CN103336885B - A kind of method solving Weapon-Target Assignment Problem based on differential evolution algorithm - Google Patents

Abstract

The present invention a kind ofly solves the method for Weapon-Target Assignment Problem based on differential evolution algorithm, belongs to Computer Simulation and method optimizes field.On the basis of traditional weapon-target assignment problem, this method with the addition of the detection of sensor to target, and the factors such as the detection transfer time of sensor, the interconnection constraint of target number, sensor and weapon that can simultaneously detect, the reserved observing time of weapon are considered in problem, first determine larger solution space, then according to the constraint of weapon-sensors association, intercept point constraint be set for specific objective reduce solution space; Carry out coding reconstruct to the solution space after reducing, the position can distributed on interval at it by each ammunition of weapon is optimized variable, utilizes differential evolution algorithm to find optimal solution.The present invention, closer to the Weapon-Target Assignment Problem of actual conditions, reduces optimization complicacy, improves optimization efficiency, decreases solution space and taken up space, and operates flexibly more convenient.

Description

A kind of method solving Weapon-Target Assignment Problem based on differential evolution algorithm

Technical field

The invention belongs to Computer Simulation and method optimizes field, relate to a kind of method solving Weapon-Target Assignment Problem based on differential evolution algorithm.

Background technology

Weapon-Target Assignment Problem is an important topic in warfare decision, be exactly how reasonably to distribute our troops to meet enemy troops head-on, to reach best fighting effect, belong to NP(Non-deterministicPolynomial, the uncertain problems of polynomial expression complexity) complete problem, there is no effective determinacy method for solving at present.The fundamental purpose of its research is, fast and effeciently solve the assignment problem of extensive weapon-target, to improve the automatization level that battleficld command controls, namely for multiple threat target, defence can distribute defence resource timely and effectively, effectively eliminate unfriendly target to threaten, make the loss reduction of defence side.Generally speaking troops' resource that Weapon-Target Assignment Problem relates to is only limitted to weapon and target, and as detecting the sensor of resource not at the row of troops' resource.

Be developed so far from the sixties in 20th century five, the solution of Weapon-Target Assignment Problem experienced by the improvement from traditional algorithm to intelligent optimization algorithm.The eighties was mainly limited to traditional algorithm in the past, and as implicit enumeration method, dynamic programming etc., these algorithms are comparatively simple, but comparatively loaded down with trivial details when realizing, and when especially destination number increases, speed of convergence is just very slow.Since the eighties, some optimized algorithms, as artificial neural network, chaos and heuristic search algorithm etc., provide new thinking and means for solving challenge.

Differential evolution algorithm is the optimized algorithm based on swarm intelligence theory, it can by colony between individuality cooperation and competition realize solving optimization problem.Be compared to evolution algorithm, differential evolution algorithm remains the global search strategy based on population, adopts real coding, operates and man-to-man competition surviving policy based on the simple editing of difference, reduce the complicacy of genetic manipulation.Differential evolution algorithm have memory individual optimal solution and middle group's internal information share feature, its essence is a kind of greedy genetic algorithm with blessing thought based on real coding, this algorithm is simple and easy to use, robustness is good, ability of searching optimum is strong, differential evolution algorithm has multiple strategy to select simultaneously, can adjust hunting zone and speed of convergence as required.

Summary of the invention

The object of the invention is, sensor is listed among troops' resource to be allocated, propose the Weapon-Target Assignment Problem more tallied with the actual situation, and provide a kind of method solving Weapon-Target Assignment Problem based on differential evolution algorithm.

The present invention is based on the method that differential evolution algorithm solves Weapon-Target Assignment Problem, the sensor resource of defend equipment is added Weapon-Target Assignment Problem, and solution specifically comprises following step:

Step one: the input variable and the initializing variable that obtain Weapon-Target Assignment Problem.

Input variable comprises: target number N, weapon number M, number of probes G, the priority of resource allocation S of each target _i, i=1,2 ..., N; The ammunition quantity B initial or often kind of current weapon has _j, j=1,2 ..., M; The destination number O initial or current each sensor can detect simultaneously _k, k=1,2 ..., G; The belligerent matrix of target-weapon, comprising information has: weapon sequence number, target sequence number and correspondence thereof can commit time section, can launch time section, can weapon corresponding to intercept point and target distance from; The detectable matrix of target-sensor, comprising information has: the target in the detectable time period of sensor sequence number, target sequence number and correspondence thereof, detectable time period and the distance between sensor; The interconnection constraint of weapon and sensor, the weapon comprising association and sensor information, the temporal information associated.

Initializing variable comprises: the property value C of each target _ni, the property value C of each weapon _mj, the property value C of each sensor _gk, a jth weapon M _jto i-th target N _ithe estimated value P of kill probability _ij, detection T transfer time of each sensor _d_ G _k, wherein, i=1,2 ..., N, j=1,2 ..., M, k=1,2 ..., G; Reserve T observing time _d; Given kill probability value P _given; Given priority of resource allocation is not worth S _given.

Step 2: determine solution space, comprises step 2.1 ~ step 2.3.

Step 2.1: that determines target can by weapon attacking time shaft, specifically: find out N from the belligerent matrix of target-weapon _ican by weapon M _jsection MT effective time attacked _ij, by MT _ijsplice in chronological order, form target N _ican by the time shaft MT of weapon attacking _i, time shaft MT _ion the information that stores of each time period have: target, weapon, can attack time section initial time, can the attack time section end time; Wherein, i=1,2 ..., N, j=1,2 ..., M.

Step 2.2: that determines target can by sensor detection time axle, specifically: find out N from the detectable matrix of sensor-target _ican by sensor G _ksection GT effective time of detection _ik, by GT _iksplice according to time sequencing, form target N _ibe detected time shaft GT _i, time shaft GT _ion the information that stores of each time period have: target, sensor, detectable time period initial time, detectable end time time period; Wherein, i=1,2 ..., N, k=1,2 ..., G.

Step 2.3: the be blocked time shaft determining target, specifically: by target N _ican by weapon attacking time shaft MT _iwith can by sensor detection time axle GT _iarranging according to time sequencing, by time shaft there being overlapping part merge, obtaining target N _ibe blocked time shaft T _i.N number of time shaft corresponding to N number of target constitutes initial solution space A ₁.

Use binary digit coding each target of mode record each time period in can attack weapon and detectable sensor information.The coding binary figure place recording weapon information in each time period is identical with the number of weapon, from right to left, each represents the weapon that weapon sequence number is serial number, the value of every is 1 or 0,1 represents that corresponding weapon can be attacked target within this time period, and 0 represents that corresponding weapon can not be attacked target within this time period.The coding binary figure place recording sensor information in each time period is identical with the number of sensor, from right to left, each representative sensor sequence number is the sensor of serial number, the value of every is 1 or 0,1 represents that respective sensor can detect target within this time period, and 0 represents that respective sensor can not detect target within this time period.

Step 3: reduce initial solution space, comprise step 3.1 and step 3.2.

Step 3.1: reduce solution space according to weapon-sensors association constraint.Reject the weapon attacking information not meeting the interconnection constraint of weapon-sensor, then can not be rejected in solution space by the time period of any weapon attacking, obtain the solution space A reduced ₂.

Can not be rejected in solution space by the time period of any weapon attacking, specifically: to the target N of interconnection constraint having weapon-sensor _i, for the interconnection constraint of every bar weapon-sensor, perform and operate as follows:

If weapon M _jwith sensor G _kthere is interconnection constraint, traversal target N _ibe blocked time shaft T _ion all time periods, to can by weapon M _jcertain time period of attacking, judge whether meet sensor G in this time period _krequirement, if do not meet, then at time shaft T _ion by weapon M in this time period _jto target N _ithe information of attacking is rejected, if meet, order gets the next time period, if can by weapon M in this time period _jattack and meet again sensor G _krequirement, then continue to get subsequent time period, until can not by M _jattack or do not meet sensor G _krequirement; Now, judge all can by weapon M _jattack and meet again sensor G _kthe time period T of requirement _i.x1~ T _i.x2, whether cover weapon M _jwith sensor G _ksection correlation time, if not, then at time shaft T _ion by time period T _i.x1~ T _i.x2interior weapon M _jto target N _ithe information of attacking is rejected.

Step 3.2: reduce solution space for specific objective arranges intercept point constraint.Traversal target is priority of resource allocation S _ibe more than or equal to given priority level value S _giventarget, an intercept point is at least set; After having traveled through all targets, the more state of modern weapons and sensor, and surplus resources; Then, rejected from solution space the time period of distributing intercept point, the attack information being the weapon of zero by residue ammunition is rejected in solution space; Finally can not be rejected in solution space by the time period of any weapon attacking, obtain the solution space A reduced ₃.

If target N _ipriority of resource allocation be not worth S _i>=S _given, then the operation of (1) and (2) below performing:

(1) weapon and ammunition is distributed.To target N _itime shaft T can be blocked _ion all time period T _i.x, to calculate in each time period weapon to target N _iaverage kill probability x is positive integer, x=1,2 ..., represent the xth time period that can be blocked on time shaft.Elect section sometime as need to arrange intercept point time period, time period T _i.xwith probability selected.Intercept point is set to, weapon M on this intercept point by by the starting point of time period selected _l, l ∈ 1,2 ..., M} is to target N _ithe ammunition number distributed is B _l* (S _i/ Σ S _i) * (qp), wherein p is M _lto N _iall can the length of attack time section, q is the length of current slot.If had plural weapon can to target N by the time period selected _iattack, carry out the distribution of ammunition number from the weapon that kill probability is high, the distribution ammunition number of gained rounds up.

(2) sensor is distributed.If there is the interconnection constraint of weapon and particular sensor, be then target N according to interconnection constraint _idistributing sensor and detection time, if onrelevant constraint, is target N according to nearby principle _iassign sensor; If the selected time period is (time1 ~ time2), then the detection time distributing sensor is (time1 ~ time2+T _d_ G _k).

Step 4: coding reconstruct is carried out to solution space, comprises step 4.1 ~ step 4.3.

Step 4.1: reconstruct solution space.To solution space A ₃the time shaft that can carry out tackling according to weapon is reconstructed, and obtains the solution space A after reconstructing ₄.To weapon M _j, at solution space A ₃middle travel time section, obtains all containing M _jthe time period of attack information, all time periods obtained are spliced according to time sequencing, form weapon M _jcan commit time axle TM _j.j＝1,2,……,M。

Step 4.2: divide solution space.By weapon M _jcan commit time axle TM _jon each time period need divide with the time by minimum interception, and need to reject being less than minimum interception with the time period of time, if final time shaft TM _jon live part by ML _jindividual minimum interception need form with the time period, j=1, and 2 ..., M.Described minimum interception need comprise reserved T observing time with the time _d, and T launch time of weapon _fwith can attack time T _lbetween linear relationship.

Step 4.3: solution-vector coding.To weapon M _j, by TM _jon ML _jthe individual time period is mapped to ML _jthe real number point set of individual real number composition 1,2 ..., ML _jon.Get interval (0, ML _j] as weapon M _jammunition can distribute interval.j=1,2,……,M；.

Step 5: utilize differential evolution algorithm to find optimal solution, comprise step 5.1 ~ step 5.7.

Step 5.1: selected optimized variable and objective function.Selected weapon M _jeach ammunition be optimized variable in the position that it can distribute on interval, s=1,2 ..., B _j; J=1,2 ..., M.

Objective definition function f is:

Wherein: m _ifor target N _ithe quantity of the intercept point distributed.λ _ifor scale-up factor, determine according to actual conditions. j _u∈ { j ₁..., j _arepresent: be total a kind weapon j to the allocation result of certain target interception point ₁..., j _atackle it, the kill probability value of often kind of weapon is ..., the ammunition quantity of often kind of weapon allocation is ...,

Step 5.2: produce initial population.

T is for the e in population individual x _et () is expressed as:

x _e(t)=(x _el(t),x _e2(t),...,x _eW(t)),e＝1,2,......,NP；t＝1,2,......,t _max

Wherein, x _ethe corresponding optimized variable of each component of (t), W forms individual chromosome number, i.e. the number of optimized variable; NP is population scale; t _maxit is maximum evolutionary generation.

NP individuality is produced at random according to the span of optimized variable.Each individuality in population produces by the following method:

According to weapon kill probability order from high to low, to weapon M _jall available ammunition (s=1,2 ..., B _j) be assigned to interval (0, ML by non-uniform probability _j] on.For a certain segment, if weapon can attack the target of more than 2 in this segment, then use the mode of roulette, with probability S _i/ Σ S _ione of them target selected, distributes unitedly the ammunition in this segment in selected target.By ammunition by its interval distributed and Target Assignment in the starting point of the commit time section corresponding to it, whether have resource to use conflict (sensor resource uses constraint) to the individuality inspection generated, if there is conflict, regenerate.

Add up the intercept point number of each target, calculate the fitness value of each individuality according to objective function.

Step 5.3: produce test vector

From current t for 3 the individual x of Stochastic choice population _p1(t), x _p2(t) and x _p3(t), then t+1 for population e population at individual r optimized variable corresponding to test vector h _er(t+1) be:

h _er(t+1)＝x _p1r(t)+F*(x _p2r(t)-x _p3r(t))，e≠p1≠p2≠p3

Wherein,

X _p1r(t), x _p2r(t) and x _p3rt () represents individual x respectively _p1(t), x _p2(t) and x _p3r the optimized variable of (t);

X _p2r(t)-x _p3rt () is differentiation vector; F is zoom factor.

If the distribution that the test vector produced is not in ammunition is interval, regenerate test vector according to the individual production method of population in step 5.2.

Step 5.4: produce filial generation, r filial generation v of e population at individual _er(t+1) be:

Wherein, randl _erbe the random decimal between [0,1], CR is crossover probability, and CR ∈ [0,1], rand (e) is the random integers between [1, N].

Target Assignment carried out to the filial generation generated and checks each individuality whether to have resource to use conflict, if there is conflict, regenerating according to the individual production method of population in step 5.2.

Add up the intercept point number of each target, calculate the fitness value of each individuality in filial generation according to objective function.

Step 5.5: Population Regeneration.The t+1 upgraded is for the e in population individual x _e(t+1) be:

$x_e(t+1)=\left\{\begin{array}{cc}{v}_e(t+1),& f(v_{\mathrm{el}}(t+1),\&CenterDot;\&CenterDot;\&CenterDot;,v_{\mathrm{en}}(t+1))<f(x_{\mathrm{el}}\left(t\right),\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{en}}\left(t\right))\\ x_e\left(t\right),& f(v_{\mathrm{el}}(t+1),\&CenterDot;\&CenterDot;\&CenterDot;,v_{\mathrm{en}}(t+1))\&GreaterEqual;f(x_{\mathrm{el}}\left(t\right),\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{en}}\left(t\right))\end{array}\right.$

Wherein, f (v _el(t+1) ..., v _en(t+1)) be filial generation v _e(t+1) fitness, f (x _e1(t) ..., x _en(t)) be parent x _ethe fitness of (t).

Judge whether the individuality of fitness optimum in new population meets formula: if meet, termination of iterations, returns fitness optimum individual, otherwise, go to step 5.3 execution.

Or judge whether t+1 has exceeded iteration upper limit t _maxif exceed, then termination of iterations, returns fitness optimum individual, otherwise continues to go to step 5.3 execution.

Step 5.6: distribute surplus resources.

For the optimum individual selected, if there is surplus resources, when not producing conflict, priority allocation to the high target of priority, or adopts distribution principle nearby to distribute.

Step 5.7: the allocation result after the optimization obtained is carried out Gray code, determines the intercept point position of Target Assignment and weapon, the sensor resource of distribution.

Advantage of the present invention and good effect are:

1) on the basis of traditional weapon-target assignment problem, with the addition of the detection of sensor to target, and the factors such as the detection transfer time of sensor, the interconnection constraint of target number, sensor and weapon that can simultaneously detect, the reserved observing time of weapon are considered in problem, the starting point of problem compares and tallies with the actual situation, and can solve the Weapon-Target Assignment Problem closer to actual conditions.

2) put forward the methods in the present invention is used to provide the reference of solution aspect for solving the Weapon-Target Assignment Problem adding sensor detection constraint.

3) first obtain larger solution space, then reduce solution space by partially restrained, instead of using a series of constraint all as the constraint condition in optimization method, greatly reduce the complicacy of optimization method, improve optimization efficiency.

4) use the mode of binary digit coding to record in solution space certain time period upper assailable weapon and detectable sensor information, or be used for the detectable sensor information in restructuring of record solution space.Compared with general array or vectorial location mode, decrease solution space and taken up space; In practical operation, compared with array or vectorial flexibly more convenient; And just can get information about very much weapon by a binary number and carry out attacking or sensor carries out the information that detects.

5) in the present invention, intercept point is set to optimized variable, reconstruct solution space by the visual angle of time shaft from goal displacement to weapon, have successfully completed the conversion from primitive solution space to the solution space inputted as differential evolution algorithm, closely fitting optimization aim, having provided support for utilizing differential evolution algorithm optimizing further.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the Weapon-Target Assignment Problem method that solves based on differential evolution algorithm in the present invention;

Fig. 2 is that target-weapon can attack time axle;

Fig. 3 is the detectable time shaft of target-sensor;

Fig. 4 is target N _ibe blocked time shaft;

Fig. 5 is that the inventive method reduces the process flow diagram of solution space according to weapon-sensors association constraint;

Fig. 6 is that the inventive method reduces the instance graph of solution space according to weapon-sensors association constraint;

Fig. 7 be specific objective arrange intercept point constraint reduce solution space process flow diagram;

Fig. 8 is reconstruct solution space legend.

Embodiment

Below in conjunction with accompanying drawing, technical scheme of the present invention is further described.

Weapon-Target Assignment Problem to be solved by this invention, on the basis of traditional Weapon-Target Assignment Problem, with the addition of the detection of sensor to target, and the factor such as the interconnection constraint of destination number, sensor and weapon, the reserved observing time of weapon that the detection transfer time of sensor, sensor can be detected take into account in problem simultaneously, the starting point of problem compares and tallies with the actual situation, and can solve closer to actual conditions.Weapon-Target Assignment Problem to be solved by this invention is described below:

A collection of offensive weapons is implemented to attack by specific flight path, a set of defend equipment implements interception to attacking target, defend equipment comprises weapon and sensor two class resource, equipment limited amount, requiring the optimize resource allocation providing defend equipment by certain rule in the dynamic case, dynamically ensureing that the comprehensive belligerent result to attacking target is best.

This Weapon-Target Assignment Problem is closing to reality situation more, and listed in by sensor among troops' resource to be allocated, target, weapon, sensor carry out the differentiation of kind by attribute; Target is N number of altogether, with N _i(i=1,2 ..., N) to distinguish, each target has the property value C that characterizes target type _ni(i=1,2 ..., N), meanwhile, each target all has a priority of resource allocation and is not worth S _i(i=1,2 ..., N), S _ibe interception order, relevant with the threaten degree of target, S _ivalue is higher, and to represent the threaten degree of target higher, and the priority level of Resourse Distribute is also higher.M altogether, weapon, with M _j(j=1,2 ..., M) to distinguish, each weapon has the property value C that characterizes weapon type _mj(j=1,2 ..., M), sensor G altogether, with G _k(k=1,2 ..., G) to distinguish, each sensor has the property value C that characterizes sensor type _gk(k=1 ..., G); A jth weapon carries ammunition B _j, j=1,2 ..., M.A kth sensor can detect O simultaneously _kindividual destination number k=1 ..., G, corresponding detection T transfer time of sensor of each attribute _d_ G _k, k=1 ..., G.Detection to refer to when sensor is by the transfer time of shifting the detection of certain target as needing during detection to another one target transfer time.

When using differential evolution algorithm to carry out the distribution of troops' resource, set the goal to one, except will distributing weapon and it hit, also need to distribute sensor to detect it, the sensor now distributed should meet the incidence relation of weapon and sensor, also to determine the ammunition that weapon uses simultaneously, and As soon as possible Promising Policy ammunition assignment constraints condition, S is not worth for priority of resource allocation _ihigher than given priority level value S _giventarget, ensure at least can distribute an intercept point.The result of final distribution is the intercept point of each Target Assignment, and the intercept point of a Target Assignment may have multiple.

Wherein, the incidence relation of weapon and sensor is: to the Resourse Distribute of same target, the incidence relation between weapon and sensor should be met, the change that this incidence relation physical presence is very many, can simplify and abstractly be: for the weapon of some attribute, one or two of section must be had between weapon and particular community sensor correlation time, require to have m particular community sensor assignment at least to the detection time section of this target covering section correlation time (namely detection time section contains section correlation time).M is at least 1, is no more than at most sensor sum.This interconnection constraint comprises two-layer constraint: one is the constraint of weapon and particular community sensor; Two is constraints of section correlation time.Weapon and sensor correlation time section mainly in order to ensure the position accurately determining target, prevent loss target.Exemplify for: if an attribute is C _m3weapon to target N ₁when attacking, require now to have m attribute to be C _g5sensor, in the process of attacking and attacking, this target is detected for carrying out at weapon.

Ammunition assignment constraints condition is: suppose to be that interception implemented by total a kind weapon, and sequence number is j to the allocation result of certain intercept point ₁, j ₂..., j _a, the ammunition of often kind of weapon to this Target Assignment is respectively ..., individual, the kill probability estimated value of often kind of weapon is respectively ..., then should meet

Wherein, P _givenrelevant with Target Attribute values, during research desirable 0.96.When ammunition is not enough, this condition may be not being met, and now meets the condition in optimization aim.

As shown in Figure 1, the method that the present invention is based on differential evolution algorithm solution Weapon-Target Assignment Problem can complete in accordance with the following steps:

Step one: the initializing variable and the input variable that obtain Weapon-Target Assignment Problem;

Input variable mainly comprises: the number N of target, weapon, sensor, M, G; The priority of resource allocation S of each target _i(i=1 ..., N); The ammunition quantity B initial or often kind of current weapon has _j(j=1 ..., M); The destination number O initial or current each sensor can detect simultaneously _k, k=1 ..., G; The belligerent matrix of target-weapon; The detectable matrix of target-sensor; Interconnection constraint between weapon and particular community sensor, the weapon comprising association and sensor information, the temporal information associated.

Often kind of weapon all to target may be formed belligerent maybe can not be formed belligerent, after judging according to space-time condition early stage, can obtain the belligerent matrix of target-weapon, the information comprised in belligerent matrix has: weapon sequence number, target sequence number and correspondence thereof can commit time section, can launch time section, can weapon corresponding to intercept point and target distance from.Owing to commit time Duan Yuke section launch time can be one_to_one corresponding and there is certain relation, therefore, get can commit time section as initial conditions during problem solving, distance be to can commit time carry out sliding-model control after the discretize result of correspondence that obtains.

Each sensor all may form detection to target maybe can not form detection, after judging according to space-time condition early stage, can obtain the detectable matrix of target-sensor, the information comprised in matrix has: the target in the detectable time period of sensor sequence number, target sequence number and correspondence thereof, detectable time period and the distance between sensor.Distance be sliding-model control is carried out to the detectable time after the discretize result of correspondence that obtains.

Initializing variable mainly comprises: in order to identify the property value C of target, weapon, each target of sensor, weapon, sensor _ni(i=1 ..., N), C _mj(j=1 ..., M), C _gk(k=1 ..., G); Weapon j is to the estimated value P of the kill probability of target i _ij(i=1 ..., N; J=1 ..., M); Detection T transfer time of each sensor _d_ G _k(k=1 ..., G); Reserve T observing time _d; Given kill probability value P _given; Given priority of resource allocation is not worth S _given.Whether reserved observing time refers to, need to observe to tackle successfully after interception, if interception implemented by unsuccessful weapon delivery again, there is reserved observing time period.

Step 2: determine larger solution space, specifically comprise step 2.1 ~ step 2.3.

Step 2.1: determine that target can by the time shaft of weapon attacking, namely target-weapon can engagement time axle.

Travel through all targets, from the belligerent matrix of the target-weapon of input, find out target N _i(i=1 ..., N) and can by weapon M _j(j=1 ..., M) and section MT effective time that attacks _ij, when i, j are identical, be likely multiple MT _ij, namely same weapon to same target may have multiple can commit time section.As shown in Figure 2, target N _ican by weapon M ₁, M ₂and M ₃section effective time of attacking is respectively MT _i1, MT _i2and MT _i3, wherein, can by weapon M ₃section effective time of attacking has two sections of MT _i3-1and MT _i3-2.By MT _ij(j=1 ..., M) to splice in chronological order, during splicing, first arrange according to the initial time of time period, identical the arranging according to the termination time of initial time, obtains target N _ican by the time shaft MT of weapon attacking _i.Time shaft MT _imay be discontinuous, this time shaft MT _ion comprise target N _ican by the information of certain several weapon attacking in section sometime.Each time period store information have: target, weapon, can attack time section initial time, can the attack time section end time.As shown in Figure 2, target N _ican by the time shaft MT of weapon attacking _icomprise some time section, such as the 2nd time period can by weapon M ₁and M ₂attack.

Step 2.2: determine the time shaft that target can be detected by sensor, i.e. the detectable time shaft of target-sensor.

Travel through all targets, from the detectable matrix of sensor-target, find out N _ican by a kth sensor G _k(k=1 ..., G) and section GT effective time that detects _ik.When i, k are identical, multiple GT may be there is _ik, namely same sensor may have multiple detectable time period to same target.As shown in Figure 3, target N _ican by sensor G ₁, G ₂, G ₃and G ₄section effective time of detection is respectively GT _i1, GT _i2, GT _i3and GT _i4.By GT _iksplice according to time sequencing, splicing principle is consistent with the principle that can be adopted by attack time axis information of target, forms target N _ibe detected time shaft GT _i.Time shaft GT _imay be discontinuous, this time shaft comprises target N _ican by the information of certain several sensor detection in section sometime.The information that each time period stores has: target, sensor, detectable time period initial time, detectable end time time period.As shown in Figure 3, target N _ithe time shaft GT that can be detected by sensor _icomprise some time section, such as the 2nd time period can by sensor G ₁and G ₂detection.

Step 2.3: determine the time shaft that target can be blocked, i.e. the initial solution space of problem.

First, step 2.1 is obtained target N _i(i=1 ..., N) can by weapon attacking time shaft MT _i, all time points on extraction time axle, obtain the most original weapon attacking time point information, sort according to time point.

Secondly, the target N that step 2.2 is obtained _i(i=1 ..., N) be detected time shaft GT _i, all time points on extraction time axle, obtain the most original sensor detection time dot information, sort according to time point.

3rd, two that above two steps are obtained orderly time point sequences are carried out fusion and are arranged, and obtain the new orderly time point sequence comprising weapon attacking and sensor detection information.The information of same time point is merged, and the information in different time points is by judging the information of adding weapon attacking or the sensor detection lacked accordingly.What now obtain is still original object time dot information.

4th, the target N obtained by the 3rd step _i(i=1 ..., N) all time point informations, obtain target N _i(i=1 ..., N) and initial time slice and information thereof, form target N _ibe blocked time shaft T _i, time shaft T _imay be discontinuous, it comprises target N _iin section sometime can by certain several weapon attacking and in this time period adoptable sensor carry out the information detected.The be blocked time shaft of all targets constitutes initial solution space A ₁.

As shown in Figure 4, can attack time axle by the target-weapon of Fig. 2, and on the detectable time shaft of target-sensor of Fig. 3, information arranges according to time sequencing, obtains being blocked time shaft.

In the inventive method, use the mode of binary digit coding to record certain time period upper assailable weapon and detectable sensor information in solution space, record the coding binary M position altogether of weapon information in each time period, represent weapon M successively from right to left ₁..., M _mif corresponding position is 1, then represent weapon corresponding to this binary digit can at this moment between can attack target in section; Be that 0 expression can not be attacked.Same, record the coding binary G position altogether of sensor information in each time period, from right to left representative sensor G successively ₁..., G _gif corresponding position is 1, then represent sensor corresponding to this binary digit can at this moment between can detect target in section, be 0 and represent that corresponding sensor can not detect target.

Step 3: reduce initial solution space, comprise step 3.1 and step 3.2.

Step 3.1, reduces solution space according to weapon-sensors association constraint.

In initial solution space, travel through all targets, for certain target N _i, whether have corresponding interconnection constraint, if having the interconnection constraint of weapon-sensor, then at solution space A about intrafascicular the searching of weapon-sensor ₁in find this target N _icorresponding solution---time shaft T _iif, weapon M _jwith sensor G _kthere is interconnection constraint, travel through all time periods on this time shaft, to can by weapon M _jcertain time period of attacking, judge whether this time period meets particular community sensor G _krequirement, if do not met, then can directly by weapon M in this time period _jto target N _ithe information of carrying out attacking is rejected, if meet particular community sensor G _krequirement, then the sequential search next time period, if the next time period can by weapon M _jattack, also meet particular community sensor G _krequirement, then again order search downwards, can not by M until find _jattack or do not meet particular community sensor G _kthe time period required.What now judge to find allly meets particular community sensor G _kthe time period T required _i.x1~ T _i.x2whether cover weapon M _jwith sensor G _ksection correlation time, if section correlation time can not be covered, then by time shaft T _ion section T continuous time _i.x1~ T _i.x2middle weapon M _jto target N _ithe information of carrying out attacking is rejected.Idiographic flow as shown in Figure 5.

After operation above, may comprise in solution space can not by the time period of any weapon attacking, and these belong to invalid solution space, therefore, also needs to be rejected in solution space by the time period of any weapon attacking.Reduce solution space further.

Solution space after note reduces is A ₂.

Below in conjunction with Fig. 5 and Fig. 6, the realization of this step is described with an example.

Target N _ibe blocked time shaft T _ion time period T _i.1, T _i.2, T _i.3, T _i.4, T _i.5respectively can by weapon M ₁, sensor G ₁, weapon M ₁and M ₂, sensor G ₁, weapon M ₂, sensor G ₁and G ₂, weapon M ₃, sensor G ₂, weapon M ₁, sensor G ₃attack or detect.Suppose that all interconnection constraints are as follows: attribute is C _m1weapon need with 1 attribute be C _g1sensors association, and attribute is C _m1weapon only have M ₁, attribute is C _g1sensor only have G ₁, then now be associated as weapon M ₁need and sensor G ₁be associated, equally, M ₃the weapon of corresponding attribute needs and 1 G ₂the sensor of corresponding attribute is associated.Weapon M ₁with sensor G ₁section correlation time, and weapon M ₃with sensor G ₂correlation time section as shown in Figure 6.Record in the temporal information of the association that correlation time, section inputted in step 1.

Interconnection constraint is got in circulation successively, and that first time circulation obtains is M ₁with G ₁interconnection constraint, then obtained in time period above by circulation, be designated as T _i.x(x ∈ 1,2,3,4,5}), parameters TimeSlice ₁be used for recording the sensor detectable time period, it is as follows to carry out step:

A. this time period T is judged _i.xcan by weapon M ₁attack, if cannot be attacked, then order gets T _i.xnext time period T _{i. (x+1)}, proceed this judgement; If can be attacked, then check the sensor detection information that this time period is corresponding, perform b;

B. G is judged ₁can detect it, if can not, then to T _i.xperform c, if can detect, then to T _i.xperform d;

C. by M in this time period ₁the information that current goal is attacked can be rejected from solution space, judge whether all time periods on the be blocked time shaft of current goal have all traveled through, if do not have, continue to get next time period T _{i. (x+1)}, to T _{i. (x+1)}perform a;

D. this time period T is recorded _i.xtemporal information, be added to TimeSlice ₁in, from then on position starts to get next time period T _{i. (x+1)}, to T _{i. (x+1)}perform e;

E. judge that can time slice by weapon M ₁attack, if cannot be attacked, then perform g, if can be attacked, then check the sensor detection information execution f that this time period is corresponding;

F. this time period upper G is judged ₁can to detecting, if can not, then perform g.If can detect, then perform d;

G. TimeSlice is judged ₁whether comprise weapon M ₁with sensor G ₁section correlation time, if can not, by TimeSlice ₁in namely by weapon M ₁attacking again can by sensor G ₁each time period of detection performs c; Otherwise, illustrate that the detection time of the sensor with interconnection constraint can cover correlation time, meet interconnection constraint, retain this time frag info, do not need to do any process, then judge whether all time periods on the be blocked time shaft of current goal have all traveled through, if no, order gets T _i.xthe next time period, then turn a perform.

Complete the process reducing solution space for an interconnection constraint to this, similar process is carried out in desirable next interconnection constraint, reduces solution space.In Fig. 6, for M ₁with G ₁interconnection constraint, first get time period T _i.1, perform a, b, d, by T _i.1temporal information to be added to TimeSlice ₁in, then get time period T _i.2; Perform e, f, d, by T _i.2temporal information to be added to TimeSlice ₁in; Continue to get time period T _i.3, perform e, g; Continue to get time period T _i.4, perform a; Continue to get time period T _i.5, perform a, b, c, at time shaft T _iupper erasing time section T _i.5information.For M ₃with G ₂interconnection constraint, time period T _i.1~ T _i.3all can not by weapon M ₃attack, time period T _i.4can by weapon M ₃attack, and can by M ₃the sensor G of association ₂detect, then T _i.4temporal information to be added to TimeSlice ₁in; Due to T _i.4do not cover M ₃with G ₂section correlation time, at time shaft T _iupper erasing time section T _i.4information.

After operation above, may comprise in solution space can not by the time period of any weapon attacking, and these belong to invalid solution space, as being labeled as the time period of none in Fig. 6, therefore, also need to be rejected in solution space by the time period of any weapon attacking.Reduce solution space further.Shown in Fig. 6, the be blocked time shaft of the target Ni after step 3.1 reduces is by T _i.1~ T _i.3composition.

Solution space after note reduces is A ₂.

Step 3.2: reduce solution space for specific objective arranges intercept point constraint.

Owing to there is constraint: those priority of resource allocation are not worth S _ibe more than or equal to given priority level value S _giventarget, at least need to arrange an intercept point.

The execution flow process of this step as shown in Figure 7, travels through all targets, checks target N _ipriority of resource allocation be not worth S _iif, S _i>=S _given, then perform the operation of following first to fourth:

The first, calculate average kill probability.To target N _itime shaft T can be blocked _ion all time period T _i.x(x=1,2 ...), to calculate in each time period weapon to the average kill probability of target =this time period interior kill probability to all weapons that this target is attacked and all weapon numbers of in/this time period, this target being attacked;

The second, the selected time period that intercept point is set.Elect section sometime as need to arrange intercept point time period, time period T _i.xwith probability selected, namely select this time period by the mode of roulette.Intercept point is set to by by the starting point of time period selected;

3rd, distribute ammunition number.Weapon M on this intercept point _l, l ∈ 1 ..., M} is to this target N _ithe ammunition number distributed is B _l* (S _i/ Σ S _i) * (qp), wherein p is M _lto N _iall can the length of attack time section, q is the length of the time period of current selection, B _lrepresent weapon M _lammunition, Σ S _irepresent and the priority of resource allocation of all targets is sued for peace.If this time period has multiple weapon can to target N _iattack, then from the weapon that kill probability is high, carry out the distribution of ammunition number, the distribution ammunition number of gained rounds up, and avoids the problem occurring distribution 0 ammunition.

4th, distribute sensor.After weapon and ammunition distribute, need to distribute sensor to target N _idetect.If there is the interconnection constraint of weapon and particular sensor, be then target N according to interconnection constraint _idistributing sensor and detection time, if onrelevant constraint, is target N according to nearby principle _iassigning sensor, is target N according to minimum principle _idistributing the detection time of sensor, can the detection time of specified sensor be selected time period length.For detection time, there is such process herein: when distributing sensor detection time, join in detection time transfer time by the detection of this sensor, the selected time period is (time1 ~ time2), then the detection time distributing sensor is (time1 ~ time2+T _d_ G _k).Such as, sensor G is assigned with _kdetected by the 30th second at the 5th second, then the actual detection time provided can be the 5th second to (30+T _d_ G _k) detect second.Such process simplifies the process distributing sensor detection time, improves the efficiency of distributing sensor, and ensure that the needs of sensor to detection transfer time.

5th, upgrade resource status.After having traveled through all targets, weapon and the detectable destination number of residue ammunition, sensor and residue thereof thereof all there occurs change, need the state upgrading corresponding weapon and sensor, and upgrade remaining available resource.

6th, reduce solution space.The time period of distributing intercept point is rejected from solution space; Check the surplus resources of weapon and sensor, if there is residue ammunition is the weapon of zero, this weapon in the time period of solution space is rejected the information that target is attacked; After above step, can not be rejected in solution space by the time period of any weapon attacking, reduce solution space.

Solution space after note reduces is A ₃.

Step 4: coding reconstruct is carried out to solution space.Each step is all operate from the angle of target time shaft above, in order to optimize conveniently, need to carry out coding reconstruct to solution space, namely operating from the angle against time axle of weapon.Step 4 comprises three sub-steps: step 4.1, reconstruct solution space; Step 4.2, divides the solution space of reconstruct; Step 4.3, encodes to solution space.For follow-up optimization is prepared.

Step 4.1: reconstruct solution space.

To solution space A ₃the time shaft that can carry out tackling according to weapon is reconstructed.Composition graphs 8, is explained as follows reconstructing method:

Step 4.1.1, obtains all time periods that weapon can carry out attacking.Travel through all weapons, for weapon M _j(j=1 ..., M), at solution space A ₃in obtain and allly comprise M _jtime period, formed reconstruct after initial solution space; Such as, in Fig. 8, for M ₁, find this weapon to target N ₁and N ₂the assailable time period.

Step 4.1.2, the fractionation initial solution space after reconstruct being carried out to the time period is merged.Need a point situation to discuss herein, these situations comprise: two time slices overlap completely, now need information to merge, and only retain a time slice; Next time slice comprises a time slice, now needs the information of a supplementary upper time slice, changes the initial time of next time slice; A upper time slice comprises next time slice, now needs a upper time slice to split into three little time slices, and supplements weapon and sensor information respectively; Next time slice is the latter half of a upper time slice, now needs the end time changing a upper time slice, supplements the information of next time slice; Part of occuring simultaneously is the forward part of next time slice, now needs two time slices to split into three, and supplements respective information respectively; Two time slices, without common factor, now do not need to change the information of time slice.Form each weapon M _jcan commit time axle TM _j, this time shaft may be discontinuous, this time shaft comprises weapon M _jthe target can hit in section sometime and spendable sensor information, j=1 ..., M.

Step 4.1.3, obtains the solution space reconstructed.Travel through all weapons, time shaft corresponding for each weapon is arranged according to time sequencing, obtain the solution space reconstructed.Solution space after note reconstruct is A ₄.

Step 4.2: divide solution space.

First, calculate minimum interception and need use the time.Minimum interception need with comprising reserved T observing time in the time _d, and T launch time of weapon _fwith can attack time T _lbetween linear relationship.T when transmitted _fwith can attack time T _lbetween be the simplest linear relationship time, minimum interception can need be expressed as with the time: minimum interception need with time=T _d+ | T _l-T _f|.

Secondly, to weapon M _j(j=1 ..., M), by time shaft TM _jon each time period need divide with the time by minimum interception, and need to reject being less than minimum interception in solution space with the time period of time.

Finally, computing time axle TM _jon the minimum interception that is cut into of live part need add up to ML with the time period _jindividual.

Step 4.3: solution-vector coding.

Travel through all weapons, for weapon M _j(j=1 ..., M):

First, M is obtained _jtime shaft TM _jon time period sum, be designated as ML _j;

Secondly, by TM _jthe time period of upper all interceptings is mapped to ML _jthe integer point set of individual real number composition 1,2 ..., ML _jon;

Finally, interval (0, ML is got _j], using this interval as weapon M _jammunition can distribute interval.

Step 5: utilize differential evolution algorithm to find optimal solution.

Step 5.1: selected optimized variable and objective function.

Selected weapon M _jeach ammunition be optimized variable in the position that it can distribute on interval.Then optimized variable has W=Σ B _j(j=1,2 ..., M) and individual.Ammunition (s=1,2 ..., B _j) the interval distributed at (0, ML _j] on, allocation result is a real number, rounds up to this real number, obtain point set 1,2 ..., ML _jon sequence number.Herein (s=1,2 ..., B _j) in s represent weapon M _jall ammunitions (from 1 to B _j), be to weapon M _jall ammunitions identify.

Optimized variable is weapon M _j(j=1 ..., M) each ammunition (s=1,2 ..., B _j; J=1,2 ..., M) and be optimized variable in the position that it can distribute on interval.

Objective function is:

Wherein: m _ifor target N _ithe quantity of the intercept point distributed.λ _ifor scale-up factor, determine according to actual conditions.Total a kind weapon j to the allocation result of certain target interception point ₁..., j _atackle it, the kill probability value of often kind of weapon is ..., the ammunition quantity of often kind of weapon allocation is ...,

Step 5.2: produce initial population according to certain principle.

If make x _et () is that t is individual for e in population, then

x _e(t)=(x _el(t),x _e2(t),...,x _eW(t)),e＝1,2,......,NP；t＝1,2,......,t _max

Wherein, x _ethe corresponding optimized variable of each component of (t), W forms individual chromosome number, i.e. the number of optimized variable; NP is population scale; t _maxit is maximum evolutionary generation.

NP individuality is produced at random according to the span of optimized variable.

Each individuality in population produces by the following method:

Weapon M is obtained successively according to weapon kill probability order from high to low _j, to weapon M _jall available ammunition (s=1,2 ..., B _j) be assigned to interval (0, ML by non-uniform probability _j] on.

For a certain segment, if weapon can zeal in this segment, then use the mode of roulette, with probability S _i/ Σ S _isome in these targets selected, distributes unitedly the ammunition in this segment in selected target.

By ammunition by its interval distributed and Target Assignment in the starting point of the commit time section corresponding to it, resource whether is had to use conflict (sensor resource uses constraint) to the individuality inspection generated, if there is conflict, even distribute by this scheme, the target number that this time period inner sensor can detect simultaneously has exceeded the quantity of the target that sensor can detect simultaneously, then regenerate.

Add up the intercept point number of each target, calculate the fitness value of each individuality according to objective function.

Step 5.3: produce test vector.

From current t for 3 the individual x of Stochastic choice population _p1(t), x _p2(t) and x _p3(t), then t+1 for population e population at individual r optimized variable corresponding to test vector h _er(t+1) be:

h _er(t+1)＝x _p1r(t)+F*(x _p2r(t)-x _p3r(t))，e≠p1≠p2≠p3

Wherein,

X _p1r(t), x _p2r(t) and x _p3rt () represents individual x respectively _p1(t), x _p2(t) and x _p3r the optimized variable of (t);

X _p2r(t)-x _p3rt () is differentiation vector; F is zoom factor.

During evolution, in order to ensure the validity of separating, must judge whether test vector meets boundary condition, if do not meet boundary condition, then regenerate test vector by random device, the individual production method generating initial population in Methods and steps 5.2 is identical.Boundary condition refers to that the distribution of ammunition is interval.

Step 5.4: produce filial generation.

Apply the t+1 of crossover operator generation in population, r filial generation v of e population at individual _er(t+1) be:

Wherein, randl _erbe the random decimal between [0,1], CR is crossover probability, and CR ∈ [0,1], rand (e) is the random integers between [1, N], and this Crossover Strategy can guarantee x _e(t+1) one-component and x is had at least _et the respective component of () is relevant.

Whether carry out Target Assignment to the filial generation generated and check each individuality to have resource to use conflict, if having, regenerate by random device, the individual production method generating initial population in Methods and steps 5.2 is identical.

Add up the intercept point number of each target, calculate the fitness value of each individuality in filial generation.

Step 5.5: Population Regeneration.By filial generation v _eand parent x (t+1) _et the fitness of () compares, the t+1 of renewal is for the e in population individual x _e(t+1) be:

$x_e(t+1)=\left\{\begin{array}{cc}{v}_e(t+1),& f(v_{\mathrm{el}}(t+1),\&CenterDot;\&CenterDot;\&CenterDot;,v_{\mathrm{en}}(t+1))<f(x_{\mathrm{el}}\left(t\right),\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{en}}\left(t\right))\\ x_e\left(t\right),& f(v_{\mathrm{el}}(t+1),\&CenterDot;\&CenterDot;\&CenterDot;,v_{\mathrm{en}}(t+1))\&GreaterEqual;f(x_{\mathrm{el}}\left(t\right),\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{en}}\left(t\right))\end{array}\right.---\left(4\right)$

Wherein, f (v _el(t+1) ..., v _en(t+1)) be filial generation v _e(t+1) fitness, f (x _el(t) ..., x _en(t)) be parent x _ethe fitness of (t).

Then, whether inspection obtains optimal result.

Judge whether the individuality of fitness optimum in new population meets formula (1), if meet, then termination of iterations, returns fitness optimum individual, otherwise, go to step 5.3 execution.

Or judge whether t+1 has exceeded iteration upper limit t _maxif exceeded the iteration upper limit, then termination of iterations.Now return fitness optimum individual, otherwise, return step 5.3 and continue to perform.

Step 5.6: distribute surplus resources.

For the optimum individual selected, if there is surplus resources, when not producing conflict, priority allocation to the high target of priority, or adopts distribution principle nearby to distribute.

Step 5.7: Gray code is optimized result.

Allocation result after the optimization obtained is carried out Gray code, namely contrary with the process of step 4.3, determine the intercept point position of Target Assignment and weapon, the sensor resource of distribution.

So far, the Resourse Distribute to weapon, sensor and target is completed.

Claims (2)

1. solve a method for Weapon-Target Assignment Problem based on differential evolution algorithm, it is characterized in that, the sensor resource of defend equipment is added Weapon-Target Assignment Problem, and solution specifically comprises the steps:

Step one: the input variable and the initializing variable that obtain Weapon-Target Assignment Problem;

Input variable comprises: target number N, weapon number M, number of probes G; The priority of resource allocation S of each target _i, i=1,2 ..., N; The ammunition quantity B initial or often kind of current weapon has _j, j=1,2 ..., M; The destination number O initial or current each sensor can detect simultaneously _k, k=1,2 ..., G; The belligerent matrix of target-weapon, comprising information has: weapon sequence number, target sequence number and correspondence thereof can commit time section, can launch time section, can weapon corresponding to intercept point and target distance from; The detectable matrix of target-sensor, comprising information has: the target in the detectable time period of sensor sequence number, target sequence number and correspondence thereof, detectable time period and the distance between sensor; The interconnection constraint of weapon and sensor, the weapon comprising association and sensor information, the temporal information associated;

Initializing variable comprises: the property value C of each target _ni, the property value C of each weapon _mj, the property value C of each sensor _gk, a jth weapon M _jto i-th target N _ithe estimated value P of kill probability _ij, detection T transfer time of each sensor _d_ G _k, wherein, i=1,2 ..., N, j=1,2 ..., M, k=1,2 ..., G; Reserve T observing time _d; Given kill probability value P _given; Given priority of resource allocation is not worth S _given;

Step 2: determine larger solution space, comprise step 2.1 ~ step 2.3;

Step 2.1: determine that target can by the time shaft of weapon attacking, specifically: from the belligerent matrix of target-weapon, find i-th target N _ican by weapon M _jsection MT effective time attacked _ij, by MT _ijsplice in chronological order, form target N _ican by the time shaft MT of weapon attacking _i, time shaft MT _ion the information that stores of each time period have: target, weapon, can attack time section initial time, can the attack time section end time; Wherein, i=1,2 ..., N, j=1,2 ..., M;

Step 2.2: determine the time shaft that target can be detected by sensor, specifically: from the detectable matrix of sensor-target, find out N _ican by sensor G _ksection GT effective time of detection _ik, by GT _iksplice according to time sequencing, form target N _ibe detected time shaft GT _i, time shaft GT _ion the information that stores of each time period have: target, sensor, detectable time period initial time, detectable end time time period; Wherein, i=1,2 ..., N, k=1,2 ..., G;

Step 2.3: by target N _ican by weapon attacking time shaft MT _iwith can by sensor detection time axle GT _iarranging according to time sequencing, by time shaft there being overlapping part merge, obtaining target N _ibe blocked time shaft T _i; N number of time shaft corresponding to N number of target constitutes initial solution space A ₁; Wherein, i=1,2 ..., N;

Use the weapon in each target of mode record each time period of binary digit coding and sensor information; The coding binary figure place recording weapon information in each time period is identical with the number of weapon, from right to left, each represents the weapon that weapon sequence number is serial number, the value of every is 1 or 0,1 represents that corresponding weapon can be attacked target within this time period, and 0 represents that corresponding weapon can not be attacked target within this time period; The coding binary figure place recording sensor information in each time period is identical with the number of sensor, from right to left, each representative sensor sequence number is the sensor of serial number, the value of every is 1 or 0,1 represents that respective sensor can detect target within this time period, and 0 represents that respective sensor can not detect target within this time period;

Step 3: reduce initial solution space, comprise step 3.1 ~ step 3.2;

Step 3.1: traversal has the be blocked time shaft of the target of the interconnection constraint of weapon-sensor, reject the weapon attacking information not meeting the interconnection constraint of weapon-sensor, concrete methods of realizing is: to the target N of interconnection constraint having weapon-sensor _i, for the interconnection constraint of every bar weapon-sensor, perform and operate as follows:

If weapon M _jwith sensor G _kthere is interconnection constraint, traversal target N _ibe blocked time shaft T _ion all time periods, to can by weapon M _jcertain time period of attacking, judge whether meet sensor G in this time period _krequirement, if do not meet, then at time shaft T _ion by weapon M in this time period _jto target N _ithe information of attacking is rejected, if meet, order gets the next time period, if can by weapon M in this time period _jattack and meet again sensor G _krequirement, then continue to get subsequent time period, until can not by M _jattack or do not meet sensor G _krequirement; Now, judge all can by weapon M _jattack and meet again sensor G _kthe time period T of requirement _i.x1~ T _i.x2, whether cover weapon M _jwith sensor G _ksection correlation time, if not, then at time shaft T _ion by time period T _i.x1~ T _i.x2interior weapon M _jto target N _ithe information of attacking is rejected;

Then can not be rejected in solution space by the time period of any weapon attacking, obtain the solution space A reduced ₂;

Step 3.2: traversal target is priority of resource allocation S _ibe more than or equal to given priority level value S _giventarget, an intercept point is at least set; After having traveled through all targets, the more state of modern weapons and sensor, and surplus resources; Then, rejected from solution space the time period of distributing intercept point, the attack information being the weapon of zero by residue ammunition is rejected in solution space; Finally can not be rejected in solution space by the time period of any weapon attacking, obtain the solution space A reduced ₃;

Step 4: coding reconstruct is carried out to solution space, comprises step 4.1 ~ step 4.3;

Step 4.1: to solution space A ₃the time shaft that can carry out tackling according to weapon is reconstructed, and obtains the solution space A after reconstructing ₄; To weapon M _j, at solution space A ₃middle travel time section, obtains all containing M _jthe time period of attack information, all time periods obtained are spliced according to time sequencing, form weapon M _jcan commit time axle TM _j, j=1,2 ..., M;

Step 4.2: by weapon M _jcan commit time axle TM _jon each time period need divide with the time by minimum interception, and need to reject being less than minimum interception with the time period of time, final time shaft TM _jby ML _jindividual minimum interception need form with the time period, j=1, and 2 ..., M; Described minimum interception need comprise reserved T observing time with the time _d, and T launch time of weapon _fwith can attack time T _lbetween linear relationship;

Step 4.3: by weapon M _jcan commit time axle TM _jon ML _jthe individual time period is mapped to ML _jthe real number point set of individual real number composition 1,2 ..., ML _jon; Get interval (0, ML _j] as weapon M _jammunition can distribute interval; J=1,2 ..., M;

Step 5: utilize differential evolution algorithm to find optimal solution, comprise step 5.1 ~ step 5.7;

Step 5.1: selected weapon M _jeach ammunition B _jsbe optimized variable in the position that it can distribute on interval, j=1,2 ..., M, s=1,2 ..., B _j; Objective definition function f is:

Wherein, m _ifor target N _ithe quantity of the intercept point distributed; λ _ifor scale-up factor; b _ju, j _u∈ { j ₁..., j _arepresent: to target N _ithe allocation result of intercept point is total a kind weapon j ₁..., j _atackle it, the kill probability value of often kind of weapon is the ammunition quantity of often kind of weapon allocation is

Step 5.2: produce initial population;

T is for the e in population individual x _et () is expressed as:

x _e(t)＝(x _e1(t),x _e2(t),…,x _eW(t))，e＝1,2,......，NP；t＝1,2,......,t _max

Wherein, x _et the corresponding optimized variable of each component of (), W forms individual chromosome number, and NP is population scale, t _maxit is maximum evolutionary generation;

Each individuality in population produces by the following method: according to weapon kill probability order from high to low, to weapon M _jall available ammunition B _js(s=1,2 ..., B _j) be assigned to interval (0, ML by non-uniform probability _j] on; If weapon can attack the target of more than 2 in a certain segment, with probability S _i/ Σ S _ione of them target selected, distributes unitedly in selected target by the ammunition in this segment; By ammunition by its distribute interval and Target Assignment in the starting point of corresponding commit time section, to generate individuality inspection whether have resource use conflict, if there is conflict, regenerate;

Add up the intercept point number of each target, calculate the fitness value of each individuality according to objective function;

Step 5.3: produce test vector, specifically: from current t for 3 the individual x of Stochastic choice population _p1(t), x _p2(t) and x _p3(t), then t+1 for population e population at individual r optimized variable corresponding to test vector h _er(t+1) be:

h _er(t+1)＝x _p1r(t)+F*(x _p2r(t)-x _p3r(t))，e≠p1≠p2≠p3

Wherein, x _p1r(t), x _p2r(t) and x _p3rt () represents individual x respectively _p1(t), x _p2(t) and x _p3r the optimized variable of (t); x _p2r(t)-x _p3rt () is differentiation vector; F is zoom factor;

If the distribution that the test vector produced is not in ammunition is interval, then regenerate test vector according to the individual production method of population in step 5.2;

Step 5.4: produce filial generation, r filial generation v of e population at individual _er(t+1) be:

Wherein, randl _erbe the random decimal between [0,1], CR is crossover probability, CR ∈ [0,1], and rand (e) is the random integers between [1, N];

Target Assignment carried out to the filial generation generated and checks each individuality whether to have resource to use conflict, if there is conflict, regenerating according to the individual production method of population in step 5.2;

Add up the intercept point number of each target, calculate the fitness value of each individuality in filial generation according to objective function;

Step 5.5: Population Regeneration: upgrade t+1 for the e in population individual x _e(t+1) be:

$x_e(t+1)=\left\{\begin{array}{cc}{v}_e(t+1),& f\left(v_{e1}\right(t+1),...,v_{en}(t+1\left)\right)<f\left(x_{e1}\right(t),...,x_{en}(t\left)\right)\\ x_e\left(t\right),& f\left(v_{e1}\right(t+1),...,v_{en}(t+1\left)\right)\&GreaterEqual;f\left(x_{e1}\right(t),...,x_{en}(t\left)\right)\end{array}\right.$

F (v _e1(t+1) ..., v _en(t+1)) be filial generation v _e(t+1) fitness, f (x _e1(t) ..., x _en(t)) be parent x _ethe fitness of (t);

Judge whether the individuality of fitness optimum in new population meets formula: if meet, termination of iterations, returns fitness optimum individual, otherwise, go to step 5.3 execution;

Or judge whether t+1 has exceeded iteration upper limit t _maxif exceed, then termination of iterations, returns fitness optimum individual, otherwise, go to step 5.3 execution;

Step 5.6: for the optimum individual selected, if there is surplus resources, when not producing conflict, priority allocation to the high target of priority, or adopts distribution principle nearby to distribute;

Step 5.7: the allocation result after the optimization obtained is carried out Gray code, determines the intercept point position of Target Assignment and weapon, the sensor resource of distribution.

2. the method solving Weapon-Target Assignment Problem based on differential evolution algorithm according to claim 1, it is characterized in that, the traversal target described in step 3.2, is priority of resource allocation S _ibe more than or equal to given priority level value S _giventarget, at least arrange an intercept point, concrete methods of realizing is: travel through all targets, check target N _ipriority of resource allocation be not worth S _iif, S _i>=S _given, then the operation of (1) and (2) below performing:

(1) weapon and ammunition is distributed: to target N _itime shaft T can be blocked _ion all time period T _i.x(x=1,2 ...), to calculate in each time period weapon to target N _iaverage kill probability then with probability select time section T _i.xas needing the time period arranging intercept point; The starting point of the time period of selection is set to intercept point, weapon M on this intercept point _l, l ∈ 1 ..., M} is to target N _ithe ammunition number distributed is B _l* (S _i/ Σ S _i) * (q/p), wherein, p is M _lto target N _iall can the length of attack time section, q is the length of the time period of current selection, B _lrepresent weapon M _lammunition; If there is plural weapon can to target N in time period of selecting _iattack, carry out the distribution of ammunition number from the weapon that kill probability is high, the distribution ammunition number of gained rounds up;

(2) sensor is distributed; If there is the interconnection constraint of weapon and particular sensor, be then target N according to interconnection constraint _idistributing sensor and detection time, if onrelevant constraint, is target N according to nearby principle _iassign sensor; If the selected time period is (time1 ~ time2), then the detection time distributing sensor is (time1 ~ time2+T _d_ G _k).