GB2618302A - Hypernetwork model-based organization architecture modeling method and space exploration algorithm - Google Patents

Abstract

A hypernetwork model-based combat system architecture modeling method. The modeling method comprises the following steps: decomposing missions of a combat system into a network of tasks that can be executed by an device system; building a system network according to the task network and corresponding relationships between task nodes and system nodes; building a command control network and combining the task network and system network to build a combat system architecture model; building a space exploration problem model of the combat system architecture; and converting decision maker problems in a combat system search problem model into formalized dynamic programming problems. The combat system architecture model built by the present modeling method can help decision makers make optimal choices.

Description

ORGANIZATION ARCHITECTURE MODELING METHOD BASED ON

HYPERNETVVORK MODEL AND SPACE SEARCH ALGORITHM

CROSS REFERENCE TO RELATED APPLICATION

[1] This patent application claims the benefit and priority of Chinese Patent Application No. 202011267527.2 filed with the China National Intellectual Property Administration (CNIPA) on November 13, 2020 and entitled "COMBAT SYSTEM ARCHI1ECTURE MODELING METHOD BASED ON HYPERNETWORK MODEL AND SPACE SEARCH ALGORITHM". Both of the aforementioned applications are incorporated by reference herein in their entireties

TECHNICAL FIELD

[2] The present disclosure relates to the technical field of combat systems, and in particular, to a combat system architecture modeling method based on a hypemetwork model and a space search algorithm.

BACKGROUND ART

[3] With systems becoming informationalized and intelligent, there have been increasingly diversified interrelations between systems. An organization is a system which is integrated by a limited number of component systems, and such component systems are independent and operational, and are interconnected over a period of time to realize a higher purpose. Since an organization is very complicated, how to study an organization is an urgent problem needing to be solved by researchers at present. Fortunately, an organization architecture (OA) provides an effective idea for solving this problem. OA reflects the configuration of components in the system and interaction between the components and an external environment. OA focuses on a physical entity, an information stmcture and a system function and is a core framework of the system. OA is involved throughout the overall process of design, demand demonstration, prototype development, application testing and field experiment. Therefore, the core elements of an organization can be configured optimally by studying the organization via the organization architecture and defining a reasonable formalized OA.

[4] OA is set of apparatuses connected by a command and control network, and each apparatus has a particular function to support the fulfilling of a particular task. OA is used to guide the construction of a specific organization. In view of the uncertainty of the potential capacity of an organization architecture, an architecture model and a space search problem model of an architecture solution are constructed, and their solving algorithm is established. The following problems existing in modeling and selection of an architecture need to be solved: first, the architecture has uncertain potential capability. In previous studies, the capability of an organization would be determined after the architecture was established. In fact, the uncertainty of the potential capability of the architecture is embodied in task uncertainty and the diversity of resource combinations on the one hand; and on the other hand, it is embodied in the impact of secondary factors because only main factors that affect the capability of the system are often considered in designing the architecture. Second, if continuous development of the architecture is selected, there may be a variety of strategies for an agent to obtain the potential capability of the architecture. Therefore, an agent should evaluate the expected rewards of these strategies in order to make the best choice. Third, several optimal architectures are selected from the space of multiple architecture solutions. In previous studies, only one architecture solution was often selected, and there was a lack of research on the selection of multiple architecture solutions. In view of the above problems, there is an urgent need to construct a novel architecture model and a space search problem of an organization architecture solution, and a dynamic search algorithm of an architecture solution space is proposed to solve the above problems.

SUNIMARY

1051 An objective of the present disclosure is to provide a combat system architecture modeling method based on a hypernetwork model and a space search algorithm. A combat system architecture model that permits selection of multiple architecture solutions can be constructed by using this modeling method, such that an agent can make the best choice. The algorithm is a parallel search algorithm based on decision indicators. Furthermore, the algorithm is a polynomial time algorithm having a reward obviously superior to that of other reference algorithm, which is optimal under the assumption of an independent combat system architecture solution space.

[6] To achieve the above objective, the present disclosure provides the following solutions: [7] A combat system architecture modeling method based on a hypernetwork model is provided, where based on a capability generation mechanism, with tasks, an apparatus systems and a command and control structure as key elements of a combat system architecture, the combat system architecture includes following heterogeneous networks: task networks, a system network a command and control network. The method includes following steps: 1081 Si, decomposing a mission of a combat system into the task networks executable by apparatus systems; [09] S2, defining correspondences between task nodes and system nodes, and constructing the system network according to the task networks and the correspondences between the task nodes and the system nodes, wherein the system nodes are apparatuses having particular functions and capable of independently fulfilling particular tasks, denoted as SY; and the task nodes are activity processes executable by the apparatus systems, denoted as TA; 1101 53, defining correspondences between the system nodes and command and control nodes, establishing the command and control network according to the correspondences, and combining the task networks and the system network to construct a combat system architecture model, where the command and control nodes are logical nodes for information processing, organization management, decision planning, and control feedback, denoted as C2; the command and control network is an organization network that links all the command and control nodes through Gc2 = (Vs EL-) instruction relationships, denoted as, and the correspondences between the system nodes and the command and control nodes are defined as a bipartite graph, denoted as Gsc ji,C2,L,sc) GC? ire? where represents the command and control network, while a command and control node, EC 2 a set of edge between P nodes, G5c the correspondences between the system nodes and the command and control nodes, i" a system node, and Esc a set of edges between nodes Vs'Y and nodes Vc2; 1111 54, constructing a space search problem model of the combat system architecture according to the combat system architecture model and agent n, where the agent n E N,N= {1, 2, ... /VI) and a nth agent is denoted as Agent n Agent n represents a solution space for the nth agent, with each solution space including a plurality of combat system architectures each used to describe a combat system, N represents an agent space, and IATI represents the number of agents; [12] S5, transforming an agent problem in the space search problem model of the combat system architecture into a formalized dynamic planning problem according to the space search problem model of the combat system architecture.

[13] Further, each task network may be abstracted as a directed graph, denoted as TA = KT rT.4,ET.4 where VTA represents a set of the task nodes, while ETA a set of edges between the task nodes, and GL1 a task network; the task network has starting task nodes, ending task nodes, and intermediate nodes; and the task network involves two logical relationships: a causal relationship and a parallel relationship.

[14] Further, the system network represents functional relationships between the system Gsr -ficts' nodes, denoted as - , where VsiT represents a set of the system node, while EsY a set of edges between the system nodes, and Li the system network. The correspondences between the task nodes and the system nodes are defined as a bipartite graph, denoted as GIS (vry 'Y EIS) , where FIT's represents an edge set between node VTA and node VsT, and (7TS represents the correspondences between the task nodes and the system nodes.

[15] Further, the command and control network is an organization network that links all the command and control nodes through the instruction relationships, denoted as Gc2 the correspondences between the system nodes and the command and control nodes are defined use _ (vs) vC' 2, Esc) as a bipartite graph, denoted as, where ESc represents a set of edges between nodes VSY and nodes Vc2, and T-T represents the correspondences between the system nodes and the command and control nodes.

[16] Further, a topological model of the combat system architecture is a heterogeneous network GA includes three types of nodes and five types of relationships, denoted as GA= (1711 Ty-C2, ET1,ESY Eis); a development cost of the combat system architecture is defined as c, and c E C, with C representing a cost space, a potential capability of the combat system architecture is defined as w, and the combat system architecture has a certain capability iv e TV for accomplishing a mission, with kV representing the capability of the combat system architecture to accomplish a mission; and the combat system architecture model is composed of the topological model, the development cost and the capability of the combat (GA,C, system architecture, denoted as [17] The present disclosure further provides a space search algorithm of a combat system architecture solution The the space search algorithm is based on dynamic planning of a combat system architecture space, enables action selection by determining defined indicators, and includes the following steps.

[18] 1) indicator defining step, for based on a classical Pandora's rule, defining decision indicators for each agent to perform various actions; Crnk + z'l -Tk uric)+ ) [19] [20] inferring [21] designing a simple and optimal search rule according to a state!If (D, and an c" rn,k f -zycf,k)d147";1, (;) j_ indicator set {11 neN,IneAlk,keK, where k represents a solution, each solution representing a combat system architecture, while K, a solution set for the nth agent, m an action, Alk an action set of the kth solution, a decision indicator for the nth action of the kth solution for the nth agent, a development cost corresponding to the nth action of the kth solution for the nth agent, (x) a reward x of each solution complying with a probability distribution, and x4-a reward of solution k; [22] 2) search algorithm performing step, for simplifying a calculation of an optimal architecture solution as determination of the indicators according to the search mle in the indicator defining step, wherein the search algorithm comprises a single-agent search algorithm and a multi-agent cooperation algorithm; and 1231 3) dividing step, for dividing the single-agent search algorithm into three phases: indicator sorting, indicator determination, and solution selection, and in Sorting procedure, calculating indicators for all actions in all architectures according to an expression yn = max keD" sorting the indicators, saving a sorting result to a vector, and calling Developing procedure to obtain the optimal architecture solution, where Y, represents a maximum reward for the nth agent, and Di, represents a developed architecture set.

[24] Further, in the search algorithm performing step, calculation of each indicator is independent, and the indicator is not affected by probability distributions of rewards of other combat system architectures.

[25] The present disclosure further provides a space search method of a combat system architecture solution, including: [26] decomposing a mission of a combat system into task networks executed by apparatus systems, [27] defining correspondences between task nodes and system nodes, and constructing a system network according to the task networks and the correspondences between task nodes and system nodes, wherein the system nodes are apparatuses having particular functions and capable of independently fulfilling particular tasks, denoted as SY; and the task nodes are activity processes executed by the apparatus systems, denoted as TA; [28] defining correspondences between system nodes and command and control nodes, establishing a command and control network according to the correspondences, and combining the task networks and the system network to construct a combat system architecture model, wherein the command and control nodes are logical nodes for information processing, organization management, decision planning, and control feedback, denoted as C2, the command and control network is an organization network that links all the command and control nodes GC 2 Q7C through instruction relationships, denoted as; the correspondences between the system nodes and the command and control nodes are defined as a bipartite graph, denoted as Gsc (vSY,17C2,ESC C2 represents the command and control nodes, E represents a set represents the correspondences between the system nodes r nodes, vC2 [29] constructing a space search problem model of combat system architectures according to n G N = 11, 2, * * I the combat system architecture model and agent n, where the agent, and the nth agent is denoted as Agent n; Agent n represents a solution space for the nth agent, with each solution space including a plurality of combat system architectures each used to describe a combat system; N represents an agent space, and N represents a number of agents; [30] transforming an agent problem in the space search problem model of the combat system architectures into a formalized dynamic planning problem according to the space search problem model of the combat system architectures; [31] performing dynamic planning of a combat system architecture space by using a search algorithm, where the search algorithm includes a single-agent search algorithm and a multi-agent cooperation algorithm, and the single-agent search algorithm includes three phases: indicator sorting, indicator determination, and solution selection; [32] performing the dynamic planning of the combat system architecture space by using the single-agent search algorithm, which specifically includes: -,C2 -T C2 wherein " represents the command and control network, while represents the system nodes, and Esc a set of edges of edges between V-nodes, and the command and control between nodes VsY and nodes 131 determining decision indicators for each agent to perform various actions based on a Pandora's rule: pa] =-CH Z" k In,k - n n,k \ ifn,k X) pfl 1361 wherein a set of the decision indicators is ne AT, in e iVI,c,k e K, with k representing a solution, each solution representing one combat system architecture, representing a solution set for the nth agent, in representing an action, Al representing an action set of the kth solution, IC epresenting a decision indicator for the mth action of the kth solution for the nth agent, representing a development cost corresponding to the mth action of the kth solution for the nth agent, kfc:, (x) representing that a reward x of each solution complies with a probability distribution, and representing a reward of solution k; Fri sorting the decision indicators for each agent to perform various actions; [38] adding a combat system architecture corresponding to a maximum decision indicator to a developed architecture set, and updating an undeveloped architecture set, with a union set of the developed architecture set and the undeveloped architecture set being an agent solution space; 191 determining a reward corresponding to the space search problem model of the combat system architectures in the developed architecture set; [40] determining whether the reward is greater than or equal to a maximum decision indicator corresponding to one of combat system architectures in the undeveloped architecture set to obtain a first determination result; [41] determining a combat system architecture corresponding to a maximum reward in the = yflmaxxk developed architecture set as a final solution according to an expression keD, if the first determination result is yes, wherein represents a maximum reward for the nth agent, and Di, represents the developed architecture set; and ( ) [42] adding the combat system architecture corresponding to the maximum decision indicator in the undeveloped architecture set to the developed architecture set if the first determination result is no, updating the undeveloped architecture set, and then returning to the "determining a reward corresponding to the space search problem model of the combat system architectures in the developed architecture set-.

1431 Based on specific examples provided in the present disclosure, the present disclosure has the following technical effects: 1441 With the combat system architecture modeling method based on a hypernetwork model provided in the present disclosure, a combat system architecture model that permits selection of multiple architecture solutions can be constructed based on elements of generating the capability of a system architecture and according to a formalized definition of a combat system architecture, a multi-agent dynamic planning problem and a space search problem framework of a combat system architecture solution, such that an agent can make the best choice. The space search algorithm of a combat system architecture solution is a parallel search algorithm based on decision indicators. Furthermore, the algorithm is a polynomial time algorithm having a reward obviously superior to that of other reference algorithm, which is optimal under the assumption of an independent combat system architecture solution space.

BRIEF DESCRIPTION OF THE DRAWINGS

[45] To explain the technical solutions in examples of the present disclosure or in the prior art more clearly, the accompanying drawings required in the examples will be described below in brief Apparently, the accompanying drawings in the following description show merely some examples of the present disclosure, and other drawings may be derived from these accompanying drawings by a person of ordinary skill in the art without creative efforts.

[46] FIG I is a schematic diagram of a task network in an example of the present disclosure.

Fr] FIG 2 is a schematic diagram of a system network in an example of the present

disclosure.

[48] FIG. 3 is a schematic diagram of a command and control network in an example of the

present disclosure.

1491 FIG. 4 is a schematic diagram of mapping relationships of three types of networks in an example of the present disclosure, in which FIG. 4a is a "task-system" bipartite graph, and FIG. 4b is a "system-command and control" bipartite graph.

[50] FIG. 5 is a schematic diagram of a transition relationship between possible states of architecture effects in an example of the present disclosure.

1511 FIG. 6 is a diagram of experimental data analysis in an analysis experiment conducted on a sequence searching algorithm based on decision indicators in an example of the present disclosure.

DETAILED DESCRIPTION OF THE EMBODIMENTS

1521 The technical solutions in the examples of the present disclosure will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described examples are merely a part rather than all of the examples of the present disclosure. All other examples derived from the examples of the present disclosure by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.

1531 To make the objective, features, and advantages of the present disclosure clearer and more comprehensible, the present disclosure will be further described in detail below in conjunction with the accompanying drawings and specific examples.

1541 Example

1551 A combat system architecture modeling method based on a hypemetwork model is provided, where based on a capability generation mechanism, with tasks, an apparatus system and a command and control structure as key elements of the combat system architecture, the combat system architecture is composed of the following heterogeneous networks: task networks, a system network a command and control network The method specifically includes the following steps.

1561 In step Si, a mission of a combat system is decomposed into task networks executable by an apparatus system.

1571 In step S2, a correspondence between a task node and a system node is defined, and a system network is constructed according to the task networks and the correspondences between task nodes and system nodes, where a system node is an apparatus haying a particular function and capable of independently fulfilling a particular task, denoted as Si A task node is an activity process executed by the apparatus system, denoted as TA; 1581 In step 53, correspondences between the system nodes and command and control nodes are defined, and a command and control network is established according to the correspondences, and the task networks and the system network are combined to construct a combat system architecture model. A command and control node is a logical node for information processing, organization management, decision planning, and control feedback, denoted as C2. The command and control network is an organization network that links all the command and control nodes through instruction relationships, denoted as G 2 = (r" 2) . The correspondences between the system nodes and the command and control nodes are defined as a bipartite graph, cisr _ (1"cy v-C2,A SC) denoted as, where Esc represents an edge set between nodes Vs' and nodes Vc2.

[59] In step S4, a space search problem model of a combat system architecture is constructed according to the combat system architecture model and agent n, where the agent n 17 N, N = (I, 2, ... N and the nth agent is denoted as Agent n.

[60] In step S5, an agent problem in the space search problem model of a combat system architecture is transformed into a formalized dynamic planning problem according to the space search problem model of a combat system architecture.

1611 In this example, a potential capability of a combat system architecture is the capability of a combat system developed based on the combat system architecture to accomplish a particular mission, denoted as W. [62] Mathematically, the uncertainty of the potential capability of an architecture may be represented by a probability distribution, i.e., W complies with a particular probability distribution. In addition, it is believed herein that a hypernetwork is a heterogeneous network that links a plurality of types of nodes. A combat system architecture based on a hypemetwork is composed of the following heterogeneous networks: task networks, a system network, and a command and control network. To accomplish a mission of a combat system, the mission is decomposed into a series of executable tasks, called task networks, as shown in FIG. 1, which is an example of a task network. In the combat system architecture, an apparatus system with a particular function is used to fulfill particular tasks, and therefore, a relationship between system nodes (such as unmanned aerial vehicles, tanks, and warships) is affected by task nodes, as shown in FIG. 2, which is an example of a system network and shows a logical relationship between system functions. FIG. 4 shows a correspondence between the task network and the system network, in which FIG. 4a is a "task-system" bipartite graph, and FIG. 4b is a "system-command and control" bipartite graph.

1631 A command and control node is used to process a instruction information between the superior and the subordinate. For example, a node receives task information from a superior command and control node, processes the information and then passes subtask information to a subordinate command and control node. In another aspect, a node may exchange information with a command and control node at the same level. Therefore, the command and control network refers to an organization network that links all the command and control nodes through an instruction relationship, as shown in FIG. 3, which is an example of a command and control network 164] In this example, a topological model based on a combat system architecture can be obtained qualitatively. If there are more nodes and the linkage relationship is more complicated, the development cost of the combat system architecture may be higher and the potential capability of the architecture may be greater.

1651 In this example, the capability of a combat system is measured by a reward which is a comprehensive measurement of the development cost of a combat system architecture and the revenue from the implementation of reconnaissance strategies by a cluster. The reward x of each solution complies with a probability distribution W (x), and the rewards of different solutions are independent of each other. Specifically, k" c Kn'K' =11' 2'...21K' th K representing the number of solutions in the solution space for Agent n. In addition, the solution spaces for some Agents may overlap, and in this case, KJ n K 0,1, e. Alternatively, they may do not KK n =0 overlap, and in this case, ' . The reward of each architecture is uncertain in advance and can be obtained through different actions. For the same solution k, Agent n may use action a;: e = 11, 2, ... H} solutions in an undeveloped solution space, and finally a solution is selected as a final option from all the developed solution spaces. The goal of an Agent is to select an architecture having the highest expected reward and the lowest accumulated search cost.

[66] In this example, states of a combat system architecture are the forms of the combat system architecture in the development process, including an undeveloped state and a developed state. FIG. 5 illustrates a state transition relationship of a combat system architecture. As shown, the unknown state indicates that a solution has not been developed, and its reward is unknown; and the known state indicates that a solution has been developed, and its reward value is known. On this basis, based on the formal description of the space search problem model of a combat system architecture, a set of two types of binary decision variables is specifically defined: ci. ine 11,a e A,c1, E Dd" =1 When Agent n uses act on a to develop architecture k, aA, and en d" = 0 d" n rn cli/1 k c K, , where d k k 2 the development cost is a 'k otherwise, 'Lk rn represents whether the nth Agent uses action a to develop architecture k or not. When Agent n eventually uses action m to develop architecture k as a final architecture solution, in development. Continuous explanation is performed among where s".represents whether the nth Agent is eventually chosen to develop architecture k with action m or not; otherwise s, = O represents a capacity value. The formal description of the space search problem model of a combat system architecture is as follows: CSoSAS: Max E neN knelin 4 e4,7 d" <1,n E N,k " E K ",a kn E A An_ (a) kn.4 n E N, k" E K ",a E 4 (b) A, .0 sn k, =1, n E N) At, eic a e di? " E {0,1}, ii E N,k, E C (d) EktnEN kn EK",a," EA," (e) 167] An objective function is the sum of a maximum reward of a developed architecture and a lowest accumulated search cost. In particular, constraint (a) ensures that for the solution space for any Agent, a solution has either been developed or not. Constraint (b) represents that if an Agent finally chooses a solution, the solution must have been developed. Constraint (c) represents that each Agent chooses only one solution in the end. Constraint (d) represents the value space of two types of decision variables. Constraint (e) represents the cost of each Agent performing different actions, for each solution.

168] In this example, the search problem of the combat system architecture is transformed into a formalized dynamic planning problem. By taking Agent n for example, in dynamic planning, the solution space A:n thereof is firstly divided into two mutually exclusive sets: one is an ever-increasing developed architecture set D e A: - n, while the other one is a decreasing undeveloped architecture set D G it, L), VD,c it,. The combat system architecture solution space is known in advance, which is obtained by an agent for particular missions and particular areas, and focus on finding a solution that maximizes the objective function. For each decision making, Agent n can select whether to choose and develop an unknown solution from the set Du or stop searching and choose a final solution from the set D" . If the Agent selects to continue searching, there will be Mk types of actions to develop architecture k. If the Agent ti stops searching, a solution with the highest reward is selected from the developed solution space: y" = max -wk k el) "

D

[69] At any time, the states of Agent n are statistically defined as ( "'j"). The undeveloped architecture space for each Agent is defined as 15 - -52," *, } , and the highest reward of the developed solution space for each Agent is Y Y1'* * "Y\I. Thus, the states of the system are defined as (D. Besides, a state evaluation function W(D'4 is defined as an expected discount value that can be obtained according to an optimum strategy from a time when the maximum known reward is y and the undeveloped architecture set is b. For each subset and the maximum known reward y, the state evaluation function tu (.5,y) needs to satisfy a basic iteration relation 0SoSAS iv (5, 0= max { ?Pi (A, . (52 * 0 Sum, where y) = max {-e + - )1' dTc,( h-{11,xk)df [70] Variable Vim (D"' Y) represents the state evaluation function after action m is performed in the state (D, 4 cni,k represents the cost of all Agents performing action mk to develop architecture kn, Cm.h.=1 r,eA tie A I. Further, an Agent needs to compare rewards produced by different actions, and choose and perform the action having the maximum expected reward.

By taking action in for example, if reward x' , the current highest reward will not change, and the expected state evaluation value is," xk-Y, the current highest -" , reward will be updated to x, and the expected state evaluation value is (2, n-', + -(k) x) GT4 (1),TA,ET4 11] The task network may be abstracted as a directed graph, denoted as where VTA represents a task node set, and ETA represents a set of edges between nodes Each task network has starting task nodes, ending task nodes, and intermediate nodes. The task network involves two logical relationships: a causal relationship and a parallel relationship [72] The system network represents a functional relationship between system nodes, denoted as E sy) , where VsY represents a system node set, between system nodes. The correspondences between the task defined as a bipartite graph, denoted as (-5 _ (I/ 7-51 -sy Ers and EsY represents a set of edges nodes and the system nodes are where ETs represents a set of edges between node VTA and node VsY [73] The command and control network is an organization network that links all the command and control nodes through instruction relationships, denoted as c ic -2 /17C,E 2) The correspondences between the system nodes and the command and control nodes are defined Gsc _ (r-si Tc(22 Esc) as a bipartite graph, denoted as, where Esc represents a set of edges between node VsY and node Vc2.

1741 A topological model of the combat system architecture is a heterogeneous network GA composed of three types of nodes and five types of relationships, denoted as GA-(17'1 ET1 E SY E/S ES( ) A development cost of the combat system architecture is defined as c, and. A potential capability of the combat system architecture is defined as w, and the combat system architecture has a certain capability Iv e 14/ for accomplishing a mission. The combat system architecture model is composed of the topological model, the development cost and the capability of the combat system architecture, denoted as (GA,(7,W) Fs] Example 2: The present disclosure further provides a space search algorithm of a combat system architecture solution. The search algorithm is based on dynamic planning of a combat system architecture space, enables action selection by determining a defined indicator, and includes the following steps.

1761 In step 1), an indicator is determined. Particularly, based on a classical Pandora's rule, decision indicators for each agent to perform various action are defined: xdW"ci (1-) thereby, the following expression is derived: ZmjdWrnk (x k) 17] A simple and optimal search rule is designed according to a state and an indicator set 1.? N E " G 1781 In step 2), the search algorithm is performed. Particularly, the calculation of an optimal solution is simplified as determination of an indicator according to the search rule in step 1), where the search algorithm includes a single-agent search algorithm and a multi-agent cooperation algorithm.

179] In step 3), the single-agent search algorithm is divided into three phases: indicator sorting, indicator determination, and solution selection. In Sorting procedure, indicators for all actions in all architectures are calculated according to an expression y, -max 1'k' and sorted. In keD" addition, a sorting result is saved to a vector, and Developing procedure is called to obtain an optimal architecture solution.

1801 In step 2), the calculation of each indicator is independent, and the indicator is not affected by a probability distribution of rewards of other combat system architectures.

1811 In Sequence Searching procedure, an optimal architecture solution can be calculated after K iterations at most. According to the designed rule, the current maximum sampled reward is compared with the maximum indicator in each iteration. If the maximum sampled reward is not less than the maximum indicator, the searching is stopped, and architecture m with the current maximum sampled reward is used as the selected architecture. Otherwise, the procedure is executed for continuous searching according to corresponding architecture index i and action a. If the sampled reward of architecture i is obtained, the variable is updated, where represents removing architecture i in the set.

1821 In Executing procedure, if the action taken is consultation, it is determined whether the reward of architecture i can be obtained through a relevant organization, that is, whether it is true. Furthermore, "-" means sampling, which means sampling a probability distribution.

1831 In this example, a Sequence Searching algorithm based on decision indicators is a polynomial time algorithm. The time complexity of the algorithm depends on the time complexity of a Sorting algorithm. In the algorithm, an Agent performs corresponding actions in sequence based on the magnitudes of architecture indicators, and this sequence will not change during the entire search process. Therefore, the complexity of the algorithm proposed in the present disclosure is equal to the complexity of the Sorting algorithm, and therefore, it is tenable to take the algorithm as a polynomial time algorithm.

[84] Each solution selected by the Sequence Searching algorithm based on decision indicators is conditionally optimal, and the algorithm has local optimality. In the combat system architecture problem, the selection of each solution can be mapped to a classic Pandora problem. In the Pandora problem, the reward of each project complies with a probability distribution. The actual reward of a project is unknown before the project is run, and thus needs to be obtained by sampling. In the combat system, each developed architecture can be regarded as a project k, which has a sampled reward rk. Once the sampled reward of solution k is obtained, the three projects are moved into the searched set D. The search problem model of the combat system architecture uses a search strategy based on indicators, that is, if an Agent wants to search for a new solution, an unsearched solution with the highest indicator is selected, otherwise a searched solution with the maximum sampled reward is selected. It proves that this search strategy can effectively solve the Pandora problem and obtain the best expected reward.

[85] The present disclosure further provides a space search method of a combat system architecture solution, including the following steps.

[86] In step Si, a mission of a combat system is decomposed into task networks executed by an apparatus system.

[87] In step S2, correspondences between task nodes and system nodes are defined, and a system network is constructed according to the task networks and the correspondences between the task nodes and the system nodes, where the system node is an apparatus having a particular function and capable of independently fulfilling a particular task, denoted as SY; and the task node is an activity process executed by the apparatus system, denoted as TA.

[88] In step S3, correspondences between system nodes and command and control nodes are defined, and a command and control network is established according to the correspondences, and the task networks and the system network are combined to construct a combat system architecture model, where the command and control node is a logical node for information processing, organization management, decision planning, and control feedback, denoted as C2. The command and control network is an organization network that links all the command and GC2 (T,c2, C2 E) control nodes through instruction relationships, denoted as; and the correspondences between system nodes and command and control nodes are defined as a bipartite graph, denoted as control network, while C 2 edges between T7nodes, represents the command and represents command and control nodes, EC2 represents a set of Gsr represents the correspondences between the system nodes and sc), where." 72 vC2 the command and control nodes, r-sr represents the system nodes, and ESC represents a set of the edge between node VsY and node Vc2.

[89] In step S4, a space search problem model of a combat system architecture is constructed according to the combat system architecture model and agent n, where the agent E N,N = (I, 2,...,INI} , and the nth agent is denoted as Agent n; Agent n represents a solution space for the nth agent, with each solution space including a plurality of combat system architectures each used to describe a combat system, N represents an agent space, and represents the number of agents.

[90] In step S5, an agent problem in the space search problem model of the combat system architecture is transformed into a formalized dynamic planning problem according to the space search problem model of the combat system architecture.

1911 In step S6, dynamic planning of a combat system architecture space is performed by using a search algorithm, where the search algorithm includes a single-agent search algorithm and a multi-agent cooperation algorithm, and the single-agent search algorithm includes three phases: indicator sorting, indicator determination, and solution selection.

1921 Particularly, dynamic planning of the combat system architecture space is performed by using the single-agent search algorithm, which specifically includes the following steps.

[93] First, decision indicators for each agent to perform various actions are determined based on a Pandora's rule: cc cc where a decision indicator set is {"2k IneN,meM,,ke with k representing a solution, each solution representing one combat system architecture, Ka representing a solution set for the nth agent, 111 representing an action, kik representing an action set of the kth solution, representing a decision indicator for the nth action of the kth solution for the nth agent, representing a development cost corresponding to the nth action of the kth solution for the W" representing that a reward x of each solution complies with a probability nth agent, "1,k distribution, and representing a reward of the kth solution; 1941 Then, the decision indicators for each agent to perform various actions are sorted.

[95] Further, the combat system architecture corresponding to a maximum decision indicator is added to a developed architecture set, and an undeveloped architecture set is updated, where a union set of the developed architecture set and the undeveloped architecture set is an agent solution space.

[96] Further, a reward corresponding to the space search problem model of a combat system architecture in the developed architecture set is determined.

1971 Further, it is determined whether the reward is greater than or equal to a maximum decision indicator corresponding to a combat system architecture in the undeveloped architecture set to obtain a first determination result.

[98] If the first determination result is yes, a combat system architecture corresponding to a maximum reward in the developed architecture set is determined as a final solution according to = max x" $ a formula keD, , where Y, represents a maximum reward for the nth agent, and represents the developed architecture set.

1991 And, if the first determination result is no, the combat system architecture corresponding to the maximum decision indicator in the undeveloped architecture set is added to the developed architecture set, and the undeveloped architecture set is updated. The method then proceeds to the step of determining a reward corresponding to the space search problem model of a combat system architecture in the developed architecture set.

[100] As described below, simulation experiments is conducted on the sequence searching algorithm based on decision indicator.

[101] Experiment settings are as follows. It is assumed that to accomplish a certain mission, such as border patrol, continuous reconnaissance, and electromagnetic interference, an unmanned aerial vehicle cluster needs to be sent to the target area to perform tasks. For such a new mission, how to build an unmanned aerial vehicle cluster and how to plan a task sequence for the cluster are problems that a commander needs to solve. A mission capacity-oriented technical solution is generally based on a top-down idea and finally are converted into a solution method based on a multi-agent system: firstly, the mission is decomposed into task networks; secondly, unmanned aerial vehicles with particular functions can fulfill particular tasks, and therefore, a mapping relationship between a mission domain and unmanned aerial vehicles can be constructed; furthermore, there are command and control relationships between unmanned aerial vehicles, and thus, a command and control network can be constructed; finally, a multi-agent system model is established, where each Agent has a task list, particular functions, and command and control relationships. Such a multi-Agent system is an apparatus system architecture solution. To maximize the combat effectiveness of the system, it is necessary to select an optimal architecture solution.

11021 To evaluate the performance of a Greedy Search based Dynamic Planning (GSDP) algorithm, the following statistical indicators are defined: (1) Average effectiveness Reward, which is the objective function described by the following expressions: CSoSAS: Max E fl kk ILS' dcl s.t.

c N.Jc, c K,a;: c d" > S/ c c c k =LneN E {0,1},n eN,k, eK,a eA (d) cleracNiceK"d; (e) where the effectiveness of a simulation is a difference between the reward and the accumulated cost of an architecture; (2) Number of developed architectures, which is an average number of developed architectures; (3) Running Time, which is the recorded running time of a procedure. The average effectiveness is used to evaluate the performance of the algorithm, and the number of consultations and the number of known architectures are used to analyze the search process. [103] To compare the performance of the GSDP algorithm, four reference algorithms are designed under the framework of organization architecture search (OAS) problem. (t) A Random Algorithm (RA) is designed, which allows an Agent randomly to select an action at each moment.

Specifically, an architecture k is first randomly selected from set K. If k e, that is, the architecture has been developed, the search is ended, and the reward of the architecture is obtained. If k c, that is, the architecture k is undeveloped and the reward is unknown, an action is randomly selected, the process then proceeds to continue to perform the random action until the search is ended. (2) A Traversal Development Algorithm (TDA) is designed, which allows an Agent to develop all architectures and obtain the sampled reward of each architecture. For an undeveloped architecture k, an Agent may choose the least costly action for development et; = arg minc, e Al. When the Agent has completed the development of all architectures, the architecture with the highest reward among all the developed architectures is selected as a final solution. (3) A General Development Algorithm (GEA) is similar to an algorithm that uses indicators in determination. An evaluation indicator for a local search algorithm is a difference between the highest expected value and the development cost, i.e., max {Ea {x,,)-cd I d e D, a e 2,4) When the highest reward from the developed architectures has exceeded this indicator, search is stopped and the architecture with the highest reward is selected.

[104] Experimental results: [105] Four scenarios are designed to evaluate the extendibility of an architecture space. The number of architecture solutions may be K={20, 100, 1000, 10000}, and there are 3 actions to develop each architecture. The cost of each action complies with three uniform distributions t/i(1, 3), t72(0.5, 4), and t/3(1.5, 2.5), and the reward of each architecture complies with a probability distribution Wk(wk)-U(ak, bk), where ak-U(50,60), bk-U(90,100). Thc number of Agents and the performance indicator of each algorithm when the number NoS of solutions selected (NoS is the number of the selected solutions, that is, how many solutions the multi-Agent system finally chooses) is Ito 10 are investigated. FIG. 6 is a diagram of experimental data analysis in an analysis experiment conducted on the sequence searching algorithm based on decision indicators.

[106] Scenario B]: an optimal architecture is selected by searching in a solution space containing K = 20 architectures, as shown in Table pl below, with local traversal algorithm (LEA) representing a local traversal algorithm.

U071 Table p1

NoS 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Mean 13.56 16.56 17.40 17.99 17.85 17.90 18.28 18.37 18.44 17.88 GSDP Standard 1.2(1 1.51 1.11 0 69 0 83 0.74 0.73 0 59 0 71 0.59 Deviation Mean 11.53 14.43 14.80 13 98 14.38 14.48 14.20 14.21 14.72 14.90 LEA Standard 2.24 2.43 2.87 2.81 2.67 2.77 2.68 2.32 2.52 2.64 Deviation Mean 2.625852 2.6566 2.3810 2.314156 2.1531 2.0466 1.98063 1.8857 1.6913 1.562792 0710617 7 958864 857963 2867099 FS 714161 618735 068185 23986 179939 7953895 0 4110 4787 0861 9439 310 94904 1970

RA

2.625852 2.6566 2.3810 2.314156 2.1531 2.0466 1.98063 1.8857 1.6913 1.562792 Standard 0710617 958864 857963 2867099 714161 618735 068185 23986 179939 7953895 Deviation 7 4110 4787 g 0861 9439 310 94904 1970 0 Mean -93.86 -33.47 -8.73 0.19 5.35 7.89 9.44 10.54 11.47 12.55

GEA

Standard 0.38 4.88 7.33 1.60 1.19 0.81 0.86 0.77 0.60 0.50 Deviation 11081 Scenario B2: an optimal architecture is selected by searching in a solution space containing K =104 architectures, as shown in Table p2 below. [1 P n 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 GSDP Mean 11.57 18_48 18_81 18_89 19_00 19_12 19_07 19_16 19_13 19_12 19_10 Standard 1.50 0.71 0.51 0.57 0.50 0.49 0.45 0.43 0.41 0.48 0.36 Deviation LEA Mean 12.67 14.45 14.61 15.18 14.95 15.63 14.54 15.20 14.72 15.28 14.46 Standard 2_43 2.50 2.99 2_96 2_92 2.87 2_71 2_69 2_52 2.80 2_76 Deviation RA Mean 10' -2.06 -1.79 -1.74 -1.91 -1.89 -1.83 -1.69 -1.44 -1.54 -1.54 -1.49 Standard 1_13 0.99 0.98 1_12 1_00 1.06 1_10 0_88 0_90 0.84 0_91 Deviation III' GEA Mean 103 -10.7 -5.52 -2.72 -1.82 -1.35 -1.08 -0.90 -0.77 -066 -0.59 -0.53 Standard 0.11 55_40 24_02 16_12 11_01 8.22 6_94 4_64 3_47 2.15 0_11 Deviation MO] Scenario B3: an optimal architecture is selected by searching in a solution space containing K = 06 architectures, as shown in Table p3 below.

I' 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 19.4 19.4 19.4 19.5 19.5 GSDP Mean 10_74 18_72 19_06 19_28 19_28 19.38 0 5 2 2 0 Standard 2.39 0 61 0.50 0.37 0.36 0.36 0.31 0.32 0.28 0.25 0.29 Deviation 15.0 14.7 14.8 14.9 14.5 Mean 12.82 14.20 15.19 14.98 14.76 15.01 3 1 0 9

LEA

Standard 2.56 2.86 7.78 2.64 2.92 2.78 3.04 2.68 2.64 2.71 2.80 Deviation -1.9 -1.8 -1.7 -1.8 -1.6 Mean 10 -2.04 -1.87 -1.88 -1.78 -1.96 -2.05 6 5 2 0 9 RA Standard Deviation 1.07 0.93 0.97 1.02 0.91 0.96 1.08 0.93 0.81 0.94 0.94 -0.9 -0.7 -0.6 -0.6 -0.5 Mean 105 -10.77 -5.50 -2.75 -1.83 -1.38 -1.10 GEA 2 9 9 1 5 Standard 63.3 48.8 35.4 23.4 0.02 634.21 254.39 158.82 103.92 82.19 0.03 Deviation 1 / 1 0 [112] In this example, with the combat system architecture mode ing method based on a hypernetwork model provided in the present disclosure, a combat system architecture model that permits selection of multiple architecture solutions can be constructed based on elements of generating the capability of a system architecture and according to a formalized definition of a combat system architecture, a multi-agent dynamic planning problem and a space search problem framework of a combat system architecture solution, such that an agent can make the best choice. The space search algorithm of a combat system architecture solution is a parallel search algorithm based on decision indicators. Furthermore, the algorithm is a polynomial time algorithm having a reward obviously superior to that of other reference algorithm, which is optimal under the assumption of an independent combat system architecture solution space.

[113] Several specific examples are used for illustration of the principles and implementation of the present disclosure herein. The descriptions of the above-mentioned examples are merely used for assisting in understanding the method of the present disclosure and its core ideas. In addition, those of ordinary skill in the art can make various modifications in terms of the specific implementation and the scope of application in accordance with the ideas of the present disclosure. In conclusion, the content of the present specification shall not be construed as a limitation to the present disclosure.

[114] The above embodiments are provided merely for the purpose of describing the present disclosure and are not intended to limit the scope of the present disclosure. The scope of the present disclosure is defined by the appended claims. Various equivalent replacements and modifications made without departing from the spirit and scope of the present disclosure should all fall within the scope of the present disclosure.

Claims (8)

1. WHAT IS CLAIMED IS: I. A combat system architecture modeling method based on a hypernetwork model, wherein based on a capability generation mechanism, with tasks, apparatus systems and a command and control structure as key elements of a combat system architecture, the combat system architecture comprises following heterogeneous networks: task networks, a system network, and a command and control network, the method comprising following steps: decomposing a mission of a combat system into the task networks executable by apparatus systems; defining correspondences between task nodes and system nodes, and constructing the system network according to the task networks and the correspondences between the task nodes and the system nodes, wherein the system nodes are apparatuses having particular functions and capable of independently fulfilling particular tasks, denoted as SY; and the task nodes are activity processes executable by the apparatus systems, denoted as TA; defining correspondences between the system nodes and command and control nodes, establishing the command and control network according to the correspondences, and combining the task networks and the system network to construct a combat system architecture model, wherein the command and control nodes are logical nodes for information processing, organization management, decision planning, and control feedback, denoted as C2; the command and control network is an organization network that links all the command and control nodes Gc2 _ KFC2, EC2 through instruction relationships, denoted as; the correspondences between the system nodes and the command and control nodes are defined as a bipartite graph, denoted as 6,SC SC), wherein Esc represents a set of edges between nodes VsY and nodes Vc2; constructing a space search problem model of the combat system architecture according to n the combat system architecture model and agent n, wherein the agent n c N, N = {1, 2, and a nth agent is denoted as Agent n; and transforming an agent problem in the space search problem model of the combat system architecture into a formalized dynamic planning problem according to the space search problem model of the combat system architecture.
2. 2. The combat system architecture modeling method according to claim 1, wherein each task network s capable of being abstracted as a directed graph, denoted as GT1 (1/T1, ET 1) , wherein VTA represents a set of the task nodes, and ETA represents a set of edges between the task nodes; the task network has starting task nodes, ending task nodes, and intermediate nodes; and the task network involves two logical relationships: a causal relationship and a parallel relationship.
3. 3. The combat system architecture modeling method according to claim 1, wherein the system network represents functional relationships between the system nodes, denoted as GSY (17SY Esr) , wherein Vs' represents a set of the system node, and Es' represents a set of edges between the system nodes; the correspondences between the task nodes and the system 01-8 (v vsy,ETs nodes are defined as a b partite graph, denoted as, wherein ETs represents a set of edges between nodes VTA and nodes Vs'.
4. 4 The combat system architecture modeling method according to claim 1, wherein the command and control network is an organization network that links all the command and control G 2, the correspondences C (T, nodes through the instruction relationships, denoted as between the system nodes and the command and control nodes are defined as a bipartite graph, Gsc)denoted as, wherein Esc represents a set of edges between nodes Vs and nodes VC2.The combat system architecture modeling method according to claim 1, wherein a topological model of the combat system architecture is a heterogeneous network GA comprising three types of nodes and five types of relationships, denoted as GA -(17771, vSY ve 2, ETA, ESI E T
5. S SC; a development cost of the combat system architecture is defined as c, and c c C; a potential capability of the combat system architecture is defined as w, and the combat system architecture has a certain capability 11) e W for accomplishing a mission; and the combat system architecture model comprises the topological model, the development cost and the capability of the combat system architecture, denoted as (GA,C,w)
6. 6. A space search algorithm of a combat system architecture solution, which is based on the combat system architecture modeling method according to any one of claims 1 to 5, wherein the space search algorithm is based on dynamic planning of a combat system architecture space, enables action selection by determining defined indicators, and comprises following steps: indicator defining step, for based on a classical Pandora's rule, defining decision indicators for each agent to perform various actions; )-( xdff"7.k) deriving: )6./1c, (x, ) designing a simple and optimal search rule according to a state W (A and an indicator set (*kin e N,mk EYI4,ku Ku).search algorithm performing step, for simplifying a calculation of an optimal architecture solution as determination of the indicators according to the search rule in the indicator defining step, wherein the search algorithm comprises a single-agent search algorithm and a multi-agent cooperation algorithm; and dividing step, for dividing the single-agent search algorithm into three phases: indicator sorting, indicator determination, and solution selection, and in Sorting procedure, calculating y indicators for all actions in all architectures according to an expression nkeD, sorting the indicators, saying a sorting result to a vector, and calling Developing procedure to obtain the optimal architecture solution
7. 7. The space search algorithm according to claim 6, wherein in the search algorithm performing step, calculation of each indicator is independent, and the indicator is not affected by probability distributions of rewards of other combat system architectures.
8. 8. A space search method of a combat system architecture solution, comprising decomposing a mission of a combat system into task networks executed by apparatus systems; defining correspondences between task nodes and system nodes, and constructing a system network according to the task networks and the correspondences between task nodes and system nodes, wherein the system nodes are apparatuses having particular functions and capable of independently fulfilling particular tasks, denoted as SY; and the task nodes are activity processes executed by the apparatus systems, denoted as TA; defining correspondences between system nodes and command and control nodes, establishing a command and control network according to the correspondences, and combining the task networks and the system network to construct a combat system architecture model, wherein the command and control nodes are logical nodes for information processing, organization management, decision planning, and control feedback, denoted as C2, the command and control network is an organization network that links all the command and control nodes GC 2 Q7C2 EC'2) through instruction relationships, denoted as; the correspondences between the system nodes and the command and control nodes are defined as a bipartite graph, denoted as Gsc (vs, ,T-Esc) , wherein Gc 2 represents the command and control network, while2T-C C2 represents the command and control nodes, E represents a set of edges between v nodes, SY:Grepresents the correspondences between the system nodes and the command and control nodes, V) represents the system nodes, and Esc a set of edges between nodes VsY and nodes Vc2; constructing a space search problem model of combat system architectures according to the combat system architecture model and agent n, wherein the agent " N = ti'2'.*, and the nth agent is denoted as Agent n; Agent n represents a solution space for the nth agent, with each solution space comprising a plurality of combat system architectures each used to describe a combat system; N represents an agent space, and N represents a number of agents; transforming an agent problem in the space search problem model of the combat system architectures into a formalized dynamic planning problem according to the space search problem model of the combat system architectures; performing dynamic planning of a combat system architecture space by using a search algorithm, wherein the search algorithm comprises a single-agent search algorithm and a multi-agent cooperation algorithm, and the single-agent search algorithm comprises three phases: indicator sorting, indicator determination, and solution selection; performing the dynamic planning of the combat system architecture space by using the single-agent search algorithm, which comprises: determining decision indicators for each agent to perform various actions based on a Pandora's rule: 171 xt1W' ) = cc, (x, -z?" cat (x wherein a set of the decision indicators is J,,ti "kin e AV, m e Alk,keKn ith representing a solution, each solution representing one combat system architecture, representing a solution set for the nth agent, 171 representing an action, M representing an action set of the kth solution, zmir representing a decision indicator for the mth action of the kth solution for the nth agent, eiCC,Ai representing a development cost corresponding to the mth action of the kth solution for the nth agent, Win ti-h. (r)representing that a reward x of each solution complies with a probability distribution, and xi, representing a reward of solution k; sorting the decision indicators for each agent to perform various actions; adding a combat system architecture corresponding to a maximum decision indicator to a developed architecture set, and updating an undeveloped architecture set, with a union set of the developed architecture set and the undeveloped architecture set being an agent solution space, determining a reward corresponding to the space search problem model of the combat system architectures in the developed architecture set; determining whether the reward is greater than or equal to a maximum decision indicator corresponding to one of combat system architectures in the undeveloped architecture set to obtain a first determination result; determining a combat system architecture corresponding to a maximum reward in the y = max X, kED.developed architecture set as a final solution according to an expression, if the first determination result is yes, wherein represents a maximum reward for the nth agent, and D, represents the developed architecture set, and adding the combat system architecture corresponding to the maximum decision indicator in the undeveloped architecture set to the developed architecture set if the first determination result is no, updating the undeveloped architecture set, and then returning to the "determining a reward corresponding to the space search problem model of the combat system architectures in the developed architecture set".