CN108549402A - Unmanned aerial vehicle group method for allocating tasks based on quantum crow group hunting mechanism - Google Patents

Abstract

The present invention relates to a kind of unmanned aerial vehicle group method for allocating tasks based on quantum crow group hunting mechanism, including：Establish the unmanned aerial vehicle group Task Assignment Model from multiple starting points to multiple tasks, including unmanned plane model number, starting and terminal point and distribution model；Initialize quantum crow group；Fitness calculating carried out to every quantum crow according to fitness function, the position of the corresponding quantum crow of calculated fitness function minimum value saves as global optimum's food position；Update the quantum position and position of every quantum crow；Fitness calculating is carried out to every quantum crow according to fitness function, determine the hiding food position of every quantum crow, optimal food position so far is found simultaneously, and global optimum's food position is exported if reaching greatest iteration algebraically, is mapped as task allocation matrix.The present invention solves discrete multiple constraint object function Solve problems, and discrete quantum crow algorithm has fast convergence rate, the high advantage of convergence precision as Evolution Strategies.

Description

Task allocation method for UAV swarm based on quantum crow swarm search mechanism

technical field

The invention relates to a task allocation method for unmanned aerial vehicles, in particular to a task allocation method for unmanned aerial vehicles based on a quantum crow swarm search mechanism, which belongs to the field of autonomous control of unmanned aerial vehicles.

Background technique

UAV is also called unmanned aerial vehicle (Unmanned Aerial Vehicle, UAV). During its use, it does not need to carry an operator, provides lift with aerodynamic force, and can fly autonomously through remote control or under the control of a predetermined program. Fly, and perform specific tasks by carrying mission equipment. UAVs have the advantages of small size, flexible use, good concealment, and strong adaptability. They can complete some specific tasks and tasks that humans cannot reach and engage in in various harsh, dangerous, and extreme environments. The development, production and use costs of UAVs are much lower than those of manned aircraft, so they have broad application space in both military and civilian fields.

UAV task allocation is one of the key technologies of UAV autonomous control, and it is an important factor for UAV to realize intelligent, autonomous flight and task execution. UAV task assignment means that in the whole process of task execution, a certain task assignment method is used to determine whether to perform tasks and what tasks to perform for UAVs. Reasonable task assignments can ensure that the cost of UAVs is minimized on the one hand. On the other hand, it can best complete various tasks.

After searching the existing technical literature, it was found that Tang Chuanlin et al.

(2011, Vol.18, No.10, pp.28–31) published a task allocation model in "Multi-UCAV Ground Attack Target Allocation Based on Game Theory", using game theory algorithms to seek optimal task allocation, However, the algorithm model is complex, the accuracy is not high enough, and the amount of calculation is large. Mehmet "Approximating the optimal mapping for weapon target assignment by fuzzy reasoning" published in "Information Sciences" (2014, Vol.255, No.10, pp.28–31) uses fuzzy reasoning to solve the weapon-target assignment problem, However, the reasoning process is complicated, the calculation load is large, and the practicability is not high. With the development of intelligent heuristic computing technology, intelligent optimization technology has been applied to the problem of multi-UAV task assignment. In "Control and Decision Making" (2010, Vol.25, No.9, pp.1359-1364), "Multi-UAV Task Allocation Method Based on Particle Swarm Algorithm" was published by Li Wei et al. The task allocation problem of UAV, but the particle swarm algorithm is easy to fall into local optimum, and the convergence accuracy needs to be improved. In the "Acta Aerospace Sciences" (2014, Vol.25, No.9, pp.1626-631) proposed by Li Yan et al. "Cooperative air combat firepower allocation based on SA-DPSO hybrid optimization algorithm" combines simulated annealing algorithm and particle swarm optimization algorithm Combined with UAV task allocation, this method has a better convergence speed, but it is easy to fall into the curse of dimensionality, and the optimization performance is not enough.

Because the above UAV task assignment methods are nonlinear solution methods, it is very easy to fall into local extremum during the solution process, and it is difficult to obtain the global optimal solution. However, the existing UAV task assignment methods seldom consider various evaluation indicators and constraints comprehensively in the task assignment of UAV swarms, so their application range is limited. From this point of view, it is very valuable to find new task allocation methods to improve the performance of UAV combat.

Contents of the invention

In view of the above-mentioned prior art, the technical problem solved by the present invention is to provide a UAV swarm task assignment method based on the quantum crow swarm search mechanism that considers multiple starting points and multiple destinations and is suitable for discrete problems.

In order to solve the above-mentioned technical problems, the present invention provides a method for assigning unmanned aerial vehicle swarm tasks based on the quantum crow swarm search mechanism, comprising the following steps:

Step 1: Initialize the maximum iteration algebra as T _max , and establish a UAV group task assignment model from multiple starting points to multiple tasks: Assume that there are U types of UAVs that perform Q tasks from M starting points;

Let the coordinates of the mth starting point of the UAV be Where 1≤m≤M, the coordinates of the qth task of the UAV are Among them, 1≤q≤Q, all UAVs are divided into L types according to the starting point and model, where L=U×M, that is, the same type of UAV has the same starting point and belongs to the same model;

According to the starting point of the model of the l-th UAV, the coordinates of the starting point of the l-th UAV are obtained as Where l=1,2,...,L, then the distance between the starting point of the l-type UAV and the q-th task is D _l,q and satisfies:

The task assignment matrix of the UAV group is represented by the assignment matrix A={a _l,q |a _l,q ∈{0,1}} _L×Q with L rows and Q columns. tasks, then a _l,q ＝1, otherwise a _l,q ＝0;

Assuming that each UAV has D types of weapons, the probability that the l-th UAV uses the d-th weapon is The cost of using the d-th weapon for the l-type UAV is δ _l,d , where 1≤d≤D, 1≤l≤L, and the lethality rate of the d-th weapon for the q-th task is Where 1≤d≤D, 1≤q≤Q, the damage probability matrix of the UAV is set P={P _l,q,d |P _l,q,d ∈[0,1]} _L×Q×D , P _l,q,d is the damage degree of the l-type UAV to the q-th task using the d-th weapon and it satisfies: The damage threshold of the qth task is W _q , where 1≤q≤Q, assuming the value of the qth task is V _q , the number of the l-type UAVs is B _l , and the attack on the qth task The maximum formation number of UAVs is C _q , the maximum range of type l UAVs is R _l , the maximum range of all UAVs is O _max , the flight speed of type l UAVs is Z _l , and all unmanned The maximum flight time of the aircraft is Z _max ;

The UAV task allocation model is represented by target value benefit function, flight distance function, ammunition consumption cost function and target coverage function respectively:

(1) The normalized target value benefit function is: Among them, A is the task assignment matrix, A＝{a _l,q |a _l,q ∈{0,1}} _L×Q , P _l,q,d is the type l UAV using d weapons to attack the qth The damage probability of a task, d is the type of weapon used, the type of weapon used by the l-th type UAV needs to be set in advance, V _q is the value of the qth task, N is the number of people who actually participate in the task in the task allocation number of man-machines, is the maximum task value, and max is the function to obtain the maximum value;

(2) The normalized flight distance function is: Where λ ₁ , λ ₂ are the weights of the two factors, λ ₁ + λ ₂ = 1, λ ₁ , λ ₂ ≥ 0, is the length of the longest path, R _l is the maximum range of the l-type unmanned aerial vehicle;

(3) The normalized ammunition consumption cost function is: Among them, the type of weapon used by the l-type UAV needs to be set in advance, δ _max is the maximum cost,

(4) The normalized target coverage function is:

The UAV swarm task assignment model satisfies the following constraints:

(1) Mission force constraints: the number of dispatched drones of each type cannot exceed the number of owned drones of this type,

(2) Combat radius constraint of UAV: ensure that the flight distance of UAV is within its combat radius, a _l,q ×(D _l,q -R _l )≤0(l=1,2,... ,L;q=1,2,...,Q);

(3) Constraints on the damage degree of the target: the damage degree of the UAV performing the task q to the task q is not less than the damage threshold of the task, P _{l, q, d} are the damage probability of the qth mission using d weapons for the l-type UAV, d is the type of weapon used, and the type of weapon used by the l-type UAV needs to be set in advance;

(4) Constraint on the number of drones attacking the target: the number of drones attacking the qth task does not exceed the maximum number of formations, that is,

(5) Constraints on the range of the attack target: that is, the range of the attack mission does not exceed the given maximum range, O _max is the maximum range of all drones;

(6) Constraint on the time of attacking the target: the time of attacking the task does not exceed the given maximum time, Z _max is the maximum flight time of all drones;

Determine the population size K of the quantum crow group, the dimension of the optimization problem J=L×Q, arrange the UAV group task assignment matrix A={a _l,q |a _l,q ∈{0 ,1}} The elements in _L×Q , use Correspondingly record the elements in the UAV swarm task assignment matrix A;

Step 2: Initialize the quantum crow group:

Set the quantum position of the i-th quantum crow Each dimension of is set to Among them, 1≤i≤K, 1≤j≤J, and measure the quantum position of the i-th quantum crow to obtain the position of the i-th quantum crow Initialize the hidden food position of the i-th quantum crow as Where 1≤i≤K, t is the number of iterations, initially set t=0;

For the j-th dimension of the quantum position of the i-th quantum crow Make a measurement to get the jth dimension of the position of the i-th quantum crow Where 1≤i≤K, 1≤j≤J, is a random number that satisfies the uniform distribution;

Step 3: Calculate the fitness of each quantum crow according to the fitness function, and store the position of the quantum crow corresponding to the minimum value of the calculated fitness function as the global optimal food position

Step 4: Update the quantum position and position of each quantum crow:

The i-th quantum crow randomly selects another quantum crow s in the quantum crow group, and then follows the quantum crow s to find the food location hidden by the quantum crow s. Quantum crow i updates the quantum position through strategy 1, otherwise quantum crow i updates the quantum position through strategy 2;

Strategy 1 satisfies: the i-th quantum crow updates its position according to the hidden food position of the quantum crow s, and the update equation of the j-th dimension quantum rotation angle of the i-th quantum crow is: Among them, e ₁ is a constant, which determines the degree of influence on the evolution of the quantum crow by guiding the position of the quantum crow, and H is the flight length;

Strategy 2 satisfies: the i-th quantum crow updates its position according to its own hidden food position and the optimal food position, and the update equation of the j-th dimension quantum rotation angle of the i-th quantum crow is Among them, e ₂ and e ₃ are constants, which determine the influence degree of guiding the position of the quantum crow to the evolution of the quantum crow;

The evolution process of the quantum position is as follows:

Among them, ζ=0.15/J is the mutation probability, and abs() is the absolute value function;

The position of the quantum crow is obtained by measuring the quantum position of the quantum crow, and the measurement rules are as follows:

Where 1≤i≤K, 1≤j≤J, is a random number that satisfies the uniform distribution;

Step 5: Calculate the fitness of each quantum crow according to the fitness function, determine the hidden food position of each quantum crow, and find the optimal food position up to this iteration algebra;

Set the position of the i-th quantum crow Assign a value to the task allocation matrix A, according to Perform fitness calculations;

Use the greedy selection strategy to select the hidden food position of the quantum crow, if but otherwise

Step 6: If the maximum iteration number T _max is reached, the algorithm is terminated, and step 7 is executed; otherwise, set t=t+1, and return to step 4 to continue;

Step 7: Output the global optimal food position and map it into a task assignment matrix.

The present invention is a kind of unmanned aerial vehicle swarm task assignment method based on quantum crow swarm search mechanism, also includes:

The fitness evaluation process in Step 3 is as follows: First, the position of the i-th quantum crow in the t-th generation Assign a value to the task assignment matrix A, where the j-th dimension of the i-th quantum crow of the t-th generation Assigned to a _{l, q} , according to Perform fitness calculations, where

c ₁ , c ₂ , c ₃ , c ₄ , c ₅ , c ₆ are penalty coefficients, ω ₁ , ω ₂ , ω ₃ , ω ₄ are weighting factors, ω ₁ + ω ₂ + ω ₃ + ω ₄ = 1, 0≤ω ₁ ,ω ₂ ,ω ₃ ,ω ₄ ≤1, max is the function to obtain the maximum value, and min is the function to obtain the minimum value.

Beneficial effects of the present invention: the present invention aims at the deficiencies of the existing UAV swarm task assignment method, proposes a UAV swarm task assignment model considering multiple starting points and multiple destinations, and proposes a quantum crow group suitable for discrete problems The search mechanism is used to solve the task assignment problem of UAV swarms. Compared with the prior art, the present invention fully considers the situation of performing multiple tasks from multiple starting points encountered in the process of task assignment of the UAV group, and simultaneously considers the target value benefit function, flight distance function, and ammunition consumption cost. Function and target coverage Function multiple targets, with the following advantages:

(1) The present invention solves the problem of solving discrete multi-constrained objective functions, and designs a novel discrete quantum crow algorithm as an evolutionary strategy, utilizes linear weights to process different objective functions, and the method designed has the advantages of fast convergence speed and high convergence precision advantage.

(2) Compared with the existing UAV swarm task assignment method, the present invention can effectively solve the multi-target constraint requirement for the UAV swarm, which shows that the method has wider applicability.

(3) The simulation results show that the UAV swarm task assignment method proposed in the present invention can obtain a more reasonable UAV task assignment scheme than the Particle Swarm Optimization (PSO) method, thus illustrating the effectiveness of the method sex.

Description of drawings

Figure 1 is a flow chart of UAV swarm task assignment;

Fig. 2 is the flowchart of quantum crow position update;

Figure 3 shows the convergence curves of the two methods for multi-UAV task assignment.

Detailed ways

The specific embodiments of the present invention will be described below in conjunction with the accompanying drawings.

As shown in Figure 1, the technical solution of the present invention comprises the following steps:

Step 1: Establish a UAV group task assignment model from multiple starting points to multiple tasks, assuming that there are U types of UAVs performing Q tasks from M starting points.

Let the coordinates of the mth starting point of the UAV be Where 1≤m≤M, the coordinates of the qth task of the UAV are where 1≤q≤Q. All UAVs can be divided into L types according to their origin and model, where L=U×M, that is, UAVs of the same type have the same origin and belong to the same model.

According to the starting point of the type l UAV, the coordinates of the starting point of the lth type UAV can be obtained as Where l=1,2,...,L, then the distance D _l,q between the starting point of the l-type UAV and the q-th task is

The task assignment matrix of the UAV group can be expressed by the assignment matrix of L rows and Q columns A={a _l,q |a _l,q ∈{0,1}} _L×Q , if the lth type of UAV executes the qth task, then a _l,q =1, otherwise a _l,q =0.

Assuming that each UAV has D types of weapons, the probability that the l-th UAV uses the d-th weapon is The cost of the l-type UAV using the d-th weapon is δ _l,d , where 1≤d≤D, 1≤l≤L. The lethality rate of the d-th weapon to the q-th task is Where 1≤d≤D, 1≤q≤Q, the damage probability matrix of the UAV is set P={P _l,q,d |P _l,q,d ∈[0,1]} _L×Q×D , P _l,q,d is the damage degree of the l-type UAV using the d-th weapon to the q-th task. The damage threshold of the qth task is W _q , where 1≤q≤Q. Suppose the value of the qth task is V _q , the number of the l-type UAVs is B _l , the maximum number of formations of UAVs attacking the qth task is C _q , and the maximum range of the l-type UAVs is is R _l , the maximum range of all UAVs is O _max , the flight speed of type l UAV is Z _l , and the maximum flight time of all UAVs is Z _max .

The UAV task assignment model can be expressed by target value benefit function, flight distance function, ammunition consumption cost function and target coverage function.

(1) The normalized target value benefit function is: Among them, A is the task assignment matrix, A＝{a _l,q |a _l,q ∈{0,1}} _L×Q , P _l,q,d is the type l UAV using d weapons to attack the qth The damage probability of a task, d is the type of weapon used, and the type of weapon used by the l-type UAV needs to be set in advance. V _q is the value of the qth task, N is the number of UAVs that actually participate in the task in the task allocation, is the maximum task value, and max is the function to obtain the maximum value.

(2) The normalized flight distance function is: Where λ ₁ , λ ₂ are the weights of the two factors, λ ₁ + λ ₂ = 1, λ ₁ , λ ₂ ≥ 0, is the length of the longest path, R _l is the maximum range of type l UAV.

(3) The normalized ammunition consumption cost function is: Among them, the type of weapon used by the l-type UAV needs to be set in advance, δ _max is the maximum cost,

(4) The normalized target coverage function is:

In addition, the following constraints should also be met:

(1) Mission force constraints. That is, the number of dispatched drones of each type cannot exceed the number of owned drones of this type.

(2) UAV combat radius constraints. That is, it is necessary to ensure that the flight distance of the UAV is within its combat radius. a _l,q ×(D _l,q −R _l )≦0 (l=1,2,...,L; q=1,2,...,Q).

(3) Constraints on target damage. That is, the damage degree of the UAV performing the task q to the task q should not be less than the damage threshold of the task. P _{l, q, d} are the damage probability of the qth mission using d weapons for the l-type UAV, and d is the type of weapon used. The type of weapon used by the l-type UAV needs to be set in advance.

(4) Constraints on the number of UAVs attacking the target. The number of UAVs attacking the qth task does not exceed its maximum formation number, that is,

(5) Constraints on the range of the attack target. That is, the range of the attack mission does not exceed the given maximum range. O _max is the maximum range of all drones.

(6) Constraints on the time to attack the target. That is, the time of the attack task does not exceed the given maximum time. Z _max is the maximum flight time of all drones.

Then, determine the population size K of the quantum crow group, the dimension of the optimization problem J=L×Q, and arrange the UAV group task assignment matrix A={a _l,q |a _l,q ∈{ 0,1}} The elements in _L×Q , use Correspondingly record the elements in the UAV swarm task assignment matrix A.

Step 2: Initialize the quantum crow group.

Set the quantum position of the i-th quantum crow Each dimension of is set to Among them, 1≤i≤K, 1≤j≤J, and measure the quantum position of the i-th quantum crow to obtain the position of the i-th quantum crow Initialize the hidden food position of the i-th quantum crow as where 1≤i≤K. t is the number of iterations, initially set t = 0.

For the j-th dimension of the quantum position of the i-th quantum crow Make a measurement to get the jth dimension of the position of the i-th quantum crow Where 1≤i≤K, 1≤j≤J, is a random number that satisfies a uniform distribution.

Step 3: Calculate the fitness of each quantum crow according to the fitness function, and store the position of the quantum crow corresponding to the minimum value of the calculated fitness function as the global optimal food position

The fitness evaluation process is as follows:

First, the position of the i-th quantum crow in the t-th generation Assign a value to the task assignment matrix A, where the j-th dimension of the i-th quantum crow of the t-th generation Assigns to a _l,q . according to Perform fitness calculations, where

c ₁ , c ₂ , c ₃ , c ₄ , c ₅ , c ₆ are penalty coefficients, ω ₁ , ω ₂ , ω ₃ , ω ₄ are weighting factors, ω ₁ + ω ₂ + ω ₃ + ω ₄ = 1, 0≤ω ₁ ,ω ₂ ,ω ₃ ,ω ₄ ≤1, max is the function to obtain the maximum value, and min is the function to obtain the minimum value.

Step 4: Update the quantum position and location of each quantum crow.

As shown in Figure 2, the i-th quantum crow randomly selects another quantum crow s in the quantum crow group, and then follows the quantum crow s to find the food position hidden by the quantum crow s. The quantum crow s finds that the perceived probability of being followed is μ. like Quantum crow i updates the quantum position through strategy 1, otherwise quantum crow i updates the quantum position through strategy 2.

Strategy 1: The i-th quantum crow updates its position according to the hidden food position of the quantum crow s. The update equation of the quantum rotation angle of the jth dimension of the i-th quantum crow is Among them, e ₁ is a constant, which determines the influence degree of guiding the position of the quantum crow to the evolution of the quantum crow, and H is the flight length.

Strategy 2: The i-th quantum crow updates its position according to its hidden food position and the optimal food position. The update equation of the quantum rotation angle of the jth dimension of the i-th quantum crow is Among them, e ₂ and e ₃ are constants, which determine the influence degree of guiding the position of the quantum crow to the evolution of the quantum crow.

The evolution process of the quantum position is as follows:

Among them, ζ=0.15/J is the mutation probability, and abs() is the function for obtaining the absolute value.

The position of the quantum crow is obtained by measuring the quantum position of the quantum crow. The measurement rules are as follows:

Where 1≤i≤K, 1≤j≤J, is a random number that satisfies a uniform distribution.

Step 5: Calculate the fitness of each quantum crow according to the fitness function, determine the hidden food position of each quantum crow, and find the optimal food position so far.

Set the position of the i-th quantum crow Assign a value to the task assignment matrix A. according to Perform fitness calculations.

Use the greedy selection strategy to select the hidden food position of the quantum crow, if but otherwise

Step 6: If the maximum number of iterations is reached, the algorithm is terminated and Step 7 is executed; otherwise, set t=t+1, and return to Step 4 to continue.

Step 7: Output the global optimal food position and map it into a task assignment matrix.

Specific examples are as follows:

Its model parameters are set as follows:

The number of UAV models U=4, the number of UAV starting points M=3, the coordinates of the starting points are (368,319,150), (264,44,264) and (296,242,347.5), the number of UAV missions Q=10, the first The coordinates of the task are (264,715,800), the task value is 5, and the damage threshold is 0.5; the coordinates of the second task are (225,605,670), the task value is 5, and the damage threshold is 0.5; the coordinates of the third task are (168,538,340), the task value is 2, and the damage threshold is 0.5; the coordinates of the fourth task are (180,455,670), the task value is 1, and the damage threshold is 0.5; the coordinates of the fifth task are (120,400,600), The task value is 2, and the damage threshold is 0.5; the coordinates of the sixth task are (96,304,233), the task value is 5, and the damage threshold is 0.5; the coordinates of the seventh task are (10,451,233), and the task value is 5 , the damage threshold is 0.5; the coordinates of the eighth task are (162,660,233), the task value is 5, the damage threshold is 0.5, the coordinates of the ninth task are (110,561,45), the task value is 5, the damage The degree thresholds are all 0.5; the coordinates of the 10th task are (105,473,1830), the task value is 5, and the damage thresholds are all 0.5. UAV weapon types D = 2, wherein the first type and the second type of UAV use the second type of weapon, the third type and the fourth type of UAV use the first type of weapon, the first The cost of the first type of weapon is 5 units, and the cost of the second type of weapon is 3 units. The probability of the first type of drone choosing the first type of weapon is 0.67, and the probability of choosing the second type of weapon is 0.78; The probability of choosing the first weapon for a type of drone is 0.67, and the probability of choosing the second weapon is 0.78; the probability of choosing the first weapon for the third type of drone is 0.92, and the probability of choosing the second weapon is 0.92. The probability is 0.92; the probability of choosing the first weapon for the fourth type of drone is 0.92, and the probability of choosing the second weapon is 0.92. The kill rate of the first weapon to the first task and the second task is 0.92; the kill rate of the first weapon to the third task, the fourth task and the fifth task is 0.8; The kill rate of the 6th task, the 7th task and the 8th task is 0.94; the kill rate of the 1st weapon to the 9th task and the 10th task is 0.6. The kill rate of the second weapon for the first task, the second task, the third task, the fourth task and the fifth task is 0.8; the second weapon is for the sixth task and the seventh task The kill rate of the 8th task and the 8th task is 0.92; the kill rate of the 2nd weapon to the 9th task is 0.97; the kill rate of the 2nd weapon to the 10th task is 0.6. The number of Type 1 drones is 5, and the maximum range is 300. The number of Type 2 drones is 6, and the maximum range is 900. The number of Type 3 drones is 6, and the maximum range is 900. The number of Type 4 drones is 15, and the maximum range is 1700. The number of Type 5 drones is 3, and the maximum range is 300. The number of Type 6 drones is 5, and the maximum range is 900. The number of Type 7 drones is 6 and the maximum range is 900. The number of Type 8 drones is 4 and the maximum range is 1700. The number of Type 9 drones is 5, and the maximum range is 300. The number of Type 10 drones is 10, and the maximum range is 900. The number of Type 11 drones is 5 and the maximum range is 900. The number of Type 12 drones is 10, and the maximum range is 1700. The maximum formation number of UAVs attacking missions is 8. Weights λ ₁ =1, λ ₂ =0, objective function weights ω ₁ =0.322, ω ₂ =0.214, ω ₃ =0.1856, ω ₄ =0.2784. Penalty coefficient c ₁ =c ₂ =c ₃ =c ₆ =50, c ₄ =c ₅ =0. For the above coordinates, the unit of the voyage is km.

The parameters of the UAV swarm task assignment method based on the quantum crow swarm search mechanism are set as follows: the population size K=20, the maximum number of iterations is 200, the perception probability μ=0.1, the degree of influence on the evolution of the quantum crow e ₁ =0.06, e ₂ =0.03, e ₃ =0.01, flight length H=2.

For the parameter setting of the UAV swarm task assignment method based on the particle swarm optimization algorithm, see "Multiple Unmanned Aerial Vehicles Based on Particle Swarm Optimization Algorithm" published in "Control and Decision" (2010, Vol. Man-Machine Task Allocation Method", other parameters are the same as those of the UAV swarm task allocation method based on the quantum crow swarm search mechanism.

As shown in Figure 3, under the above parameter setting conditions, the convergence curves for the two methods to achieve multi-UAV task allocation, the present invention has a faster convergence effect.

The results of the UAV swarm task assignment method based on the quantum crow swarm search mechanism are shown in the table:

Table 1 Model allocation of UAVs corresponding to tasks at each starting point

Among them, M1 represents the first starting point, M2 represents the first starting point, and M3 represents the third starting point. Q1 to Q10 represent the 1st to 10th tasks respectively. U1 means UAV type 1, U2 means UAV type 2, U3 means UAV type 3, U4 means UAV type 4, and 0 means there is no UAV to perform the task from the starting point.

The invention solves the problem that the traditional algorithm has a slow search speed and a large amount of calculation, and it is difficult to find the optimal task assignment of the UAV group, and the existing UAV group task assignment design based on intelligent computing seldom considers various evaluation indicators and constraints comprehensively. , its scope of application is limited. A task assignment model considering UAV swarm is proposed, and a discrete quantum crow swarm search mechanism is proposed to solve the task assignment problem of UAV swarm. The steps required for this method are as follows: the first step is to establish a UAV swarm task allocation model from multiple starting points to multiple tasks, including the number of UAV models, starting point and ending point, and allocation model. The second step is to initialize the quantum crow group. The third step is to calculate the fitness of each quantum crow according to the fitness function, and store the position of the quantum crow corresponding to the minimum value of the calculated fitness function as the global optimal food position. The fourth step is to update the quantum position and location of each quantum crow. The fifth step is to calculate the fitness of each quantum crow according to the fitness function, determine the hidden food position of each quantum crow, and find the optimal food position so far, and output the global optimal food if it reaches the maximum iteration algebra location, mapped to a task assignment matrix.

Claims (2)

1. a kind of unmanned aerial vehicle group method for allocating tasks based on quantum crow group hunting mechanism, which is characterized in that including following step Suddenly：

Step 1：Initialization greatest iteration algebraically is T_max, establish the unmanned aerial vehicle group task distribution from multiple starting points to multiple tasks Model：Assuming that there is the unmanned plane of U kind models to execute Q task from M starting point；

If the coordinate of m-th of starting point of unmanned plane isWherein 1≤m≤M, q-th of task of unmanned plane Coordinate isWherein 1≤q≤Q is divided into L type to all unmanned planes according to starting point and model, wherein L=U × M, i.e. same type unmanned plane starting point having the same and belong to same model；

According to the starting point of the model of l type unmanned planes, the starting point coordinate for obtaining l type unmanned planes is Wherein l=1,2 ..., L, then the starting point of l types unmanned plane is D at a distance from q-th of task_l,qAnd meet：

The task allocation matrix of unmanned aerial vehicle group L row Q row allocation matrix A={ a_l,q|a_l,q∈{0,1}}_L×QIt indicates, if first of type Number unmanned plane execute q-th of task, then a_l,q=1, otherwise a_l,q=0；

If each unmanned plane has the D kind weapons, l type unmanned planes to be using the probability of d kind weaponsL type unmanned planes make Cost with d kind weapons is δ_l,d, wherein 1≤d≤D, 1≤l≤L, the killing rate of d kinds q-th of task of weapon pair areWherein 1≤d≤D, 1≤q≤Q, if unmanned plane injures probability matrix P={ P_l,q,d|P_l,q,d∈[0,1]}_L×Q×D, P_l,q,dIt is degree of injuring and satisfaction of the l types unmanned plane using d kind q-th of task of weapon pair： Degree of the injuring threshold value of q-th of task is W_q, wherein 1≤q≤Q, if the value of q-th of task is V_q, possess l type unmanned planes Quantity is B_l, the unmanned plane maximum formation number to the attack of q-th of task is C_q, the ultimate run of l type unmanned planes is R_l, entirely The ultimate run of portion's unmanned plane is O_max, the flying speed of l type unmanned planes is Z_l, the maximum flight time of whole unmanned planes is Z_max；

Unmanned plane Task Assignment Model is respectively by target value revenue function, flying distance function, consumption bullet amount cost function and mesh Mark coverage rate function representation：

(1) normalization target value revenue function is：Wherein A is task Allocation matrix, A={ a_l,q|a_l,q∈{0,1}}_L×Q, P_l,q,dQ-th of task is injured using d kinds weapon for l types unmanned plane Probability, d are set in advance for the type needs of the weapon used using the unmanned plane of the type of weapon, l types, V_qIt is q-th The value of business, N are the unmanned plane number of actual participation task in task distribution, It is worth for maximum task, max is to seek max function；

(2) normalization flying distance function is：Its Middle λ₁, λ₂For the weight of two factors, λ₁+λ₂=1, λ₁,λ₂>=0,For the length of longest path, R_lFor the ultimate run of l type unmanned planes；

(3) normalization consumption bullet cost function is：What the unmanned plane of wherein l types used The type needs of weapon are set in advance, δ_maxFor maximum cost,

(4) normalization target coverage degree function is：

The unmanned aerial vehicle group Task Assignment Model meets following constraints：

(1) task troops constrain：The unmanned plane of each type sets out number and no more than possesses the type unmanned plane number,

(2) unmanned plane combat radius constrains：Ensure the flying distance of unmanned plane within its combat radius, a_l,q×(D_l,q-R_l)≤ 0 (l=1,2 ..., L；Q=1,2 ..., Q)；

(3) to the constraint of target Damage degree：Degree of injuring injuring not less than the task of the unmanned plane of execution task q to task q Threshold value is spent,P_l,q,dFor l types without The man-machine probability of injuring using d kinds q-th of task of weapon pair, d are the type using weapon, the force that the unmanned planes of l types uses The type needs of device are set in advance；

(4) constraint of the unmanned plane number of target of attack：It is no more than its maximum to the unmanned plane number of q-th of task attack to form into columns Number, i.e.,

(5) constraint of the voyage of target of attack：I.e. the voyage of strike mission is no more than given ultimate run,O_maxFor the ultimate run of all unmanned planes；

(6) constraint of the time of target of attack：The time of strike mission is no more than given maximum time,Z_maxFor the maximum flight time of all unmanned planes；

Population scale K, the dimension J=L × Q of optimization problem for determining quantum crow group, are incremented by, mode incremental q is arranged according to l Unmanned aerial vehicle group task allocation matrix A={ a_l,q|a_l,q∈{0,1}}_L×QIn element, useIt is right The element in unmanned aerial vehicle group task allocation matrix A should be recorded；

Step 2：Initialize quantum crow group：

By the quantum position of i-th quantum crowBe set as per one-dimensionalIts In 1≤i≤K, 1≤j≤J, and the quantum position of i-th quantum crow is measured, obtains the position of i-th quantum crowThe hiding food position of i-th quantum crow of initialization isWherein 1≤i≤K, t are iterations, set t=0 at the beginning；

To the jth dimension of the quantum position of i-th quantum crowIt measures, obtains the jth of the position of i-th quantum crow DimensionWherein 1≤i≤K, 1≤j≤J,Be meet it is equally distributed with Machine number；

Step 3：Fitness calculating is carried out to every quantum crow according to fitness function, calculated fitness function is minimum The position for being worth corresponding quantum crow saves as global optimum's food position

Step 4：Update the quantum position and position of every quantum crow:

I-th quantum crow randomly selects another quantum crow s in quantum crow group, is sent out followed by quantum crow s The food position now hidden by quantum crow s, quantum crow s have found that the perception probability being followed is μ, ifAmount Sub- crow i carries out the update of quantum position by strategy 1, and otherwise quantum crow i carries out the update of quantum position by strategy 2；

Strategy 1 meets：Food position that i-th quantum crow is hidden according to quantum crow s carries out the update of position, i-th The renewal equation at quantum rotation angle of the jth dimension of quantum crow isWherein e₁It is normal Number determines the influence degree for guiding the position of the quantum crow to develop the quantum crow, and H is Flight Length；

Strategy 2 meets：I-th quantum crow carries out position more according to itself hiding food position and optimal food position Newly, the renewal equation at the quantum rotation angle of the jth dimension of i-th quantum crow isWherein e₂,e₃For constant, determines and guide the quantum crow The influence degree that develops to the quantum crow of position；

The evolution process of quantum position is as follows：

Wherein ζ=0.15/J is mutation probability, and abs () is to seek ABS function；

The position of quantum crow is obtained to quantum crow quantum position measurement, measurement rules are as follows：

Wherein 1≤i≤K, 1≤j≤J,It is to meet equally distributed random number；

Step 5：Fitness calculating is carried out to every quantum crow according to fitness function, determines hiding for every quantum crow Food position, while finding to the optimal food position of current iteration algebraically；

By the position of i-th quantum crowIt is assigned to task allocation matrix A, according toCarry out fitness calculating；

The hiding food position of quantum crow is chosen using greedy selection strategy, ifThenOtherwise

Step 6：If reaching greatest iteration algebraically T_max, algorithm termination, execution step 7；Otherwise, t=t+1, return to step are enabled Four continue；

Step 7：Global optimum's food position is exported, task allocation matrix is mapped as.

2. a kind of unmanned aerial vehicle group method for allocating tasks based on quantum crow group hunting mechanism according to claim 1, It is characterized in that：The process of fitness evaluation described in step 3 is as follows：

First by the position of t i-th quantum crow of generationIt is assigned to task allocation matrix A, wherein t i-th quantum crow of generation The jth of crow is tieed upIt is assigned to a_l,q, according to Fitness calculating is carried out, wherein

c₁,c₂,c₃,c₄,c₅,c₆For penalty coefficient, ω₁,ω₂,ω₃,ω₄For weighted factor, ω₁+ω₂+ω₃+ω₄=1,0≤ ω₁,ω₂,ω₃,ω₄≤ 1, max are to seek max function, and min is to seek minimum value function.