A platoon cooperative task allocation method based on hierarchical mixed auction algorithm

Abstract

The application provides a formation cooperative task allocation method based on a hierarchical mixed auction algorithm, wherein a manned aircraft first makes a tactical decision by quantifying attribute values of tasks in a battlefield to determine priorities of different types of tasks, and then the tasks are distributed to corresponding types of unmanned aircrafts for task auction allocation solution, so as to improve the solution rate of the algorithm and the rationality of the allocation scheme, the target function is improved by introducing a coverage factor and a penalty term, the rationality of the task allocation is optimized, and a task re-allocation mechanism is formulated, namely, the manned aircraft performs real-time dynamic re-allocation of the tasks according to the residual resources of the unmanned aircrafts and the execution cost. The application takes advantage of the human intelligence of the manned aircraft pilots, can not only quickly find the unmanned aircrafts most suitable for the selected tasks, but also realize dynamic task allocation in a complex and changeable environment, balance the task load of the unmanned aircrafts, and reasonably use the unmanned aircrafts in the unmanned aircraft cluster as much as possible.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of manned / unmanned cooperative combat and combat decision-making, and belongs to the technical field of multi-aircraft cooperative air combat, and specifically relates to a manned / unmanned aircraft task allocation method based on a hierarchical decision-making mechanism and a contract net auction algorithm. BACKGROUND

[0002] At present, the technical level of unmanned aerial vehicles is still insufficient in terms of autonomous cognitive ability, communication interference and reliability when performing tasks in a dynamic and complex environment. Therefore, a manned / unmanned aircraft mixed formation cooperative combat mode is generated, in which the efficient on-site judgment and decision-making ability of the manned aircraft can complement the technical and tactical advantages of the unmanned aircraft, and the tactical level command and control of the formation is completed by the manned aircraft command and control system and the unmanned aircraft autonomous control system in cooperation, so as to fully exert the respective advantages of the two. The manned / unmanned aircraft task allocation problem (AP) is essentially a combinatorial optimization problem, which belongs to the Non-Polynomial Complete (NPC) problem. For a given combat task, combat aircraft group and weapon resources, based on battlefield situation information and combat effectiveness evaluation function, a suitable combat aircraft needs to be found for each combat sub-task to perform the task, and finally the overall combat effectiveness value of all combat aircraft performing the task needs to be maximized. The result of task allocation is affected by multiple conditions such as battlefield situation change, problem solving scale, combat effectiveness objective function and weapon resource constraints. According to the different structures of task allocation planning, multi-aircraft cooperative task allocation can be divided into centralized allocation and distributed allocation.

[0003] Centralized task allocation transmits the situation awareness information of the entire battlefield to the information processing center, generates a combat task allocation scheme for the entire combat aircraft group through the information processing center allocation strategy, and then issues the scheme to each combat aircraft for execution of the corresponding combat task. This method was first proposed and applied to classical mathematical models for solving various problems, such as linear programming, covariance control, dynamic programming and game theory. When solving simple task allocation problems, optimal solution methods such as exhaustive method and mixed integer linear programming (MILP) are usually used. The time complexity of the optimal solution method is generally high, and it is only suitable for solving small-scale allocation tasks. When the battlefield scale is large, the search space of the solution is too large, and the computational cost increases exponentially, so this method is no longer applicable. In centralized allocation, each combat aircraft needs to maintain communication with the information processing center at all times, and the limitation of the communication scale will greatly limit the size of the task that the combat aircraft group can handle, and it is easy to cause a single point failure problem of the entire combat task, that is, once the central information processing center fails, the task allocation fails, and it can no longer meet the current battlefield environment which is constantly expanding and changing.

[0004] Distributed task allocation architecture disperses the role of information processing center to different warplanes, and each warplane calculates separately, so as to reduce the demand for communication bandwidth and the processing capacity of the information processing center, overcome the shortcomings of centralized task allocation, and better meet the dynamic and complex battlefield environment. The traditional distributed task allocation algorithm is mainly based on the auction algorithm and the contract net protocol based on market mechanism. The former is based on the consistency of situation awareness of different allocation units, and the latter is based on the consistency of task space. Compared with the centralized method, the task planning time can be obviously shortened. In the cooperative task allocation of manned / unmanned aerial vehicles, the manned aerial vehicle acts as the role of contract party A, and the unmanned aerial vehicle acts as the role of competing merchant. The bidding algorithm generally includes two types. The first type is the traditional bidding algorithm, that is, there is a central system as the bidding manager responsible for this auction, which is responsible for calculating the bid of each bidder and selecting the bidder with the highest bid. The second type is to select the bidding manager from each intelligent agent participating in the bidding, and exercise the role of the central system manager. Although the bidding algorithm can solve the consistency problem, the solving process is time-consuming. There are two new consistency algorithms, which can realize fast calculation and solve the consistency problem of robust distributed algorithms for task allocation problem, which are consensus-based auction algorithm (CBAA) and consensus-based bundle algorithm (CBBA). Among them, the CBBA algorithm is a derivative of the CBAA algorithm for solving multi-task allocation problem. Both algorithms use market-based decision-making strategy as a distributed task selection mechanism, and use local communication consistency strategy as a conflict avoidance mechanism, and prove that the proposed algorithm has the characteristics of fast convergence under reasonable assumptions and network connectivity.

[0005] There are various methods for cooperative task allocation of manned / unmanned aerial vehicles, each with advantages and disadvantages. The method based on market mechanism competition adopts the bidding and bidding competition mechanism, which can coordinate the tasks among multiple unmanned aerial vehicles, has the advantages of less communication amount and good robustness. However, this kind of method also has the following problems, such as not considering the priority of the task before the task auction, and having limited processing capacity for sudden situations such as new target appearing or unmanned aerial vehicle damage during task execution. It is often limited to single-machine single-task allocation problem, and has insufficient multi-machine multi-task allocation capability. SUMMARY

[0006] In order to overcome the prior art, the application provides a formation cooperative task allocation method based on a hierarchical hybrid auction algorithm. Based on a hierarchical decision mechanism and a contract net auction algorithm, the application designs a task allocation algorithm for manned / unmanned aircraft cooperative combat, i.e. a hierarchical decision algorithm. In the task bundle construction stage, the manned aircraft first makes a tactical decision by quantifying the values of four attributes of tasks in the battlefield, i.e. value, threat, interference and unknown, determines the priority of different types of tasks, and then assigns the tasks to the corresponding type of unmanned aircraft for task auction allocation solution, thereby improving the solution rate of the algorithm and the rationality of the allocation scheme. In addition, the target function is improved by introducing a coverage factor and a penalty term, further optimizing the rationality of the task allocation. Moreover, for the sudden situation of new tasks and damaged unmanned aircraft, a task re-allocation mechanism is formulated, i.e. the manned aircraft performs real-time dynamic re-allocation of tasks according to the remaining resources of the unmanned aircraft and the execution cost.

[0007] The technical scheme adopted by the application to solve the technical problems is that before the unmanned aircraft executes a task, the manned aircraft autonomously acquires initial information of the task area. The manned aircraft makes a decision according to four attributes of the task, i.e. value, threat value, interference value and unknown value, filters out a task type that is urgently needed to be executed at present or has the highest priority in the whole combat, and generates a reasonable task auction sequence. Then, the related unmanned aircraft uses the contract net auction algorithm to bid for the task based on the improved target function on the basis of the task execution decision of the manned aircraft. The above steps are repeated until all the tasks are allocated or the unmanned aircrafts are all allocated with tasks.

[0008] The technical scheme adopted by the application to solve the technical problems is:

[0009] 1) Phase one: generating a task auction sequence

[0010] Step 1: initializing a task attribute set, an unmanned aircraft type set and a task execution constraint set;

[0011] The task type is set as four types of attack task T atc , reconnaissance task T rec , interference task T jam and decoy task T dec , and the unmanned aircraft type corresponds to four types of attack unmanned aircraft U atc , reconnaissance unmanned aircraft U rec , interference unmanned aircraft U jam and decoy unmanned aircraft U dec . The attack unmanned aircraft executes the attack task and has reconnaissance capability; the reconnaissance unmanned aircraft executes the reconnaissance task; the interference unmanned aircraft executes the interference task; the decoy unmanned aircraft mainly executes the decoy task and has reconnaissance capability; the specific task execution constraint set is shown as follows: Figure 1 ​

[0012] Step 2: initialize the number of UAVs n and the number of tasks m, and the attribute values of each task;

[0013] Set the UAV set U = {U1, U2, U3, …, U n}, containing n UAVs, and the task set T = {T1, T2, T3, …, T m}, containing m tasks; the task T j has five attributes: position (X j , Y j ), value V(T j ), threat value Th(T j ), interference value J(T j ) and unknown value X(T j ), j = 1, 2, 3, …, m, and set the attribute values corresponding to each task;

[0014] Step 3: Set the resource space of the task;

[0015] Construct the resource vector of the task Tas:

[0016]

[0017] Where, j = 1, 2, 3, …, m, m is the number of tasks, Tas j represents the resources required to complete the task T j , which has four parameters; represents the amount of ammunition required to complete the task T j , represents the amount of reconnaissance required to complete the task T j , represents the amount of interference required to complete the task T j , represents the amount of decoys required to complete the task T j ;

[0018] Step 4: Set the resource space of the UAV;

[0019] Since each UAV has limited resources that can be executed, based on the principle of resource constraints, construct the resource vector of each UAV Res as:

[0020]

[0021] Where, i = 1, 2, 3, …, n, n is the number of UAVs, Res i represents the resource vector of the UAV U i , which has five parameters; represents the amount of ammunition of the UAV U i , represents the amount of reconnaissance of the UAV Ui reconnaissance volume, U-shaped drone i The amount of bait, U-shaped drone i The amount of interference, U-shaped drone i The amount of oil;

[0022] Construct the resource vector Req required for the drone to perform its mission:

[0023]

[0024] Where i represents the drone number, j = 1, 2, 3, ..., m, and m is the number of missions. U-shaped drone i Execute task T j The required and available resources, this resource vector has five parameters; U-shaped drone i Execute task T j The amount of ammunition that can be provided U-shaped drone i Execute task T j The amount of reconnaissance that can be provided. U-shaped drone i Execute task T j The amount of interference that can be provided U-shaped drone i Execute task T j The amount of bait that can be provided U-shaped drone i Complete task T j Required amount of oil;

[0025] Step 5: Initialize the weight coefficients, normalization coefficients, and iteration counts for each attribute value;

[0026] Step 6: Calculate the total attribute values ​​of the mission within the combat area;

[0027] Step 7: Determine the priority of tasks based on the decision function;

[0028] Manned aircraft according to mission T j Value V(T) j Threat value Th(T) j Interference value J(T) j ) and unknown value X(T) j The decision function Class uses four attributes (Class, Class, Class, Class) to determine the priority of tasks. {T} for:

[0029] Class {T} =max[Kv • V score ,K th • Th score ,K j • J score ,K x • X score ] (5)

[0030] The current environment attribute is determined according to the output of the decision function, when K v • V score is the largest of the four values, the environment of the task area is determined as a high-value environment, and the UAV will first perform an attack task, and the attack-type UAV first performs the task auction; when K th • Th score is the largest of the four, the environment of the task area is defined as a high-threat environment, and the UAV will first perform a decoy task, and the decoy-type UAV first performs the task assignment; when K j • J score is the largest, the task area environment is defined as a high-interference environment, and the UAV first performs an interference task, and the interference-type UAV first performs the auction task; when K x • X score is the largest, the task area environment is defined as a high-unknown environment, and the UAV first performs a reconnaissance task, and the reconnaissance-type UAV first performs the auction task;

[0031] Step 8: Sort the UAVs and tasks in descending order according to the resources required by the tasks;

[0032] After determining the type of task to be performed first, according to the UAV task constraint relationship in Figure 1 , select the UAVs that have the ability to perform the type of task and sort them in descending order according to the UAV resources, and sort the type of task in descending order according to the task resources; other tasks of different types are auctioned according to the task type attribute value from high to low;

[0033] 2) Phase two: UAV task auction

[0034] After the manned aircraft determines the auction order of the tasks, the decision result is fed back to the UAV, which then executes the auction algorithm; before each bidding, each task will check its own resource vector Tas j , if one of the parameters in its resource vector s = 1, 2, 3, 4, it means that the demand of the task has been completed, and there is no need to bid for new UAVs for this demand;

[0035] Suppose the task sequence of UAV U i is , which represents the UAV U iThe tasks performed are T1, T2 and T3;

[0036] Step 9: Calculate the benefits of the UAV completing the task;

[0037] Benefits(U i j ) represents the benefits of the UAV U i completing the task T j ; the benefits come from three aspects: one is the value attribute V(T j ) of the target, such as the benefits generated by destroying high-value targets such as base stations and airports; two is the threat value attribute Th(T j ) of the target, such as the benefits generated by destroying high-threat targets such as enemy air defense radars or anti-aircraft guns; three is the interference value attribute J(T j ) of the target, such as the benefits generated by interfering with enemy radars or communication systems;

[0038] Calculate the task benefits Benefits(U i (T j ) of the UAV U i completing the task T j ;

[0039] Step 10: Calculate the cost of the UAV completing the task;

[0040] The cost Cost includes fuel consumption cost, time cost and destruction cost; in the uniform cruise phase of the UAV, the flight distance is close to a linear relationship with the fuel consumption, so the distance cost is used to replace the fuel consumption cost and the time cost, and the calculation formula of the cost Cost(U i (T j ) of the UAV U i completing the task T j is:

[0041] Cost(U i (T j )) = K d D(U i , T j ) + K p P i j (8)

[0042] Where D(U i , T j ) represents the distance cost of the UAV U i completing the task T j , and P i j represents the probability of the UAV U i being destroyed when completing the task T j ; ​

[0043] Step 11: Calculate the net benefit of the UAV completing the task;

[0044] Rwd(U i (T j )) represents the net benefit of the UAV U i completing the task T j , which is calculated as follows:

[0045] Rwd(U i (T j )) = Benefits(U i (T j )) - Cost(U i (T j )) (9)

[0046] Step 12: Each UAV constructs a local task sequence based on the task objectives to be allocated;

[0047] Based on the market mechanism, the UAV has self-interest and always chooses the task that is most beneficial to itself. Based on the auction order of the tasks determined by the manned aircraft, the relevant heterogeneous UAV U i bids for the task in the type of task it performs according to the net benefit Rwd(U j (T i )) after completing T j ; then establishes its own local task sequence;

[0048] Step 13: Calculate the task area size of the UAV;

[0049] The value of is used to describe the size of the task area of the UAV U i , the size of the value is proportional to the actual range of the task area of the UAV U i ; the value of is taken as the average value of the distance between all adjacent tasks in the task sequence of the UAV U i ;

[0050] Step 14: Calculate the target coverage factor;

[0051] In the task allocation process, some targets may not be in the local task sequence of the UAV U i , but within the range that can be covered by the task area of the UAV; therefore, in the following dynamic task allocation, the cost of the UAV executing these targets is lower; the coverage factor can make the UAV individuals more evenly execute the tasks of the task area, avoid concentrating in the high-value or low-cost target area, and be beneficial to the dynamic task allocation and further optimization of the target function;

[0052] Step 15: Calculate the penalty clause;

[0053] To balance the use of individuals within a drone swarm and fully utilize the resources carried by each drone, a penalty term P is proposed for the objective function:

[0054]

[0055] Where μ is a negative constant, L i For U drones i Number of tasks executed For allocation scheme X n×m The average number of missions performed by all drones in the country. U-shaped drone i task set The number of tasks contained therein; when At that time, the penalty item is 0, and no penalty is incurred; during the auction process, L i and The larger the difference, the greater the penalty to the profit obtained from the auction task. This can limit the number of tasks that the drone can perform, thereby balancing the revenue of each drone system and optimizing the overall effect of the algorithm.

[0056] Step 16: Calculate the objective function;

[0057] The benefits generated by the UAV's mission sequence are quantified using the UAV objective function G, as calculated below:

[0058]

[0059] Among them, G(X) n×m ) represents the task allocation scheme X n×m The objective function includes four components: mission benefit, mission cost, coverage factor, and penalty clause. The coverage factor mainly considers the associated costs between missions. Specifically, a new mission discovered by the UAV is associated with its own local mission sequence. If the new mission is within the UAV's combat path, the subsequent dynamic mission allocation has a lower cost to the UAV.

[0060] Step 17: Check if any new tasks have appeared in the task area. If so, return to step 8; if no new tasks have appeared, proceed to step 18.

[0061] Step 18: Determine if any drones have disappeared. If a drone has disappeared, return to step 8; if no drones have disappeared, proceed to step 19.

[0062] Step 19: Optimize the allocation algorithm; increment the iteration count by 1 after optimization.

[0063] The final goal of the allocation algorithm is to optimize the objective function and find the optimal or suboptimal task allocation scheme X of the objective function n×m The mathematical expression of the optimized objective function is as follows:

[0064] max[G(X n×m )](18)

[0065] Step 20: Determine whether the iteration number M is reached. If the iteration number M is reached, output the optimal task allocation scheme X obtained in step 19 n×m ; if the iteration number M is not reached, go to step 5.

[0066] In step 5, set the weight coefficients a1, a2, a3, and a4; wherein a1 is the weight coefficient corresponding to the value V(T j ), a2 is the weight coefficient corresponding to the threat value Th(T j ), a3 is the weight coefficient corresponding to the interference value J(T j ), a4 is the weight coefficient corresponding to the unknown value X(T j ), and a1, a2, a3, and a4>0.

[0067] Set the normalization coefficients K v , K th , K j , and K x ; wherein K v is the normalization coefficient corresponding to the value, K th is the normalization coefficient corresponding to the threat value, K j is the normalization coefficient corresponding to the interference value, and K x is the normalization coefficient corresponding to the unknown value.

[0068] Set the normalization coefficients of the task attributes to the revenue function K a , K b , and K c ; wherein K a represents the weight of the revenue function to the value, K b represents the weight of the revenue function to the threat, and K c represents the weight of the revenue function to the interference; set the normalization coefficients K d and K p ; wherein K d represents the normalization coefficient of the distance cost, and K p represents the normalization coefficient of the destroyed cost.

[0069] Set the iteration termination number M of the contract network auction algorithm. When the iteration number is reached, stop the auction and output the final task allocation result.

[0070] In step 6, the total attribute values of the tasks in the combat area are as follows:

[0071]

[0072] Where m is the sum of all missions within the operational area, and V score This indicates the total value of all missions within the operational area; Th score J represents the total threat value for all missions in the operational area; score X represents the total interference value for all missions in the operational area; score This represents the total unknown value for all missions in the operational area.

[0073] In step 9, the UAV U is calculated. i Complete task T j Task Benefits (U i (T j )) of public

[0074] The formula is:

[0075] Benefits (U i (T j ))=K a V(T j )+K b Th(T j )+K c J(T j (6)

[0076]

[0077] Among them, V(T) j ),Th(T j ), J(T j These three values ​​reflect the attributes of the current task. These attributes will change as the environment changes. 'a' is the weighting coefficient of the objective function on the benefit function.

[0078] In step 13, assuming the drone U i The task sequence is but The calculation is as follows:

[0079]

[0080] Where s represents the number of tasks, (x i ,y i ) and (x i+1 ,y i+1 D(T) represents the coordinates of the corresponding task. i ,T i+1 ) represents task T i and Task T i+1 The distance between them.

[0081] In the step 14, the coverage factor Cover is defined as:

[0082]

[0083] wherein, is the unmanned aerial vehicle U i the set of coverable tasks, sum(T) is the total number of tasks in the assignment scheme X n×m cov denotes the set of covered tasks, denotes the shortest distance between the tasks T j and the unmanned aerial vehicle U i in the local task sequence, (x i ,y i ) and (x j ,y j ) are the coordinate positions of the corresponding tasks; when , the task T j is within the coverage range of the unmanned aerial vehicle U i and belongs to the set of covered tasks T cov .

[0084] The beneficial effects of the present application are that since the auction sequence of the traditional contract net auction algorithm is randomly generated, it indicates that the auction of tasks is out of order, without considering the type of tasks and the priority of different types of tasks, which not only affects the solving rate of the algorithm but also leads to unreasonable task allocation. The manned / unmanned aerial vehicle cooperative contract net auction algorithm proposed in the present application can take advantage of the human intelligence of manned aircraft pilots, not only can quickly find the most suitable unmanned aerial vehicle for the selected task, but also can realize dynamic task allocation in complex and variable environment. By considering the coverage factor and the penalty clause, the objective function set is more conducive to maximizing the overall benefit, thereby optimizing the overall task execution efficiency. In addition, the introduction of the penalty clause can balance the task load of the unmanned aerial vehicles, and reasonably use the unmanned aerial vehicles in the unmanned aerial vehicle cluster as much as possible. In the environment with the advantage of unmanned aerial vehicle cluster, in the mode of manned / unmanned aerial vehicle cooperative combat, the present application can efficiently and quickly classify and order the allocation method of the heterogeneous unmanned aerial vehicle cluster cooperating with the manned aircraft. BRIEF DESCRIPTION OF DRAWINGS

[0085] Figure 1 Constraint relationship of four kinds of heterogeneous unmanned aerial vehicles.

[0086] Figure 2 Manned / unmanned aerial vehicle cooperative task allocation algorithm flow.

[0087] Figure 3 Combat scene initial situation diagram.

[0088] Figure 4 ​Task allocation result diagram in high value environment.

[0089] Figure 5 Task allocation total revenue curve with iteration number. DETAILED DESCRIPTION

[0090] The application will be further described below in combination with the drawings and examples.

[0091] The overall implementation process of the manned / unmanned aircraft task allocation method based on the hierarchical decision mechanism and the contract net auction algorithm is as shown in the figure. Figure 2 The technical solution will be further described clearly and completely in combination with the drawings and specific implementation cases:

[0092] 1) Phase one: generating task auction sequence

[0093] Step 1: initializing task attribute set, unmanned aircraft type set and task execution constraint set;

[0094] In the application, the task types are set as attack task T atc , reconnaissance task T rec , interference task T jam and decoy task T dec . The unmanned aircraft types are divided into attack unmanned aircraft U atc , reconnaissance unmanned aircraft U rec , interference unmanned aircraft U jam and decoy unmanned aircraft U dec . The attack unmanned aircraft mainly executes attack task and has reconnaissance capability; the reconnaissance unmanned aircraft executes reconnaissance task; the interference unmanned aircraft executes interference task; the decoy unmanned aircraft mainly executes decoy task and has reconnaissance capability. The specific task execution constraint set is as shown in the figure. Figure 1

[0095] Step 2: initializing unmanned aircraft number n and task number m, and attribute value of each task;

[0096] The unmanned aircraft set U containing n unmanned aircrafts is set as U = {U1, U2, U3,..., U n}, the task set T containing m tasks is set as T = {T1, T2, T3,..., T m}. Wherein n = 10, m = 20. Among all the 10 unmanned aircrafts, U1-U4 are set as attack unmanned aircrafts, U5-U7 are set as reconnaissance unmanned aircrafts, U8-U9 are set as decoy unmanned aircrafts, and U 10 is set as interference unmanned aircraft. According to actual combat scene, all the unmanned aircrafts in the application take off from the same initial position (1000, 1000) to execute tasks. The task T j ​(j = 1, 2, 3,..., m) have five attributes: position (X j , Y j ), value V(T j ), threat value Th(T j ), interference value J(T j ) and unknown value X(T j ), and the values of each attribute corresponding to each task are set. In the first combat scenario, different types of tasks are distinguished by different colors and shapes. Among them, the red triangle represents the attack task; the blue rectangle represents the reconnaissance task; the green circle represents the decoy task; and the yellow pentagram represents the interference task. The initial situation diagram of the specific combat scenario is shown in Figure 3 .

[0097] Step 3: Set the resource space of the task;

[0098] Construct the resource vector of the task Tas:

[0099]

[0100] Wherein, j = 1, 2, 3,..., m, m is the number of tasks, Tas j represents the resources required for completing the task T j , which has four parameters; represents the amount of ammunition required for completing the task T j , represents the amount of reconnaissance required for completing the task T j , represents the amount of interference required for completing the task T j , represents the amount of decoy required for completing the task T j . In the case of the present application, the position of the task and the first four values in the resource vector of the task are randomly generated. In this case, 20 resource vectors of tasks are randomly generated, and the specific task information is shown in Table 1:

[0101] Table 1 Initial information of each type of task

[0102]

[0103]

[0104] Step 4: Set the resource space of the UAV;

[0105] Under the actual situation of the battlefield, since the tasks executable by each UAV are limited, based on the principle of resource constraint, the resource vector Res of each UAV is constructed:

[0106]

[0107] where i = 1, 2, 3, …, n, n is the number of UAVs. Res i represents the resource vector of the UAV U i , which has five parameters; represents the oil amount of the UAV U i , represents the ammunition amount of the UAV U i , represents the reconnaissance amount of the UAV U i , represents the decoy amount of the UAV U i , represents the jamming amount of the UAV U i ; the initial information of the UAVs is shown in Table 2, in which the initial positions of the UAVs are the same, conforming to the actual environment of the battlefield, and the key resource vectors of the UAVs of the same type are different to facilitate distinction.

[0108] The resource vector Req required by the UAV to perform a task is constructed:

[0109]

[0110] where i represents the UAV number, j = 1, 2, 3, …, m, m is the number of tasks, represents the resources required and provided by the UAV U i to perform a task T j , and the resource vector has five parameters; represents the ammunition amount provided by the UAV U i to perform a task T j , represents the reconnaissance amount provided by the UAV U i to perform a task T j , represents the jamming amount provided by the UAV U i to perform a task T j , represents the decoy amount provided by the UAV U i to perform a task T j , represents the oil amount required by the UAV U i to complete a task T j ;

[0111] Table 2: Initial information of each UAV

[0112]

[0113] Step 5: Initialize the attribute value weight coefficients, normalization coefficients, and iteration times;

[0114] Set the weight coefficients a1, a2, a3, a4, respectively corresponding to the value V(T j ), threat value Th(T j ), interference value J(T j ) and unknown value X(T j ), and a1, a2, a3, a4>0. In the experiment, set a1=a2=a3=a4=1.

[0115] Set the normalization coefficients K v , K th , K j , K x ; wherein K v is the normalization coefficient corresponding to the value, K th is the normalization coefficient corresponding to the threat value, K j is the normalization coefficient corresponding to the interference value, and K x is the normalization coefficient corresponding to the unknown value. In the experiment, set K v =K th =K j =K x =0.6.

[0116] Set the normalization coefficients K a , K b , K c of the task attribute to the revenue function; wherein K a represents the weight of the revenue function to the value, K b represents the weight of the revenue function to the threat, K c represents the weight of the revenue function to the interference; set the normalization coefficients K d , K p ; wherein K d represents the normalization coefficient of the distance cost, and K p represents the normalization coefficient of the destroyed cost. In the experiment, set K a =0.5, K b =K c =K d =K p =0.8.

[0117] Set the number of iterations M of the contract net auction algorithm, when the number of iterations is reached, stop the auction, and output the final task allocation result. In the experiment, set the number of iterations M=30.

[0118] Step 6: Calculate the total attribute value of the tasks in the combat area;

[0119]

[0120] Wherein, m is the sum of all tasks in the combat area. V scoreTotal value of all tasks in the operational area; Th score Total threat value of all tasks in the operational area; J score Total interference value of all tasks in the operational area; X score Total unknown value of all tasks in the operational area.

[0121] Step 7: Determine the priority of the task according to the decision function;

[0122] The manned aircraft makes decisions according to the four attributes of the value V(T j ), the threat value Th(T j ), the interference value J(T j ) and the unknown value X(T j ) of the task T j ), and determines the priority of the task; the decision function Class {T} is:

[0123] Class {T} = max[K v ·V score ,K th ·Th score ,K j ·J score ,K x ·X score ] (5)

[0124] According to the output of the decision function, the current environment attribute is determined; when the value of K v ·V score is the largest of the four values, the environment of the task area is determined as a high-value environment, and the unmanned aerial vehicle will first perform an attack task, and the attack-type unmanned aerial vehicle will first perform a task auction; when the value of K th ·Th score is the largest of the four, the environment of the task area is defined as a high-threat environment, and the unmanned aerial vehicle will first perform a decoy task, and the decoy-type unmanned aerial vehicle will first perform a task assignment; when the value of K j ·J score is the largest, the environment of the task area is defined as a high-interference environment, and the unmanned aerial vehicle will first perform an interference task, and the interference-type unmanned aerial vehicle will first perform an auction task; when the value of K x ·X score is the largest, the environment of the task area is defined as a high-unknown environment, and the unmanned aerial vehicle will first perform a reconnaissance task, and the reconnaissance-type unmanned aerial vehicle will first perform an auction task. In the example of the present application, the battlefield attribute values K v ·V score = 96.49, K th ·Th score = 81.47, K j ·J score = 63.24, Kx ·X score = 90.58. At this time, it can be determined that the current environment is a high-value environment, and then the UAV preferentially performs the attack task.

[0125] Step 8: According to the required resources of the task, the UAVs and the tasks are ranked in descending order;

[0126] After determining the type of the first executed task, according to the UAV task constraint relationship in the middle, the UAVs suitable for performing the task of this type are sorted in descending order according to the UAV resources, and the tasks of this type are sorted in descending order according to the task resources. Other tasks of different types are sequentially auctioned according to the task type attribute value from high to low. Figure 1

[0127] 2) Phase two: UAV task auction

[0128] After the manned aircraft determines the auction order of the task, the decision result is fed back to the UAV, and the UAV executes the auction algorithm. Before each bidding, each task will check its own resource vector Tas j If a parameter in the resource vector of the task is s = 1, 2, 3, 4, it indicates that the demand of the task has been completed, and there is no need to bid for the demand to the new UAV.

[0129] Suppose the task sequence of the UAV U i is It indicates that the UAV U i executes the tasks T1, T2 and T3.

[0130] Step 9: Calculate the benefits of the UAV completing the task;

[0131] Benefits(U i (T j )) represents the benefits of the UAV U i completing the task T j ; the benefits come from three aspects: one is the value attribute V(T j ) of the target, such as the benefits generated by destroying high-value targets such as base stations and airports; two is the threat value attribute Th(T j ) of the target, such as the benefits generated by destroying high-threat targets such as enemy air defense radars or anti-aircraft guns; three is the interference value attribute J(T j ) of the target, such as the benefits generated by interfering with the enemy radar or communication system;

[0132] The calculation of the task benefits Benefits(U i (T j )) of the UAV U i completing the task T j is as follows:​

[0133] Benefits(U i j ))=K a V(T j )+K b Th(T j )+K c J(T j ) (6)

[0134]

[0135] Where, V(T j ), Th(T j ), J(T j ) three values reflect the current task attributes, when the environment changes, these attributes will also change, a is the target function weight coefficient of the benefit function.

[0136] Step 10: calculate the cost of unmanned aerial vehicle to complete the task;

[0137] The cost Cost includes fuel consumption cost, time cost, destroyed cost; the flight distance and fuel consumption of the unmanned aerial vehicle in the uniform cruise stage is close to linear relationship, the distance cost is used to replace the fuel consumption cost and time cost in the application, the unmanned aerial vehicle U i complete the task T j The cost Cost(U i (T j )) calculation formula is:

[0138] Cost(U i (T j ))=K d D(U i ,T j )+K p P i j (8)

[0139] Where, D(U i ,T j ) represents the distance cost of the unmanned aerial vehicle U i complete the task T j , P i j represent the probability of the unmanned aerial vehicle U i complete the task T j be destroyed.

[0140] Step 11: calculate the net benefit of unmanned aerial vehicle to complete the task;

[0141] Rwd(U i (T j ​) represents a UAV U i the net benefit of completing task T j . The calculation is as follows:

[0142] Rwd(U i (T j )) = Benefits(U i (T j )) - Cost(U i (T j )) (9)

[0143] Step 12: Each UAV constructs a local task sequence according to the task targets to be allocated;

[0144] Since the attack task is auctioned first, the attack UAVs fly out of the base first; and since the attack UAVs have the ability to perform reconnaissance tasks, the attack UAVs will also compete for reconnaissance tasks near their tasks; considering the case of cross-task execution, U1 first bids for tasks T3, T2, T13 with a higher price, and U2 puts tasks T5, T10, T11 into its local task queue, resulting in two reconnaissance UAVs having no tasks to perform. The algorithm of the present application adds an adaptive restriction penalty clause in the objective function, and all UAVs except UAVs U6 and U7 can execute approximately three tasks in the auction process.

[0145] Step 13: Calculate the task area size of the UAV;

[0146] The value of is used to describe the size of the task area of UAV U i , the size of the value is proportional to the actual range of the task area of UAV U i ; The value of is taken as the average value of the distance between all adjacent tasks in the task sequence of UAV U i , assuming that the task sequence of UAV U i is The calculation of is as follows:

[0147]

[0148] where s represents the number of tasks, (x i , y i ) and (x i+1 , y i+1 ) represent the coordinate positions of the corresponding tasks, and D(T i , T i+1 ) represents the distance between task T i and task T i+1 .

[0149] Step 14: Calculate the target coverage factor;

[0150] During the task allocation process, some targets may not be present on the UAV. i Local task sequence However, within the coverage area of ​​the drone's mission area, the cost of the drone performing these objectives is relatively low in the subsequent dynamic task allocation; the coverage factor can enable individual drones to perform tasks more evenly within the mission area, avoiding concentration in high-value or low-cost target areas, which is beneficial for dynamic task allocation and further optimization of the objective function.

[0151] Cover factor is defined as:

[0152]

[0153] in, It is a drone U i The set of tasks that can be covered, sum(T) is the allocation scheme X. n×m The total number of tasks in T cov Indicates the coverage task set. Indicates task T j and U drones i The shortest distance between tasks in the local task sequence, (x i ,y i ) and (x j ,y j ) represents the coordinates of the corresponding task; when At that time, task T j In drone U i Within the coverage area and belonging to the coverage task set T cov .

[0154] Step 15: Calculate the penalty clause;

[0155] To balance the use of individuals within a drone swarm and fully utilize the resources carried by each drone, a penalty term P is proposed for the objective function:

[0156]

[0157] Where μ is a negative constant, L i For U drones i Number of tasks executed For allocation scheme X n×m The average number of missions performed by all drones in the country. U-shaped drone i task set The number of tasks contained therein; when When the penalty term is 0, no penalty is generated; in the auction process, L i and The greater the difference, the greater the penalty on the profit obtained by the auction task, which can limit the number of tasks executed by the unmanned aerial vehicle, thereby balancing the benefits of each unmanned aerial vehicle system and optimizing the overall effect of the algorithm.

[0158] Step 16: Calculate the objective function;

[0159] The unmanned aerial vehicle objective function G is used to quantify the benefits generated by the unmanned aerial vehicle executing the task sequence, and is calculated as follows:

[0160]

[0161] Wherein, G(X n×m ) represents the objective function of the task allocation scheme X n×m , which includes task revenue, task cost, coverage factor and penalty clause; the coverage factor mainly considers the association cost between tasks, specifically, the new task discovered by the unmanned aerial vehicle is associated with its own local task sequence, if the new task is within the combat path of the unmanned aerial vehicle, the subsequent dynamic task allocation has a lower cost for the unmanned aerial vehicle.

[0162] Step 17: Determine whether there is a new task in the task area, if yes, return to step 8; if no, continue to the next step;

[0163] Due to the complex and variable environment of the combat task area, some targets may not be detected during the initialization detection process of manned aircraft. The unmanned aerial vehicle can discover these targets according to its own radar when executing tasks. When a new task is detected in the combat task area, the unmanned aerial vehicle that discovers the new task transmits the information of the task to the manned aircraft through broadcasting, and the manned aircraft immediately determines the type of the task and the resources required to complete the task, and then searches for unmanned aerial vehicles that can execute the task of this type (generally searches for unmanned aerial vehicles near the newly discovered task, following the principle of proximity), and then uses the auction algorithm to allocate tasks among these unmanned aerial vehicles. The specific process is that these unmanned aerial vehicles immediately check their local queues and resource remaining, and under the condition of permission, they will bid for the task according to the received information, otherwise they will not bid, and then the winner inserts the task into the appropriate position of its local task sequence according to the attribute value, or rearranges the local task queue according to the remaining resources.

[0164] Step 18: Determine whether there is an unmanned aerial vehicle disappearing, if yes, return to step 8; if no, continue to the next step;

[0165] In actual combat environment, the UAVs are at risk of being destroyed in the process of performing tasks, once the UAVs are destroyed, the tasks carried by the UAVs cannot be completed, so the manned aircraft will monitor the UAVs in a timely manner, if a UAV does not send state information to the manned aircraft within the specified time, the manned aircraft will determine that the UAV is destroyed, and the unexecuted tasks are recovered and redistributed. The distribution mechanism is similar to the task distribution mechanism when a new target is found, the manned aircraft first determines the type of unexecuted task and the resources required to complete the task, and then finds the UAVs that can perform the task of this type (generally finds in the vicinity of the newly found task, the principle of proximity), and then uses the auction algorithm to distribute the tasks among the found UAVs. The specific process is that these UAVs immediately check their local queues and resource remaining, and in the case of conditions permitting, they will bid for the tasks according to the received information, otherwise they will not bid, and then the winner inserts the tasks into the appropriate position of the local task sequence according to the attribute value, or rearranges the local task queue according to the remaining resources.

[0166] Step 19: optimization distribution algorithm, after optimization, the iteration number is increased by 1;

[0167] The final purpose of the distribution algorithm is to optimize the objective function, and find the optimal or suboptimal task distribution scheme X of the objective function n×m The mathematical expression of the optimized objective function is as follows:

[0168] max[G(X n×m )](18)

[0169] Step 20: judge whether the iteration number M is reached, if the iteration number M is reached, output the optimal task distribution scheme X obtained in step 19 n×m ; if the iteration number M is not reached, go to step 5.

[0170] The effectiveness of the hierarchical decision auction algorithm is analyzed by selecting a typical combat case, and the feasibility and effectiveness of the manned / unmanned aircraft hierarchical decision auction algorithm are proved, and the specific distribution results are shown in Table 3. At the same time, in order to show the superiority of the algorithm, the control variable method is used to limit the simulation conditions, and the algorithm is compared with the Consensus-Based Bundle Algorithm (CBBA) in terms of total task distribution income, iteration number and task distribution completion time under different combat scale dimensions.

[0171] Table 3 Task distribution results in high-value battlefield environment

[0172]

[0173]

[0174] Table 3 is the task allocation result in the high value environment, the attack unmanned aerial vehicle first allocates tasks, because it has the ability to perform reconnaissance tasks, so it will preempt the reconnaissance unmanned aerial vehicle task, and may cause some reconnaissance unmanned aerial vehicles not to be allocated tasks.

[0175] The combat scale in the multi-task allocation stage is determined by the number of warplanes N and the number of tasks M, and other information is kept unchanged. The iteration number and task allocation operation time of 6 groups of simulation results under different combat scales are collected by changing the number of warplanes N and the number of tasks M. In order to avoid the influence of random factors on the simulation results, the battlefield information parameters of the two algorithms under the same combat scale are completely consistent, and the simulation result data collection adopts the method of taking average value by multiple simulations. The simulation results are shown in Table 4:

[0176] Table 4 Task allocation time efficiency results under different combat scales

[0177]

[0178] (1) Iteration number analysis

[0179] Generally, the iteration number of the algorithm of the present application is lower than that of the CBBA algorithm. The algorithm of the present application avoids unnecessary competitive auction by sorting the priority of the task type according to the battlefield attribute and sorting the ability attribute of the unmanned aerial vehicle and the task before the task auction, thereby reducing the iteration number of the algorithm.

[0180] (2) Analysis of allocation operation time

[0181] The allocation operation time of the algorithm of the present application is lower than that of the CBBA algorithm, and the average operation time is reduced by about 67.38%. The hierarchical decision auction algorithm is based on the determination of the task type order, and the allocation operation time generally increases with the expansion of the allocation number scale. In addition, for the same reason, the sensitivity of the allocation operation time of the CBBA algorithm to the task number scale is not high, and it is greatly affected by the number of warplanes. Finally, compared with small number of task allocation, the iteration number and operation time of multi-task allocation process are greatly improved, which is mainly due to the complexity of the problem model.

[0182] The task allocation total revenue curve of the two algorithms with the iteration number is as follows: Figure 5The figure shows. Here we set the allocation scene of ten UAVs to twenty tasks. The red solid line represents the total revenue of the task with the change of the iteration number when using the hierarchical decision auction algorithm; the blue dotted line represents the total revenue of the task with the change of the iteration number when using the CBBA algorithm. It can be seen that the hierarchical decision algorithm converges faster than the CBBA algorithm with the increase of the iteration number, and the revenue almost reaches the highest when the iteration number is 25, while the iteration number of the CBBA algorithm is close to 35. This is because the task sequence of the CBBA algorithm is randomly generated, and there are two stages of auction and consistency in the task auction process, so the iteration is relatively slow. The task allocation algorithm of the present application has a reasonable task sequence generated in priority, and then the UAV is auctioned, the task is reasonably allocated in order to avoid the task conflict, that is, there is no consistency stage, so the iteration is fast. At the same time, the objective function of the algorithm adopts the coverage factor and the penalty term, so that the task allocation is more reasonable and effective, so the revenue of the task allocation will be higher than that of the CBBA algorithm. The above experimental results verify the advantages of the hierarchical decision auction algorithm, and show that the algorithm has faster convergence speed, higher revenue value, and better performance in dynamic complex environment, so that the task allocation is more reasonable and effective.

Claims

1. A platoon cooperative task allocation method based on a hierarchical hybrid auction algorithm, characterized in that It comprises the following steps: 1) phase one: generating task auction sequence Step 1: initialize the task attribute set, the unmanned aerial vehicle type set and the task execution constraint set; Setting the task type as an attack task T atc , a reconnaissance task T rec , a jamming task T jam , and a decoy task T dec Four types, the unmanned aerial vehicle type corresponds to four types of attack unmanned aerial vehicle U atc , reconnaissance unmanned aerial vehicle U rec , jamming unmanned aerial vehicle U jam , and decoy unmanned aerial vehicle U dec ; the attack unmanned aerial vehicle executes the attack task, and has the reconnaissance capability; the reconnaissance unmanned aerial vehicle executes the reconnaissance task; the jamming unmanned aerial vehicle executes the jamming task; the decoy unmanned aerial vehicle mainly executes the decoy task, and has the reconnaissance capability; Step 2: initialize the unmanned aerial vehicle quantity n and the task quantity m, and the attribute value of each task; Let U = {U1, U2, U3, ..., U4} be a set of n drones. n Let T = {T1, T2, T3, ..., Tn} be a set of m tasks. m };Task T j There are five attributes: position (X) j ,Y j ), Value V(T) j Threat value Th(T) j Interference value J(T) j ) and unknown value X(T) j ), j = 1, 2, 3, ..., m, set the attribute values ​​corresponding to each task; Step 3: set the resource space of the task; Construct the resource vector Tas of the task: wherein j = 1, 2, 3,..., m, m is the number of tasks, Tas j denotes the task T j resources required to be completed, with four parameters; denotes the task T j amount of ammunition required to be completed, denotes the task T j amount of reconnaissance required to be completed, denotes the task T j amount of interference required to be completed, denotes the task T j amount of decoy required to be completed; Step 4: set the resource space of the unmanned aerial vehicle; Since the task resources executable by each unmanned aerial vehicle are limited, based on the principle of resource constraint, the resource vector Res of each unmanned aerial vehicle is constructed as: wherein i = 1, 2, 3,..., n, n is the number of UAVs, Res i denotes the resource vector of the UAV U i , which has five parameters; denotes the ammunition of the UAV U i , denotes the reconnaissance of the UAV U i , denotes the decoy of the UAV U i , denotes the interference of the UAV U i , denotes the oil of the UAV U i ; Construct the resource vector Req required by the unmanned aerial vehicle to execute the task: wherein i represents the number of the UAV, j = 1, 2, 3,..., m, m is the number of tasks, represents the UAV U i performs the task T j the required and the available resources, the resource vector having five parameters; represents the UAV U i performs the task T j the available amount of ammunition, represents the UAV U i performs the task T j the available amount of reconnaissance, represents the UAV U i performs the task T j the available amount of jamming, represents the UAV U i performs the task T j the available amount of decoy, represents the UAV U i performs the task T j the required amount of fuel; Step 5: initialize the attribute value weight coefficient, the normalization coefficient and the iteration number; Step 6: calculate the total attribute value of the task in the combat area; Step 7: determine the priority of the task according to the decision function; Manned aircraft according to mission T j Value V(T) j Threat value Th(T) j Interference value J(T) j ) and unknown value X(T) j The decision function, Class, uses four attributes to determine task priority. {T} for: Class {T} = max[K v · V score , K th · Th score , K j · J score , K x · X score ] (5) wherein, K v is the normalized coefficient corresponding to the value, K th is the normalized coefficient corresponding to the threat value, K j is the normalized coefficient corresponding to the interference value, K x is the normalized coefficient corresponding to the unknown value, V score represents the total value of all tasks in the combat zone; Th score represents the total threat value of all tasks in the combat zone; J score represents the total interference value of all tasks in the combat zone; X score represents the total unknown value of all tasks in the combat zone, and the current environmental attribute is determined according to the output of the decision function, when K v ·V score is the largest of the four values, the environment of the task area is determined as a high-value environment, and the unmanned aerial vehicle will first perform an attack task, and the attack-type unmanned aerial vehicle first performs the task auction; when K th ·Th score is the largest of the four, the environment of the task area is defined as a high-threat environment, and the unmanned aerial vehicle will first perform a decoy task, and the decoy-type unmanned aerial vehicle first performs the task assignment; when K j ·J score is the largest, the environment of the task area is defined as a high-interference environment, and the unmanned aerial vehicle first performs an interference task, and the interference-type unmanned aerial vehicle first performs the auction task; when K x ·X score is the largest, the environment of the task area is defined as a high-unknown environment, and the unmanned aerial vehicle first performs a reconnaissance task, and the reconnaissance-type unmanned aerial vehicle first performs the auction task; Step 8: arrange the unmanned aerial vehicles and the tasks in descending order according to the resources required by the tasks; After determining the type of the task to be executed first, according to the unmanned aerial vehicle task constraint relationship in Fig. 1, the unmanned aerial vehicles with the ability to execute the task are selected and arranged in descending order according to the unmanned aerial vehicle resources, and the tasks are arranged in descending order according to the task resources; the tasks of other types are arranged in the task auction in turn according to the task type attribute value from high to low; 2) phase two: unmanned aerial vehicle task auction After the man-machine determines the auction order of the tasks, the decision result is fed back to the UAV, and the UAV executes the auction algorithm; before each bidding, each task checks its own resource vector Tas j If a parameter in its resource vector s = 1, 2, 3, 4, it indicates that the demand of the task for this item has been completed, and there is no need to bid for a new UAV for this demand. Assume the task sequence of the UAV U i is The tasks performed by the UAV U i are T1, T2 and T3; Step 9: calculate the income of the unmanned aerial vehicle to complete the task; Benefits(U i (T j )) represents the benefits of the UAV U i in completing the task T j ; the benefits come from three aspects: one is the value attribute V(T j ) of the target, such as the benefits generated by destroying high-value targets such as base stations, airports, etc.; two is the threat value attribute Th(T j ) of the target, such as the benefits generated by destroying high-threat targets such as enemy air defense radars or anti-aircraft guns; three is the jamming value attribute J(T j ) of the target, such as the benefits generated by causing interference to the enemy radar or communication system; Compute drone U i Task completion T j Task benefits Benefits(U i (T j )) Step 10: calculate the cost of the unmanned aerial vehicle to complete the task; The cost Cost includes fuel consumption cost, time cost and destruction cost; in the uniform cruise stage, the flight distance and the fuel consumption are in a linear relationship, the distance cost is used to replace the fuel consumption cost and the time cost, and the UAV U i Completes the task T j The calculation formula of the cost Cost(U i (T j )) is as follows: Cost(U i , T j ) = K d D(U i , T j ) + K p P i j (8) Wherein, D(U i , T j ) represents the distance cost of the unmanned aerial vehicle U i to complete the task T j , P i j represents the probability that the unmanned aerial vehicle U i may be destroyed to complete the task T j , K d represents the normalization coefficient of the distance cost, K p represents the normalization coefficient of the destroyed cost; Step 11: calculate the net income of the unmanned aerial vehicle to complete the task; Rwd(U i (T j ) denotes the net reward of the UAV U i for completing the task T j , calculated as follows: Rwd(U i (T j )) = Benefits(U i (T j )) - Cost(U i (T j )) (9) Step 12: each unmanned aerial vehicle constructs a local task sequence according to the task target to be allocated; Based on market mechanism, UAVs have self-interest, always choose the most beneficial task for itself, on the basis of manned aircraft to determine the auction order of the task, the relevant heterogeneous UAVs U i According to the net income Rwd(U j (T i ) after completing T j , the task bidding of the task type it performs; then establish the local task sequence respectively; Step 13: calculate the task area size of the unmanned aerial vehicle; use The value used to describe the U drone i The size of the task area, The value is related to the U drone i The actual range of the task area is directly proportional to the actual size of the task area; The value is taken as U of the drone i The average distance between all adjacent tasks in the task sequence; Step 14: calculate the target coverage factor; During task assignment, some targets can not be in the UAV's U i Local task sequence However, they are within the range covered by the UAV's task area, so the cost of the UAV to execute these targets is low in the following dynamic task assignment; the coverage factor can make the UAV individuals execute the tasks of the task area more evenly, avoid concentrating in high-value or low-cost target areas, and be conducive to dynamic task assignment and further optimization of the target function; Step 15: calculate the penalty clause; In order to balance the use of individuals in the unmanned aerial vehicle cluster and make full use of the resources carried by each unmanned aerial vehicle, a penalty term P of the objective function is proposed: Where μ is a negative constant, L i For U drones i Number of tasks executed For allocation scheme X n×m The average number of missions performed by all drones in the country. U-shaped drone i task set The number of tasks contained therein; when At that time, the penalty item is 0, and no penalty is incurred; during the auction process, L i and The larger the difference, the greater the penalty to the profit obtained from the auction task. This can limit the number of tasks that the drone can perform, thereby balancing the revenue of each drone system and optimizing the overall effect of the algorithm. Step 16: calculate the objective function; The unmanned aerial vehicle objective function G is used to quantify the benefits generated by the unmanned aerial vehicle executing the task sequence, and the calculation is as follows: wherein G(X n×m ) represents an objective function of the task allocation scheme X n×m , which includes four items of task revenue, task cost, coverage factor and penalty term; the coverage factor mainly considers the associated cost between tasks, specifically, the new task discovered by the UAV is associated with its own local task sequence, if the new task is within the battle path of the UAV, the subsequent dynamic task allocation has a lower cost for the UAV; Step 17: judge whether there is a new task in the task area, if yes, return to step 8; if there is no new task, enter step 18; Step 18: judge whether there is a disappearance of unmanned aerial vehicle, if yes, return to step 8; if there is no disappearance of unmanned aerial vehicle, enter step 19; Step 19: optimize the allocation algorithm, and add 1 to the iteration number after optimization; The final goal of the allocation algorithm is to optimize the objective function, find the optimal or suboptimal task allocation scheme X of the objective function n×m The mathematical expression of the optimized objective function is as follows: max[G(X n×m )] (18) Step 20: Determine whether the iteration number M is reached. If the iteration number M is reached, output the optimal task allocation scheme X obtained in step 19. n×m If the iteration number M is not reached, go to step 5.

2. The formation cooperative task allocation method based on the hierarchical hybrid auction algorithm according to claim 1, characterized in that: In step 5, weight coefficients a1, a2, a3, a4 are set; wherein a1 is a weight coefficient corresponding to the value V(T j ), a2 is a weight coefficient corresponding to the threat value Th(T j ), a3 is a weight coefficient corresponding to the interference value J(T j ), a4 is a weight coefficient corresponding to the unknown value X(T j ), and a1, a2, a3, a4 > 0. Setting the normalization coefficient K v ,K th ,K j ,K x ; wherein K v is the normalization coefficient corresponding to the value, K th is the normalization coefficient corresponding to the threat value, K j is the normalization coefficient corresponding to the interference value, K x is the normalization coefficient corresponding to the unknown value; Setting the normalization coefficient K of the payoff function to the task attribute a ,K b ,K c ; wherein K a represents the weight of the payoff function to the value, K b represents the weight of the payoff function to the threat, K c represents the weight of the payoff function to the interference; setting the normalization coefficient K d ,K p ; wherein K d represents the normalization coefficient of the distance cost, K p represents the normalization coefficient of the destruction cost; The iteration termination number M of the contract network auction algorithm is set, and when the iteration number is reached, the auction is stopped, and the final task allocation result is output.

3. The formation cooperative task allocation method based on the hierarchical hybrid auction algorithm according to claim 2, characterized in that: In step 6, the total attribute value of the task in the combat area is: where m is the sum of all tasks in the theater of operations, V score represents the total value of all tasks in the theater of operations; Th score represents the total threat value of all tasks in the theater of operations; J score represents the total jamming value of all tasks in the theater of operations; X score represents the total unknown value of all tasks in the theater of operations.

4. The formation cooperative task allocation method based on the hierarchical hybrid auction algorithm according to claim 2, characterized in that: In step 9, the UAV U i Completes the task T j The formula of the task benefits Benefits(U i (T j )) is: Benefits(U i (T j ))=K a V(T j )+K b Th(T j )+K c J(T j ) (6) where V(T j ), Th(T j ), J(T j ) are three values reflecting the properties of the current task, which will change when the environment changes, and a is the weight coefficient of the target function to the reward function.

5. The formation cooperative task allocation method based on the hierarchical hybrid auction algorithm according to claim 1, characterized in that: In step 13, assuming the task sequence of the UAV U i is then the calculation is as follows: where s represents the number of tasks, (x i ,y i ) and (x i+1 ,y i+1 ) represent the coordinate positions of the corresponding tasks, and D(T i ,T i+1 ) represents the distance between task T i and task T i+1 .

6. The formation cooperative task allocation method based on the hierarchical hybrid auction algorithm according to claim 1, characterized in that: In step 14, the cover factor Cover is defined as: wherein, is a UAV U i a coverable task set, sum(T) is the total number of tasks in the assignment scheme X n×m cov denotes a coverable task set, denotes a task T j and a UAV U i the shortest distance of a task in the local task sequence, (x i ,y i ) and (x j ,y j ) are the coordinate positions of the corresponding tasks; when , the task T j is within the coverage range of the UAV U i and belongs to the coverable task set T cov .​ 7. An electronic device, comprising: comprising: one or more processors; memory; one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the method of any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores program codes, which can be invoked by the processor to execute the method of any one of claims 1-6.

Images