6 columns represent combat execution entities, and the last 3 columns represent combat management entities. The integer part and decimal part of the individual position are taken from the first 6 columns and the last 3 columns respectively. The decimal parts of the combat management entities and combat execution entities are sorted in ascending order respectively. The sorted serial number is the corresponding platform number, and the integer part is the target cluster number.

[0156] After decoding, X P-TC Mapped to a 3D matrix It can be seen that the obtained matrix W P-TC It does not necessarily meet resource and accusation requirements and requires follow-up operations;

[0157] Step 5.1.2: For W P-TCAll non-zero elements in the row vector formed by the first row and the first to |U| columns of the first 2D matrix are set to 0 from left to right. After each setting to 0, it is determined whether equation (12) holds. If not, the elements set to 0 are set to 1 again:

[0158]

[0159] in, and They are the resource capacity satisfaction of the nth resource before and after each non-zero element is set to 0;

[0160] Then, W P-TC The above operations are performed on the |TC|-1 row vectors consisting of the 2nd to |TC| rows and the 1st to |U| columns of the first 2D matrix in ;

[0161] Step 5.1.3: Determine whether the resource capacity of all resource classes meets the requirements. If so, execute step 5.1.5; if not, P-TC All non-1 elements in the row vector formed by the first row and the first to |U| columns of the first 2D matrix are set to 1 from left to right. After each setting, it is determined whether equation (13) holds. If so, the set elements are reset to 0:

[0162]

[0163] Among them, |R'| is the number of resource types whose resource capacity does not reach the demand threshold. and They are the capacity satisfaction of the n'th resource before and after each non-1 element is set to 1;

[0164] When equation (14) is satisfied, the set-to-1 operation is stopped:

[0165]

[0166] Then, W P-TC The above operations are performed on the |TC|-1 row vectors consisting of the 2nd to |TC| rows and the 1st to |U| columns of the first 2D matrix in ;

[0167] Step 5.1.4: For W P-TC Repeat step 5.1.2 for the first 2D matrix in W P-TC 2nd to Nth fs Execute steps 5.1.2 to 5.1.4 for each 2D matrix. This sequential operation is called “redundancy removal-deficiency supplementation-redundancy removal”.

[0168] Step 5.1.5: For the matrix W P-TCExecute steps 5.1.2 to 5.1.4 for the |TC| row vectors consisting of the 1st to |TC|th rows and the |U|+1th to |U|+|M|th columns of each 2D matrix in, and finally obtain the feasible solution matrix Z P-TC , which is different from step 5.1.4, when executing step 5.13, the command capability satisfaction of the target cluster is calculated by the following formula, and it is judged whether the command capability satisfaction meets the requirements;

[0169] The target cluster command capability satisfaction degree represents the degree to which all UAV command requirements matching each target cluster are met. m The calculation formula of the target cluster command capability satisfaction is:

[0170]

[0171] Step 5.1.6: Let X P-TC =Z P-TC , we can get the combat management entity-target cluster matching relationship matrix X M-TC And the combat execution entity-target cluster matching relationship matrix X U-TC ;

[0172] Furthermore, the encoding and decoding steps in step 5.1.1 are:

[0173] The discrete artificial bee colony algorithm for multi-objective constraint processing adopts real number coding. The encoding and decoding correspond to the decision variable space and the target space respectively. The decision variable space mainly realizes the initialization and update operations of the honey source; the target space is the mapping of the decision variable space in the solution space, and realizes the feasibility of the solution after mapping, mainly realizing the non-dominated sorting and crowding distance calculation of the honey source, and completing the selection and archiving operations of the solution, as shown in the attached figure. Figure 3 .

[0174] Example

[0175] In order to verify the effectiveness of the resource intelligent combination and scheduling method for hierarchical organizational relationships proposed in the present invention, the following experiment was conducted.

[0176] As shown in Tables 1 to 4, they are target resource requirements, target location coordinates, combat execution entity resource capabilities, and combat management entity command and control capability information;

[0177] Table 1 Target resource requirements

[0178]

[0179] Table 2 Target position coordinates

[0180] Head Location Target Location Target Location Target Location TAR (36.54,26.4 <![CDATA[TAR9]]> (93.86,13.5 <![CDATA[TAR1]]> (56.89,75.4 <![CDATA[TAR2]]> (77.98,34.6 TAR (85.34,20.4 <![CDATA[TAR1]]> (25.82,15.5 <![CDATA[TAR1]]> (31.52,18.6 <![CDATA[TAR2]]> (44.84,13.7 TAR (89.34,34.3 <![CDATA[TAR1]]> (32.61,30.7 <![CDATA[TAR1]]> (61.67,9.19 <![CDATA[TAR2]]> (20.35,78.1 TAR (22.18,72.4 <![CDATA[TAR1]]> (12.73,52.3 <![CDATA[TAR2]]> (8.06,35.54 <![CDATA[TAR2]]> (90.32,28.4 TAR (62.41,65.7 <![CDATA[TAR1]]> (32.82,11.2 <![CDATA[TAR2]]> (5.38,75.07 <![CDATA[TAR2]]> (28.05,50.1 TAR (54.86,70.7 <![CDATA[TAR1]]> (30.77,89.7 <![CDATA[TAR2]]> (28.13,18.4 <![CDATA[TAR3]]> (20.06,37.4 TAR (20.56,96.8 <![CDATA[TAR1]]> (75.46,93.5 <![CDATA[TAR2]]> (74.71,88.9 <![CDATA[TAR3]]> (90.97,8.17 TAR (79.57,78.9 <![CDATA[TAR1]]> (46.92,45.1 <![CDATA[TAR2]]> (25.73,46.3 <![CDATA[TAR3]]> (45.63,76.8

[0181] Table 3 Resource Capabilities of Operational Execution Entities

[0182]

[0183]

[0184] Table 4 Battle Management Entity Command and Control Capabilities

[0185]

[0186] The following experiments were performed in the MATLAB2014a environment:

[0187] Experiment 1: Verification of the greedy clustering algorithm for the initial task cluster center

[0188] The comparison algorithms selected are the greedy clustering with random initial cluster centers (GCWRICC) algorithm and the balanced k-means (BKM) algorithm, which randomly select the initial task cluster center, and a random simulation comparison is performed with the greedy clustering algorithm with random initial cluster centers (GCWOICC algorithm) proposed in the present invention.

[0189] The number of targets is set to a random integer in the interval [30,35], and the target positions are randomly distributed in the simulation area. 50 groups of experiments are randomly performed. The simulation results are shown in the attached figure. Figure 5 shown.

[0190] From the attached Figure 5 It can be seen that in 50 random experiments, the GCWOICC algorithm has the best clustering effect in 35 times, the GCWRICC algorithm has the best clustering effect in 11 times, and the BKM algorithm has the best clustering effect in 4 times. The target position coordinate information in 50 experiments, that is, the target position coordinate information in Table 2, is taken as input information, and the GCWOICC algorithm is used for target clustering. The clustering results are shown in the attached figure. Figure 6 shown.

[0191] Experiment 2: Verification of the effectiveness of discrete artificial bee colony algorithm for multi-objective constraint processing

[0192] In order to verify the effectiveness of the multi-objective constraint processing discrete artificial bee colony algorithm (MOCHDABC algorithm) proposed in this invention, the MOCHDABC algorithm was iteratively run for 90 generations, and the non-inferior feasible solutions were recorded every 30 generations. The evolutionary performance of the MOCHDABC algorithm was determined according to the non-inferior feasible solutions of each generation. Figure 7 As shown in Figure 2, the distribution of non-inferior feasible solutions of the MOCHDABC algorithm at 30, 60, and 90 generations respectively.

[0193] From the attached Figure 7It can be seen that as the number of iterations increases, the non-inferior solution set is continuously updated and the quality of the solution is continuously improved. The number of iterations is set to 200, and a typical solution is selected to generate. The corresponding platform matching scheme is shown in Table 5.

[0194] Table 5 Resource combination and scheduling plan corresponding to typical solutions after 200 iterations

[0195]

[0196] Experiment 3: Performance verification of discrete artificial bee colony algorithm for multi-objective constraint processing

[0197] In order to verify the superiority of the MOCHDABC algorithm, the MOCHDABC algorithm is compared with the NSGA-II algorithm.

[0198] As attached Figure 8 ~Attached Fig.10 As shown in the figure, there is a comparison between the MOCHDABC algorithm and the NSGA-II algorithm under the commonly used indicators of multi-objective optimization algorithms, mainly including the coverage index, uniformity index and breadth index. Among them, the larger the coverage index and breadth index, the better, and the smaller the uniformity index, the better.

[0199] From the attached Figure 8 ~Attached Fig.10 It can be seen that the MOCHDABC algorithm is better than the NSGA-II algorithm in terms of coverage index and breadth index, but worse than the NSGA-II algorithm in terms of uniformity index. The coverage index mainly measures the quality of the solution, while the uniformity index and breadth index mainly measure the distribution of the solution. Therefore, in terms of solution quality, the MOCHDABC algorithm is better than the NSGA-II algorithm, and in terms of solution distribution, the MOCHDABC algorithm and the NSGA-II algorithm each have their own advantages and disadvantages.

[0200] It can be seen that the present invention adopts a hierarchical approach to construct the problem of air combat resource combination and scheduling, and adopts an intelligent optimization method to solve the model. High-efficiency encoding and decoding technology is introduced into the method to improve the search efficiency. Complex constraint processing technology is used to effectively improve the constraint satisfaction in the model solving process, overcome the defects of existing resource combination and scheduling, and improve the efficiency of resource combination and scheduling model construction and solution.

[0201] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A resource intelligent combination and scheduling method for hierarchical organizational relationships, characterized in that: The following steps are involved: Step 1: Define the combat management entity, combat execution entity, and combat target entity; Step 2: Construct a target clustering model based on combat target entities; Step 3: Based on the target clustering model, the target clustering result is calculated using the greedy clustering algorithm that selects the initial cluster center; Step 4: Based on the target clustering results, construct a resource combination and scheduling model based on combat management entities, combat execution entities and combat target entities; Step 5: Based on multi-objective constraint processing, discrete artificial bee colony algorithm is used to solve the resource combination and scheduling model; Step 6: Combine and schedule the current resources according to the obtained resource combination and scheduling results.

2. A resource intelligent combination and scheduling method for hierarchical organizational relationships as claimed in claim 1, characterized in that: The specific steps of step 1 are: Step 1.1: Define the combat management entity set consisting of |M| combat management entities as M = {M1, M2, …, M |M| },for Each combat management entity M in the set i The basic attribute is the charge load threshold CCM i ; Step 1.2: Define the set of combat execution entities consisting of |U| combat execution entities as U = {U1, U2, …, U |U| },for Each combat execution entity U in the set j The basic attribute is the resource capability vector Among them, |R| is the number of resource types, and the superscript u represents the combat execution entity; if 1≤n≤|R|, then Represents the combat execution entity U j The number of combat resources of the nth category, if It means U j Not having the nth category combat resources; Step 1.3: Define the combat target set consisting of |TAR| combat target entities as for Each combat target entity TAR in the collection k The basic attribute is the resource demand vector and the position vector 3. A resource intelligent combination and scheduling method for hierarchical organizational relationships as claimed in claim 2, characterized in that: The specific steps of establishing the target clustering planning model in step 2 include: Step 2.1: Set the number of targets in each target cluster to a constant or and and will contain the target number or The number of target clusters is defined as A1 and A2 respectively, and the values ​​of A1 and A2 are calculated by the following formula: Among them, |TAR| represents the number of combat targets, and |TC| represents the number of target clusters; Step 2.2: Construct the target clustering model: in, d(TAR k ,TC m ) normalized value; d(TAR k ,TC m ) represents the combat target entity TAR k and the mth target cluster TC m The target position distance between is the decision variable for whether the kth target will be classified into the mth target cluster; is the number of targets in each target cluster.

4. The resource intelligent combination and scheduling method for hierarchical organizational relationships as claimed in claim 3, characterized in that: The specific steps of step 3 are as follows: Step 3.1: Calculate the target position distance between each combat target entity in the combat target entity set according to the target position distance calculation formula, generate the distance matrix D0, and find the maximum value d in D0 max , and d max The corresponding targets TAR1 and TAR2 are used as the initial cluster centers of different target clusters; The target position distance calculation formula is: Among them, d(TAR k ,TAR k' ) represents two combat target entities TAR k and TAR k' The target position distance between They represent the abscissa of the k-th target, the abscissa of the k'-th target, the ordinate of the k-th target, and the ordinate of the k'-th target respectively; Step 3.2: Determine whether |TC|=2 holds. If so, go to step 3.5; otherwise, calculate the target position distances between all remaining targets and TAR1 and TAR2 to form matrix D1, and perform non-dominated sorting on all 2D distance vectors in D1, select the target corresponding to the distance vector with the largest sorting value, and if there is only one target, use that target as the cluster center TAR3 of the third target cluster; otherwise, execute step 3.3; Step 3.3: Calculate the crowding distance for all 2D distance vectors with the largest ranking value, and select the target corresponding to the maximum crowding distance as the cluster center TAR3 of the third target cluster; Step 3.4: Determine whether the number of initial cluster centers has reached the preset |TC|. If so, execute step 3.5; if not, return to step 3.2 to continue calculating all remaining targets and all existing cluster centers TAR1, TAR2, ..., TAR m The distance values ​​form the matrix D m-1 , until the number of initial cluster centers reaches |TC|, and then the corresponding non-dominated sorting and crowding distance calculation are performed; Step 3.5: Calculate the distances of all remaining targets to the centers of |TC| clusters respectively. For each cluster, select a target closest to the cluster to be merged. When different clusters need to merge the same target, the target is preferentially merged into the cluster with the closer distance, and the other cluster is merged into the target with the second closest distance. Step 3.6: Update the cluster center coordinates using the following formula and determine the number of targets included: Whether the number of target clusters reaches A1, or contains the target number Whether the number of target clusters reaches A2, if so, the target clusters that reach the target number limit will no longer be merged into the new target; The cluster center coordinate update formula is: Among them, a * is the updated cluster center coordinate, N is the number of targets, a n is the nth target coordinate; Step 3.7: Determine whether all targets have been incorporated into the cluster. If so, terminate. Otherwise, execute step 3.

5.

5. The resource intelligent combination and scheduling method for hierarchical organizational relationships as claimed in claim 4, characterized in that: The resource combination and scheduling model constructed in step 4 is: and Among them, η0 is the minimum redundancy of the command and control capability of the combat management entity in each target cluster; η1 is the minimum redundancy of various resource capabilities in each target cluster; The decision variable indicating whether the i-th combat management entity executes the m-th target cluster task; The decision variable indicating whether the jth combat execution entity executes the mth target cluster task.

6. The resource intelligent combination and scheduling method for hierarchical organizational relationships as claimed in claim 5, characterized in that: The specific steps of step 5 include: Step 5.1: First, randomly generate the initial honey source within the feasible interval to obtain the initial platform-target cluster matching matrix Among them, the platform refers to the combat management entity or combat execution entity. The matrix X P-TC The number of rows is the number of nectar sources N fs , the number of columns is the sum of the number of combat execution entities and combat management entities |U|+|M|, the first |U| columns represent the combat execution entity-target cluster matching relationship, and the |U|+1~|U|+|M| columns represent the combat management entity-target cluster matching relationship; Next, the matrix X is calculated using the following honey source location generation formula: P-TC The element x in row i and column j ij : in, and Represents the matrix X P-TC The upper and lower bounds of the j-th column variable, rand represents a random number in the interval (0,1); Finally, set the capacity N of the external archive ψ used to store the space for saving the previous optimal solutions rf N rf =N fs , and save the solutions corresponding to all initial nectar sources in ψ; Step 5.2: The employed bees look for a better nectar source location near the initial nectar source and use the following nectar source update formula to update the nectar source location: Among them, x ij ' is the element of the newly generated nectar source, To randomly select the element corresponding to the nectar source from ψ, x i'j is the corresponding position element of other nectar sources, RAND is the update threshold, which is generally taken as 0.5; if x ij 'Beyond , then take x ij ' is the nearest boundary value, that is: The fitness value of the new nectar source is calculated. If the new nectar source dominates the old nectar source, it is replaced; if the new nectar source is dominated, the old nectar source is kept unchanged; if they are equivalent, a random choice is made to replace or keep it unchanged; Step 5.3: After all employed bees have completed their search, they exchange information about the location of the nectar source with the observation bees. The observation bees use the tournament operator to select the best nectar source. fs Select two nectar sources from the nectar sources, and the corresponding two fitness values ​​are recorded as F1 = (f1', f2') and F2 = (f1", f2"). If F1 dominates F2, it satisfies At this time, the first nectar source is selected and updated according to the nectar source update formula; otherwise, the second nectar source is selected and updated according to the nectar source update formula; Step 5.4: When the preset iter_limit iterations are reached, if the solution quality of the nectar source has not improved, the employed bees corresponding to the nectar source will become scout bees, abandon the original nectar source, and generate a new nectar source location according to the nectar source location generation formula.

7. A resource intelligent combination and scheduling method for hierarchical organizational relationships as claimed in claim 6, characterized in that: Calculate the matrix X P-TC The steps include: Step 5.1.1: Decode the nectar source. Each nectar source is a vector with 1 row and n columns, and the individual position value range is [1,|TC|+1). After decoding, X P-TC Mapped to a 3D matrix Step 5.1.2: For W P-TC All non-zero elements in the row vector formed by the first row and the first to |U| columns of the first 2D matrix are set to 0 from left to right. After each setting, check whether the following equation is true. If not, reset the elements that were set to 0 to 1: in, and They are the resource capacity satisfaction of the nth resource before and after each non-zero element is set to 0; Next, for W P-TC The above operations are performed on the |TC|-1 row vectors consisting of the 2nd to |TC| rows and the 1st to |U| columns of the first 2D matrix in respectively; Step 5.1.3: Determine whether the resource capacity of all resource classes meets the requirements. If so, execute step 5.1.5; if not, P-TC All non-1 elements in the row vector formed by the first row and the first to |U| columns of the first 2D matrix are set to 1 from left to right. After each setting, it is determined whether the following equation is true. If so, the set elements are reset to 0: Among them, |R'| is the number of resource types whose resource capacity does not reach the demand threshold. and They are the capacity satisfaction of the n'th resource before and after each non-1 element is set to 1; Next, determine whether to stop the setting operation according to the following formula. If it is satisfied, stop. Otherwise, continue to set it to 1: Finally, for W P-TC The above operations are performed on the |TC|-1 row vectors consisting of the 2nd to |TC| rows and the 1st to |U| columns of the first 2D matrix in respectively; Step 5.1.4: For W P-TC Repeat step 5.1.2 for the first 2D matrix in W P-TC 2nd to Nth fs Execute steps 5.1.2 to 5.1.4 for each 2D matrix; Step 5.1.5: For the matrix W P-TC The |TC| row vectors consisting of the 1st to |TC|th rows and the |U|+1st to |U|+|M|th columns of each 2D matrix in execute steps 5.1.2 to 5.1.

4. When executing step 5.13, the command capability satisfaction of the target cluster is calculated by the following formula, and it is judged whether the command capability satisfaction meets the requirements, and finally the feasible solution matrix Z is obtained. P-TC ; The target cluster command capability satisfaction degree represents the degree to which all UAV command requirements matching each target cluster are met. m The calculation formula of the target cluster command capability satisfaction is: Among them, CCS m The target cluster is charged with capability satisfaction; Step 5.1.6: Let X P-TC =Z P-TC , we can get the combat management entity-target cluster matching relationship matrix X M-TC And the combat execution entity-target cluster matching relationship matrix X U-TC .

Images