Multi-objective allocation method and device based on distributed multi-node computing for clusters

Abstract

This invention relates to a multi-target allocation method and device based on distributed multi-node computing for cluster operations, belonging to the field of target allocation technology in adversarial combat missions. It addresses the limitations of traditional target allocation algorithms in cluster adversarial combat missions, such as the inability to quickly find the optimal target allocation solution and the inability to handle matching a variable number of targets due to the limitations of the aircraft's own computing power. This invention considers the actual observations of individual clusters when designing target allocation rules, utilizes communication to enable cluster individuals to perceive global target information, and under the same rules, uses different random number seeds at different stages of the target allocation algorithm to give cluster individuals differentiated search capabilities at different stages, avoiding repetitive searches. It specifies the number of calculation rounds or times based on the acceptable duration of the mission, interacts with individual calculation results, saves the optimal result within a finite time, repeats the calculation over multiple finite time periods, and outputs the calculated optimal value within a specified time.

Description

Technical Field

[0001] This invention belongs to the field of target allocation technology in combat missions, and relates to a multi-target allocation method and device for clusters. Background Technology

[0002] In aerial swarm warfare, numerous enemy and friendly aircraft are involved. To complete a strike mission, different targets must be assigned to different aircraft. The quality of this assignment and the timeliness of the computation determine the overall performance of the aircraft in swarm warfare. Traditional target assignment algorithms mostly employ centralized computation and utilize intelligent heuristic algorithms and their variations, such as ant colony optimization, particle swarm optimization, and genetic algorithms, to complete the target assignment process. Traditional target assignment algorithms have the following limitations in swarm warfare: First, they fail to effectively consider the limited computational capabilities of individual swarm units in actual use, resulting in excessively long assignment computation times and affecting operational timeliness. Second, the common problem of matching a variable number of targets in swarm warfare, such as 2 vs. 1, 3 vs. 1, or 3 vs. 2, makes it difficult to construct the solution structure in traditional target assignment algorithms, leading to the tendency to get trapped in local optima during the optimal solution process. Third, traditional target assignment algorithms fail to consider actual individual observations during the solution process and fail to integrate these observations into the target assignment process to reduce the complexity of the solution.

[0003] Therefore, the use of traditional target allocation algorithms in target allocation tasks during cluster combat operations is greatly limited. Summary of the Invention

[0004] The purpose of this invention is to address the problems of traditional target allocation algorithms in target allocation tasks during swarm combat operations, which are limited by the computing power of the aircraft itself, such as the inability to quickly find the optimal solution for target allocation, the inability to handle matching of a variable number of targets, and the insufficient integration of aircraft observations into the design of the target allocation method.

[0005] A multi-objective allocation method for clusters based on distributed multi-node computing includes the following steps:

[0006] Step 1: Based on the target position information T observed by the airborne aircraft j = [x,y,z], where x,y,z are the target's position coordinates, fusing the cluster's global perception information of the target; based on the fused global perception information, the index matrix SeeIndex of the target observed by aircraft i is obtained. i =[value j ], where value j The target index number is assigned, and the target index number is globally unique for each aircraft.

[0007] For matching non-fixed number of objectives, each objective can be defined as N cThe aircraft is locked onto; the target allocation rules are determined based on observations as follows:

[0008] Rule 1: During the target allocation process, the processing order of aircraft i is determined by the index matrix of observed targets. The aircraft that observes fewer targets will be processed first.

[0009] Rule 2: SeeIndex, the index matrix of targets observed by the aircraft. i When there is only one value, it means that only one target has been observed at present, and the aircraft is temporarily assigned to that target;

[0010] Rule 3: SeeIndex, the index matrix of targets observed by the aircraft. i When the spacecraft number is empty, the spacecraft number is placed into the empty matrix NullIndex, and the spacecraft in the empty matrix are randomly assigned.

[0011] The index matrix of all targets observed by all aircraft, their own position information, and global perception information are shared through a communication network;

[0012] Step 2: In the t-th finite time interval, perform distributed individual computation, that is, for each individual aircraft in the cluster, conduct independent aircraft node computation based on the data shared by the communication network and the target allocation rules:

[0013] Initialize the flag matrix FlagM and the target allocation count flag matrix FlagMN; the flag matrix FlagM records whether target j has been allocated, and each bit of the target allocation count flag matrix FlagMN stores the number of aircraft allocated to the corresponding target;

[0014] Random seeds RandSeed1, RandSeed2, and RandSeed3 are set based on a random seed mechanism.

[0015] According to the target allocation rules, aircraft are grouped based on the number of targets they observe, and aircraft with the same number of targets are grouped together. Processing starts from the group with the smallest number of observations: for aircraft with the same number of observations, aircraft are randomly selected using the random seed RandSeed1. If multiple aircraft simultaneously select the same target, random selection is performed based on the number of aircraft already allocated to that target. There are two scenarios: one is that the number of aircraft is greater than the number required for the target, and the other is that the number of aircraft is less than or equal to the number required for the target. For the first scenario, the drones that fail to be randomly selected are entered into the NullIndex matrix, which is used to record the index of the drones that fail to be randomly selected.

[0016] In each process of assigning an individual aircraft to a target, the index matrix SeeIndex of aircraft i is determined based on the corresponding flag matrix FlagM.i Update the index matrix; when there are still multiple targets that need to be assigned to aircraft during the continuous updating of the index matrix, for the selectable aircraft i, use the random seed RandSeed2 to randomly select the target.

[0017] After the aircraft in the current aircraft group has been processed, the sorting of the aircraft in the NullIndex matrix is ​​randomly shuffled using the random seed RandSeed3. According to the sorting of the randomly shuffled aircraft, for the unfinished target individuals in the Flag matrix FlagM, the targets are selected in turn, and the Flag matrix FlagMN continues to allocate aircraft to the targets according to the number of targets allocated.

[0018] After all targets are assigned aircraft, the expected index values ​​are calculated based on the current target assignment results.

[0019] For the t-th finite time period, multiple allocations will be performed and corresponding expected index values ​​will be obtained. The optimal expected index value of the current spacecraft node and the corresponding target allocation result will be stored within the finite time period.

[0020] Step 3: After each distributed individual computation is completed within a finite time period, the results of each individual distributed computation are shared through a communication network, and all individual aircraft will obtain unified global optimal information; with each iteration within a finite time period, the global optimal information is updated in real time.

[0021] After T finite-time distributed individual computations, it is determined whether the time constraint / computation round is satisfied. When the time constraint / distributed computation round is reached, the target allocation result corresponding to the output target index value is determined.

[0022] Furthermore, the process of fusing the cluster's global perception information of the target, as described in step one, is autonomously completed by optoelectronic equipment or the aircraft mission vehicle.

[0023] Furthermore, the initialization flag matrix FlagM and the target allocation number flag matrix FlagMN are both 1xN in size, where N is the number of targets obtained from the global perception information.

[0024] Preferably, when initializing the flag bit matrix FlagM and the target allocation number flag matrix FlagMN, each bit in the matrix is ​​assigned 0.

[0025] Furthermore, when the number of aircraft assigned to the target reaches the upper limit, the corresponding position of the flag matrix FlagM is set to 1, and it remains at 0 at other times.

[0026] A computer storage medium storing at least one instruction, which is loaded and executed by a processor to implement a cluster-oriented, distributed, multi-node computing-based multi-objective allocation method.

[0027] A multi-target allocation device for cluster-based distributed multi-node computing is provided. The device includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement the multi-target allocation method for cluster-based distributed multi-node computing.

[0028] Beneficial effects:

[0029] 1. The present invention discloses a multi-target allocation method for clusters based on distributed multi-node computing. By combining the concept of distributed computing with the communication network of clustered aircraft, it solves the problem that centralized target allocation algorithms cannot be used for a class of aircraft with weak computing power.

[0030] 2. The present invention discloses a multi-objective allocation algorithm for clusters based on distributed multi-node computing. By introducing the actual observation rules of the aircraft, the complexity of the optimization is reduced. Therefore, it can quickly solve the multi-objective allocation for a large number of cluster objectives.

[0031] 3. The present invention discloses a multi-objective allocation algorithm for clusters based on distributed multi-node computing, which can handle matching tasks with a variable number of objectives and eliminates the process of constructing solution structures in existing intelligent heuristic algorithms. Attached Figure Description

[0032] Figure 1 This is an overall flowchart of a multi-objective allocation method for cluster-based distributed multi-node computing.

[0033] Figure 2 Flowchart of distributed individual computing nodes. Detailed Implementation

[0034] In order to overcome the limitations of existing traditional target allocation algorithms in target allocation tasks in cluster combat operations, and to solve problems such as the inability to quickly find the optimal solution for target allocation due to the limitations of the aircraft's own computing power, the inability to handle the matching of a variable number of targets, and the insufficient integration of aircraft observations into the design of the target allocation method, this invention proposes a multi-target allocation method for clusters based on distributed multi-node computing.

[0035] This invention, based on the target location information obtained globally by a cluster using its own optoelectronic / infrared devices and location information, formulates target allocation rules that consider the actual observations of individual clusters. It utilizes communication to enable cluster individuals to perceive global target information. Under the same rules, different random number seeds are used in several key stages of the target allocation algorithm to give cluster individuals differentiated search capabilities at different stages, avoiding repetitive searches. The calculation rounds or times are specified according to the acceptable duration of the task, and the calculation results of individual clusters are exchanged. The optimal result within a finite time is saved, and the process is repeated over multiple finite time periods, outputting the optimal value within a specified time. This method is effectively applicable to actual target allocation tasks in combat where both sides have a large number of clusters but limited individual computing capabilities. The following detailed description of specific implementation methods further illustrates this approach.

[0036] Specific implementation method one: Combining Figure 1 This implementation method is described below.

[0037] This implementation method is a multi-objective allocation method for clusters based on distributed multi-node computing, including the following steps:

[0038] Step 1: Based on the aerial vehicle V i Observed target location information T j = [x,y,z], where x,y,z are the position coordinates of the target. The cluster's global perception information of the target is fused. This fusion process is completed autonomously by the optoelectronic equipment or the aircraft mission aircraft. The aircraft's detection image information is shared through network communication. Duplicate target information is identified through feature point extraction and matching. Then, global perception information is given and a unique index is assigned to the target. This is existing technology and will not be described in detail in this invention.

[0039] Based on the fused global perception information, the index matrix SeeIndex of the targets observed by aircraft i is obtained. i =[value j ], where value j The target index number is assigned, and the aircraft and target index number are globally unique. For matching a variable number of targets, each target can be assigned an index number of N. c The aircraft was locked onto. Target allocation rules were determined based on observations as follows:

[0040] Rule 1: During the target allocation process, the processing order of aircraft i is determined by the index matrix of observed targets. The aircraft that observes fewer targets will be processed first.

[0041] Rule 2: SeeIndex, the index matrix of targets observed by the aircraft. i When there is only one value, it means that only one target has been observed at present, and the aircraft is temporarily assigned to that target.

[0042] Rule 3: SeeIndex, the index matrix of targets observed by the aircraft. i When the spacecraft is empty, its number is placed in the empty matrix NullIndex, and the spacecraft in the empty matrix are randomly assigned.

[0043] The above rules are based on actual swarm collaborative detection. Usually, the number of targets observed by aircraft located at the edge of the swarm is less than that of those located in the middle of the swarm. The purpose of target allocation is for subsequent strikes, so the aircraft that observes the only target is directly assigned to that target.

[0044] The index matrix of targets observed by all aircraft, their own position information, and global perception information are shared through a communication network.

[0045] Step one corresponds to Figure 1 The process by which distributed computing nodes obtain initial situational information.

[0046] Step 2: In the t-th finite time interval, perform distributed individual computation, that is, for each individual aircraft in the cluster, conduct independent aircraft node computation based on the data and rules shared by the communication network:

[0047] Initialize the flag bit matrix FlagM and the target allocation number flag matrix FlagMN; where the size of both the flag bit matrix and the target allocation number flag matrix is ​​1xN, and N is the number of targets obtained from global perception information. During the initialization process, each bit in the matrix is ​​assigned 0.

[0048] Flag matrix FlagM records whether target j has been allocated. Flag matrix FlagMN stores the number of aircraft allocated to the corresponding target in each bit, with an upper limit of N. c When the number of aircraft assigned to the target reaches the upper limit, the corresponding position of the flag matrix FlagM is set to 1, and it remains 0 at other times.

[0049] Random seeds RandSeed1, RandSeed2, and RandSeed3 are set based on a random seed mechanism.

[0050] The random seed mechanism refers to a mechanism that reduces the probability of an object being selected again. In other words, if an object is selected in the current selection process, the probability of that object being selected again in subsequent selection processes gradually decreases.

[0051] According to the target allocation rules mentioned above, aircraft are grouped based on the number of targets observed by the aircraft, and aircraft with the same number of targets are grouped together. Processing starts from the group with the smallest number of aircraft observed: for aircraft with the same number of observations, aircraft are randomly selected using a random seed RandSeed1. If multiple aircraft simultaneously select the same target, random selection is performed based on the number of aircraft already allocated to that target. There are two scenarios: one is that the number of aircraft is greater than the number required by the target, and the other is that the number of aircraft is less than or equal to the number required by the target (for example, in a certain scenario, the target requires 3 aircraft to attack, but only 2 aircraft have been selected). For drones that fail in the first scenario, they are entered into the idle matrix NullIndex, which is a dynamically changing matrix used to record the index of the drones that fail.

[0052] In each process of assigning an individual aircraft to a target, the index matrix SeeIndex of aircraft i is determined based on the corresponding flag matrix FlagM. i Update the index matrix; when there are still multiple targets that need to be assigned to aircraft during the continuous updating of the index matrix, use the random seed RandSeed2 to randomly select targets for the selectable aircraft i.

[0053] After the aircraft in the current aircraft group have been processed, the aircraft in the NullIndex free matrix are randomly shuffled using the random seed RandSeed3. Based on the shuffled aircraft sorting, targets that have not been completed in the Flag matrix FlagM are selected in turn. According to the number of targets allocated, the Flag matrix FlagMN continues to allocate aircraft to the targets.

[0054] Once all targets are assigned to aircraft, meaning each aircraft has determined which target to engage, the expected performance index (SPI) is calculated based on the current target assignment. The expected performance index, set by the user, is the desired outcome and measures the quality of the distributed computation results. SPIs can include the total distance between the aircraft and the target, or the engagement time. Generally, the goal is to achieve the optimal result for that SPI, such as minimizing the total distance or engagement time.

[0055] For the t-th finite time period, multiple allocations will be performed (e.g., 100 allocations will be performed in the t-th finite time period) and the corresponding expected index values ​​will be obtained. The optimal expected index value of the current spacecraft node and the corresponding target allocation result will be stored within the finite time period.

[0056] Step Two Corresponding Figure 1 The process begins with "Starting the distributed node computing process". Figure 2 This is the detailed process flow.

[0057] Step 3: After each distributed individual computation is completed within a finite time period, the results of each individual distributed computation are shared through a communication network, and all individual aircraft will obtain unified global optimal information; with each iteration within a finite time period, the global optimal information is updated in real time.

[0058] After T finite-time distributed individual computations, it is determined whether the time constraint / computation round is met. When the time constraint / distributed computation round is reached, the target allocation result corresponding to the output target index value is the global optimal value, thereby completing the target allocation task in cluster operations and assisting subsequent decision-making.

[0059] Step 3 Corresponding Figure 1 The process after "starting the distributed node computing process" in the middle. Specific Implementation Method Two:

[0061] This embodiment is a computer storage medium that stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the multi-objective allocation method for cluster-oriented distributed multi-node computing.

[0062] It should be understood that any method described in this invention can be provided as a computer program product, software, or computerized method, which may include a non-transitory machine-readable medium on which instructions are stored, which can be used to program a computer system or other electronic device. The storage medium may include, but is not limited to, magnetic storage media, optical storage media; magneto-optical storage media include: read-only memory (ROM), random access memory (RAM), erasable programmable memory (e.g., EPROM and EEPROM), and flash memory layers; or other types of media suitable for storing electronic instructions. Specific implementation method three:

[0064] This embodiment is a multi-target allocation device for clusters based on distributed multi-node computing. The device includes a processor and a memory. It should be understood that this includes any device including a processor and a memory as described in this invention. The device may also include other units or modules that perform display, interaction, processing, control, and other functions through signals or instructions. The device described in this embodiment can be a computer, or it can be a device set in an airborne cluster confrontation system, or it can be integrated into an airborne cluster confrontation system.

[0065] The memory stores at least one instruction, which is loaded and executed by the processor to implement the aforementioned multi-objective allocation method for cluster-oriented distributed multi-node computing.

[0066] Example

[0067] To address target allocation tasks in cluster-based combat, this invention fully considers the practical engineering application background of individual computing capabilities, observation capabilities, and communication capabilities. It utilizes the communication nodes inherent in the cluster to construct a distributed computing network, and combines the actual observation capabilities of individuals to define target allocation criteria. Random seeds are used to traverse the results at different stages of optimization, thus completing a multi-target allocation method for clusters based on distributed multi-node computing. The invention is further illustrated below with specific implementation examples.

[0068] Consider the following swarm attack scenario: 60 aircraft intercept 30 targets, with two aircraft assigned to intercept each target. In the air, the interception swarm observes the positions of the 30 targets at the current moment, and uses optoelectronic equipment to perform situational awareness fusion, obtaining the target number T at each position. j =[x j ,y j ,z j The interceptors are defined as follows: j = 1, ..., 30, and i = 1, ..., 60. The observation index matrix is ​​SeeIndex. i =[j1,…,j n [,j=1…30, where n is the number of observations by aircraft i, and the maximum value is also the maximum value for group division. The NullIndex matrix is ​​set to store the indices of aircraft that did not obtain targets after processing. During the child node calculation process, the flag matrix FlagM=[value1,…,value...] 30 [value = 0 or 1, initially 0, updated in real time during iteration by the target-assigned flag matrix, FlagMN = [value1, ... value1] 30 The value of each bit in the FlagMN matrix is ​​determined by the number of aircraft that have assigned the current target. The computation time or number of rounds is defined and iterated according to the rules described in the diagram and previous text. The deadline for each computation is determined by checking if all flag bits in the matrix are 1 and if all flag bits in the target allocation matrix are 2. Distributed computation individuals output the optimal result from a finite number of iterations, and the optimal value is determined synchronously through sharing. The result matrix resultM = [value1, ..., value2] 60 `value = j` stores the allocation result after the current calculation. The corresponding allocation value `Rv` is determined by a specified evaluation formula, which can be based on factors such as minimum distance or minimum total attack time. The optimal value is ultimately used as the target allocation result for execution.

[0069] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A multi-objective allocation method for clusters based on distributed multi-node computing, characterized in that, Includes the following steps: Step 1: Based on the target location information observed by the aircraft ,in Using the target's location coordinates, the cluster's global perception information about the target is fused together; Based on the fused global perception information, the aircraft is obtained Index matrix of observed targets ,in The target index number is assigned, and the target index number is globally unique for each aircraft. For matching non-fixed quantity targets, each target can be set to be... The aircraft is locked onto; the target allocation rules are determined based on observations as follows: Rule 1: During the target allocation process, for aircraft The processing order is determined by the index matrix of the observed targets; the aircraft that observes fewer targets will be processed first. Rule 2: Index matrix of targets observed by the aircraft When there is only one value, it means that only one target has been observed at present, and the aircraft is temporarily assigned to that target; Rule 3: Index matrix of targets observed by the aircraft When the spacecraft is empty, its number is placed into the empty matrix. In the idle matrix, aircraft are randomly assigned; The index matrix of all targets observed by all aircraft, their own position information, and global perception information are shared through a communication network; Step 2: In the t-th finite time interval, perform distributed individual computation, that is, for each individual aircraft in the cluster, conduct independent aircraft node computation based on the data shared by the communication network and the target allocation rules: Initialize the flag matrix Target allocation number flag matrix Flag matrix Record target Whether the allocation is complete, and the target number of allocations flag matrix. Each value stores the number of aircraft already allocated to the corresponding target. Set a random seed based on a random seed mechanism. , , ; According to the target allocation rules, aircraft are grouped based on the number of targets they observe, with aircraft having the same number of targets grouped together. Processing begins with the group with the smallest number of observations: for aircraft with the same number of observations, a random seed is used. Aircraft are randomly selected. If multiple aircraft simultaneously select the same target, random selection is performed based on the number of aircraft already allocated to that target. Two scenarios exist: either the number of available aircraft exceeds the target's required number, or the number of available aircraft is less than or equal to the target's required number. For the first scenario, drones that fail the random selection process enter the idle matrix. The idle matrix is ​​used to record the indices of randomly failed drones; In each process of assigning individual aircraft to a target, the corresponding flag bit matrix is ​​used as a basis. For aircraft index matrix Update the index matrix; when multiple targets still need to be assigned aircraft during the continuous updating of the index matrix, selectable aircraft... Using random seeds Randomly select targets; After the aircraft in the current aircraft group have completed processing, a random seed is used. For the free matrix The aircraft are randomly shuffled, and the aircraft are sorted according to the flag matrix. For individuals who have not yet completed their target tasks, select targets sequentially and assign a number of flags to each target according to the target's criteria. Continue to allocate aircraft to targets; After all targets are assigned aircraft, the expected index values ​​are calculated based on the current target assignment results. For the t-th finite time period, multiple allocations will be performed and corresponding expected index values ​​will be obtained. The optimal expected index value of the current spacecraft node and the corresponding target allocation result will be stored within the finite time period. Step 3: After each distributed individual computation is completed within a finite time period, the results of each individual distributed computation are shared through a communication network, and all individual aircraft will obtain unified global optimal information; with each iteration within a finite time period, the global optimal information is updated in real time. After T finite-time distributed individual computations, it is determined whether the time constraint / computation round is satisfied. When the time constraint / distributed computation round is reached, the target allocation result corresponding to the output target index value is determined.

2. The multi-objective allocation method for clusters based on distributed multi-node computing as described in claim 1, characterized in that, The process of fusing the cluster's global perception information of the target, as described in step one, is autonomously completed by optoelectronic equipment or aircraft mission vehicles.

3. A multi-objective allocation method for clusters based on distributed multi-node computing, as described in claim 1 or 2, characterized in that, The initialization flag matrix Target allocation number flag matrix The size is 1xN, where N is the number of targets obtained from global perception information.

4. A multi-objective allocation method for clusters based on distributed multi-node computing as described in claim 3, characterized in that, Initialize the flag matrix Target allocation number flag matrix Each bit in the matrix is ​​assigned 0.

5. A multi-objective allocation method for cluster-based distributed multi-node computing according to claim 4, characterized in that, When the number of target-assigned aircraft reaches its maximum value, the flag matrix... The corresponding position is 1, and the value remains 0 at all other times.

6. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement a cluster-oriented multi-objective allocation method based on distributed multi-node computing as described in any one of claims 1 to 5.

7. A multi-target allocation device for clusters based on distributed multi-node computing, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement a cluster-oriented multi-objective allocation method based on distributed multi-node computing as described in any one of claims 1 to 5.

Images