Disclosure of Invention
In order to carry out multi-heterogeneous unmanned aerial vehicle system cooperative task allocation based on multi-objective optimization on the basis of considering various risks, the invention provides a multi-heterogeneous unmanned aerial vehicle system cooperative task allocation method based on conditional probability and considering a multi-objective evolutionary algorithm. According to the method, a multi-heterogeneous unmanned aerial vehicle system task planning method considering various risks is constructed by introducing conditional probability and based on a multi-objective evolutionary algorithm capable of optimizing multiple conflict targets at the same time. In the method, a chromosome coding method is provided based on the characteristic of cooperative task allocation of the heterogeneous unmanned aerial vehicle, genetic operators are constructed, the problem of cooperative task allocation under various resource conditions can be solved by an improved multi-objective evolutionary algorithm, and a logic unlocking method for maintaining population randomness and quickly unlocking at a chromosome lock point is provided for the complex condition of chromosome lock. The multi-heterogeneous unmanned aerial vehicle system collaborative task allocation based on multi-objective optimization, which is developed under the condition of considering various risks, better conforms to the actual combat situation, can be suitable for collaborative task allocation under the condition of various resources, and the configuration of the logical unlocking method can greatly improve the allocation efficiency.
In order to achieve the purpose, the invention adopts the technical scheme that:
a multi-heterogeneous unmanned aerial vehicle system cooperative task allocation method considering a multi-objective evolutionary algorithm based on conditional probability includes the steps that firstly, objective functions and constraint conditions are set based on the conditional probability according to detected enemy target information, current battlefield situation, fighting risk and fighting resources of one party, and a multi-objective optimization model of heterogeneous unmanned aerial vehicle cooperative task allocation is established. And then, solving the constructed model by utilizing the constructed multi-objective evolutionary algorithm to obtain a Pareto solution set. And finally, selecting a certain solution from the Pareto solution set by using a solution selection method, and taking a task allocation scheme corresponding to the solution as an execution scheme. The calculation flow chart of the invention is shown in fig. 1, and comprises the following steps:
step 1: according to the detected enemy target information, the current battlefield situation, the fighting risk and the fighting resources of the party, setting a target function based on the conditional probability and giving constraint conditions, and establishing a multi-target optimization model for the cooperative task allocation of the heterogeneous unmanned aerial vehicle, which is concretely as follows:
step 1-1: collecting fundamental data of enemy targets and my unmanned aerial vehicles in mission planning
For convenience of description, order
Setting the detected enemy target to contain N
T Target, jth target marked as T
j The set of objects is denoted as
Each target contains three types of subtasks: the classification task, the attack task and the evaluation task are respectively marked as C, A and V. Different enemy targets have different values and risks, and high value targets are often accompanied by high risks. Setting a target T
j The value of (A) is recorded as
The high value target was designated Hv and the low value target was designated Lv. In order to improve the success rate of the battle, the attack on the high-value target is set twice, and the attack on the low-value target is set once. Let A
i Representing the ith attack task, target
Is recorded as a task set
Setting up
And
respectively represent the unmanned aerial vehicle to the target T
j Subtasks C, A of
1 、A
2 And the starting execution time of V,
and
respectively represent the unmanned aerial vehicle to the target T
j Subtasks C, A of
1 、A
2 And the execution time of V.
Setting N in unmanned aerial vehicle cluster
U Erect unmanned aerial vehicle, the ith unmanned aerial vehicle marks as U
i The set of drones is noted
The set cluster comprises three types of heterogeneous unmanned aerial vehicles: the reconnaissance plane, the fighter plane and the ammunition plane are respectively marked as s, c and m. The types of tasks executable by the s-type unmanned aerial vehicle are C and V, the types of tasks executable by the C-type unmanned aerial vehicle are C, A and V, and the type of tasks executable by the m-type unmanned aerial vehicle is A. Let unmanned aerial vehicle U
i Is recorded as
The success rate and survival rate for different drones performing different tasks for different targets are different. Setting up
Express unmanned plane U
i Execution target T
j The success rate of the task type k of (c),
express unmanned plane U
i Execution target T
j The mortality rate of task type k, wherein,
k∈I
3 k =1,2,3 denotes task types C, a, and V, respectively. According to the capability of the unmanned aerial vehicle, when the unmanned aerial vehicle U
i Is of type s
When unmanned plane U
i Is m times
Let unmanned aerial vehicle U
i Is recorded as a task set
Wherein
Presentation to unmanned aerial vehicle U
i The number of tasks of (a) is,
express unmanned plane U
i The ith executed task in the task set.
Step 1-2: constructing an objective function of a model
For target T
j If some subtasks are not successfully executed, the target T is determined to be successful execution of the subtasks
j Is not successfully executed. There are two risks that a drone may be destroyed while performing a task and that the task is not successfully completed. To achieve maximum operational benefit at minimum operational cost, consider as objective functions the value expectation of maximizing the targets that are successfully executed and the value expectation of minimizing the drone that is destroyed. The two objective functions are constructed by introducing conditional probabilities. Order to
Express unmanned plane U
i Whether to slave target
Flying to target T
j Performing T
j If the value of the task type k is 1, execution is indicated, otherwise, execution is not performed. Wherein, V
j Representing a target T
j Of the position of (a).
(i) Maximizing value expectation of successfully executed targets
(ii) Minimizing value expectation of a crashed drone
Wherein,
to represent
The first task of the group (c) of (c),
order to
Based on the actual situation
And
(ii) minimizing the target (i) as follows
(iii) Minimizing value expectation of unsuccessfully executed targets
Let f = (f) 1 ,f 2 ) T Wherein f is 1 =J 3 ,f 2 =J 2 。
Step 1-3: constraint condition of construction model
Based on the actual situation of the collaborative task assignment and the above analysis, the following constraints are considered.
The payload of a drone is limited, and therefore, in actual combat mission allocation, the number of type a missions allocated to each drone cannot exceed the payload of that drone. The number of tasks of task types C and V allocated to the unmanned aerial vehicle capable of executing task types C and V is not limited.
Wherein, AM i Express unmanned plane U i The load capacity of (2). If the type of the unmanned aerial vehicle is s, the payload of the unmanned aerial vehicle is 0.
In order to shorten the stay time of the same unmanned aerial vehicle on the same target and reduce the risk of discovering and destroying the unmanned aerial vehicle, the attack frequency of the same unmanned aerial vehicle on the same target is assumed to be not more than 1.
Thus, different attack missions of high value targets may be assigned to different drones.
The tasks of each target need to be performed in order, i.e. timing constraints need to be met. For the same target task, task type a can only be executed after task type C is completed, and task type V can only be executed after task type a is completely completed. Thus, task allocation needs to satisfy the following constraints.
Wherein, if the target T
j For a high value target, then
Otherwise
When each unmanned aerial vehicle executes the tasks in the task set, the tasks are required to be executed according to the sequence in which the tasks in the set are distributed, that is to say
Can only be performed after its previous task is completed.
Step 1-4: multi-objective optimization model for constructing multi-heterogeneous unmanned aerial vehicle cooperative task allocation
Based on the objective function and the constraint condition constructed above, the multi-objective optimization of the cooperative task allocation of the heterogeneous unmanned aerial vehicle is to minimize the vector function f on the basis of satisfying the constraint (4) - (7).
min f=(f 1 ,f 2 ) T
s.t.(4)-(7)
Step 2: solving a multi-objective optimization model for collaborative task allocation by using the constructed multi-objective evolutionary algorithm to obtain a Pareto solution set thereof, which is specifically as follows:
step 2-1: setting parameters for improved multiobjective optimization algorithms
Setting population size S, maximum iteration number G and cross probability P c And the probability of variation P m The value of (c).
Step 2-2: initializing a population
The coding of the chromosome is in such a way that the chromosome is constituted by genes, each of which represents the assignment of a certain task. The information in the gene includes a target number corresponding to the assigned task, the assigned task type, and a number of the unmanned aerial vehicle executing the task. The corresponding genome of the same unmanned aerial vehicle forms a sub-chromosome, and all the sub-chromosomes form a complete chromosome. Randomly generating a population P with the size of S g (g = 0). The initialization procedure for generating an initial population containing S chromosomes according to the above coding scheme is given below.
Step 2-2-1: inputting target parameters and population quantity of unmanned aerial vehicle and enemy
Inputting ammunition amount A of all unmanned aerial vehicles in array form m The unmanned aerial vehicle number sets comprise an enemy target number set TS, unmanned aerial vehicle number sets AU with types of c and m, and unmanned aerial vehicle number sets CVU with types of s and c. Let the number of chromosomes in the population equal 0.
Step 2-2-2: judging the chromosome number of the current population
And if the number of chromosomes in the population is equal to S, completing population initialization and exiting the current initialization process. Otherwise, executing step 2-2-3 to step 2-2-4
Step 2-2-3: an empty Chromosome is initialized and recorded as Chromosome
Step 2-2-4: encoding chromosomal Chromosome
Step 2-2-4-1: randomly selecting a target from TS, and recording as
Step 2-2-4-2: for the selected target
Subtask C and attack task A
1 Is distributed
The genes Gene1, gene2, gene3, gene4 are initialized. Firstly, randomly selecting an unmanned aerial vehicle from CVU (continuously variable Unit), and recording the unmanned aerial vehicle as
Order to
Chromosome = Chromosome @ Gene1. Then randomly selecting an unmanned aerial vehicle from AU (all-in-one) to be recorded as
Order to
Chromosome = Chromosome @ Gene2, and updates a
m Instant command
Step 2-2-4-3: if it is
Is a high value target and A
m Not equal to 0, then the target is checked
Attack task A of
2 Distributing, otherwise, turning to step 2-2-4
Randomly selecting an unmanned aerial vehicle from AU on the premise of satisfying constraint (5) and recording the unmanned aerial vehicle as
Order to
Chromosome = Chromosome ═ Gene3. Update A
m Instant command
Step 2-2-4-4: for the selected target
Is allocated to the subtask V of
Randomly selecting an unmanned aerial vehicle from the CVU and recording the selected unmanned aerial vehicle as
Removing targets from TS
Order to
Chromosome=Chromosome∪Gene4。
Step 2-2-4-5: judging whether the current chromosome is coded
If the length of the array TS is not 0, and A m If the vector is not zero, executing the step 2-2-4-1 to the step 2-2-4-5, otherwise, adding 1 to the number of chromosomes in the population, exiting the encoding process of the current chromosome, and turning to the step 2-2-2.
Step 2-3: calculating a population P according to the objective functions (2) and (3) g Fitness value of middle chromosome
Step 2-4: generation of S-size sub-populations
Step 2-4-1: from the population P using a roulette algorithm g Two crossed parents F are selected 1 ,F 2
Step 2-4-2: selecting corresponding crossover operators according to the conditions of total resources and total task quantity, and performing crossover operation according to crossover probability
Let L denote the number of high value targets. Order RM
i Represents the parent F
i Corresponding combinations of drones with residual ammunition, RT
i Representing the parent F
i A set of enemy targets that are not allocated. Let p be
1 And p
2 Express from crossA randomly selected cross-point in the fork parent,
and
respectively represent F
1 And F
2 Cross point p of
1 And p
2 A set of drones performing task types C and V,
and
respectively represent F
1 And F
2 Cross point p of
1 And p
2 A set of drones executing task type a. Let PT
1 And PT
2 Respectively represent F
1 And F
2 The set of allocated targets.
By the parent F
1 The interleaving procedure is given as an example. If it is
I.e. the ammunition is sufficient and has a residue of
In this case, the crossover process is step 2-4-2-1 to step 2-4-2-4.
Step 2-4-2-1: inputting information of intersection and intersection parent, i = p 1 -1
Step 2-4-2-2: judging whether the cross process is finished
If i satisfies i ≦ p 2 1, executing the step 2-4-2-3 to the step 2-4-2-4, otherwise, quitting the opposite parent F 1 The interleaving operation of (1).
Step 2-4-2-3: performing cross operation on the gene at the cross point i
The following operation was performed when the task type in the gene at the intersection i is A. If it is
And is
If the corresponding target in the gene at the cross point i is removed and is not an empty set, then
Randomly selecting an unmanned aerial vehicle, otherwise, selecting from RM
1 And (4) randomly selecting an unmanned aerial vehicle from the set after removing the corresponding target in the gene at the cross point i. Replacing information of the unmanned aerial vehicle in the current gene with the information of the selected unmanned aerial vehicle, and updating the RM
1 。
The following operation is performed when the task type in the gene at the intersection i is C or V. If it is
Then the slave is random
Randomly selecting an unmanned aerial vehicle, otherwise, never belonging to
Randomly selecting one unmanned aerial vehicle from the unmanned aerial vehicles of type s or c. And replacing the information of the unmanned aerial vehicle in the current gene by using the information of the unmanned aerial vehicle.
Step 2-4-2-4: let i = i +1, go to step 2-4-2
If it is
I.e. the ammunition is sufficient and has no residue, there is
The crossover process now goes from step 2-4-2-5 to step 2-4-2-8.
Step 2-4-2-5: inputting information of intersection and intersection parent, and making i = p 1 -1
Step 2-4-2-6: judging whether the cross process is finished
If i satisfies i ≦ p 2 1, executing the step 2-4-2-7 to the step 2-4-2-8, otherwise, quitting the opposite parent F 1 The interleaving operation of (1).
Step 2-4-2-7: performing crossover operation on the gene at the crossover point i
Order cross parent F
1 ,F
2 The targets corresponding to the genes at the cross points i of (A) are respectively referred to as
If it is
Then go to step 2-4-2-8, otherwise perform the following procedure.
If it is
All are low value targets or all are high value targets, then directly at F
1 In exchange for the corresponding gene
The information of (1). If it is
In order to be a high-value target,
for low-value targets, the numbers of targets in the corresponding genes are interchanged first, and then one target is randomly selected
And the target number of the gene is changed to
If it is
In order to be a low-value target,
for high value targets, the numbers of targets in the corresponding genes are then interchanged first, and one is then randomly selected
Attack the gene and change the target number of the gene to
Step 2-4-2-8: let F 1 Cross of (p) 1 And p 2 The number of genes between which the information is updated is denoted as l, and let i = i + l, go to step 2-4-2-6
If it is
That is, the ammunition amount is insufficient, there are
The crossover process now goes from step 2-4-2-9 to step 2-4-2-13.
Step 2-4-2-9: inputting information of intersection and intersection parent, i = p 1 -1
Step 2-4-2-10: judging whether the cross process is finished
If i satisfies i ≦ p 2 1, executing the step 2-4-2-11 to the step 2-4-2-13, otherwise, exiting the parent F 1 The interleaving operation of (1).
Step 2-4-2-11: performing crossover operation on the gene at the crossover point i
Let the target in the gene at cross point i be recorded
If it is
Then the slave RT
1 ∩PT
2 In the method, a target is randomly selected, otherwise, the target is selected from the RT
1 Randomly selects a target. Marking the selected target as
When in use
At the time of low value, if
For low value targets, then target
Target replacement in all corresponding genes
Otherwise, from F
1 Randomly selecting a low value target
Target object
And
target replacement in all corresponding genes
When in use
When it is a high-value target, if
For low value targets, then first from RT
1 In which a target is randomly selected
Target object
Target replacement in all corresponding genes
Then randomly selecting an object
And replacing the target of the gene with a target of the gene
Finally, add target
The genes of task types C and V of (a); if it is
For high value targets, the target
Target replacement in all corresponding genes
Step 2-4-2-12: updating RT 1 And PT 1
Step 2-4-2-13: let F 1 Cross of (p) 1 And p 2 The number of genes whose information is updated in between is denoted as l, and let i = i + l, go to step 2-4-2-10.
Step 2-4-3: performing mutation operation on the cross filial generation according to the mutation probability
The two cross filial generations obtained by the cross operation are marked as O
1 ,O
2 . With O
1 For example, the mutation process is given. Let q be
1 And q is
2 The starting point and the termination point of the gene segment needing variation. Let O be
1 The set of low-to-medium value targets is denoted CT, O
1 The collection of C and V genes is designated G
cv ,O
1 The collection of genes of A in (A) is denoted as G
a ,O
1 Corresponding set of drones with ammunition surplus andthe set of unassigned objects are denoted respectively
And
step 2-4-3-1: input variance and O 1 Let i = q 1 -1
Step 2-4-3-2: judging whether the mutation process is finished
If i satisfies i ≦ q 2 1, executing the step 2-4-3-3 to the step 2-4-3-4, otherwise, exiting from the pair O 1 The mutation operation of (3).
Step 2-4-3-3: performing mutation operation on the gene at the mutation point i
When the temperature is higher than the set temperature
If O is
1 Is C or V, O is removed from the CVU
1 Randomly selecting an unmanned aerial vehicle from the CVU according to the unmanned aerial vehicle in the gene at the variation point i; if O is
1 Is A, then
In which O is removed
1 The gene at the mutation point i, and then
Randomly selecting an unmanned aerial vehicle and updating
Replacing O with information of selected drone
1 The variation point i of (a) is determined.
When the temperature is higher than the set temperature
First from
Randomly selecting a target, denoted T
m Then find all targets and O
1 The gene at the mutation point i of (1) is targeted to the same gene, and the target is replaced with T
m . If T
m Is a high value target, and O
1 Is a low value target, then O
1 The target in the gene at the variation point i of (a) is removed from the CT, then one target is randomly selected from the CT and all the gene information is removed, and finally ammunition is distributed to the T
m (ii) a If T is
m Is a low value target, and O
1 Is a high value target, then T is randomly deleted
m A gene of the attack task of (1).
When in use
If O is
1 The type of task in the gene at the mutation point i of (3) is C or V, then from G
cv In which O is removed
1 The gene at the mutation point i of (3), and then randomly from G
cv Selecting a gene; otherwise, from G
a In which O is removed
1 The gene at the mutation point i of (3), and then randomly from G
a One gene is selected. Contacting the selected gene with O
1 The unmanned aerial vehicle information of the gene at the variation point i is exchanged.
Step 2-4-3-4: let O be 1 Variation point q of (2) 1 And q is 2 The number of genes whose information is updated in between is denoted l, and let i = i + l, go to step 2-4-3-2
Step 2-4-4: repeating steps 2-4-1 to 2-4-3 until a sub-population of size S is obtained
Step 2-5: judging whether chromosomes in the sub-population obtained in the step 2-4 are locked or not, and unlocking the locked chromosomes
The following gives the judgment process and unlocking method for a chromosome deadlock situation.
Step 2-5-1: transforming chromosomes into a form consisting of daughter chromosomes
And arranging the genes in the chromosome from small to large according to the unmanned aerial vehicle number, and obtaining the task executed by each unmanned aerial vehicle and the execution sequence according to the sub chromosomes.
Step 2-5-2: set of established representations of completed tasks, denoted CS
Step 2-5-3: if the dimension of CS is equal to the dimension of chromosome, terminating the unlocking process, otherwise, executing the step 2-5-4 to the step 2-5-6
Step 2-5-4: determining whether the task in the current gene of each daughter chromosome can be executed
Each unmanned aerial vehicle starts to execute from the first task in the task set, if the task type is C, the task is directly executed, the task is deleted from the task set, and the task is added into the CS; if the task type is A, judging whether a task type C of a target corresponding to the task is contained in the CS, if so, directly executing, deleting the task from the task set and adding the task into the CS, otherwise, the task cannot be executed, and the unmanned aerial vehicle is in a waiting state; if the task type is V, judging whether a task type C of a target corresponding to the task and all attack tasks of the target are contained in the CS, if so, directly executing, deleting the task from a task set and adding the task into the CS, otherwise, the task cannot be executed, and the unmanned aerial vehicle is in a waiting state.
Step 2-5-5: judging whether the chromosome is locked
And judging the states of all the current unmanned aerial vehicles according to the step 2-5-4, and if all the current unmanned aerial vehicles are in the waiting state, locking the chromosome.
Step 2-5-6: fixed point unlocking at deadlocking point
And if the type of the task to be executed by the unmanned aerial vehicle at the deadlock point is A, randomly selecting the task with the type of C from the rest tasks of the unmanned aerial vehicle and exchanging the sequence with the current task. This is performed for all drones in the waiting state. And (5) turning to the step 2-5-3.
Step 2-6: binding population P
g And
obtaining a population Q with the scale of 2S
g
Step 2-7: calculating a population Q according to the objective functions (2) and (3) g Fitness value of middle chromosome
Step 2-8: slave population Q based on non-dominated quick sorting method and elite retention strategy g S chromosomes are selected to form a parent population P of the next iteration q
Step 2-9: let g = g +1,q = g
Step 2-10: if G is less than G, turning to the step 2-3, otherwise outputting a Pareto solution set
And step 3: selecting a certain solution from the Pareto solution set by using a solution selection method, and taking a task allocation scheme corresponding to the solution as an execution scheme
Order to
Representing the number of non-dominated solutions on the Pareto optimal front,
representing the ith non-dominated solution. First, calculate
Wherein alpha is
1 And alpha
2 Is a weight and satisfies alpha
1 +α
2 =1. And the decision maker takes the value of the weight according to the preference degree of the objective function. Then, for the calculated S
i And (6) sorting. Finally, the smallest S is selected
i A task allocation scheme of a non-dominant solution to which the value corresponds.
The invention has the beneficial effects that:
the invention provides the enemy target information and various resource information of our party which are currently detected, and the method can comprehensively consider various risks in actual combat, such as the fact that the unmanned aerial vehicle is destroyed and the task is not successfully completed. On the basis, the method can perform cooperative task allocation on the multi-heterogeneous unmanned aerial vehicle system according to different combat resource conditions, can simultaneously optimize a plurality of objective functions with conflict relationships, and can enable the unmanned aerial vehicle cluster to obtain a more efficient and practical task execution scheme. And various risks and various battle resource conditions are considered, so that the distribution method is more perfect. Meanwhile, the logical unlocking method constructed aiming at the dying condition of the chromosome greatly improves the optimization efficiency of the algorithm.
Detailed Description
The present invention is further illustrated by the following examples.
Considering the situation that the ammunition amount is sufficient, 5 heterogeneous unmanned planes perform tasks on 5 targets. Let unmanned aerial vehicle U
i Is recorded as a task set
Unmanned plane U
i Is recorded as
Unmanned plane U
i The loading of the ammunition is recorded as AM
i . Let the high value target mark be Hv, the low value target mark be Lv, target T
j The value of (A) is recorded as
Task set notation of objectsDo not like
Let unmanned aerial vehicle U
i Execution target T
j The success rate of task type k of is recorded as
Unmanned plane U
i Execution target T
j The mortality rate of task type k is recorded as
The invention provides a multi-heterogeneous unmanned aerial vehicle system task allocation method based on conditional probability and considering a multi-objective evolutionary algorithm, which comprises the following steps of:
step 1: according to the detected enemy target information, the current battlefield situation, the fighting risk and the fighting resources of the party, setting a target function based on the conditional probability and giving constraint conditions, and establishing a multi-target optimization model for the cooperative task allocation of the heterogeneous unmanned aerial vehicle
Step 1-1: collecting fundamental data of enemy targets and my unmanned aerial vehicles in mission planning
Collecting the type of the unmanned aerial vehicle,
AM
i The type of object,
And
the information of (a) is shown in tables 1 to 3.
Table 1 information of unmanned aerial vehicle
TABLE 2 information of the objects
TABLE 3 success rate and destroyed rate when unmanned aerial vehicle executes task
Step 1-2: constructing an objective function of a model
Order to
Express unmanned plane U
i Whether to slave target
Flying to target T
j Performing T
j If the value of the task type k is 1, the execution is indicated, otherwise, the execution is not performed. Wherein, V
j Representing a target T
j The position of (a). To obtain the maximum operational benefit at the minimum operational cost, the following objective function is constructed.
(i) Maximizing value expectation of successfully executed targets
(ii) Minimizing value expectation of a crashed drone
Wherein,
represents M
Ui The first task of the group (c) of (c),
order to
Based on the actual situation have
And
(ii) minimizing the target (i) as follows
(iii) Minimizing value expectation of unsuccessfully executed targets
Let f = (f) 1 ,f 2 ) T Wherein f is 1 =J 3 ,f 2 =J 2 。
Step 1-3: constraint condition of construction model
Based on the characteristics of the collaborative task allocation and the above analysis, the following constraints are considered.
The payload of a drone is limited, and therefore, in the actual combat mission allocation, the number of type a missions allocated to each drone cannot exceed the payload of that drone. The number of tasks of task types C and V assigned by the drones capable of executing task types C and V is not limited.
Wherein, AM i Express unmanned plane U i The amount of loading of. If the type of the unmanned plane is s, the loading capacity is0。
In order to shorten the stay time of the same unmanned aerial vehicle on the same target and reduce the risk of discovering and destroying the unmanned aerial vehicle, the attack frequency of the same unmanned aerial vehicle on the same target is assumed to be not more than 1.
Thus, different attack missions of high value targets may be assigned to different drones.
The tasks of each target need to be performed in order, i.e. timing constraints need to be met. For tasks of the same goal, task type A can only be executed after task type C is completed, whereas task type V can only be executed after task type A is completed. Thus, task allocation needs to satisfy the following constraints.
when each unmanned aerial vehicle executes the tasks in the task set, the tasks are required to be executed according to the sequence in which the tasks in the set are distributed, that is to say
Can only be performed after its previous task is completed.
Step 1-4: multi-objective optimization model for constructing multi-heterogeneous unmanned aerial vehicle cooperative task allocation
Based on the objective function and the constraint conditions constructed above, the multi-objective optimization of the cooperative task allocation of the heterogeneous unmanned aerial vehicles is to minimize a vector function f on the basis of satisfying the constraints (11) - (14).
min f=(f 1 ,f 2 ) T
s.t.(11)-(14)
And 2, step: solving a multi-objective optimization model of cooperative task allocation by using a constructed multi-objective evolutionary algorithm to obtain a Pareto solution set thereof
Step 2-1: setting improved multi-objective optimization algorithm parameters
The population size S =100, the maximum number of iterations G =200, and the crossover probability P are set c =0.8 and mutation probability P m =0.2。
Step 2-2: initializing a population
The encoding of the chromosome is such that the chromosome is constructed by genes, each of which represents the assignment of a certain task. The information in the gene includes a target number corresponding to the assigned task, the assigned task type, and the number of the unmanned aerial vehicle executing the task. Randomly generating a population P with the size of 100 g (g = 0). The initialization procedure for generating an initial population containing 100 chromosomes according to the above coding scheme is given below.
Step 2-2-1: inputting unmanned aerial vehicle and enemy target parameters
Inputting ammunition amount A of all unmanned aerial vehicles in array form m The unmanned aerial vehicle number sets comprise an enemy target number set TS, unmanned aerial vehicle number sets AU with types of c and m, and unmanned aerial vehicle number sets CVU with types of s and c. Let the number of chromosomes in the population equal 0.
Step 2-2-2: judging the chromosome number of the current population
And if the number of chromosomes in the population is equal to 100, finishing population initialization and exiting the current initialization process. Otherwise, executing step 2-2-3 to step 2-2-4
Step 2-2-3: an empty Chromosome is initialized and recorded as Chromosome
Step 2-2-4: encoding chromosomal Chromosome
Step 2-2-4-1: randomly selecting a target from TS, and recording as
Step 2-2-4-2: for the selected target
Subtask C and attack task A of (1)
1 Is distributed
The genes Gene1, gene2, gene3, gene4 are initialized. Firstly, randomly selecting an unmanned aerial vehicle from the CVU, and recording the unmanned aerial vehicle as
Order to
Chromosome = Chromosome ═ Gene1. Then, randomly selecting an unmanned aerial vehicle from AU (all-purpose aircraft) to be recorded as
Order to
Chromosome = Chromosome U Gene2, and renews A
m Instant command
Step 2-2-4-3: if it is
Is a high value target and A
m Not equal to 0, then the target is checked
Attack task A of
2 Distributing, otherwise, switching to step 2-2-4-4 to randomly select an unmanned aerial vehicle from AU on the premise of satisfying the constraint (12) and recording as
Order to
Chromosome = Chromosome @ Gene3. Update A
m Instant command
Step 2-2-4-4: for the selected target
Is allocated to the subtask V of
Randomly selecting an unmanned aerial vehicle from the CVU and recording the selected unmanned aerial vehicle as
Removing targets from TS
Order to
Chromosome=Chromosome∪Gene4。
Step 2-2-4-5: judging whether the current chromosome is coded
If the length of the array TS is not 0, and A m If the vector is not zero, executing the step 2-2-4-1 to the step 2-2-4-5, otherwise, adding 1 to the number of chromosomes in the population, exiting the encoding process of the current chromosome, and turning to the step 2-2-2.
Step 2-3: calculating a population P according to the objective functions (10) and (9) g Fitness value of mesochromosome
Step 2-4: generation of a size 100 sub-population
Step 2-4-1: from the population P using a roulette algorithm g Two crossed parents F are selected 1 ,F 2
Step 2-4-2: selecting corresponding crossover operators according to the conditions of total resources and total task quantity, and performing crossover operation according to crossover probability 0.8
Order RM
i Represents the parent F
i Corresponding combinations of drones with residual ammunition, RT
i Represents the parent F
i A set of enemy targets that are not allocated. Let p be
1 And p
2 Representing randomly selected intersection points from the intersecting parents,
and
respectively represent F
1 And F
2 Cross point p of
1 And p
2 A set of drones performing task types C and V,
and
respectively represent F
1 And F
2 Cross point p of
1 And p
2 A set of drones executing task type a. Let PT
1 And PT
2 Respectively represent F
1 And F
2 The set of allocated targets.
By the father generation F
1 The interleaving procedure is given as an example.
I.e. the ammunition is sufficient and has a residue of
In this case, the crossover process is step 2-4-2-1 to step 2-4-2-5.
Step 2-4-2-1: inputting information of intersection and intersection parent, i = p 1 -1
Step 2-4-2-2: judging whether the cross process is finished
If i satisfies i ≦ p 2 1, executing the step 2-4-2-3 to the step 2-4-2-4, otherwise, quitting the opposite parent F 1 The interleaving operation of (3).
Step 2-4-2-3: performing cross operation on the gene at the cross point i
The following operation was performed when the task type in the gene at the intersection i is A. If it is
And is provided with
If the corresponding target in the gene at the cross point i is removed and is not an empty set, then
Randomly selecting an unmanned aerial vehicle, otherwise, selecting from RM
1 And (4) randomly selecting an unmanned aerial vehicle from the set after removing the corresponding target in the gene at the cross point i. Replacing information of the unmanned aerial vehicle in the current gene with the information of the selected unmanned aerial vehicle, and updating the RM
1 。
The following operation is performed when the task type in the gene at the intersection i is C or V. If it is
Then randomly get from
Randomly selecting an unmanned aerial vehicle, otherwise, never belonging to
Randomly selecting one unmanned aerial vehicle from the unmanned aerial vehicles of type s or c. And replacing the information of the unmanned aerial vehicle in the current gene by using the information of the unmanned aerial vehicle.
Step 2-4-2-4: let i = i +1, go to step 2-4-2
Step 2-4-3: performing mutation operation on the cross filial generation according to the mutation probability of 0.2
The two cross filial generations obtained by the cross operation are marked as O
1 ,O
2 . With O
1 For example, the mutation process is given. Let q be
1 And q is
2 Starting and ending of gene segments requiring mutationAnd (6) stopping the point. Let O be
1 The set of low and medium value targets is denoted CT, O
1 The collection of genes in C and V is denoted G
cv ,O
1 The collection of genes of A in (A) is denoted as G
a ,O
1 The corresponding set of drones with ammunition remaining and the set of unassigned targets are denoted respectively
And
step 2-4-3-1: input variation points and O 1 Let i = q 1 -1
Step 2-4-3-2: judging whether the variation process is finished
If i satisfies i ≦ q 2 1, executing the step 2-4-3-3 to the step 2-4-3-4, otherwise, exiting from the pair O 1 The mutation operation of (3).
Step 2-4-3-3: performing mutation operation on the gene at the mutation point i
If O is
1 Is C or V, O is removed from the CVU
1 Randomly selecting an unmanned aerial vehicle from the CVU according to the unmanned aerial vehicle in the gene at the variation point i; if O is
1 Is A, then
In which O is removed
1 The gene at the mutation point i, and then
Randomly selecting an unmanned aerial vehicle and updating
Replacing O with information of selected drone
1 The variation point i of (a) is determined.
Step 2-4-3-4: let O be 1 Variation point q of 1 And q is 2 The number of genes between which information is updated is denoted as lAnd let i = i + l go to step 2-4-3-2
Step 2-4-4: repeating steps 2-4-1 to 2-4-3 until a sub-population of
size 100 is obtained
Step 2-5: judging whether chromosomes in the sub-population obtained in the step 2-4 are locked or not, and unlocking the locked chromosomes
The following gives the judgment process and unlocking method for a chromosome deadlock situation.
Step 2-5-1: transforming chromosomes into a form consisting of daughter chromosomes
And arranging the genes in the chromosome from small to large according to the unmanned aerial vehicle number, and obtaining the task executed by each unmanned aerial vehicle and the execution sequence according to the sub chromosomes.
Step 2-5-2: constructing an empty set of completed tasks, denoted CS
Step 2-5-3: if the dimension of CS is equal to the dimension of chromosome, terminating the unlocking process, otherwise, executing the step 2-5-4 to the step 2-5-6
Step 2-5-4: determining whether the task in the current gene of each sub-chromosome can be executed
Each unmanned aerial vehicle starts to execute from the first task in the task set, if the task type is C, the unmanned aerial vehicle directly executes, deletes the task from the task set and adds the task into the CS; if the task type is A, judging whether a task type C of a target corresponding to the task is contained in the CS, if so, directly executing, deleting the task from the task set and adding the task into the CS, otherwise, the task cannot be executed, and the unmanned aerial vehicle is in a waiting state; if the task type is V, whether the task type C of the target corresponding to the task and all attack tasks of the target are contained in the CS is judged, if yes, the task is directly executed, the task is deleted from the task set and added into the CS, otherwise, the task cannot be executed, and the unmanned aerial vehicle is in a waiting state.
Step 2-5-5: judging whether the chromosome is locked
And judging the states of all the current unmanned aerial vehicles according to the step 2-5-4, and if all the current unmanned aerial vehicles are in the waiting state, locking the chromosome.
Step 2-5-6: fixed point unlocking at lock dead point
And if the type of the task to be executed by the unmanned aerial vehicle at the deadlock point is A, randomly selecting the task with the type of C from the rest tasks of the unmanned aerial vehicle and exchanging the sequence with the current task. This is performed for all drones in the waiting state. And (5) turning to step 2-5-3.
Step 2-6: binding population P
g And
obtaining a population Q of
size 200
g
Step 2-7: calculating a population Q according to the objective functions (10) and (9) g Fitness value of middle chromosome
Step 2-8: slave population Q based on non-dominated quick sorting method and elite retention strategy g 100 chromosomes are selected to form a parent population P of the next iteration q 。
Step 2-9: let g = g +1,q = g
Step 2-10: if g is less than 200, turning to step 2-3, otherwise outputting a Pareto solution set
And 3, step 3: selecting a certain solution from the Pareto solution set by using a solution selection method, and taking a task allocation scheme corresponding to the solution as an execution scheme
Order to
Representing the number of non-dominant solutions on the Pareto optimal frontier,
representing the ith non-dominant solution. First, calculate
Wherein the weight takes the value of alpha
1 =0.5 and α
2 =0.5. Then, for the calculated S
i And (6) sorting. Finally, selectingSelecting the minimum S
i A task allocation scheme of a non-dominant solution to which the value corresponds.
The Pareto optimal front end obtained by optimization is shown in fig. 2, and fig. 3 and fig. 4 show the change of the optimal values of two objective functions in each iteration along with the number of iterations. Based on the solution selection strategy, the 15 th non-dominated solution on the Pareto frontier is obtained as the selected solution, and the corresponding solution information and the specific task allocation scheme are shown in table 4. To test the computational efficiency of the method, 15 experiments were performed with CPU run times as shown in fig. 5.
TABLE 4 Pareto optimal frontier information of the 15 th non-dominated solution and corresponding task assignment scheme
To verify the validity of the structured logical unlocking method, 100 chromosomes were randomly generated without considering timing constraints. These 100 chromosomes were subjected to 15 experiments to determine whether to lock and unlock using the constructed logical unlocking method. The unlocked CPU run time is shown in fig. 6. As can be seen from the figure, the logic unlocking method can effectively solve the locking condition.
The invention provides a multi-heterogeneous unmanned aerial vehicle system cooperative task allocation method considering a multi-objective evolutionary algorithm based on conditional probability. Based on the method, on the premise of comprehensively considering the risk of being destroyed by the unmanned aerial vehicle and the risk of incomplete task, cooperative task allocation can be carried out on the heterogeneous unmanned aerial vehicle group under various battle resource situations, and a plurality of conflict targets can be simultaneously optimized. In addition, the logic unlocking mode constructed in the method greatly improves the solving efficiency. The multi-heterogeneous unmanned aerial vehicle system task allocation method considering the operational risk and various operational situations provides a task allocation scheme more suitable for actual operational situations for pre-war task allocation, and can effectively improve the solving efficiency of task allocation.
The above-mentioned embodiments only express the embodiments of the present invention, but not should be understood as the limitation of the scope of the invention patent, it should be noted that, for those skilled in the art, many variations and modifications can be made without departing from the concept of the present invention, and these all fall into the protection scope of the present invention.