Mission decision-making method for integrated reconnaissance and attack of UAV swarm under target uncertainty

Abstract

The present invention discloses a task decision method for integrated reconnaissance and attack of a drone cluster under target uncertainty conditions. The drone cluster includes a reconnaissance subgroup and at least two attack subgroups. The task decision method includes receiving reconnaissance task information and determining a task subarea of ​​each reconnaissance drone in the reconnaissance subgroup; the reconnaissance drone determines the target in the task subarea; determines the task priority of all targets; and assigns all targets to different attack subgroups by bidding, and assigns targets to attack drones in the attack subgroup by bidding. The task decision method of this scheme solves the problem that the prior art cannot perform efficient dynamic task allocation under target uncertainty conditions.

Description

Technical Field

[0001] The present invention relates to aircraft cluster coordination technology, and in particular to a task decision method for unmanned aerial vehicle cluster reconnaissance and attack integration under target uncertainty conditions. Background Art

[0002] Mission decision-making is one of the key technologies for multi-aircraft to achieve cluster coordination and efficiently complete the given tasks. Due to the difference between multi-aircraft and traditional single-aircraft in the coupling mechanism of mission planning and maneuvering, not only the flight control and maneuvering strategies at the individual level of aircraft should be considered, but also the decision-making at the cluster level (such as the allocation of confrontation strategies, tasks or targets). Therefore, this problem has a typical "decision-execution" hierarchical feature. Among them, the decision-making layer performs target evaluation and task allocation based on the information of the own cluster and the information of the mission target; so that the execution layer executes the strategy determined by the decision-making layer. However, the current research on how to plan the mission and maneuvering strategies of aircraft clusters in confrontation scenarios is not sufficient. This is mainly because the confrontation targets are usually a type of non-cooperative targets that do not communicate at the information level and do not cooperate in maneuvering behaviors. The state information and maneuvering behaviors of the targets are uncertain, which seriously restricts the actual performance of the existing aircraft cluster task decision-making methods.

[0003] There are two points that need to be improved in the current research on cluster task decision-making. First, it usually assumes that all target information is completely known. However, in actual environments, the location and configuration information of the target may be unknown. Therefore, there is more than one type of task, and at least two types of tasks should be included: collaborative reconnaissance and target attack. Second, in most studies, the heterogeneous characteristics of heterogeneous multi-UAV clusters are only reflected in the fact that drones have different uses and categories. However, there may also be numerical differences in performance between drones of the same type. Therefore, it is necessary to further study the task decision-making problem of multi-UAV clusters with stronger heterogeneity.

[0004] In order to cope with large-scale dynamic real-time task decision-making scenarios, distributed task allocation algorithms, such as multi-agent algorithms, contract net protocols (CNP), auction algorithms, etc., have the characteristics of strong robustness, strong local update capability, strong scalability, and suitability for large-scale allocation. However, for the problem of making multi-aircraft task decisions to cope with target uncertainty, the main difficulty currently faced is: how to carry out efficient dynamic task allocation of multiple aircraft under target uncertainty conditions. The first problem for aircraft clusters to achieve confrontation tasks is how each aircraft can clarify its own task goals. This is a typical cluster task allocation problem, which belongs to the decision-making layer research in the "decision-execution" architecture. The number, information and completion progress of tasks may change with the changes in the situation of both parties, which puts higher requirements on the real-time and readjustment capabilities of task allocation algorithms. At the same time, how to determine appropriate allocation criteria based on task requirements, target uncertainty and the characteristics of one's own aircraft to evaluate the advantages and disadvantages of different allocation schemes is also an important issue that has a direct impact on the results. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for making decisions on the integrated reconnaissance and strike of a swarm of unmanned aerial vehicles under target uncertainty conditions, which solves the problem that the prior art cannot perform efficient dynamic task allocation under target uncertainty conditions.

[0006] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0007] A task decision method for an unmanned aerial vehicle cluster integrating reconnaissance and attack under target uncertainty is provided. The unmanned aerial vehicle cluster includes a reconnaissance subgroup and at least two attack subgroups. The task decision method includes the following steps:

[0008] S1, receiving reconnaissance mission information, and using the improved CNP algorithm with uncertain contracts to determine the mission sub-area of ​​each reconnaissance UAV in the reconnaissance sub-group;

[0009] S2, the reconnaissance UAV conducts reconnaissance on its mission sub-area and determines the targets in the mission sub-area; then, a fuzzy comprehensive evaluation method is used to determine the mission priority of all targets reconnaissanced by all reconnaissance UAVs;

[0010] S3, the attack subgroup calculates the first level of compatibility between itself and the target with the highest priority among all unassigned targets, and assigns the target with the highest priority to the attack subgroup with the highest level of compatibility;

[0011] S4, determine whether there is an unassigned target among all targets, if so, update the attack capability of the attack subgroup that accepts the target and return to step S3, otherwise go to step S5;

[0012] S5, the attack drone receives all targets assigned to its attack subgroup and calculates its second-layer compatibility with the unassigned target with the highest mission priority;

[0013] S6. According to the defense capability of the target and the attack capability of the attack drone, the unassigned target with the highest mission priority in the attack subgroup is assigned to multiple attack drones with the highest adaptability in the second layer;

[0014] S7. Determine whether there are unassigned targets among all the targets undertaken by the attack subgroup. If so, update the attack capability of the attack drone that accepts the target and return to step S6. Otherwise, complete the task decision.

[0015] Furthermore, the drone cluster includes three attack subgroups, namely a missile subgroup, a laser subgroup and an electromagnetic pulse subgroup. The properties of the attack drones in each attack subgroup are:

[0016] MU Q / LU R / EU S ={Uloc,vol,sur,misatt,lasatt,EMPatt,vel}

[0017] Among them, MU Q ,LU R and EU S They represent the attack drones in the missile subgroup, laser subgroup and electromagnetic pulse subgroup respectively; Uloc is the current position of the attack drone; vol is the mission capacity of the attack drone; sur is the survivability of the attack drone; misatt is the missile attack capability of the attack drone; lasatt is the laser attack capability of the attack drone; EMPatt is the electromagnetic pulse attack capability of the attack drone; vel is the flight speed of the drone;

[0018] The properties of the target are:

[0019] AT N ={Tloc,misdef,lasdef,EMPdef,pro,latest,need,risk}

[0020] Among them, AT N is the attribute of the target; Tloc is the target location; misdef is the target's defense capability against missiles; lasdef is the target's defense capability against lasers; EMPdef is the target's defense capability against electromagnetic pulses; pro is the benefit of completing the task; latest is the latest time limit for the task; need is the number of drones required to complete the task; risk is the risk level of the task.

[0021] Furthermore, the expression of the highest level of adaptability is:

[0022]

[0023] Among them, J R1 The highest level of adaptability; is the average misattack capability of all attacking drones in the missile subgroup, is the average lasatt capability of all attack drones in the laser subgroup, is the average EMPatt capability of all attack drones in the EMP subgroup.

[0024] Furthermore, the expression for calculating the second-layer adaptability is:

[0025] J R2 =C 1 ·(constant+αC 2 +βC 3 -γC 4 -δC 5 )

[0026] Among them, J R2 is the second layer adaptability; C 1 is the task capacity constraint; C 2 The survivability constraint of the attacking drone; C 3 is task adaptability; C 4 is the flight time; C 5 is the latest time constraint of the task; constant is a positive number; α, β, γ, δ are C 2 , C 3 , C 4 and C 5 The weight coefficient of .

[0027] Furthermore, the task capacity constraint C 1 for:

[0028]

[0029] Among them, remaining task volume is the task capacity of attacking drones; required task volume is the task capacity corresponding to the target with the highest priority in the current task;

[0030] UAV survivability constraints C 2 The expression is:

[0031] C 2 =sur-risk

[0032] Task adaptability C 3 The expression is:

[0033] C 3 =max(misatt-misdef,lasatt-lasdef,EMPatt-EMPdef)

[0034] Flight time C 4 The expression is:

[0035]

[0036] Among them, ||·|| 2 is the 2-norm;

[0037] Task latest time constraint C 5 The expression is:

[0038] C 5 =tasktime-latest

[0039] Among them, tasktime is the start time of the target after the current attack drone accepts the target with the highest priority.

[0040] Furthermore, the method for determining the mission sub-area of ​​each reconnaissance drone in the reconnaissance sub-group includes:

[0041] S11. After receiving the task information, the reconnaissance drone calculates its own benefit function according to the attributes of the reconnaissance drone and bids for the task when its task capacity meets the set requirements;

[0042] S12, the reconnaissance drone with the highest profit function among all bids wins the bid and accepts the reconnaissance mission, and then updates its own mission capacity;

[0043] S13. The reconnaissance drone receiving the reconnaissance mission determines the size of its mission sub-area for reconnaissance in the target potential area according to its reconnaissance performance and flight speed, wherein the mission sub-area is a rectangle;

[0044] S14. According to the geometric relationship of the rectangle, the coordinates of the four vertices of the task sub-area are determined:

[0045]

[0046] Among them, poi e is the entry point of the task sub-area; x and y are the coordinates of the entry point of the task sub-area; poi 2 ,poi 3 ,poi 4 are the second, third, and fourth vertices in the counterclockwise direction from the entry point of the rectangular task sub-region; a is the length of the task sub-region; w i RT for the task sub-area i Width;

[0047] S15, determine whether the target potential area has been assigned to the reconnaissance drone, if so, go to step S2, otherwise, update poi 4 It is the entry point for the next reconnaissance mission and returns to step S11.

[0048] Furthermore, the reconnaissance drone adopts an "S"-shaped parallel search mode; the length of the mission sub-area is equal to the length of the target potential area, and its width N i for:

[0049]

[0050] Among them, f R (i) is the profit function of the reconnaissance drone; b is the width of the target potential area; Δd k For reconnaissance drone RU k Width; vel(RU k ) is a reconnaissance drone RU k Flight speed; vel(RU j ) is a reconnaissance drone RU j Flight speed; RU k and RU j are the kth and jth reconnaissance UAVs in the reconnaissance subgroup; P is the total number of reconnaissance UAVs in the reconnaissance subgroup; [·] indicates the rounding up symbol.

[0051] Furthermore, the method for determining the task priority of the target includes:

[0052] According to the reconnaissance and strike mission of the drone swarm, set the comment set:

[0053] u={urg,imp,exe,com}

[0054] Among them, urg represents the urgency of the task; imp represents the importance of the task; exe represents the execution time of the task; com represents the complexity of the task;

[0055] For each evaluation factor in the evaluation set, multiple levels of evaluation are set to form an evaluation factor set; then, a weight vector of the evaluation factor set of the drone cluster's reconnaissance and strike mission is determined;

[0056] The bilateral Gaussian function is used as the membership function for single factor evaluation, and the membership degree of the evaluation factor is obtained:

[0057] r w =[r w1 ,r w2 ,...,r wm ],w=1,2,...,n

[0058] Among them, r o is the membership degree; r w1 、r w2 、r wm The probability of making the first evaluation on the wth factor, the probability of making the second evaluation on the wth factor, and the probability of making the mth evaluation on the wth factor; n is the degree of membership of different evaluation indicators to the wth factor;

[0059] The membership degree is used to construct the evaluation matrix R of the reconnaissance and attack mission of the drone cluster, and the parameter matrix B is calculated:

[0060]

[0061] in, is the operator; a is the weight vector; b 1 ,b 2 ,b p ,b m are the parameters corresponding to the first evaluation, the parameters corresponding to the second evaluation, the parameters corresponding to the pth evaluation, and the parameters corresponding to the mth evaluation; r wp is the possibility of making the p-th evaluation on the w-th factor; a w is any element in the weight vector; m is the number of elements in the evaluation factor set;

[0062] According to the parameter matrix B, calculate the priority p of the drone cluster reconnaissance and attack mission:

[0063]

[0064] Where G = [g 1 ,g 2 ,...,g m ] is the priority weight vector of each comment in the comment set; g 1 ,g 2 ,...,g m are the elements in the priority weight vector respectively; g j and b j are the priority weight vector of the comments in the comment set and the j-th element in the parameter matrix, respectively.

[0065] Furthermore, the UAV cluster reconnaissance and strike integrated task decision method under target uncertainty also includes:

[0066] A1. Determine whether the reconnaissance drone detects a new target when the attack drone is performing the target mission; if so, proceed to step A2; otherwise, all attack drones continue to perform the existing mission;

[0067] A2. Use fuzzy comprehensive evaluation method to determine the task priority of new goals;

[0068] A3. Lock the target that the attack drone is executing, release all targets that the attack drone is not executing, and use the new target and all unexecuted targets as unassigned targets;

[0069] A4. Execute steps S3 to S7.

[0070] Furthermore, the reconnaissance subgroup and the attack subgroup are each composed of a plurality of drones with different performance attributes.

[0071] The beneficial effects of the present invention are as follows: when allocating the target task of reconnaissance and attack, the improved CNP algorithm of uncertain contracts can adaptively determine the size of the task sub-area, which can improve the load balance and collaborative reconnaissance efficiency of the allocation scheme. For the target attack task allocation process, the fuzzy comprehensive evaluation method is first used to determine the priority of the target, thereby solving the task timing problem. After that, after the drone sub-group is used to undertake the target task, the attack drone is used in the sub-group to undertake the target task. This method makes full use of the heterogeneous characteristics of the drone cluster with multiple sub-groups, which can reduce the computing cost and improve the rationality of the allocation scheme.

[0072] In addition, the task redistribution mechanism in this scheme can effectively deal with emergencies during mission execution and improve the robustness and adaptability of cluster task planning. Comparative simulation shows that compared with the classic CNP method, this scheme can improve the efficiency of cluster reconnaissance and balance the load of each reconnaissance drone in reconnaissance task allocation, and can obtain a better attack task allocation scheme in attack task allocation.

[0073] When the attack drone is performing a target mission, if a new target is received, this solution can respond to emergencies during the mission through a reallocation mechanism to ensure that all targets are executed efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a flow chart of the decision-making method for the integrated reconnaissance and strike mission of UAV swarms under target uncertainty conditions.

[0075] Figure 2 This is a schematic diagram of the drone cluster structure.

[0076] Figure 3 Schematic diagram of collaborative task allocation in a scenario where the initial target is unknown.

[0077] Figure 4 Schematic diagram of regional reconnaissance in the "S" shaped parallel search pattern.

[0078] Figure 5 Schematic diagram of the i-th task sub-area.

[0079] Figure 6Schematic diagram of the membership functions of the four evaluation sets.

[0080] Figure 7 The simulated map in the simulation example, (a) is the real map, and (b) is the map at the beginning.

[0081] Figure 8 Results of reconnaissance task allocation; (a) is a schematic diagram of the task sub-area determined by the improved CNP algorithm of the uncertain contract using this scheme, and (b) is a schematic diagram of the task sub-area determined by the classic CNP algorithm.

[0082] Fig. 9 This is a comparison chart of the reconnaissance time of a single reconnaissance drone.

[0083] Fig.10 This is a comparison chart of the task performance indicators of the reconnaissance subgroup.

[0084] Fig.11 Schematic diagram of the four normalized attributes of the attack task.

[0085] Fig.12 Schematic diagram of the profit function of three rounds of negotiation among missile subgroups.

[0086] Fig.13 Schematic diagram of the profit function of each round of negotiation among laser subgroups.

[0087] Fig.14 Schematic diagram of the profit function of each round of negotiation among EMP subgroups.

[0088] Fig.15 Schematic diagram of the task allocation results for missile subgroups.

[0089] Fig.16 Schematic diagram of the task allocation results for the laser sub-swarm.

[0090] Fig.17 Schematic diagram of the task allocation results of the EMP subgroup.

[0091] Fig.18 A comparison chart of normalized evaluation indicators. DETAILED DESCRIPTION

[0092] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0093] like Figure 2As shown in the figure, the UAV cluster in this scheme consists of one reconnaissance subgroup and three attack subgroups, namely missile subgroup, laser subgroup and electromagnetic pulse subgroup. Each subgroup consists of multiple UAVs with different performance attributes. Compared with the heterogeneous multi-agent swarms considered in most existing task allocation studies, the heterogeneous UAV cluster in this scheme has a more complex structure, stronger heterogeneity, better robustness, higher adaptability and stronger collaborative task capability.

[0094] Assume that there are multiple fixed targets in the site area (potential target area), and the exact location, information and number of the targets are unknown. The drone cluster needs to detect the site area before attacking the discovered targets. The collaborative task allocation scenario in this scheme is as follows: Figure 3 shown.

[0095] In this scheme, the attributes of the attacking drones in each attack subgroup are:

[0096] MU Q / LU R / EU S ={Uloc,vol,sur,misatt,lasatt,EMPatt,vel}

[0097] Among them, MU Q ,LU R and EU S They represent the attack drones in the missile subgroup, laser subgroup and electromagnetic pulse subgroup respectively; Uloc is the current position of the attack drone; vol is the mission capacity of the attack drone; sur is the survivability of the attack drone; misatt is the missile attack capability of the attack drone; lasatt is the laser attack capability of the attack drone; EMPatt is the electromagnetic pulse attack capability of the attack drone; vel is the flight speed of the drone;

[0098] The properties of the target are:

[0099] AT N ={Tloc,misdef,lasdef,EMPdef,pro,latest,need,risk}

[0100] Among them, AT N is the attribute of the target; Tloc is the target location; misdef is the target's defense capability against missiles; lasdef is the target's defense capability against lasers; EMPdef is the target's defense capability against electromagnetic pulses; pro is the benefit of completing the task; latest is the latest time limit for the task; need is the number of drones required to complete the task; risk is the risk level of the task.

[0101] refer to Figure 1 , Figure 1 A task decision method for integrated reconnaissance and attack of a swarm of unmanned aerial vehicles under target uncertainty conditions is shown, and the task decision method includes steps S1 to S7.

[0102] In step S1, the reconnaissance mission information is received, and the mission sub-area of ​​each reconnaissance UAV in the reconnaissance sub-group is determined by using the improved CNP algorithm with uncertain contract;

[0103] In one embodiment of the present invention, a method for determining a mission sub-area of ​​each reconnaissance drone in a reconnaissance sub-group includes:

[0104] S11. After receiving the task information, when its task capacity meets the set requirements, the reconnaissance drone calculates its own profit function according to its attributes and bids for the task; the expression of the profit function is:

[0105]

[0106] Among them, f R (i) is the profit function of the reconnaissance drone; vel(RU K ) is the UAV RU K Flight speed; vel(RU j ) is a reconnaissance drone RU j The flight speed of the reconnaissance drones in the reconnaissance subgroup.

[0107] S12, the reconnaissance drone with the highest profit function among all bids wins the bid and accepts the reconnaissance mission, and then updates its own mission capacity;

[0108] S13. The reconnaissance drone that receives the reconnaissance mission determines its mission sub-area in the target potential area based on its reconnaissance performance and flight speed (the schematic diagram of the mission sub-area can be found in Figure 5 ) size, and the task sub-area is a rectangle.

[0109] During implementation, this scheme prefers that the reconnaissance drone adopts the "S" type parallel search mode, the schematic diagram of which can be referred to Figure 4 The length of the task sub-region is equal to the length of the target potential region, and its width N i for:

[0110]

[0111] Where b is the width of the target potential area; Δd k For reconnaissance drone RU k Width; RU k and RU j are the kth and jth reconnaissance UAVs in the reconnaissance subgroup; [·] indicates the rounding up symbol.

[0112] S14. According to the geometric relationship of the rectangle, the coordinates of the four vertices of the task sub-area are determined:

[0113]

[0114] Among them, poi e is the entry point of the task sub-area; x and y are the coordinates of the entry point of the task sub-area; poi 2 ,poi 3 ,poi 4 are the second, third, and fourth vertices in the counterclockwise direction from the entry point of the rectangular task sub-region; a is the length of the task sub-region; w i RT for the task sub-area i Width;

[0115] S15, determine whether the target potential area has been assigned to the reconnaissance drone, if so, go to step S2, otherwise, update poi 4 It is the entry point for the next reconnaissance mission and returns to step S11.

[0116] In step S2, the reconnaissance UAV conducts reconnaissance on its mission sub-area and determines the targets in the mission sub-area; then, a fuzzy comprehensive evaluation method is used to determine the task priority of all targets reconnaissanced by all reconnaissance UAVs.

[0117] In one embodiment of the present invention, a method for determining a task priority of a target includes:

[0118] According to the reconnaissance and strike mission of the drone swarm, set the comment set:

[0119] u={urg,imp,exe,com}

[0120] Among them, urg represents the urgency of the task; imp represents the importance of the task; exe represents the execution time of the task; com represents the complexity of the task;

[0121] For each evaluation factor in the evaluation set, multiple levels of comments are set to form an evaluation factor set, which can be specifically referred to in Table 1; then the weight vector of the evaluation factor set of the drone cluster's reconnaissance and strike mission is determined;

[0122] Table 1 Definition of comment set

[0123] Comments Evaluation factor set Emergency Comments <![CDATA[v urg ={Urgent, General, Not Urgent}. ]]> Importance Comments <![CDATA[v imp ={important, general, unimportant}. ]]> Execution Time Comments <![CDATA[v exe ={short, medium, long}. ]]> Complexity Critique Set <![CDATA[v com ={complex, general, simple}. ]]>

[0124] The bilateral Gaussian function is used as the membership function for single factor evaluation to obtain the membership degree of the evaluation factor (the membership degrees of the four evaluation sets in the comment set of this scheme can be referred to Figure 6 ):

[0125] r w =[r w1 ,r w2 ,...,r wm ],w=1,2,...,n

[0126] Among them, r o is the membership degree; r w1 、r w2 、r wm The probability of making the first evaluation on the wth factor, the probability of making the second evaluation on the wth factor, and the probability of making the mth evaluation on the wth factor; n is the degree of membership of different evaluation indicators to the wth factor;

[0127] The membership degree is used to construct the evaluation matrix R of the reconnaissance and attack mission of the drone cluster, and the parameter matrix B is calculated:

[0128]

[0129] in, is the operator; a is the weight vector; b 1 ,b 2 ,b p ,b m are the parameters corresponding to the first evaluation, the parameters corresponding to the second evaluation, the parameters corresponding to the pth evaluation, and the parameters corresponding to the mth evaluation; r wp is the possibility of making the p-th evaluation on the w-th factor; a w is any element in the weight vector; m is the number of elements in the evaluation factor set;

[0130] According to the parameter matrix B, calculate the priority p of the drone cluster reconnaissance and attack mission:

[0131]

[0132] Where G = [g 1 ,g 2 ,...,g m ] is the priority weight vector of each comment in the comment set; g 1 ,g 2 ,...,g m are the elements in the priority weight vector respectively; g j and b j are the priority weight vector of the comments in the comment set and the j-th element in the parameter matrix, respectively.

[0133] This scheme adopts a fuzzy comprehensive evaluation method to integrate the urgency, importance, execution time and complexity of the UAV cluster reconnaissance and attack mission, and can autonomously evaluate the mission priority. The weight set and priority weight vector can be customized and configured according to the scenario and needs, so that the mission priority evaluation has multi-preference characteristics and can respond to different evaluation needs.

[0134] In step S3, the attack subgroup calculates a level of compatibility between itself and the target with the highest priority among all unassigned targets, and assigns the target with the highest priority to the attack subgroup with the highest level of compatibility;

[0135] During implementation, the expression of the highest level of adaptability preferred by this solution is:

[0136]

[0137] Among them, J R1 The highest level of adaptability; is the average misattack capability of all attacking drones in the missile subgroup, is the average lasatt capability of all attack drones in the laser subgroup, is the average EMPatt capability of all attack drones in the EMP subgroup.

[0138] In step S4, it is determined whether there is an unassigned target among all targets. If so, the attack capability of the attack subgroup that accepts the target is updated and the process returns to step S3. Otherwise, the process goes to step S5.

[0139] In step S5, the attack drone receives all targets assigned to its attack subgroup and calculates its second-layer compatibility with the unassigned targets with the highest mission priority;

[0140] During implementation, the solution preferably calculates the expression for the second-layer adaptability as follows:

[0141] J R2 =C 1 ·(constant+αC 2 +βC 3 -γC 4 -δC 5 )

[0142] Among them, J R2 is the second layer adaptability; C 1 is the task capacity constraint; C 2 The survivability constraint of the attacking drone; C 3 is task adaptability; C 4 is the flight time; C 5 is the latest time constraint of the task; constant is a positive number; α, β, γ, δ are C2 , C 3 , C 4 and C 5 The weight coefficient of .

[0143] Task capacity constraint C 1 for:

[0144]

[0145] Among them, remaining task volume is the task capacity of attacking drones; required task volume is the task capacity corresponding to the target with the highest priority in the current task;

[0146] UAV survivability constraints C 2 The expression is:

[0147] C 2 =sur-risk

[0148] Task adaptability C 3 The expression is:

[0149] C 3 =max(misatt-misdef,lasatt-lasdef,EMPatt-EMPdef)

[0150] Flight time C 4 The expression is:

[0151]

[0152] Among them, ||·|| 2 is the 2-norm;

[0153] Task latest time constraint C 5 The expression is:

[0154] C 5 =tasktime-latest

[0155] Among them, tasktime is the start time of the target after the current attack drone accepts the target with the highest priority.

[0156] Attack drones are only available in J R2 >0, and bid when J R2 = 0, it is considered that the bid is abandoned, so the task allocation will strictly meet C 1 constraint.

[0157] In step S6, the unassigned targets with the highest mission priority in the attack subgroup are assigned to multiple attack drones with the highest second-layer adaptability according to the defense capability of the target and the attack capability of the attack drone;

[0158] In step S7, it is determined whether there are unassigned targets among all the targets undertaken by the attack subgroup. If so, the attack capability of the attack drone that accepts the target is updated and the process returns to step S6. Otherwise, the task decision is completed.

[0159] During implementation, the task decision method of the drone cluster reconnaissance and attack integration under the condition of target uncertainty in this scheme also includes:

[0160] A1. Determine whether the reconnaissance drone detects a new target when the attack drone is performing the target mission; if so, proceed to step A2; otherwise, all attack drones continue to perform the existing mission;

[0161] A2. Use fuzzy comprehensive evaluation method to determine the task priority of new goals;

[0162] A3. Lock the target that the attack drone is executing, release all targets that the attack drone is not executing, and use the new target and all unexecuted targets as unassigned targets;

[0163] A4. Execute steps S3 to S7.

[0164] In the process of target task allocation in this scheme, there is a first-level leader in the attack subgroup as the manager of the first-level negotiation (the process of allocating tasks to the attack subgroup), and each subgroup has a second-level leader as the manager of the second-level negotiation (the process of allocating attack subgroup tasks to attack drones). The manager can be any attack drone in the attack subgroup, and the manager of the second-level negotiation can also participate in the bidding as an executor.

[0165] The effect of the task decision-making method of this scheme is explained below with reference to a specific simulation example:

[0166] A. Simulation Condition Settings

[0167] The experiments were performed in the MATLAB environment on a desktop computer with an Intel(R) Core(TM) i7-10700 CPU, 2.90 GHz and 16 GB RAM.

[0168] Without loss of generality, the simulation experiment takes heterogeneous UAV clusters as the object and simulates the scenario of heterogeneous UAV clusters coordinating ground reconnaissance and strike. The application of the decision-making layer task allocation method in UAV clusters will be given later to prove the applicability and superiority of the task decision method of this scheme. The map of this scenario is 300km×300km, including 10 targets and 6 detection warning areas. Each relative target has different position coordinates and defense configurations. The attribute parameters of each target and detection warning area are shown in Tables 2 and 3. Among the above parameters, only the target position is known, and the rest of the information needs to be obtained through reconnaissance detection.

[0169] Table 2 Target parameter settings

[0170]

[0171]

[0172] Table 3 Detection warning area parameter settings

[0173] area Center coordinates (km) Radius(km) Region 1 (165,39) 18 Area 2 (270,60) 15 Area 3 (240,210) 45 Region 4 (105,105) 21 Area 5 (186,171) 42 Area 6 (69,195) 50

[0174] Based on the above parameter settings, the real simulation map environment is as follows Figure 7 However, at the beginning of the mission, the target is considered to be hidden in the fog of war, and its number, location, and defense configuration information are unknown. Therefore, the simulation map from the perspective of the friendly heterogeneous drone cluster at the beginning is as follows: Figure 7 As shown in (b).

[0175] In the simulation experiment, the friendly heterogeneous UAV cluster contains 35 UAVs, which are divided into 4 subgroups, namely the reconnaissance subgroup, missile subgroup, laser subgroup and EMP subgroup. Among them, the missile subgroup, laser subgroup and EMP subgroup have target attack capabilities. The reconnaissance subgroup contains 5 UAVs. The attribute parameters of the reconnaissance subgroup are shown in Table 4. The three attack subgroups have different attack methods, and the attribute parameters are shown in Table 5.

[0176] Table 4. Parameter settings of reconnaissance subgroup

[0177] Drone Number Scan line width (km) Flight speed (km / h) Starting location (km) Reconnaissance No. 1 20 3600 (150，1) Reconnaissance 2 20 2400 (150，1) Reconnaissance No. 3 20 1200 (150，1) Reconnaissance 4 20 2400 (150，1) Reconnaissance No. 5 20 3600 (150，1)

[0178] Table 5 Parameter settings of three types of attack subgroups

[0179]

[0180]

[0181] B. Simulation Results and Analysis

[0182] In order to verify the effectiveness of the task decision method of this scheme, simulation verifications are carried out on reconnaissance task allocation, attack task allocation and dynamic task reallocation respectively.

[0183] (1) Regional reconnaissance mission allocation simulation

[0184] This section compares the improved CNP algorithm based on uncertain contracts with the classic CNP algorithm to prove the superiority of the improved CNP algorithm of this scheme. Figure 8 The comparison shows that the improved CNP algorithm based on uncertain contracts can adaptively adjust the size of the reconnaissance sub-area (task sub-area) according to the performance of the UAV.

[0185] The mission time of each reconnaissance drone of the two algorithms is as follows: Fig. 9 The overall reconnaissance time, average reconnaissance time and variance of the reconnaissance time of the two algorithms are shown in Fig.10 As shown. Based on the above comparison, the improved CNP algorithm based on uncertain contracts in this scheme can adaptively determine the size of the reconnaissance sub-area according to the performance of the UAV, and make the reconnaissance time allocation of each UAV more balanced. The total reconnaissance time and the average reconnaissance time of each UAV of the improved CNP algorithm based on uncertain contracts are both smaller than those of the traditional CNP algorithm, which proves that the improved CNP algorithm based on uncertain contracts in this scheme can effectively improve the efficiency of collaborative reconnaissance. The variance of the reconnaissance time of the improved CNP algorithm is much lower than that of the traditional CNP algorithm, which proves that the algorithm can effectively solve the problem of unbalanced load of UAVs.

[0186] (2) Simulation results of attack task allocation

[0187] The 10 targets discovered by reconnaissance correspond to 10 target attack tasks. First, the task priority is evaluated through the fuzzy comprehensive evaluation module. The influencing factors take into account the task urgency, task importance, task execution time and task complexity. The four attributes of each task are normalized as follows: Fig.11 shown.

[0188] The weight vectors of the four evaluation sets are set to a = [0.2, 0.3, 0.4, 0.1]. Finally, the priority evaluation results of the attack tasks are shown in Table 6. The design of the weight vector shows that the priority is most affected by the execution time, followed by the importance, then the urgency, and finally the complexity of the task. This task priority evaluation focuses more on the time cost and importance of the task. The final task evaluation results are consistent with the theoretical analysis.

[0189] Table 6 Task priority evaluation results

[0190] Task Number Task Priority Task 7 95 Task 8 92 Task 6 87 Task 4 84 Task 10 79 Task 5 79 Task 2 77 Task 1 72 Task 9 70 Task 3 68

[0191] Table 7. First-layer negotiation allocation results

[0192] Attack subgroup number Assigning a task number 1 7，10，5 2 4，2，1 3 8，6，9，3

[0193] Table 8 Attack task allocation Layer 2 negotiation results

[0194]

[0195]

[0196] Then, the CNP allocation of the two-layer negotiation is performed based on the priority evaluation results. The results of the first-layer negotiation are shown in Table 7. The second-layer negotiation is performed in parallel in each subgroup. Finally, the attack task allocation results are shown in Table 8.

[0197] During the two-level negotiation process, the two-level benefit function of each task in the three subgroups is as follows: Figure 12 to Figure 14 As shown in the figure, the attack task allocation results of the three attack subgroups are shown in Figures 15 to 17 shown.

[0198] This section compares the improved CNP algorithm with the classic CNP algorithm, and selects two evaluation indicators, namely, average attack effectiveness and allocation time. Average attack effectiveness refers to the average value of the difference between the attack capability of the drone and the defense capability of the assigned target, which can indicate the attack effect of the cluster allocation scheme. Allocation time refers to the time required for the algorithm to complete all assigned attack tasks, which can indicate the computational cost and speed of the allocation decision. The evaluation indicators for the comparison of the two algorithms are shown in Table 9. The comparison results of the normalized evaluation indicators are shown in Table 9. Fig.18 shown.

[0199] The comparison between the improved CNP algorithm and the classic CNP algorithm shows that the improved CNP algorithm has obvious advantages in average attack effectiveness and allocation time. Specifically, the average attack effectiveness of the improved CNP algorithm is 55.86% higher than that of the traditional CNP algorithm, and the task allocation time is shortened by 45.61%.

[0200] Table 9 Performance comparison of two algorithms

[0201] Evaluation Metrics Improved CNP algorithm Classic CNP algorithm Average attack effectiveness 55.60714 35.67857 Allocation time(s) 0.341 0.627

[0202] (3) Simulation results of dynamic task reallocation

[0203] This section simulates the dynamic reallocation process when the reconnaissance UAV discovers a new emergency task during patrol. The information of the new emergency task is shown in Table 10.

[0204] Table 10 Emergency task information generated by simulation

[0205]

[0206]

[0207] During the task execution, the time when the reconnaissance drone discovers the emergency task is dynamically variable. In the simulation, this time is set to the time when the entire drone cluster completes the first four tasks originally assigned (i.e., Task 7, Task 4, Task 8, and Task 6 listed in Table 5), and these four tasks no longer participate in the reallocation. The priority of the emergency task set in this section is greater than the attack task. Since the completed tasks are fixed and do not participate in the reallocation, the newly added emergency tasks are inserted at the front of the remaining task list according to the priority, so there is no need to re-evaluate the emergency tasks.

[0208] Table 11 Layer 1 negotiation redistribution results

[0209] Drone subgroup number Task Number 1 7,12,10,5 2 4,11,2,1 3 8,6,13,9,3

[0210] Table 12 Layer 2 negotiation reallocation results

[0211]

[0212] Then, the tasks are reallocated through the two-layer allocation mechanism of steps S3 to S7 of this solution. The results of the first-layer negotiation reallocation are shown in Table 11. The results of the second-layer negotiation reallocation are shown in Table 12. The reallocation algorithm was run 10 times, and the average task reallocation time was 0.404ms. The comparison of the allocation time of the initial allocation and the reallocation is shown in Table 13.

[0213] Table 13 Comparison of allocation time between initial allocation and redistribution

[0214] Initial allocation time of the two-layer negotiation CNP algorithm Double-layer negotiation CNP algorithm reallocation time 0.341s 0.404s

[0215] According to the above comparative analysis, the two-tier allocation mechanism proposed in this scheme has good dynamic allocation performance and can be reallocated according to the real-time situation changes on the battlefield. Since the reallocation process does not require task evaluation again, dynamic reallocation can be achieved with a lower time cost. The reallocation time is only 18% longer than the initial allocation time.

Claims

1. A task decision method for UAV swarm reconnaissance and attack under target uncertainty, characterized by: The drone swarm includes a reconnaissance subgroup and at least two attack subgroups. The task decision method includes the following steps: S1, receiving reconnaissance mission information, and using the improved CNP algorithm with uncertain contracts to determine the mission sub-area of ​​each reconnaissance UAV in the reconnaissance sub-group; S2, the reconnaissance drone conducts reconnaissance on its mission sub-area and determines the target in the mission sub-area; Then, the fuzzy comprehensive evaluation method is used to determine the task priorities of all targets reconnaissance by all reconnaissance UAVs; S3, the attack subgroup calculates the first level of compatibility between itself and the target with the highest priority among all unassigned targets, and assigns the target with the highest priority to the attack subgroup with the highest level of compatibility; S4, determine whether there is an unassigned target among all targets, if so, update the attack capability of the attack subgroup that accepts the target and return to step S3, otherwise go to step S5; S5, the attack drone receives all targets assigned to its attack subgroup and calculates its second-layer compatibility with the unassigned target with the highest mission priority; S6. According to the defense capability of the target and the attack capability of the attack drone, the unassigned target with the highest mission priority in the attack subgroup is assigned to multiple attack drones with the highest adaptability in the second layer; S7. Determine whether there are unassigned targets among all the targets undertaken by the attack subgroup. If so, update the attack capability of the attack drone that accepts the target and return to step S6. Otherwise, complete the task decision.

2. The task decision method according to claim 1, characterized in that: The drone cluster includes three attack subgroups, namely, a missile subgroup, a laser subgroup and an electromagnetic pulse subgroup. The properties of the attack drones in each attack subgroup are: MU Q / READ R / EU S ={Uloc,vol,sur,misatt,lasatt,EMPatt,vel} Among them, MU Q ,LU R and EU S They represent the attack drones in the missile subgroup, laser subgroup and electromagnetic pulse subgroup respectively; Uloc is the current position of the attack drone; vol is the mission capacity of the attack drone; sur is the survivability of the attack drone; misatt is the missile attack capability of the attack drone; lasatt is the laser attack capability of the attack drone; EMPatt is the electromagnetic pulse attack capability of the attack drone; vel is the flight speed of the drone; The properties of the target are: AT N ={Tloc,misdef,lasdef,EMPdef,pro,latest,need,risk} Among them, AT N is the attribute of the target; Tloc is the target location; misdef is the target's defense capability against missiles; lasdef is the target's defense capability against lasers; EMPdef is the target's defense capability against electromagnetic pulses; pro is the benefit of completing the task; latest is the latest time limit for the task; need is the number of drones required to complete the task; risk is the risk level of the task.

3. The task decision method according to claim 2, characterized in that: The expression of the highest level of adaptability is: Among them, J R1 The highest level of adaptability; is the average misattack capability of all attacking drones in the missile subgroup, is the average lasatt capability of all attack drones in the laser subgroup, is the average EMPatt capability of all attack drones in the EMP subgroup.

4. The task decision method according to claim 2, characterized in that: The expression for calculating the second-layer adaptability is: J R2 =C1·(constant+αC2+βC3-γC4-δC5) Among them, J R2 is the second-layer adaptability; C1 is the mission capacity constraint; C2 is the attack drone survivability constraint; C3 is the mission adaptability; C4 is the flight time; C5 is the mission latest time constraint; constant is a positive number; α, β, γ, δ are the weight coefficients of C2, C3, C4 and C5 respectively.

5. The task decision method according to claim 4, characterized in that: The task capacity constraint C1 is: Among them, remaining task volume is the task capacity of attacking drones; required task volume is the task capacity corresponding to the target with the highest priority in the current task; The expression of UAV survivability constraint C2 is: C2=sur-risk The expression of task adaptability C3 is: C3=max(misatt-misdef,lasatt-lasdef,EMPatt-EMPdef) The expression of flight time C4 is: Among them, ||·||2 is the 2-norm; The expression of the task latest time constraint C5 is: C5=tasktime-latest Among them, tasktime is the start time of the target after the current attack drone accepts the target with the highest priority.

6. The task decision method according to any one of claims 1 to 4, characterized in that: The method for determining the mission sub-area of ​​each reconnaissance drone in the reconnaissance sub-group includes: S11. After receiving the task information, the reconnaissance drone calculates its own benefit function according to the attributes of the reconnaissance drone and bids for the task when its task capacity meets the set requirements; S12, the reconnaissance drone with the highest profit function among all bids wins the bid and accepts the reconnaissance mission, and then updates its own mission capacity; S13. The reconnaissance drone receiving the reconnaissance mission determines the size of its mission sub-area for reconnaissance in the target potential area according to its reconnaissance performance and flight speed, wherein the mission sub-area is a rectangle; S14. According to the geometric relationship of the rectangle, the coordinates of the four vertices of the task sub-area are determined: Among them, poi e is the entry point of the task sub-region; x and y are the coordinates of the entry point of the task sub-region; poi2, poi3, poi4 are the second, third and fourth vertices in the counterclockwise direction from the entry point of the rectangular task sub-region; a is the length of the task sub-region; w i RT for the task sub-area i Width; S15, determine whether the target potential areas have been allocated to the reconnaissance drone. If so, go to step S2, otherwise, update poi4 as the entry point of the next reconnaissance mission and return to step S11.

7. The task decision method according to claim 6, characterized in that: The reconnaissance drone adopts an "S" type parallel search mode; the length of the mission sub-area is equal to the length of the target potential area, and its width N i for: Among them, f R (i) is the profit function of the reconnaissance drone; b is the width of the target potential area; Δd k For reconnaissance drone RU k Width; vel(RU k ) is a reconnaissance drone RU k Flight speed; vel(RU j ) is a reconnaissance drone RU j Flight speed; RU k and RU j are the kth and jth reconnaissance UAVs in the reconnaissance subgroup; P is the total number of reconnaissance UAVs in the reconnaissance subgroup; [·] indicates the rounding up symbol.

8. The task decision method according to claim 1, characterized in that: Methods for determining the task priorities of goals include: According to the reconnaissance and strike mission of the drone swarm, set the comment set: u={urg,imp,exe,com} Among them, urg represents the urgency of the task; imp represents the importance of the task; exe represents the execution time of the task; com represents the complexity of the task; For each evaluation factor in the evaluation set, multiple levels of evaluation are set to form an evaluation factor set; then, a weight vector of the evaluation factor set of the drone cluster's reconnaissance and strike mission is determined; The bilateral Gaussian function is used as the membership function for single factor evaluation, and the membership degree of the evaluation factor is obtained: r w =[r w1 ,r w2 ,...,r wm ],w=1,2,...,n Among them, r o is the membership degree; r w1 、r w2 、r wm The probability of making the first evaluation on the wth factor, the probability of making the second evaluation on the wth factor, and the probability of making the mth evaluation on the wth factor; n is the degree of membership of different evaluation indicators to the wth factor; The membership degree is used to construct the evaluation matrix R of the reconnaissance and attack mission of the drone cluster, and the parameter matrix B is calculated: in, is the operator; a is the weight vector; b1, b2, b p ,b m are the parameters corresponding to the first evaluation, the parameters corresponding to the second evaluation, the parameters corresponding to the pth evaluation, and the parameters corresponding to the mth evaluation; r wp is the possibility of making the p-th evaluation on the w-th factor; a w is any element in the weight vector; m is the number of elements in the evaluation factor set; According to the parameter matrix B, calculate the priority p of the drone cluster reconnaissance and attack mission: Where G = [g1, g2, ..., g m ] is the priority weight vector of each comment in the comment set; g1,g2,...,g m are the elements in the priority weight vector respectively; g j and b j are the priority weight vector of the comments in the comment set and the j-th element in the parameter matrix, respectively.

9. The task decision method according to claim 1, characterized in that: Also includes: A1. Determine whether the reconnaissance drone detects a new target when the attack drone is performing the target mission; if so, proceed to step A2; otherwise, all attack drones continue to perform the existing mission; A2. Use fuzzy comprehensive evaluation method to determine the task priority of new goals; A3. Lock the target that the attack drone is executing, release all targets that the attack drone is not executing, and use the new target and all unexecuted targets as unassigned targets; A4. Execute steps S3 to S7.

10. The task decision method according to claim 1, characterized in that: The reconnaissance subgroup and the attack subgroup are both composed of multiple drones with different performance attributes.

Images