Multi-task allocation algorithm preferential method for complex task scene

Abstract

The invention provides a complex task scene-oriented multi-task allocation algorithm preferential method, which comprises the following steps of: on the basis of a task chain and an OODA theory, considering node attributes, heterogeneity and connection edge directivity on the basis of a task chain model, and establishing an effective task network model; according to the effective task network model, establishing a dynamic reconstruction strategy after random failure and deliberate attack failure; initializing an effective task network model, performing failure analysis on unmanned cluster nodes and connecting edges, and realizing task success rates of different task allocation algorithms by utilizing an adjacent matrix and an arrival matrix; according to the task success rate numerical simulation results of different task allocation algorithms, the advantages and disadvantages of the allocation algorithms are evaluated, and preferential analysis of the task allocation algorithms in different task scenes is realized. The task success rate of the cluster in a complex task environment is greatly improved, the task allocation and execution process is simulated and optimized based on the task success rate of the effective task network, the control decision of the cluster is effectively supported, and the task execution efficiency of the cluster is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle clusters, and in particular relates to a method for optimizing multiple task allocation algorithms for complex task scenarios. Background Art

[0002] An unmanned swarm is a new type of task execution force formed by the collaborative and organic integration of multiple unmanned equipment systems and related support systems with interconnected and interactive functions.

[0003] Unmanned swarms are composed of heterogeneous unmanned systems, including drones, unmanned vehicles, and unmanned underwater vehicles. Unmanned swarms usually perform complex tasks, which can be broken down into different subtasks, and different tasks require different systems to complete. Unmanned swarms are complex and highly coordinated, and the mission success rate is the core factor that reflects whether they can run smoothly and normally. The mission success rate of unmanned swarms is not only related to the stability and performance of a single device, but also directly reflects the mission execution efficiency of the entire swarm.

[0004] Due to the characteristics of unmanned clusters, such as node heterogeneity, directed edges, and reconfigurable task networks, traditional evaluation methods for cluster task success rates based on complex networks are difficult to reflect their characteristics. At the same time, the computational efficiency of multi-agent or task execution simulation methods is low, making it difficult to achieve dynamic evaluation of unmanned cluster task success rates for rapid task response.

[0005] In the evaluation of unmanned cluster task allocation algorithms, the task success rate indicator is a key evaluation criterion used to measure the performance of the algorithm under a specific task. This indicator mainly reflects the proportion of unmanned clusters that successfully complete assigned tasks during the execution of tasks. In simple terms, the task success rate is usually expressed as the ratio of the number of completed tasks to the total number of tasks.

[0006] Specifically, the following aspects need to be considered when evaluating the task success rate indicator: (1) Task completion definition: clarify under what circumstances the task is considered to be successfully completed. For example: accurate arrival at the target location, correct execution of specific operations, etc. (2) Time factor: consider whether the task is completed within the specified time, which is especially important for some tasks with high real-time requirements. (3) Task value: Tasks of different values ​​may affect the allocation results of the allocation algorithm, and then affect the task success rate. Therefore, factors such as task value should be considered comprehensively during the evaluation. (4) Internal and external factors: Changes in the external environment (such as weather, terrain, etc.) and the physical and electromagnetic interference of the task target and the reliability of the equipment itself will also affect the task success rate. (5) Communication and coordination: The task success rate of the unmanned cluster is closely related to the communication efficiency and coordination ability between its nodes. A good communication and coordination mechanism can improve the overall task success rate.

[0007] Through the comprehensive evaluation of the above aspects, the task success rate indicator can effectively reflect the pros and cons of the unmanned cluster task allocation algorithm and provide an important reference for algorithm improvement and optimization. However, traditional task allocation strategy evaluation indicators mainly focus on the algorithm's running time and resource consumption, and often ignore the impact of the task success rate on it. Summary of the invention

[0008] The purpose of the present invention is to solve the shortcomings of the prior art in ignoring the task success rate in the optimization of unmanned cluster task allocation, and to provide a method for optimizing multiple task allocation algorithms for complex task scenarios, so that the unmanned cluster can effectively utilize cluster resources, reasonably allocate tasks, maximize target value, and greatly improve the task success rate of the cluster in a complex task environment. Based on the task success rate of the effective task network, the task allocation and execution process is simulated and optimized, which effectively supports the control decision-making of the cluster and improves the cluster task execution efficiency.

[0009] To achieve the above purpose, the technical solution provided by the present invention is:

[0010] A method for selecting the best of multiple task allocation algorithms for complex task scenarios, including:

[0011] Step 1: Based on the task chain and OODA theory, consider the node attributes, heterogeneity and edge directionality on the basis of the task chain model to establish an effective task network model;

[0012] Step 2: Based on the effective task network model, a dynamic reconstruction strategy after random failure and intentional attack failure is established;

[0013] Step 3: Initialize the effective task network model, analyze the failure of unmanned cluster nodes and edges, and use the adjacency matrix and arrival matrix to achieve the task success rate of different task allocation algorithms;

[0014] Step 4: Through the numerical simulation results of task success rates of different task allocation algorithms, the advantages and disadvantages of the allocation algorithms are evaluated, and the optimal analysis of task allocation algorithms under different task scenarios is realized.

[0015] As a further improvement of the present invention, the expression of the effective task network model is:

[0016] eon={A,V,E}

[0017] In the formula, A represents the set of adjacency matrices in the task chain, V represents the set of nodes in the task chain, and E represents the set of edges in the task chain; specifically:

[0018] A={A TS ,A SD ,A DW ,A WT}

[0019] In the formula, A TS A represents the adjacency matrix set of the detection equipment that discovers the mission target and obtains certain information data. SD A represents the adjacency matrix set of the detection device that uploads the detected information to the decision-making device. DW A represents the set of adjacency matrices of a decision-making device receiving control instructions from another decision-making device. WT A set of adjacency matrices representing the task objectives affecting the equipment;

[0020] V=(S,D,W,T)

[0021] In the formula, S represents the detection node, D represents the decision node, W represents the influencing node, and T represents the communication link;

[0022]

[0023] s i (i=1,2,...,I)

[0024] d j (j=1,2,...,J)

[0025] w m (m=1,2,...,M)

[0026] t n (n=1,2,...,N)

[0027] In the formula, represents the edge set of the network between the nth target node and the i-th detection node, represents the set of edges in the network between the i-th detection node and the j-th decision node, represents the set of edges in the network between the j-th decision node and the m-th influencing node, represents the edge set of the network between the mth influencing node and the nth target node, s i represents the i-th detection node, I represents the number of i-th detection nodes, and d j represents the j-th decision node, J represents the number of j-th decision nodes, and w m represents the mth influencing node, M represents the number of mth influencing nodes, t n represents the nth target node, and N represents the number of nth target nodes.

[0028] As a further improvement of the present invention, the step 2 comprises:

[0029] Step (21) according to the effective task network model, considering the performance and characteristics of the unmanned cluster equipment, analyzing the performance of the unmanned cluster equipment, establishing an unmanned cluster node model, and setting node attribute indicators;

[0030] Step (22) is based on the unmanned cluster node model, considers device communication and resource constraints, establishes an unmanned cluster node edge model, and sets edge attribute indicators.

[0031] As a further improvement of the present invention, the step (21) is specifically:

[0032] (211) For the detection node S which mainly includes the detection equipment, there are significant performance differences between different detection equipment; the attribute model of the detection node S is established as:

[0033]

[0034] In the formula, s i represents the node in the detection node S, I represents the number of the i-th detection node, Represents the detection node s i The failure rate, Represents the detection node s i The repair rate, Represents the detection node s i The degree, Represents the detection node s i The x-axis coordinate of Represents the detection node s i The y-axis coordinate of cluster k Represents node s i is a member of cluster K, Represents node s i The maximum communication distance, Indicates the maximum detection distance;

[0035] (212) For decision node D, the main performance indicators include response time, throughput and accuracy. The detection of decision nodes and the results of task allocation jointly affect the cycle of the entire OODA loop and the generation and effect of task chains. The communication capability of decision nodes directly affects their ability to publish tasks to other nodes. The attribute model of decision node D is established as follows:

[0036]

[0037] Where, d j represents the j-th decision node, J represents the number of j-th decision nodes, Indicates d j The failure rate of the node, Indicates d jThe repair rate of the node, Indicates d j The degree of the node, represents the x-axis position coordinate of the j-th decision node, Indicates the y-axis position coordinate of the j-th decision node, cluster k Indicates that the jth decision node is a member of cluster K, represents the maximum communication distance of the jth decision node;

[0038] (213) Different target nodes T exhibit different characteristics. The main performance indicators include: target value, detection difficulty and task completion difficulty. The attribute model of the target node is established as follows:

[0039]

[0040] Where, t n represents the nth target node, N represents the number of nth target nodes, Indicates t n The failure rate of the node, Indicates t n The repair rate of the node, Indicates the x-axis position coordinate of the nth target node, Indicates the y-axis position coordinate of the nth target node, cluster k Indicates that the nth target node is a member of cluster K, Indicates the target value.

[0041] As a further improvement of the present invention, the step (22) is specifically:

[0042] side Considering the allocation results and detection probability of detection tasks, the unmanned cluster node connection model Set up as follows:

[0043]

[0044] In the formula, represents the nth target node t n With the i-th detection node s i The distance represents the result of the detection task allocation, represents the nth target node t n For the i-th detection node s i The detection probability of

[0045] Unmanned cluster node connection model Set up as follows:

[0046]

[0047] In the formula, represents the i-th detection node s i With the jth decision node d j The distance between represents the i-th detection node s i With the jth decision node d j The success rate of communication tasks between represents the i-th detection node s i With the jth decision node d j Packet loss rate between

[0048] Unmanned cluster node connection model Set up as follows:

[0049]

[0050] In the formula, represents the jth decision node d j and the mth influencing node w m The distance represents the jth decision node d j and the mth influencing node w m The success rate of communication tasks, represents the jth decision node d j and the mth influencing node w m Packet loss rate;

[0051] Unmanned cluster node connection model Set up as follows:

[0052]

[0053] In the formula, Indicates the mth influencing node w m and the nth target node t n The distance between Indicates the allocation result of the affected task. Indicates the mth influencing node w m For the nth target node t n The impact on the probability of task completion.

[0054] As a further improvement of the present invention, the step three comprises:

[0055] Step (31) initializes the simulation times, simulation time, node edge attributes, and connection relationships between some nodes of the effective task network model to obtain the adjacency matrix A SD ,A DW ;

[0056] Step (32) considers the task value and the impact cost factors and uses a variety of task allocation algorithms to allocate the unmanned cluster impact tasks and detection tasks to obtain the adjacency matrix A TS ,A WT ;

[0057] Step (33) If the node and the edge exist, that is, no failure occurs, then α() = 1; if the node and the edge do not exist, that is, failure occurs, then α() = 0, where α() represents the existence of the node and the edge; the node characteristic function is as follows:

[0058]

[0059] In the formula, s i represents the i-th detection node, d j represents the jth decision node, w m represents the mth influencing node;

[0060] The edge existence indicator function is as follows:

[0061]

[0062] In the formula, represents the set of edges in the network between the i-th detection node and the j-th decision node, represents the edge set of the network between the nth target node and the i-th detection node, represents the set of edges in the network between the j-th decision node and the m-th influencing node, Represents the edge set of the network between the mth influencing node and the nth target node;

[0063] Step (34) Use the adjacency matrix and the arrival matrix to calculate the number of valid task chains N eol (t), and integrate the task success criterion N eol (t)≥τ i Calculate the task success rate; where: τ i Represents the minimum number of valid task chains required for task i to succeed.

[0064] As a further improvement of the present invention, step (33) further includes a random failure step and a deliberate attack step, specifically:

[0065] The random failure step includes: determining the number of node and edge failures of the unmanned cluster through Monte Carlo simulation according to the failure rate and distribution of the nodes of the unmanned cluster node model and the edge failure rate of the unmanned cluster node edge model, randomly removing the failed nodes and edges of the unmanned cluster node model, and adding the failed nodes and edges to the failed node list and the failed edge list respectively; wherein the detection node s i , detection node dj and the detection node w m The failure rate is determined by the parameter Failure rate Failure rate It can be described by the Poisson distribution of

[0066] The deliberate attack steps include: determining the attack mode, arranging the nodes in the unmanned cluster in descending order according to the node degree, then removing the corresponding number of failed nodes and their edges, and adding the failed nodes to the failed node list; taking CNA as a special case to study the node failure and dynamic reconstruction strategy; setting the node degree attack strategy to attack only the first 50% of the nodes, setting Represents node s i ,d j ,w m degree;

[0067] Specifically:

[0068] When the list of failed nodes is not empty, the process of adjusting or reconstructing the connections and relationships between different clusters within the cluster is adopted. Inter-cluster reconstruction I is a rule-based reconstruction strategy;

[0069] When the list of failed nodes is not empty, intra-cluster reconstruction II is used to modify the connections and relationships between nodes in the same cluster, optimize the internal structure and coordination of the unmanned cluster, and make the communication, resource allocation, and task allocation between nodes more efficient;

[0070] When the list of failed nodes is not empty, repair and generate reconstruction III repairs or adds new nodes or edges within the cluster, which involves deploying additional clusters or activating dormant nodes to improve the overall ability and capacity of the cluster;

[0071] When the task changes or the cluster structure changes on a large scale, task redistribution IV is triggered, which involves the redistribution of tasks between nodes in the cluster to optimize task allocation, maximize cluster efficiency, and ensure the completion of priority task goals.

[0072] As a further improvement of the present invention, step (34) is specifically:

[0073] Step (341) obtains the adjacency matrix A between nodes of different types through the effective task network model = {A TS ,A SD ,A DW ,A WT};

[0074] Step (342) considers the node and edge attributes and existence, and verifies the existence of the adjacency matrix elements:

[0075]

[0076] If there is a connection between nodes, otherwise,

[0077] The existence probability of an adjacency matrix element is:

[0078]

[0079] Among them, if the failure and generation of nodes obey exponential distribution, the probability of the existence of unmanned cluster nodes is:

[0080]

[0081] The existence probability of each edge of the unmanned cluster is expressed as:

[0082]

[0083] In the formula, represents the allocation result of the detection task, represents the i-th detection node s i With the jth decision node d j Effective communication between Indicates the mth influencing node w m and the nth target node t n effective communication between; wherein:

[0084]

[0085]

[0086] After all the adjacency matrix elements have been verified, go to step (343), otherwise go to step (342) to continue verification;

[0087] Step (343) calculates the number of valid task chains:

[0088] Calculate the arrival matrix A TT , A TT =A TS ×A SD ×A DW ×A WT ;

[0089] Calculate the arrival matrix A TT The number of effective task chains N eol , N eol =Trace(A TT ), A TS Indicates that the dimension is N×I, A SD Indicates the dimension is I×J, A DW Indicates the dimension is J×M, A WT Indicates that the dimension is M×N;

[0090] Task success rate calculation i=1, determine whether the task chain is a valid task chain

[0091] If N eol ≥τ i Indicates that the task is completed effectively, the task successfully forms a valid task chain, and the number of successful tasks Num task-success =Num task-success +1,i=i+1;

[0092] If N eol <τ i If i=n t Then go to the next step, otherwise go to this step; n t is the number of T nodes;

[0093] Calculate the task success rate based on the number of simulations

[0094] As a further improvement of the present invention, the dynamic reconstruction strategy of step 2 includes: rule-based dynamic reconstruction and task-based dynamic reconstruction;

[0095] The unmanned swarm task allocation strategy is evaluated by using the task success rate indicator that integrates the actual task completion process and completion results; the task success rate indicator is closely combined with the effective task chain to form effective task completion as the task success standard;

[0096] The step three uses the unmanned cluster adjacency matrix and the arrival matrix to calculate the number of valid task chains.

[0097] The advantages of the present invention are:

[0098] 1. The present invention selects the best task allocation algorithms under different task scenarios, so that the unmanned cluster can effectively utilize cluster resources, reasonably allocate tasks, maximize the target value, and greatly improve the task success rate of the cluster in complex task environments.

[0099] 2. The method of the present invention is universal and can be extended to the optimization of various types of unmanned cluster task allocation strategies. It only needs to clarify the equipment and environmental parameters to complete the modeling, simulation, analysis and optimization of large-scale unmanned cluster task processes. It can effectively support the control decision-making of the cluster and improve the task execution efficiency of the cluster. It provides solid and reliable decision-making support for mission actions and provides theoretical and methodological support for the rational scheduling of equipment and resources.

[0100] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0102] Figure 1 : A flowchart of a method for selecting the best of multiple task allocation algorithms for complex task scenarios provided by the present invention;

[0103] Figure 2 : Flowchart of the task allocation algorithm provided by the present invention;

[0104] Figure 3 : Node distribution diagram of the unmanned cluster provided by the present invention;

[0105] Figure 4 : Model diagram of the effective task network model of the unmanned cluster provided by the present invention;

[0106] Figure 5 : The impact diagram of random failure and intentional attack on unmanned cluster provided by the present invention;

[0107] Figure 6 : Schematic diagram of the rule-based unmanned cluster reconstruction strategy provided by the present invention;

[0108] Figure 7 : Schematic diagram of the task-based unmanned cluster reconstruction strategy provided by the present invention;

[0109] Figure 8 :The impact of different reconstruction strategies on the success rate of unmanned swarm tasks;

[0110] Fig. 9 : Task success rate analysis chart of three unmanned cluster task allocation algorithms. DETAILED DESCRIPTION

[0111] Embodiments of the present invention are described in detail below. The embodiments are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0112] In order to solve the problem of ignoring the task success rate when selecting the best unmanned cluster task allocation, and the problem that there is a large gap between the selection of the best index and the actual process results due to the large number of task allocation algorithms in the task scenario, please refer to Figure 1The embodiment of the present invention provides a method for selecting the best of multiple task allocation algorithms for complex task scenarios. Based on the task success rate of the effective task network, the task allocation and execution process is simulated and optimized. The unmanned cluster of the embodiment of the present invention is composed of 240 nodes, including 40 detection nodes, 20 decision nodes, 120 influence nodes, and 60 target nodes. The unmanned cluster task scene is a square area of ​​200×200. The distribution of nodes in the task scene is as follows: Figure 2 As shown, the S node is represented by blue, the D node is represented by red, the W node is represented by yellow, and the T node is represented by purple, and the location distribution is determined by its coordinates. The specific implementation method of the embodiment of the present invention is as follows Figure 3 The method of the embodiment comprises the following steps:

[0113] Step 1: Based on the task chain and OODA theory, an effective task network model is established by considering node attributes, heterogeneity and edge directionality on the basis of the task chain model. The embodiment of the present invention characterizes the heterogeneity, directionality, diversity and emergence of unmanned clusters through an effective task network model.

[0114] The expression of the effective task network model of the embodiment of the present invention is:

[0115] eon={A,V,E}

[0116] In the formula, A represents the set of adjacency matrices in the task chain, V represents the set of nodes in the task chain, and E represents the set of edges in the task chain; specifically:

[0117] A={A TS ,A SD ,A DW ,A WT}

[0118] In the formula, A TS A represents the adjacency matrix set of the detection equipment that discovers the mission target and obtains certain information data. SD A represents the adjacency matrix set of the detection device that uploads the detected information to the decision-making device. DW A represents the set of adjacency matrices of a decision-making device receiving control instructions from another decision-making device. WT A set of adjacency matrices representing the task objectives affecting the equipment;

[0119] V=(S,D,W,T)

[0120] In the formula, S represents the detection node, D represents the decision node, W represents the influencing node, and T represents the communication link;

[0121]

[0122] s i(i=1,2,...,I)

[0123] d j (j=1,2,...,J)

[0124] w m (m=1,2,...,M)

[0125] t n (n=1,2,...,N)

[0126] In the formula, represents the edge set of the network between the nth target node and the i-th detection node, represents the set of edges in the network between the i-th detection node and the j-th decision node, represents the set of edges in the network between the j-th decision node and the m-th influencing node, represents the edge set of the network between the mth influencing node and the nth target node, s i represents the i-th detection node, I represents the number of i-th detection nodes, and d j represents the j-th decision node, J represents the number of j-th decision nodes, and w m represents the mth influencing node, M represents the number of mth influencing nodes, t n represents the nth target node, and N represents the number of nth target nodes.

[0127] In an ideal state, all nodes in the unmanned cluster can be connected to each other to form a fully connected network. However, in actual mission scenarios, different types of nodes that make up the system are limited by their own attributes and operating resources. According to the function, the nodes are divided into four categories: S, D, W, and T. Let s i (i=1,2,...,I) represents the i-th detection node, d j (i=1,2,...,J) represents the jth decision node, w m (m=1,2,...,M) represents the mth influencing node, t n (n=1,2,...,N) represents the nth target node, I,J,M,N represent the number of nodes of different types; the effective task network model is as follows Figure 4 As shown. Where A={A TS ,A SD ,A DW ,A WT} represents the set of adjacency matrices in the network, V represents the set of nodes in the network, where nodes are divided into four categories V = (S, D, W, T), and E represents the set of edges in the network, where edges are divided into five categories Due to resource sharing and information integration, Figure 4In our unmanned cluster, real-time information data exchange can be achieved between nodes of the same type, so the communication and interactive edges between nodes of the same type are not considered.

[0128] Step 2: According to the effective task network model, a dynamic reconstruction strategy after random failure and intentional attack failure is established. The embodiment of the present invention considers the performance attribute differences of unmanned cluster equipment and the characteristics of communication task allocation to establish an unmanned cluster node model and an edge model; it integrates the internal and external influencing factors such as the environment in which the unmanned cluster task is located (weather, temperature, humidity, etc.), the impact of the task target, and the success rate of the equipment's own task, and divides it into two categories: random failure and intentional attack, to achieve unmanned cluster node and edge failure analysis.

[0129] The dynamic reconstruction strategy of the embodiment of the present invention includes: rule-based dynamic reconstruction and task-based dynamic reconstruction. The embodiment of the present invention establishes an unmanned cluster node and edge model for an effective task network model, taking into account the unmanned cluster device attributes and resource constraints, and gives a task simulation time.

[0130] Step 2 of the embodiment of the present invention includes:

[0131] Step (21) is to analyze the performance of the unmanned cluster equipment based on the effective task network model, consider the performance and characteristics of the unmanned cluster equipment, establish an unmanned cluster node model, and set node attribute indicators.

[0132] For the detection nodes mainly including detection devices in the embodiments of the present invention, there may be significant performance differences between different detection devices. Represents node s i The maximum detection distance, if the distance to the target exceeds Then the detection node cannot detect the target. The detection probability of the detection node refers to the probability that the detection node successfully detects the target within its perception range. The attribute model of the detection node S is established as follows:

[0133] (211) For the detection node S which mainly includes the detection equipment, there are significant performance differences between different detection equipment; the attribute model of the detection node S is established as:

[0134]

[0135] In the formula, s i represents the node in the detection node S, I represents the number of the i-th detection node, Represents the detection node s i The failure rate, Represents the detection node s i The repair rate, Represents the detection node s i The degree, Represents the detection node s i The x-axis coordinate of Represents the detection node s i The y-axis coordinate of cluster k Represents node s i is a member of cluster K, Represents node s i The maximum communication distance, Indicates the maximum detection distance.

[0136] (212) For decision node D, the main performance indicators include response time, throughput and accuracy. The detection of decision nodes and the results of task allocation jointly affect the cycle of the entire OODA loop and the generation and effect of task chains. The communication capability of decision nodes directly affects their ability to publish tasks to other nodes. The attribute model of decision node D is established as follows:

[0137]

[0138] Where, d j represents the j-th decision node, J represents the number of j-th decision nodes, Indicates d j The failure rate of the node, Indicates d j The repair rate of the node, Indicates d j The degree of the node, represents the x-axis position coordinate of the j-th decision node, Indicates the y-axis position coordinate of the j-th decision node, cluster k Indicates that the jth decision node is a member of cluster K, represents the maximum communication distance of the jth decision node;

[0139] (213) Different target nodes T exhibit different characteristics. The main performance indicators include: target value, detection difficulty and task completion difficulty. The attribute model of the target node is established as follows:

[0140]

[0141] Where, t n represents the nth target node, N represents the number of nth target nodes, Indicates t n The failure rate of the node, Indicates t n The repair rate of the node, Indicates the x-axis position coordinate of the nth target node, Indicates the y-axis position coordinate of the nth target node, clusterk Indicates that the nth target node is a member of cluster K, Indicates the target value.

[0142] Step (22) is based on the unmanned cluster node model, considers device communication and resource constraints, establishes an unmanned cluster node edge model, and sets edge attribute indicators.

[0143] Step (22) of the embodiment of the present invention is specifically as follows:

[0144] side Considering the allocation results and detection probability of detection tasks, the unmanned cluster node connection model Set up as follows:

[0145]

[0146] In the formula, represents the nth target node t n With the i-th detection node s i The distance represents the result of the detection task allocation, represents the nth target node t n For the i-th detection node s i The detection probability of

[0147] Unmanned cluster node connection model Set up as follows:

[0148]

[0149] In the formula, represents the i-th detection node s i With the jth decision node d j The distance between represents the i-th detection node s i With the jth decision node d j The success rate of communication tasks between represents the i-th detection node s i With the jth decision node d j Packet loss rate between

[0150] Unmanned cluster node connection model Set up as follows:

[0151]

[0152] In the formula, represents the jth decision node d j and the mth influencing node w m The distance represents the jth decision node dj and the mth influencing node w m The success rate of communication tasks, represents the jth decision node d j and the mth influencing node w m Packet loss rate;

[0153] Unmanned cluster node connection model Set up as follows:

[0154]

[0155] In the formula, Indicates the mth influencing node w m and the nth target node t n The distance between Indicates the allocation result of the affected task. Indicates the mth influencing node w m For the nth target node t n The impact on the probability of task completion.

[0156] The embodiment of the present invention implements clustering and division according to the tasks and regional distribution for the same unmanned cluster, and collaboratively completes different subtasks.

[0157] For the influencing node (W), different types of W have different capabilities, and the main performance indicators include influence accuracy, maximum task completion distance and completion probability. The attribute model of W is established as follows:

[0158]

[0159] in, Indicates w m The two-dimensional position coordinates of cluster k Indicates w m is a member of cluster K, and Indicates w m The maximum communication distance and maximum task completion distance.

[0160] Step 3: Initialize the effective task network model, analyze the failure of unmanned cluster nodes and edges, and use the adjacency matrix and arrival matrix to achieve the task success rate of different task allocation algorithms. Step 3 of the embodiment of the present invention uses the unmanned cluster adjacency matrix and arrival matrix to calculate the number of effective task chains. The embodiment of the present invention expands the traditional task network model and considers the comprehensive impact of internal and external factors such as random failures, intentional attacks, task allocation results, key performance indicators, and task resource limitations during the execution of unmanned cluster tasks.

[0161] Step three of the embodiment of the present invention includes:

[0162] Step (31) Initialize the simulation times, simulation time, node edge attributes, and connection relationships between some nodes of the effective task network model to obtain the adjacency matrix A SD ,A DW ;

[0163] Step (32) considers the task value and the impact cost factors and uses a variety of task allocation algorithms to allocate the unmanned cluster impact tasks and detection tasks to obtain the adjacency matrix A TS ,A WT ;

[0164] Step (33) If the node and the edge exist, that is, no failure occurs, then α() = 1; if the node and the edge do not exist, that is, failure occurs, then α() = 0, where α() represents the existence of the node and the edge; the node characteristic function is as follows:

[0165]

[0166] In the formula, s i represents the i-th detection node, d j represents the jth decision node, w m represents the mth influencing node;

[0167] The edge existence indicator function is as follows:

[0168]

[0169] In the formula, represents the set of edges in the network between the i-th detection node and the j-th decision node, represents the edge set of the network between the nth target node and the i-th detection node, represents the set of edges in the network between the j-th decision node and the m-th influencing node, Represents the set of edges in the network between the mth influencing node and the nth target node.

[0170] Among them, step (33) of the embodiment of the present invention also includes a random failure step and a deliberate attack step, specifically:

[0171] The random failure step includes: according to the failure rate and distribution of the nodes of the unmanned cluster node model and the edge failure rate of the unmanned cluster node edge model, the number of node and edge failures of the unmanned cluster is determined by Monte Carlo simulation, the failed nodes and edges of the unmanned cluster node model are randomly removed, and the failed nodes and edges are added to the failed node list and the failed edge list respectively; wherein, the detection node s i , detection node d j and the detection node w m The failure rate is determined by the parameter Failure rate Failure rate It can be described by a Poisson distribution.

[0172] The deliberate attack steps include: determining the attack mode, arranging the nodes in the unmanned cluster in descending order according to the node degree, then removing the corresponding number of failed nodes and their edges, and adding the failed nodes to the failed node list; taking CNA as a special case to study the node failure and dynamic reconstruction strategy; setting the node degree attack strategy to attack only the first 50% of the nodes, setting Represents the detection node s i The degree, Represents the detection node d j The degree, Represents the detection node w m The degree.

[0173] Specifically:

[0174] When the list of failed nodes is not empty, the process of adjusting or reconstructing the connections and relationships between different clusters within the cluster is adopted. Inter-cluster reconstruction I is a rule-based reconstruction strategy;

[0175] When the list of failed nodes is not empty, intra-cluster reconstruction II is used to modify the connections and relationships between nodes in the same cluster, optimize the internal structure and coordination of the unmanned cluster, and make the communication, resource allocation, and task allocation between nodes more efficient;

[0176] When the list of failed nodes is not empty, repair and generate reconstruction III repairs or adds new nodes or edges within the cluster, which involves deploying additional clusters or activating dormant nodes to improve the overall ability and capacity of the cluster;

[0177] When the task changes or the cluster structure changes on a large scale, task redistribution IV is triggered, which involves the redistribution of tasks between nodes in the cluster to optimize task allocation, maximize cluster efficiency, and ensure the completion of priority task goals.

[0178] When the task changes or the cluster structure changes on a large scale, task redistribution IV (task-based reconstruction strategy) is triggered: it involves the redistribution of tasks between nodes in the cluster. This can happen for a variety of reasons, such as changes in task requirements, node failures, or the need to balance workloads. The purpose of task redistribution is to optimize task allocation, maximize cluster efficiency, and ensure that priority task goals are achieved. Half of the simulation time is selected as the trigger condition for task redistribution. The impact of different dynamic reconstruction strategies on unmanned clusters can be found in Figure 8Among them, the success rate of unmanned cluster tasks under different strategies was compared with that under no strategy. All reconstruction strategies effectively improved the success rate of unmanned cluster tasks, and as the number of reconstruction strategies adopted increased, the success rate of unmanned cluster tasks was significantly improved. It is further inferred that when the four reconstruction strategies are adopted at the same time, the success rate of unmanned cluster tasks is improved the most.

[0179] Step (34) Use the adjacency matrix and the arrival matrix to calculate the number of valid task chains N eol (t), and integrate the task success criterion N eol (t)≥τ i Calculate the task success rate; where: τ i Represents the minimum number of valid task chains required for task i to succeed.

[0180] Step (34) of the embodiment of the present invention is specifically as follows:

[0181] Step (341) obtains the adjacency matrix A between nodes of different types through the effective task network model = {A TS ,A SD ,A DW ,A WT};

[0182] Step (342) considers the node and edge attributes and existence, and verifies the existence of the adjacency matrix elements:

[0183]

[0184] If there is a connection between nodes, otherwise,

[0185] The existence probability of an adjacency matrix element is:

[0186]

[0187] The verification element existence of the embodiment of the present invention includes the following sub-steps:

[0188] If the failure and generation of nodes follow an exponential distribution, the probability of an unmanned cluster node existing is:

[0189]

[0190] The existence probability of each edge of the unmanned cluster is expressed as:

[0191]

[0192] In the formula, represents the allocation result of the detection task, represents the i-th detection node s iWith the jth decision node d j Effective communication between Indicates the mth influencing node w m and the nth target node t n effective communication between; wherein:

[0193]

[0194] After all the adjacency matrix elements have been verified, go to step (343), otherwise go to step (342) to continue verification;

[0195] Step (343) calculates the number of valid task chains:

[0196] Calculate the arrival matrix A TT , A TT =A TS ×A SD ×A DW ×A WT ;

[0197] Calculate the arrival matrix A TT The number of effective task chains N eol , N eol =Trace(A TT ), A TS Indicates that the dimension is N×I, A SD Indicates the dimension is I×J, A DW Indicates the dimension is J×M, A WT Indicates that the dimension is M×N;

[0198] Task success rate calculation i=1, determine whether the task chain is a valid task chain

[0199] If N eol ≥τ i Indicates that the task is completed effectively, the task successfully forms a valid task chain, and the number of successful tasks Num task-success =Num task-success +1,i=i+1;

[0200] If N eol <τ i If i=n t Then go to the next step, otherwise go to this step; n t is the number of T nodes;

[0201] Calculate the task success rate based on the number of simulations

[0202] Step 4: Through the numerical simulation results of the task success rate of different task allocation algorithms, the advantages and disadvantages of the allocation algorithm are evaluated, and the optimal analysis of the task allocation algorithm under different task scenarios is realized. For the specific analysis of the various task allocation algorithms of unmanned swarms in the embodiment of the present invention, please refer to Fig. 9 Among them, the Hungarian algorithm (HA), the contract network algorithm (CNA) and the ant colony algorithm (ACA) were selected. According to the task success rate, the ant colony algorithm is more stable, and the Hungarian algorithm and the contract network algorithm have higher task success rates in the early stages of the task. Therefore, the ant colony algorithm is suitable for long-term tasks, while the Hungarian algorithm and the contract network algorithm are suitable for short-term tasks.

[0203] The embodiment of the present invention adopts a task success rate indicator that integrates the actual task completion process and completion results to evaluate the unmanned cluster task allocation strategy; the task success rate indicator is closely combined with the effective task chain to form effective task completion as the task success standard.

[0204] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be included in the protection scope of the present invention.

Claims

1. A method for selecting the best of multiple task allocation algorithms for complex task scenarios, characterized in that: include: Step 1: Based on the task chain and OODA theory, consider the node attributes, heterogeneity and edge directionality on the basis of the task chain model to establish an effective task network model; Step 2: Based on the effective task network model, a dynamic reconstruction strategy after random failure and intentional attack failure is established; Step 3: Initialize the effective task network model, analyze the failure of unmanned cluster nodes and edges, and use the adjacency matrix and arrival matrix to achieve the task success rate of different task allocation algorithms; Step 4: Through the numerical simulation results of task success rates of different task allocation algorithms, the advantages and disadvantages of the allocation algorithms are evaluated, and the optimal analysis of task allocation algorithms under different task scenarios is realized.

2. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 1, characterized in that: The expression of the effective task network model is: eon={A,V,E} In the formula, A represents the set of adjacency matrices in the task chain, V represents the set of nodes in the task chain, and E represents the set of edges in the task chain; specifically: A={A TS ,A SD ,A DW ,A WT } In the formula, A TS A represents the adjacency matrix set of the detection equipment that discovers the mission target and obtains certain information data. SD A represents the adjacency matrix set of the detection device that uploads the detected information to the decision-making device. DW A represents the set of adjacency matrices of a decision-making device receiving control instructions from another decision-making device. WT A set of adjacency matrices representing the task objectives affecting the equipment; V=(S,D,W,T) In the formula, S represents the detection node, D represents the decision node, W represents the influencing node, and T represents the communication link; s i (i=1,2,...,I) d j (j=1,2,...,J) w m (m=1,2,...,M) t n (n=1,2,...,N) In the formula, represents the edge set of the network between the nth target node and the i-th detection node, represents the set of edges in the network between the i-th detection node and the j-th decision node, represents the set of edges in the network between the j-th decision node and the m-th influencing node, represents the edge set of the network between the mth influencing node and the nth target node, s i represents the i-th detection node, I represents the number of i-th detection nodes, and d j represents the j-th decision node, J represents the number of j-th decision nodes, and w m represents the mth influencing node, M represents the number of mth influencing nodes, t n represents the nth target node, and N represents the number of nth target nodes.

3. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 1, characterized in that: The second step comprises: Step (21) according to the effective task network model, considering the performance and characteristics of the unmanned cluster equipment, analyzing the performance of the unmanned cluster equipment, establishing an unmanned cluster node model, and setting node attribute indicators; Step (22) is based on the unmanned cluster node model, considers device communication and resource constraints, establishes an unmanned cluster node edge model, and sets edge attribute indicators.

4. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 3, characterized in that: The step (21) is specifically: (211) For the detection node S which mainly includes the detection equipment, there are significant performance differences between different detection equipment; the attribute model of the detection node S is established as: In the formula, s i represents the node in the detection node S, I represents the number of the i-th detection node, Represents the detection node s i The failure rate, Represents the detection node s i The repair rate, Represents the detection node s i The degree, Represents the detection node s i The x-axis coordinate of Represents the detection node s i The y-axis coordinate of cluster k Represents node s i is a member of cluster K, Represents node s i The maximum communication distance, Indicates the maximum detection distance; (212) For decision node D, the main performance indicators include response time, throughput and accuracy. The detection of decision nodes and the results of task allocation jointly affect the cycle of the entire OODA loop and the generation and effect of task chains. The communication capability of decision nodes directly affects their ability to publish tasks to other nodes. The attribute model of decision node D is established as follows: Where, d j represents the j-th decision node, J represents the number of j-th decision nodes, Indicates d j The failure rate of the node, Indicates d j The repair rate of the node, Indicates d j The degree of the node, represents the x-axis position coordinate of the j-th decision node, Indicates the y-axis position coordinate of the j-th decision node, cluster k Indicates that the jth decision node is a member of cluster K, represents the maximum communication distance of the jth decision node; (213) Different target nodes T exhibit different characteristics. The main performance indicators include: target value, detection difficulty, and task completion difficulty; The attribute model of the target node is established as: In the formula, t n represents the nth target node, N represents the number of nth target nodes, Indicates t n The failure rate of the node, Indicates t n The repair rate of the node, Indicates the x-axis position coordinate of the nth target node, Indicates the y-axis position coordinate of the nth target node, cluster k Indicates that the nth target node is a member of cluster K, Indicates the target value.

5. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 3, characterized in that: The step (22) is specifically: side Considering the detection task allocation results and detection probability, the unmanned cluster node connection model Set up as follows: In the formula, Represents the nth target node t n With the i-th detection node s i The distance represents the result of the detection task allocation, Represents the nth target node t n For the i-th detection node s i The detection probability of Unmanned cluster node connection model Set up as follows: In the formula, represents the i-th detection node s i With the jth decision node d j The distance between represents the i-th detection node s i With the jth decision node d j The success rate of communication tasks between represents the i-th detection node s i With the jth decision node d j Packet loss rate between Unmanned cluster node connection model Set up as follows: In the formula, represents the jth decision node d j and the mth influencing node w m The distance represents the jth decision node d j and the mth influencing node w m The success rate of communication tasks, represents the jth decision node d j and the mth influencing node w m Packet loss rate; Unmanned cluster node connection model Set up as follows: In the formula, Indicates the mth influencing node w m and the nth target node t n The distance between Indicates the allocation result that affects the task. Indicates the mth influencing node w m For the nth target node t n The impact on the probability of task completion.

6. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 2, characterized in that: The step three comprises: Step (31) initializes the simulation times, simulation time, node edge attributes, and connection relationships between some nodes of the effective task network model to obtain the adjacency matrix A SD ,A DW ; Step (32) considers the task value and the influencing cost factors and uses a variety of task allocation algorithms to allocate the unmanned cluster impact tasks and detection tasks to obtain the adjacency matrix A TS ,A WT ; Step (33) If the node and the edge exist, that is, no failure occurs, then α() = 1; if the node and the edge do not exist, that is, failure occurs, then α(·) = 0, where α(·) represents the existence of the node and the edge; the node characteristic function is as follows: In the formula, s i represents the i-th detection node, d j represents the jth decision node, w m represents the mth influencing node; The edge existence indicator function is as follows: In the formula, represents the set of edges in the network between the i-th detection node and the j-th decision node, represents the edge set of the network between the nth target node and the i-th detection node, represents the set of edges in the network between the j-th decision node and the m-th influencing node, Represents the edge set of the network between the mth influencing node and the nth target node; Step (34) Use the adjacency matrix and the arrival matrix to calculate the number of valid task chains N eol (t), and integrate the task success criterion N eol (t)≥τ i Calculate the task success rate; where: τ i Represents the minimum number of valid task chains required for task i to succeed.

7. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 6, characterized in that: Step (33) also includes a random failure step and a deliberate attack step, specifically: The random failure step includes: determining the number of node and edge failures of the unmanned cluster through Monte Carlo simulation according to the failure rate and distribution of the nodes of the unmanned cluster node model and the edge failure rate of the unmanned cluster node edge model, randomly removing the failed nodes and edges of the unmanned cluster node model, and adding the failed nodes and edges to the failed node list and the failed edge list respectively; wherein the detection node s i , detection node d j and the detection node w m The failure rate is determined by the parameter Failure rate Failure rate It can be described by the Poisson distribution of The deliberate attack steps include: determining the attack mode, arranging the nodes in the unmanned cluster in descending order according to the node degree, then removing the corresponding number of failed nodes and their edges, and adding the failed nodes to the failed node list; taking CNA as a special case to study the node failure and dynamic reconstruction strategy; setting the node degree attack strategy to attack only the first 50% of the nodes, setting Represents node s i ,d j ,w m degree; Specifically: When the list of failed nodes is not empty, the process of adjusting or reconstructing the connections and relationships between different clusters within the cluster is adopted. Inter-cluster reconstruction I is a rule-based reconstruction strategy; When the list of failed nodes is not empty, intra-cluster reconstruction II is used to modify the connections and relationships between nodes in the same cluster, optimize the internal structure and coordination of the unmanned cluster, and make the communication, resource allocation, and task allocation between nodes more efficient; When the list of failed nodes is not empty, repair and generate reconstruction III repairs or adds new nodes or edges within the cluster, which involves deploying additional clusters or activating dormant nodes to improve the overall ability and capacity of the cluster; When the task changes or the cluster structure changes on a large scale, task redistribution IV is triggered, which involves the redistribution of tasks between nodes in the cluster to optimize task allocation, maximize cluster efficiency, and ensure the completion of priority task goals.

8. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 6, characterized in that: Step (34) is specifically: Step (341) obtains the adjacency matrix A between nodes of different types through the effective task network model = {A TS ,A SD ,A DW ,A WT }; Step (342) considers the node and edge attributes and existence, and verifies the existence of the adjacency matrix elements: If there is a connection between nodes, otherwise, The existence probability of an adjacency matrix element is: Among them, if the failure and generation of nodes obey exponential distribution, the probability of the existence of unmanned cluster nodes is: The existence probability of each edge of the unmanned cluster is expressed as: In the formula, represents the allocation result of the detection task, represents the i-th detection node s i With the jth decision node d j Effective communication between Indicates the mth influencing node w m and the nth target node t n effective communication between; wherein: After all the adjacency matrix elements have been verified, go to step (343), otherwise go to step (342) to continue verification; Step (343) calculates the number of valid task chains: Calculate the arrival matrix A TT , A TT =A TS ×A SD ×A DW ×A WT ; Calculate the arrival matrix A TT The number of effective task chains N eol , N eol =Trace(A TT ), A TS Indicates that the dimension is N×I, A SD Indicates the dimension is I×J, A DW Indicates the dimension is J×M, A WT Indicates that the dimension is M×N; Task success rate calculation i=1, determine whether the task chain is a valid task chain If N eol ≥τ i Indicates that the task is completed effectively, the task successfully forms a valid task chain, and the number of successful tasks Num task-success =Num task-success +1,i=i+1; If N eol <τ i If i=n t Then go to the next step, otherwise go to this step; n t is the number of T nodes; Calculate the task success rate based on the number of simulations 9. The method for selecting the best of multiple task allocation algorithms for complex task scenarios according to claim 1, characterized in that: The dynamic reconstruction strategy of step 2 includes: rule-based dynamic reconstruction and task-based dynamic reconstruction; The unmanned swarm task allocation strategy is evaluated by using the task success rate indicator that integrates the actual task completion process and completion results; the task success rate indicator is closely combined with the effective task chain to form effective task completion as the task success standard; The step three uses the unmanned cluster adjacency matrix and the arrival matrix to calculate the number of valid task chains.

Images