CN115730700A - Self-adaptive multi-target task planning method, system and equipment based on reference point - Google Patents
Self-adaptive multi-target task planning method, system and equipment based on reference point Download PDFInfo
- Publication number
- CN115730700A CN115730700A CN202211201299.8A CN202211201299A CN115730700A CN 115730700 A CN115730700 A CN 115730700A CN 202211201299 A CN202211201299 A CN 202211201299A CN 115730700 A CN115730700 A CN 115730700A
- Authority
- CN
- China
- Prior art keywords
- task
- population
- training
- helicopter
- solution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a self-adaptive multi-target task planning method, a system and equipment based on a reference point, wherein the method comprises the following steps: acquiring known task parameters, calling the constructed amphibious helicopter training task planning model, calculating population scale, constructing an initial population, and performing cross operation, mutation operation and local optimization search operation on a parent population to generate a child population; and combining the generated offspring population and the parent population to calculate a standardized solution set, and generating a reference point set and a corresponding set of the Liji solution set on the standardized solution set by using a reference point adaptive generation method based on population distribution. Constructing a new generation population by using a clustering method based on an interior angle measurement method and a scaling sorting method; and (5) performing iterative optimization until the population iteration times reach the maximum iteration times, and outputting training task planning scheme data of the amphibious helicopter. The rapid planning of the training task of the amphibious helicopter is realized more efficiently.
Description
Technical Field
The invention belongs to the technical field of task planning, and relates to a self-adaptive multi-target task planning method, system and equipment based on a reference point.
Background
In an amphibious task, a helicopter taking a marine forward base as a take-off and landing platform (LHD) can cross obstacles on a water beach by virtue of the characteristics of flexible deployment and quick concealment, and task resources are delivered at a shallow depth and a deep depth of a landing field to form local advantages at the same time, so that the sudden and quick decision of the amphibious task is achieved. However, because helicopters have the natural defects of complex guarantee, low output intensity, limited carrying capacity, easy interference and the like, the helicopters are more limited in use and have higher task risk. Therefore, the mission of the amphibious helicopter needs to be finely planned so as to fully exert the advantages thereof and simultaneously reduce the loss as much as possible.
Currently, mission planning research on aircraft grouping mainly focuses on the field of UAV (Unmanned Aerial Vehicle) cluster action planning, which can be roughly divided into two parts, namely mission allocation research and route planning research. The content of the task allocation research mainly comprises allocation decision of tasks among different resources and platforms and optimal scheduling of task execution sequence, and the current mature models in the field comprise a Traveling Salesman Problem (TSP) model, a Vehicle Routing Problem (VRP) model and a mixed integer linear programming problem (MILP) model. Traditional training task planning models generally aim to achieve the maximum results, sustain the minimum losses, complete tasks in the shortest time, and make the most of human resources.
In reality, most training task planning problems are Multi-objective Optimization problems (MOP). At the present stage, a Multi-object evolution Algorithm (MOEA) is proved to be the most effective way to solve the MOP, and can be roughly divided into three categories, namely, based on Pareto dominance, based on indexes and based on decomposition, according to the solution idea. However, the traditional training mission planning method cannot directly solve the technical problem of rapid planning of training missions of the amphibious helicopter in reality.
Disclosure of Invention
Aiming at the problems in the traditional method, the invention provides a reference point-based adaptive multi-target task planning method, a reference point-based adaptive multi-target task planning system and computer equipment, which can be used for rapidly planning training tasks of an amphibious helicopter in an efficient and practical manner.
In order to achieve the above purpose, the embodiment of the invention adopts the following technical scheme:
on one hand, the self-adaptive multi-objective task planning method based on the reference point is provided, and comprises the following steps:
acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, solution space dimensionality, simplex segmentation parameters and the maximum iteration times;
calling the constructed amphibious helicopter training task planning model; constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and an optimization objective function of the amphibious helicopter training mission planning model comprises a time-consuming objective function of a minimized landing phase, a minimized manpower loss objective function and a target function of ground threat of a minimized helicopter;
calculating the population scale of the amphibious helicopter training task planning model based on the task parameters and constructing an initial population;
performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population;
combining the generated child population and the parent population to obtain a combined population, updating IDEAL points and NADIR points according to elements in a solution set corresponding to the combined population, and calculating to obtain a standardized solution set according to the IDEAL points and the NADIR points;
generating a reference point set and a corresponding set of a Liji solution set on the standardized solution set by using a reference point self-adaptive generation method based on population distribution;
distributing each individual in the standardized solution set to a lyre solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each lyre solution set based on a scaling sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation of population;
and returning to the step of performing cross operation, mutation operation and local optimization search operation on the parent population to generate the offspring population, and outputting training task planning scheme data of the amphibious helicopter until the iteration number of the population reaches the maximum iteration number.
In another aspect, a system for adaptive multi-objective task planning based on reference points is provided, including:
the parameter acquisition module is used for acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, solution space dimensionality, simplex segmentation parameters and the maximum iteration times;
the model calling module is used for calling the constructed amphibious helicopter training task planning model; constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and an optimization objective function of the amphibious helicopter training mission planning model comprises a time-consuming objective function of a minimum landing phase, a minimum manpower loss objective function and a ground threat objective function of the minimum helicopter;
the population construction module is used for calculating the population scale of the amphibious helicopter training mission planning model based on the mission parameters and constructing an initial population;
the child generation module is used for performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population;
the merging calculation module is used for merging the generated child population and the parent population to obtain a merged population, updating an IDEAL point and an NADIR point according to elements in a solution set corresponding to the merged population, and calculating to obtain a standardized solution set according to the IDEAL point and the NADIR point;
the set generation module is used for generating a reference point set and a corresponding set of a Liji solution set on the standardized solution set by using a reference point self-adaptive generation method based on population distribution;
the new population module is used for allocating each individual in the standardized solution set to a sharp root solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each sharp root solution set based on a scaled sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation population;
and the iteration control module is used for outputting the training task planning scheme data of the amphibious helicopter when the population iteration times reach the maximum iteration times.
In yet another aspect, a computer device is provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above-mentioned reference point-based adaptive multi-objective task planning method when executing the computer program.
In still another aspect, a computer-readable storage medium is provided, on which a computer program is stored, which, when being executed by a processor, implements the steps of the above-mentioned reference point-based adaptive multi-objective task planning method.
One of the above technical solutions has the following advantages and beneficial effects:
according to the self-adaptive multi-target task planning method, the self-adaptive multi-target task planning system and the self-adaptive multi-target task planning equipment based on the reference points, the main task mode and the characteristics of the helicopter in the amphibious landing operation are extracted and analyzed, the amphibious helicopter training task planning model is provided on the basis of reasonably assuming the real landing training scene and is expressed as a multi-target optimization model, and the aim is to minimize the time consumption of the landing operation, personal casualties of own parties and the threat of the own helicopter to the blue-party airproofing fire. The training mission planning model of the amphibious helicopter considers the specific practical constraints of two aspects of training resources, training and guaranteeing the full-link mission stage time sequence, which can be used by multiple platforms. The reference point self-adaptive generation method based on population distribution is applied to the population sorting selection mechanism, the optimization capability of irregular populations is effectively enhanced on the premise of not increasing the computational complexity, a new initial solution generation, crossing, variation and local optimization method is provided on the population updating iteration mechanism, the search efficiency of a code definition space is better improved, and the optimization efficiency and the population diversity maintenance are well expressed, so that the aim of quickly planning training tasks of the amphibious helicopter is fulfilled.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments or the conventional technologies of the present application, the drawings used in the descriptions of the embodiments or the conventional technologies will be briefly introduced below, it is obvious that the drawings in the following descriptions are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart diagram illustrating a method for adaptive multi-objective task planning based on reference points in one embodiment;
FIG. 2 is a schematic illustration of a flight mission and deck preparation phase of a helicopter formation in one embodiment;
FIG. 3 is a schematic flow diagram of an iterative optimization solution to a model in one embodiment;
FIG. 4 is a schematic diagram of an example of chromosome coding in one embodiment;
FIG. 5 is a schematic illustration of an example helicopter launch plan in one embodiment;
FIG. 6 is a schematic diagram of a process for interleaving one chromosome code with another chromosome code in one embodiment;
FIG. 7 is a schematic view of a hyperplane constructed with points closest to the coordinate axes as Nadir points;
FIG. 8 is a schematic view of a hyperplane constructed with the point with the largest target on each coordinate axis as the Nadir point;
FIG. 9 is a diagram illustrating the results obtained after clustering operations are performed on the normalized solution sets in one embodiment;
FIG. 10 is a schematic diagram of a constructed pseudo-linear hyperplane and a set of reference points in one embodiment, wherein (a) is the constructed pseudo-linear hyperplane and (b) is the generated set of reference points;
FIG. 11 is a polar coordinate space Γ in one embodiment 3 A schematic diagram of the segmentation;
FIG. 12 is a diagram illustrating the result of clustering the normalized solution set according to another embodiment;
FIG. 13 is a block diagram illustrating an example of an adaptive multi-objective mission planning system based on reference points in accordance with an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the present application in the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.
It should be appreciated that reference throughout this application to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present invention. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
One skilled in the art will appreciate that the embodiments described herein may be combined with other embodiments. The term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
Training task analysis: the helicopter marshalling composed of the general helicopter and the special helicopter can execute various tasks including personnel, equipment and material vertical delivery, near-distance air support, low-altitude abstinence control, striking on sea surface targets, anti-submarine mine sweeping and the like in an amphibious landing task. In the amphibious task process, a helicopter taking off from an LHD can land vertically in the depth and depth of a target coast, a delivery machine descends manpower to cooperate with a plane to land, landing team members are assisted to rapidly advance on the target coast, or assault is carried out on target side wings or rear important points, so that a local task target is achieved. At present, the helicopter is limited by the dynamic strength and the load and can only deliver the light task unit in a small scale, so that the vertical landing unit mainly bears the cooperative task of matching with the plane landing unit in a large-scale amphibious landing task.
According to the analysis, the vertical landing task can be abstracted into the distribution of the general helicopter load (airplane landing manpower) of the offshore platform to different points of the target, and the selection and the calculation of factors such as the impact point, the impact opportunity, the airplane landing field and the helicopter route control point are specifically included. The main body for executing the vertical landing task is a general helicopter for the offshore platform, and the helicopter also comprises a special helicopter for the offshore platform, which is used for accompanying navigation, reconnaissance and shielding.
In the implementation process of the vertical landing task, consideration needs to be given to influences on vertical landing task efficiency such as cooperation of the LHD and a ship aircraft of the helicopter, intra-formation cooperation of the helicopter and the aircraft landing manpower, inter-formation cooperation between task formations of different targets in different wave times and the same wave time, inter-work-type cooperation between a vertical landing unit and a plane landing unit, and the like, and specifically includes influences on helicopter takeoff efficiency by platform guarantee capability, influences on manpower delivery efficiency by load, speed and flight of the general helicopter, influences on helicopter formation flight safety by operating field information control right, air control right and target air defense suppression effect, and the like.
By virtue of its characteristic flight characteristics, a helicopter dedicated to offshore platforms can achieve a flight altitude far below that of fixed-wing aircraft and a maneuvering speed far above that of wheeled or tracked equipment. Therefore, in a near-distance air support task implemented mainly by a helicopter special for an offshore platform, a special helicopter marshalling taking off from an LHD deck or standing by in a forward airspace can be quickly maneuvered to a target airspace in a concealed manner by utilizing ultra-low-altitude flight, and has irreplaceable important effects on implementing accurate target assault on important targets such as target vehicles, radar stations, target locations, communication hubs, front sentries, simple works, target sites, surface ships, ground units and the like.
Similar to the vertical landing task, the near-distance air support task can be abstractly understood as the distribution of the helicopter load special for the offshore platform to different targets, and specifically comprises the selection and calculation of the task target, the attack opportunity, the helicopter air route control point and other elements. The near-distance air support task is executed by a helicopter special for an offshore platform, and may also comprise an unmanned aerial vehicle for providing operation field situation reconnaissance and the like. In the task implementation process, the influence of factors such as ship surface guarantee capability, special helicopter load, speed, range, air defense targets and operating field situation on the attack efficiency of the helicopter targets also needs to be considered.
In one embodiment, as shown in fig. 1, the present application provides a method for adaptive multi-objective task planning based on reference points, comprising steps S11 to S18:
s11, acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, the solution space dimensionality, the simplex segmentation parameters and the maximum iteration times.
It is understood that before the optimization calculation is started, each preset relevant task coefficient, such as but not limited to the total number of platforms carrying the takeoff and landing platform of the helicopter and the total number of wave motions of the helicopter, and relevant known parameters of the optimization calculation, such as a solution space dimension, a simplex partition parameter, a maximum iteration number, and the like, may be obtained.
S12, calling the constructed amphibious helicopter training task planning model; the constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and the optimization objective functions of the amphibious helicopter training mission planning model comprise a time-consuming objective function of a minimum landing phase, a minimum manpower loss objective function and a ground threat objective function of a minimum helicopter.
It can be understood that amphibious combined landing is the most complex action pattern, and the training mission planning model comprises too many elements, and appropriate specification assumptions are necessary to make the model relatively simple. For convenience of description, the attacking party of the attacking and defending parties can be called the red party, and the defending party can be called the blue party. The model assumes 1 as: the training tasks of the red party landing are completed by a plane landing team, a vertical landing team and a helicopter together, wherein the training tasks of the plane landing team are formulated by a plan participated before the red party landing, and the helicopter mainly undertakes the vertical landing task and a close distance support task.
The model assumes 2 as: the vertical landing tasks are completed by the general helicopters and the special helicopters together, the types and the number of the vertical landing teams delivered to each target of the blue party and the required number of the two types of helicopters are formulated in a participation plan before action, the short-distance support tasks are completed by the special helicopters only, and the grouping of each short-distance support task is composed of a fixed number of special helicopters.
The model assumes 3 as: the helicopter can not influence the training tasks executed by the helicopter due to failure or damage, the LHD can not be attacked, and the helicopter can stably provide sufficient oil and ammunition supply and various mechanical service guarantees, and in addition, a command information system for logging in the team can provide stable and effective command and communication guarantees for all tasks.
The model assumes 4 as: the condition that the LHD is close to the landing beach (for example 50 km) is selected for research, the time for carrying out the task of ascending and flying the task of the helicopter is short, so that the helicopter is supposed to take double-wave cycle running, and the helicopter group with the same task runs at the same or close to the same wave as much as possible.
The model assumes 5 as: the method comprises the steps of assuming that each helicopter has the same ship surface allocation and transportation, service support, take-off and departure, formation flight (between LHD and landing area) and approach and landing time, assuming that the helicopter flies at a constant speed at an ultra-low altitude in the landing area, wherein the flight time depends on the flight path length of the helicopter in the landing area, and further assuming that each vertical landing task grouping and each close range support task grouping respectively have the same and fixed landing time and ground strike time.
Model dependent variable definition: the relevant variable symbols in the model are given in table 1.
TABLE 1
The model problem constraint comprises two main categories of training resource constraint and task phase time sequence constraint, wherein the training resource constraint comprises the following steps:
(1) And (3) restricting the number of vertical login teams: according to the model assumption 4, in order to deliver more vertical landing teams simultaneously in one wave, the vertical landing teams should be deployed on different LHDs in a scattered manner, and since the total number of vertical landing teams that can be input by the red party in the landing movement is limited, the number of vertical landing tasks that each LHD can support must also meet the limit of the number of vertical landing teams loaded by the LHD, specifically,the following constraints should be satisfied:
meanwhile, the total number of all vertical login teams delivered by all vertical login task plans in one login movement also meets the following constraint:
(2) The number of helicopters is restricted: according to model assumption 2, the sum of a certain helicopter required by the task formation of helicopters circularly moved by adjacent waves on a certain LHD is limited by the total number of helicopters loaded by the LHD:
the number of dedicated helicopters required for each individual training mission in equation (3)And number of helicopters in generalThe ratio between them is determined by the task type c, when c =2 the helicopter group performs the close range support task, the task group is only composed of special helicopters, therefore the number of general helicoptersWhen c =1, the helicopter group performs a vertical landing task, and the task group is composed of a general helicopter and a special helicopter for which the accompanying shield is provided, and the number of the special helicopters is generallyNumber of general helicopters should not be less than1/2, it is therefore assumed in this application that:
in the above formula (4)Indicating the number of helicopters required to round up x, and in addition, for each vertical landing missionDepending on the number of various vertical landing teams required for implementing an air assault on Lan Fangdi p nodes in the plan for participation before the action of the red party
(3) LHD guarantees resource constraints: according to model assumption 4, limited to the number of takeoff positions on each LHD flight deck, the number of helicopters grouped by the same wave-time takeoff helicopter mission should satisfy the following formula (6):
in addition, the helicopter consists assigned to the same training mission should behave as much as possible in the same or in adjacent wave groups, and therefore in the formulaThe constraint should be satisfied:
in the above formula (7), n w And (3) representing the minimum wave number required by all the helicopter marshalling in the c-th training task distribution of the target p, namely rounding up of the quotient of the sum of the two types of helicopters required by the task marshalling divided by all the takeoff positions of all the LHDs in the red.
Task phase timing constraints include:
(1) Task phase timing constraints based on execution order: a complete helicopter training task comprises 9 parts, namely a ship surface dispatching stage, a service support stage, a take-off and departure stage, a formation flight stage (cruise) from an LHD to a coastline, a formation flight stage (ultra-low altitude penetration) from the coastline to a target node, a task execution stage, a formation flight stage (ultra-low altitude penetration) from the target node to the coastline, a formation flight (cruise) from the coastline to the LHD, an approach landing stage and the like, wherein the start time and the end time of each stage meet the following constraints in each independent helicopter training task:
in the formula (9)And the consumption time of the kth sub-stage of executing the c-type training task on the p-th node target of the landing field on the first LHD of the red party in the w-th wave is shown. Wherein the consumed time of the ship surface allocation and transportation stage, the aircraft service support stage, the takeoff and departure stage and the approach and landing stageAndfor a hypothetical fixed value, the task execution phase is time consumingHelicopter formation flight time determined by task typeAndthe optimal planned route length is divided by the flight speed of the helicopter formation.
It is noted that although each helicopter training task in the present application contains 9 sub-phases, not all helicopter consists assigned to training tasks will go through these 9 phases in their entirety. For example, in the stage 1 ship surface dispatching, if all task groups of w-th and w + 2-th wave times circulating on a certain LHD need the same number of helicopters, the helicopter in w-th wave motion can directly perform service support on a takeoff position after landing, and then directly lift off to form the task group of w + 2-th wave time to execute training tasks. Thus in equation (9)Satisfying the following formula.
T in formula (10) 1 Indicating that the hypothetical surface deployment is time consuming.
(2) Task stage timing constraints based on LHD deck scheduling: due to the limited space of the LHD flight deck, the flight service support of the helicopter on the LHD usually occupies a takeoff position, so that a helicopter team ready for entering and landing must wait for the emptying of the LHD flight deck (namely, the helicopter takes off next wave time) to land. Therefore, for a helicopter formation going out on the same LHD at adjacent 2 waves, its take-off and landing times must satisfy the following constraints:
in the formula (11), the reaction mixture is,representing the landing time of the helicopter consist performing the c-th training task on the p-th node target of the landing site on the w' -th wave on the first LHD in the red,the departure time of a helicopter consist performing the same task on the w "th wave run on the same LHD is indicated.
As shown in fig. 2, if it is assumed that the helicopter mission groups running on the same wave on all LHDs in the red have to take off simultaneously, after all the helicopter groups on the w wave fall, the helicopter group No. (1) with shorter preparation time for performing service duty on the takeoff position waits for the helicopter group No. (2) requiring deck adjustment to replace the helicopters to take off together after the completion of the service duty. After the helicopters (1) and (2) are completely grouped to lift off, the helicopters (3) and (4) which return to the ground after the task is executed for the w-1 th wave can land on the deck after being emptied.
Model optimization objective: on one hand, as the manpower which can be accommodated by the beach head is limited, the land team on the red side can only be put in wave by wave, and in order to prevent the reverse impact which can be initiated by the blue side at any time, the land team on the first wave must quickly occupy and consolidate the landing field, so that the subsequent team can safely and quickly land; on the other hand, the blue party, who faces firm work, is at a disadvantage in terms of firepower and protection due to the lack of heavy equipment for the red party first wave upland team. Therefore, the Hongfang team must reasonably plan the manpower application and reduce the self casualties as much as possible on the premise of ensuring that the scheduled training tasks are completed on time. And (4) combining the analysis, selecting the time consumed by training, manpower consumption, ground threat of the helicopter and the like in the landing stage of the landing field opened by the red party as the optimization target of the model.
(1) Minimizing time consumption in a login stage: in the amphibious landing action, a defense party usually organizes power and manpower to perform counter-impact when an attacking party is unstable, whether the defense party quickly occupies the land and effectively consolidates the land before the defense party initiates the counter-impact is an important factor for determining whether a landing action is successful or not. Therefore, the total time consumed by all training teams of the red party to finish all established tasks of the stage of opening the landing field is selected as the time-consuming objective function of the landing stage by the model:
z 3 =T fin -T 0 (12)
t in formula (12) fin The method comprises the following steps that the end time of a login stage is shown, namely the time when the first wave land team of the red side finishes opening and consolidating various tasks required by a beach head login field; t is 0 And (4) corresponding to the starting moment of the landing action, namely the starting moment of the first wave helicopter in the red party.
(2) Minimizing manpower loss: the manpower loss of the red square comprises plane landing manpower loss and vertical landing manpower loss, and the manpower loss objective function can be specifically expressed as follows:
in the formula (13), the reaction mixture is,represents the manpower lost in unit time of all training teams of the red party training at the p-th blue party node at the time point t,represents the starting moment of the countermeasure at the p-th blue-party node, T p Expressed in minutes for the duration of challenge. The application utilizes the existing Lanchester equation to calculate the manpower loss of the red and blue parts.
(3) Minimizing ground threat to helicopters: in order to avoid the model from being too complex, the influence of the damaged red helicopter on the manpower grouping of two training tasks and the subsequent available helicopter on the LHD is not considered, and only the blue Fang Fangkong threat accumulation in the helicopter navigation process is used for quantitatively representing the loss of the helicopter:
in formula (14), k Scale The scale factor is expressed in terms of a scale factor,andand E (x, y, t) represents the sum of all ground threat values of each helicopter in the red side and each helicopter in the blue side in unit time at the coordinate (x, y) at the t moment. The calculation of E (x, y, t) is properly simplified, only the air defense fire coefficient at each node of the blue side, the distance between the red side helicopter and the node and the shielding effect of the terrain on the fire of the blue Fang Fangkong are considered, and the E (x, y, t) can be calculated according to the formula (15).
Eta in formula (15) SAM A boolean variable indicating whether or not the node p and the coordinate (x, y) are occluded by the terrain is 1 when not occluded and 0 when occluded;representing the air defense power coefficient at the blue node p; d (x, y, p) represents a planar distance between the node p and the coordinates (x, y). And (3) the position (x, y) of each helicopter in the red at the time t is obtained by solving a path planning based on a dijkstra (Dikstra) algorithm in the later decoding operation.
S13, calculating the population scale of the amphibious helicopter training task planning model based on the task parameters and constructing an initial population;
s14, performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population;
s15, merging the generated offspring population and the parent population to obtain a merged population, updating IDEAL points and NADIR points according to elements in a solution set corresponding to the merged population, and calculating according to the IDEAL points and the NADIR points to obtain a standardized solution set;
s16, generating a reference point set and a corresponding set of a Liji solution set on the standardized solution set by using a reference point self-adaptive generation method based on population distribution;
s17, distributing each individual in the standardized solution set to a radix-based solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each radix-based solution set based on a scaled sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation of population;
and S18, returning to the step of executing cross operation, mutation operation and local optimization searching operation on the parent population to generate the offspring population, and outputting training task planning scheme data of the amphibious helicopter until the iteration number of the population reaches the maximum iteration number.
It can be understood that the traditional planning method is inefficient in solving the problem that the objective function and the constraint are nonlinear, and is sensitive to the weight and the optimization order. With the continuous development of the multi-objective optimization problem in the direction of actualization and complication, the multi-objective evolutionary algorithm becomes a main means for solving the problem.
According to the self-adaptive multi-target task planning method based on the reference points, the main task mode and characteristics of the helicopter in the amphibious landing operation are extracted and analyzed, the amphibious helicopter training task planning model is provided on the basis of reasonably assuming a real landing training scene and is expressed as a multi-target optimization model, and the aim is to minimize the time consumption of the landing operation, personal casualties of own parties and the threat of blue-party airproofness and firepower to own helicopter. The training mission planning model of the amphibious helicopter considers the specific practical constraints of two aspects of training resources, training and guaranteeing the full-link mission stage time sequence, which can be used by multiple platforms. The reference point self-adaptive generation method based on population distribution is applied to the level of a population sorting selection mechanism, the optimization capability of irregular populations is effectively enhanced on the premise of not increasing the computational complexity, a new initial solution generation, intersection, variation and local optimization method is provided in the level of a population updating iteration mechanism, the search efficiency of a coding definition space is better improved, and the optimization efficiency and the population diversity maintenance are well expressed, so that the purpose of quickly planning training tasks of a more efficient and realistic amphibious helicopter is achieved.
In order to express the sequential relation between each training task as simply and clearly as possible, the application adopts a discrete task list coding form similar to VRP. According to model constraint 2, when the total number of LHDs and the total number of oscillations are given, the total number of helicopters required by all training tasks in the chromosome codes is also determined, so that the training tasks of different numbers of helicopters required in the amphibious helicopter training task planning model can lead to different lengths of the chromosome codes. In addition, according to the analysis, all training tasks are difficult to be contained in the corresponding chromosome codes of any decision variables of the amphibious helicopter training task planning model, so that the traditional genetic operator is difficult to realize the cross operation of chromosomes with different lengths, and the problem definition domain is difficult to be effectively searched by virtue of the chromosomes only containing limited coding information. Aiming at the problems, the population updating iterative mechanism of the evolutionary algorithm is improved, the cross operation and variation operation methods suitable for codes with different lengths are redesigned, and an initial solution generation method and a local search method are innovatively provided to improve the optimization efficiency of algorithm processing.
The known parameters include the solution space dimension (denoted m), the simplex segmentation parameter (denoted H), and the maximum number of iterations of the algorithm (denoted g) Iter ). First, the population size | P is calculated according to equation (16) g |:
Then, an Initial Population P is constructed by an Initial Population Generation Method (Initial Population Generation Method based on transformation 1st-Class Neighborhods, IPGM-T1N) based on the search of Class 1 neighborhood 0 (ii) a Second, for the parent population P g Performing crossover, mutation and local optimization operationsTo do so, a progeny population o ″, is produced g+1 (ii) a Thirdly, the generated o ″) g+1 With parent population P g Merging to obtain a merged population P' g+1 According to the combined population P' g+1 Corresponding solution set z' g+1 Element update z in (1) Ideal Parameter and z Nadir Parameters and further calculating according to formula (17) to obtain a normalized solution set
In the above formula (17), m represents a solution space dimension,andrespectively representing the original solution set z and the normalized solution setOf (a). The fourth step is thatOn the basis, a reference point set Lambda and a corresponding set omega of a litykoff solution set are obtained by using the existing reference point generation method based on population distribution, wherein Lambda = { Lambda = (lambda) } 1 ,…,λ n },Ω={ω 1 ,…,ω n },n=|P g L. Fifthly, the standardized solution is collected by using the existing clustering method based on the Inner Angle Measure (IAM)Each of which isAssign to a Riji solution set omega λ E.g. omega, based on the existingScaling-based sequencing Method (SSM) at each omega λ The non-dominant sorting operation is carried out, and the individuals corresponding to the elements with the top rank are selected to form a new generation population P g+1 . Finally, repeating the second step to the fifth step until the iteration number of the population reaches the maximum iteration number g Iter And outputting the finally optimized amphibious helicopter training task planning scheme data.
In one embodiment, several main component relationships of the proposed model optimization solution algorithm are shown in FIG. 3. The chromosome coding mode adopted in the population iteration process is as follows: decomposing all training tasks meeting the condition of setting the number of helicopter marshalls into a single task list with the number of helicopter marshalls equal to 1 and adopting discrete double-row chromosome coding; adding a gap gene in the chromosome coding of the task list,
specifically, as shown in FIG. 4, all conditions that satisfy the set number of helicopter consists (i.e., all conditions satisfy the set number of helicopter consists)) The training task is decomposed into the number of helicopter marshallingTakes a discrete two-line chromosomal coding:
σ in formula (18) i Denotes a chromosome-encoding gene, and the first row c and the second row p have the same meanings as in table 1;represents the length of the chromosome, i.e., the total number of helicopter training missions.
According to the constraint of the formula (3), if the total number of certain helicopters on the LHD is not enough to ensure the continuous calling of training tasks of two adjacent waves, limited helicopter resources must be allocated to a certain wave,while the other wave can only schedule tasks that require other types of helicopters or unsettled tasks. Therefore, it is necessary to encodeAdding empty genes to indicate the idle state of some takeoff sites of a certain LHD in a certain wave, as used in this applicationTo represent this idle state and assume that the number of idle takeoff positions is the greatest common factor of the number of helicopters required for all training missions.
According to model hypothesis 2, the neighborhood where each gene in the chromosome code is located is classified into 2 types: genes corresponding to vertical landing tasks (hereinafter referred to as class 1 training tasks) with variable helicopter number are required to be positioned in class 1 neighborhoods; genes corresponding to close range support tasks (hereinafter referred to as class 2 training tasks) requiring a fixed number of helicopters are located in class 2 neighborhoods. It is well known that the variation in chromosome length is caused by the variation in class 1 training tasks in the code.
The chromosome decoding mode adopted in the population iteration process is as follows: converting chromosome codes of the task list into an action list representing the action plan of each helicopter; obtaining a list of initiating moments of vertical login tasks and close-range support tasks of each target through the going-out list; according to the play list, the launch time list and the predetermined training plan of the plane landing team, the improved Lankaster model is used for calculating the variation of the force of each target node of the landing field along with the time to obtain a landing field node state vector, the progress direction of the landing action is deduced through the landing field state transfer model, and the achievement situation of the training purpose and the training task efficiency index set are determined.
It will be appreciated that with respect to the algorithmic decoding portion, the input parameters for a particular decoding operation are a discrete task listThe decoding process is roughly as follows: first, the discrete coding of the training task sequence is characterizedConversion into a play list representing the play plan of each helicopterBy passingThe obtained list of the initiation moments of the vertical login task and the close range support task of the target of the red party to the blue partyCombining with a predetermined training plan of a plane landing team, calculating the change situation of each target node of a landing field against manpower of two parties along with time through a Lanchester model improved in the field, further obtaining a landing field node state vector delta, deducing the progress direction of a landing action through a landing field state transfer model, and finally determining whether a red party can finally achieve the action purpose and a training task efficiency index set
In one embodiment, the process of converting to an action list further comprises the following steps:
calculating the number of all vertical landing teams delivered by each vertical landing task in the chromosome code of the task list and the number of various helicopters required by each vertical landing task and the close distance support task;
selecting 1 class training tasks meeting training resource constraint in a task list according to the number of all vertical landing teams and the number of various helicopters, assigning values to elements in an action list according to a set assignment formula, and adding the executed tasks to an executed task list according to the original list sequence; the class 1 training task is a vertical login task;
after all the class-1 training tasks meeting the training resource constraint are inserted into the action list, selecting class-2 training tasks meeting the training resource constraint in the task list according to the number of all vertical landing squadrons and the number of all helicopters, and correspondingly updating the action list and the executed task list; the class 2 training task is a close range support task
Specifically, inAndthe solving stage of (2): first, the chromosomal coding of the task list is computedThe number u of all vertical Landing teams (ALUs) posted by each vertical Landing task i And the number of helicopters hi required for each vertical landing mission and close range support mission. Second step according to u i And hi select training task listThe class 1 training task satisfying the constraints of the formulas (1) to (7) is given by the set assignment formula (19)Assigning values to the elements in (1) and adding the executed tasks to the list in the original list orderIn (1). Thirdly, inserting all class 1 training tasks meeting the constraintThen, the search is performed according to the same principle as the second step of this paragraphClass 2 training tasks satisfying the constraints and updating accordinglyAnd
assuming 3 LHDs in the red and sufficient helicopters, 3 waves of movement are accumulated, the chromosome coding in FIG. 4Transcribable to a helicopter out-of-flight schedule as shown in FIG. 5 Is a w-row of the image data,a matrix of the columns is formed,the task executed by the helicopter taking off from the jth takeoff position of the jth LHD in the w wave is represented, and the value meaning of the task is shown as formula (19):
further, the list of the initiation momentsThe solving process of (2) may include the following processes:
according to the task targets of each training task in the action list and the distribution condition of target air defense nodes in a landing field, marking out a route set with the minimum risk from the helicopter to each task target by utilizing Dijkstra algorithm;
and solving by using the action list and the airline set to obtain an initiating moment list of the task execution stage of each training task.
Specifically, inThe solving stage of (2): first, according to the helicopter's plan of actionTarget p of each training task i And planning the helicopter to each task target p by utilizing Dijkstra (Dijkstra) algorithm in combination with the known or assumed distribution condition of target air defense fire in landing fields i Of the least risky course setSecondly, solving a list of starting moments (for a vertical landing task, the moment when the vertical landing squad starts to land, and for a close distance support task, the moment when a special helicopter starts to attack on the ground) of task execution stages of each training task
Further, the process of determining the training task performance index set may specifically include the following processing:
setting the value of a corresponding landing field node state vector to be zero at the moment of initiating a landing action;
solving the bearing risk of each helicopter in the marshalling task flight process based on the setting;
according to the initiation time list and a predetermined training plan of a plane landing team, landing field nodes which are confronted at each simulation time are found, the labor loss and the time consumption of each landing field node for confronting both parties are calculated, corresponding landing field node state vectors are updated according to landing field state transfer conditions, numerical integration operation is carried out on the actual force vectors of manpower of both parties at each target node by setting a simulation step length until the landing field node state vectors meet a predetermined target of a landing action opening landing field stage or the condition that the landing field node state vectors cannot achieve the training target is determined;
and (4) counting the total consumption time of the stage of opening the landing field, the accumulated manpower loss of the vertical landing sub-team and the plane landing team and the bearing risk of all helicopters executing the training tasks to form a training task efficiency index set.
Specifically, inThe solving stage of (2): first, assume that T is initiated at login action 0 Time of day toLanding site node status vector Δ (p) =0. Secondly, solving the bearing risk of each helicopter in the red square in the flying process of the grouping task; thirdly, initiating time according to the training task of each helicopterCombining the preset training plan of the plane landing team, finding out the landing site node which is confronted at each simulation moment, calculating the manpower loss and confrontation time consumption (i.e. time consumption) of each node in confrontation with both parties by using formulas (12) - (15), and taking t as the t step And performing numerical integration operation on the actual force vectors of the manpower of the red and blue parties at each target node for setting a simulation step length until a predetermined target at a landing place stage for the login action of the red party is met or the red party is determined to be unable to achieve the action target. Finally, calculating to obtain the total consumption time z of the stage of opening the landing field 1 Cumulative human loss z for a laterove and a level rove 2 And all the helicopters performing the training task are subjected to a risk z 3 Form a target vector
In an embodiment, regarding the process of constructing the initial population in step S13, the method may specifically include the following processing steps:
searching all 1-class training task combinations meeting the vertical login squad number constraint to form a 1-class training task set;
determining the total number of tasks contained in the target chromosome according to the number constraint of the helicopters; the target chromosome is a chromosome containing elements in the class 1 training task set;
constructing an arrangement set which comprises all arrangement forms of the corresponding elements in the class 1 training task set in the chromosome with the length being the total task number based on the total task number and the task number contained in the corresponding elements in the class 1 training task set;
for each element in the permutation set, taking the residual space in the chromosome corresponding to the element as a class 2 neighborhood and randomly assigning values to obtain a permutation integrity code; the class 2 neighborhood is the neighborhood where the corresponding gene of the class 2 training task is located;
carrying out chromosome decoding on the permutation complete code to obtain a corresponding permutation target value;
if the permutation target value is a feasible solution, combining the permutation complete code and the corresponding permutation target value as an individual in the initial population;
and returning to the step of determining the total number of tasks contained in the target chromosome according to the number constraint of the helicopters until each task contained in each element in the traversal arrangement set is reached, and outputting an initial population containing an initial coding set and an initial solution set.
It can be understood that, since information including all Class 1 training tasks cannot be guaranteed in a single chromosome code, and different Class 1 training tasks in each chromosome directly determine the difference of the code length, in order to make an algorithm perform a full search on Class 1 Neighborhoods in a feasible region, and to make each chromosome code in an Initial Population include as many training tasks as possible while satisfying the constraints, an Initial Population Generation Method (Initial Population Generation Method on transforming 1st-Class neighbor nodes, IPGM-T1N) based on the search Class 1 Neighborhoods is designed, a feasible code set including all different Class 1 training task permutation combinations (i.e., all possible lengths) is generated and used as the Initial Population, and the diversity of the Initial Population is better guaranteed.
Specifically, the input is the population size | P |, which is defined as | P in equation (16) g The | are the same. Firstly, searching all 1-class training task combinations meeting the constraint of the total number of vertical landing teams, namely formula (2),form a class 1 training task set C S . Second step, for C S The ith element C S (i) First, the inclusion C is determined according to the constraint of equation (3) S (i) Target chromosome of (2)Total number of tasks involved(i.e., chromosome coding length) and then based onAnd C S (i) Number of tasks | C in S (i) The | construction includes C S (i) In a length ofIs a set of all permutations in the chromosome ofAccording to the definition of the arrangement,the number of the medium elements can be calculated by the formula (20):
the third step, forEach element ofFirstly, the residual space in the corresponding chromosome is taken as a 2-type neighborhood, and random assignment is carried out to obtain the permutation complete codeThen calling the decoding algorithm of the former to obtain the pairCorresponding permutation target valueIf it isTo be a feasible solution, thenAndcombined as initial population P 0 Of the individual. Finally, the second step and the third step in the embodiment are repeated until i = | C S (i) Until | the output contains the initial encoding set C 0 And an initial solution set Z 0 Initial population P of 0 。
In an embodiment, the process of performing the crossover operation on the parent population in step S13 may specifically include the following steps:
carrying out length transformation on the two chromosome codes to be crossed, and determining the position of a gene point which can be crossed and can carry out double-point crossing operation in the two chromosome codes to be crossed after transformation;
selecting two optional points in the crossed gene point positions in one chromosome code as selected gene segments, and combining the rest unselected gene segments with the other chromosome code to form a candidate gene library;
selecting genes from a candidate gene library and selecting gene segments to form a complete code, restoring the code length to the original length, and decoding to obtain an initial solution;
if the initial solution is a feasible solution, finishing the cross operation and outputting a new individual in the filial generation population, otherwise, repeating the selection and gene recombination processing of the selected gene segment until the initial solution is the feasible solution.
It can be understood that, for the crossover operation, aiming at the problem that the crossover operation is difficult to perform between chromosomes of different lengths, the present embodiment designs a double-point crossover method based on chromosome coding length transformation, which ensures the generation of new offspring population individuals.
Specifically, the algorithm input for the crossover operation encodes the two chromosomes to be crossedAndfirstly, for inputAndlength conversion is carried out, and expanded codes of chromosomes to be crossed are determinedAndthe gene site of the gene site capable of double-point crossing operation. Then, atOptionally two of the cross-linkable gene loci of (a) are selected gene segmentsAnd will start aThe remaining genes andcombining to form candidate gene libraryFinally, from the candidate gene librarySelection inSelection of appropriate genes and selected gene fragmentsForm complete codesRestore it to original lengthDecoding to obtain an initial solutionIf it isIf the solution is feasible, the cross operation is ended, and the new individual in the offspring population O is outputIf it isIf not, the gene fragment selection and recombination phases are repeated untilSo far as feasible.
Further, in the transformation stage of the coding length of the chromosome, specifically, in the first step, the number h of helicopters is grouped by using the greatest common factor of the number of helicopters required for all the training tasks as a unit min Assigning values and calculating the extended coding length of the chromosome according to the values
Since each chromosome in the population encodes all the required alignment for training tasksThe total number of the helicopters is always equal to the total number of helicopters which can ensure the action of all LHDs in the total action fluctuation, so that the code length of each expanded chromosome can be ensured to beSecond step, for input to be cross-codedAndaccording to which each gene is subjected to length conversion before being encodedAndthe occupied length of the material is as follows:
for codingAndthe extension code is obtained after length conversionAndthird step, construct length andandare identical to each otherSet of cross-linkable front end pointsAnd set of cross-linkable back endpointsFor theMiddle (i) Full Each gene position, if and only ifAndall are the first coding gene corresponding to one of the training tasks in the respective extension codesThe rest gene positions take the value of 0, and the same is true if and only ifAndall are the last coding gene corresponding to one training task in respective extension codesThe rest gene positions are 0.
Further, in the gene fragment selection stage: first, a set of cross-capable front end points is randomly selectedAnd set of cross-linkable back endpointsGene site position i with the median value of 1 Front And i Tail Require i Front And i Tail Satisfy i Front <i Tial Otherwise, repeating the random selection operation (i) Front =1,i Tail =L Full The existence of the condition of (1) ensures the condition i Front <i Tial Must be satisfied) as the front and back end points of the selected gene fragment. A second step ofAs selected gene fragmentThird step, use ofThe gene composition candidate gene library of (1)And deleteAnd (c) a gene that is identical to a repeat in the selected gene fragment.
Further, in the gene recombination stage: first, build and expand codesAndequal length child spreading codeSelecting gene segmentsExpanding code as descendantsI th of (1) Front To the ith Tail A bit. Second, expanding the offspringExhibition codeRemoving the selected gene segmentAll remaining gene sites outsideAs the space to be inserted, for the ith in the space to be inserted Cross Individual gene sites, searching candidate gene library in orderAll elements in (1) are selected to have a length corresponding to a single training task of all the constraints previously mentionedAs a descendant of the spread codeI th of (1) Cross To the ith Cross +l Curr -position 1, and deleting the corresponding coding in the candidate gene library. The third step is to order i Corss ←i Corss +|σ Curr Repeat the second step of this paragraph untilUntil now. Fourthly, expanding and coding the newly generated offspringPerforming the inverse operation of the chromosomal length transform to obtain the normal form of the offspring codeDecoding to obtainFinally, ifIf feasible, outputting the offspring individualsOtherwise, the candidate gene library isThe first training task in (a) is moved to the end, and the aforementioned second to fourth steps of this paragraph are repeated (likewise, i) Front =1,The existence of a situation warrants the necessity of loop termination).
As shown in fig. 6, a specific flow diagram of the interleaving operation between one code and another code as shown in fig. 4 is given.
In an embodiment, the process of the change operation in step S13 may specifically include the following processing:
generating a random number through 0-1 binomial distribution with the probability of the set mutation probability, if the random number is 1, performing mutation, and otherwise, outputting the input initial chromosome code and the initial solution as a new individual;
randomly selecting two genes in the input initial chromosome codes to carry out position exchange, and carrying out decoding operation on newly generated chromosome codes after exchange to obtain a new solution;
if the new solution is feasible, outputting the offspring individuals, otherwise, performing the position exchange operation again until the corresponding new solution is feasible.
It is understood that this example provides a mutation method based on the exchange of positions of genes in the code, and specifically, the algorithm input of the mutation operation is the initial chromosomal codeInitial solutionAnd setting the mutation probability P Mut . First, the passing probability is P Mut The 0-1 binomial distribution of (A) generates a random number B (1,P) Mut ) If 1, mutation is performed, otherwise, mutation will be performedAnd outputting as a new individual. Second, randomly selecting input chromosome codingThe two genes in the gene pair are subjected to position exchange, and the newly generated chromosome is subjected to position exchangePerforming a decoding operation to obtain a new solutionIf it isIf it is not feasible, the second step of this paragraph is restarted until the next step is completedSo far as feasible. Finally, outputting the offspring individuals
In an embodiment, the process of the local optimization search operation in step S13 may specifically include the following processing:
assigning the current chromosome code as an input chromosome code, and then assigning the optimal code and the optimal solution corresponding to the optimal individual as an input chromosome code and a solution individual corresponding to the input chromosome code respectively;
setting the tabu list as an empty set and calculating the number of 2 types of neighborhood genes in the input chromosome code;
randomly selecting one class 2 neighborhood gene in the complete permutation code, randomly selecting a target sequence number as the target sequence number segment of the class 2 neighborhood gene for assignment, and decoding the updated current permutation complete code to obtain a current solution;
if the current solution is not feasible, repeating the random selection and arrangement of the 2 types of neighborhood genes in the complete codes and the decoding operation until the corresponding current solution is feasible;
judging the quality relation between the current solution and the optimal solution, if the pareto dominance of the first solution individual is determined, assigning a Boolean variable with an initial value of 0 to be 1, and respectively updating the optimal code and the optimal solution into a completely arranged code and the current solution;
if the Boolean variable value is 1, assigning values to the positions of the 2 types of neighborhood genes in the chromosome codes and the target sequence number segments and inserting the values into a tabu list;
and assigning the initial value of the Boolean variable to be 0 again, returning to the process of randomly selecting one type 2 neighborhood gene in the complete permutation code and randomly selecting a target sequence number to assign the value to the target sequence number segment of the type 2 neighborhood gene, decoding the updated current permutation complete code to obtain the current solution until the iteration number reaches the algorithm iteration number, and outputting a new offspring population individual.
It will be appreciated that since the model in the foregoing assumes that 2-class training task grouping helicopters are fixed in number in 2, the close range support task for a target that is grouped by a larger group of dedicated helicopters beyond the fixed number is coded as multiple 2-class training tasks occurring adjacently. Therefore, unlike class 1 training tasks, considering the case of fire assault in large fleet grouping or multiple formation for a target, class 2 training tasks in chromosome coding may not only occur but also occur multiple times, which makes class 2 neighborhood codes of length n have P n The domain scope is greatly increased by the possible assignment. In order to solve the problem, a local optimization method based on tabu search for the 2-type neighborhood is provided, and the algorithm search efficiency is improved well.
Specifically, the algorithm input amount of the local optimization search is input chromosome codingAnd its corresponding decoded solution bodyTabu list depth D T And the number of algorithm iterations N. First, the current code is assigned asOptimal encoding corresponding to optimal individualsAnd an optimal solutionAre respectively assigned asAnddefining the tabu list as empty set phi, calculating input codeNumber of middle 2-type neighborhood genesSecond, randomly selecting and arranging complete codesA class 2 neighborhood gene ofAnd randomly selecting a target sequence number p as the target sequence number segment of the geneAssigning value, and then coding the updated current codePerforming decoding operation to obtain current solutionIf it isIf not, the second step of this paragraph is repeated untilSo far as feasible.
Thirdly, judging the current solutionAnd the optimal solutionIf the first solution is the first solutionPareto dominance (Pareto Dominate)(is marked asIf and only ifWherein) Then the initial value is 0 Boolean variable eta T Assign value to 1 and encode the optimumAnd an optimal solutionAre respectively updated toAndin addition, if the list L is forbidden T Has not inserted 2 kinds of neighborhood geneAny assignment of (a) will also be applied to the Boolean variable η T The value is assigned to 1.
The fourth step, if the Boolean variable eta T If the value is 1, then the neighborhood genes of class 2 are assignedInserting a tabu list L at position i in a chromosome and its target sequence number segment assignment p T In the specific insertion process, if the tabu list L is forbidden T Length | L T I has reached the assumed upper bound D T If so, delete the tabu list L first T The stack bottom element of (1), then insert { i, p } from the stack top if | L T |<D T Then insert { i, p } directly into L T In (1). Finally, the Boolean variable eta is given again T Assigning an initial value of 0, repeating the second step to the fourth step until the algorithm iterates for N times, and outputting new filial generation population individualsIn one embodiment, the reference points are used as the basis of spatial decomposition and environmental evolution in the multi-objective evolutionary algorithm and have a decisive influence on the optimizing performance of the algorithm. Taking a certain initial population generated in the foregoing (assuming that the simplex coefficient H = 8) as an example, firstly, limitations of uniformly distributed reference points generated based on two traditional methods when processing population data of a mission planning model trained by an amphibious helicopter are compared, and then a reference point adaptive generation method based on population distribution is provided and a calculation result example is given.
As shown in fig. 7, an intercept between a hyperplane formed by Nadir points (lowest points) selected based on the principle of minimum euclidean distance from coordinate axes and a coordinate axis of a solution space may be smaller than 0, that is, it cannot be guaranteed that all of the population solution sets are converted into the first quadrant. In addition, since the hyperplane constructed by this method has randomness to the intercept of each coordinate axis of the solution space, the effect of the normalization process on the individual target values of the solution space is not satisfactory.
The conventional document also discloses a method for dealing with the situation shown in fig. 7, that is, constructing a hyperplane by using the maximum value in each target dimension direction of the solution space as a Nadir point, but this method is easily affected by outliers. As shown in FIG. 8, z exists due to the solution space 1 Outliers in the direction that are significantly different from other values cause the intercept of the hyperplane in the direction to be significantly increased, so that the solution space individual cannot form a relatively uniform mapping relationship with the reference point generated by the method. As shown in FIG. 9, the linear hyperplane is at z, subject to outliers 1 The intercept on the axis is significantly larger than the z for most individuals in the normalized solution 1 The value in the axial direction leads to that the individuals in the solution set are intensively mapped on the reference point on one side of the hyperplane when the clustering operation is carried out according to the clustering criterion based on the interior angle measurement method, so that excellent individuals can be lost in the subsequent selection operation of the algorithm, and the optimization efficiency of the algorithm is reduced.
According to the above analysis, there are two ideas to improve the uniformity of mapping between solution-centered individuals and reference points on the hyperplane: firstly, the outliers which have large influence on the hyperplane intercept in the solution space are deleted, so that the hyperplane can better fit most individuals in solution concentration. Secondly, the distribution of the reference points on the hyperplane is changed, so that the reference points are gathered to the region with the solution concentration and the individual density. Because discrete codes are adopted in the method, the target value of the population individual in the solution space can be obviously changed due to the small difference of the population individual in the codes, and therefore the method of deleting the outlier in the first thought can cause the algorithm to lose the potential optimal solution; the second idea is widely applied to some clustering-based reference point generation methods, but the clustering method for solution set individuals in the methods has higher calculation complexity and is not beneficial to efficient calculation.
The reference point self-adaptive generation method based on the population distribution can better solve the problems. Further, regarding the process of generating the reference point set by using the reference point adaptive generation method based on the population distribution on the normalized solution set in the step S16, the process includes:
constructing a central vector of the standardized solution set based on median values of included angles between all individuals and each dimension direction vector in the standardized solution set;
constructing a pseudo linear hyperplane which is perpendicular to the central vector and passes through the maximum sum of all dimension target values in the standardized solution set to obtain the intercept of the pseudo linear hyperplane on each coordinate axis of the standardized solution space;
in each dimension direction of the standardized solution space, dividing a polar coordinate space defined by an included angle into H subspaces according to n different intervals according to the distribution condition of the included angle between an individual in the standardized solution set and a corresponding dimension direction vector; n is a partition coefficient, and H is a simplex coefficient;
projecting the boundary angle between the subspaces to a coordinate axis of the direction of the dimension direction vector in the standardized solution space according to the selected length to obtain H-1 separation points in a target interval on the coordinate axis;
and generating a reference point set based on H-1 separation points on each coordinate axis of the normalized solution space.
Specifically, the input variables are simplex coefficient H, partition coefficient n and standardized solution setFirst, based on the normalized solution set calculated by the formula (17)All individuals inAnd direction vector e of each dimension i Angle of (theta) i Median value ofConstruct a standardized solution setCentral vector e of Mid . Second, construct a vector e perpendicular to the center Mid And is subjected to standardized solution setThe sum of all the dimensional target values in the group is maximumThe pseudo-linear hyperplane (see fig. 10 (a)) is obtained, and the intercept of the pseudo-linear hyperplane on each coordinate axis of the normalized solution space is obtainedThirdly, in each dimension direction of the standardized solution space, according to the standardized solution setChinese medicinal compositionAnd the dimension direction vector e i Included angle theta i Distribution of (a) angle of inclination theta i The defined polar coordinate space Γ i is divided into H subspaces according to n different pitchesMaking the same pitch subspaceFormed combined subspaceWherein the population solution sets the number of individuals is equal, whereinThe fourth step is to divide the subspaceThe boundary angle between them is selected according to the lengthProjection into the normalized solution space e i On the coordinate axis of the direction, the coordinate axis is obtainedH-1 separation points in the intervalFinally, based on H-1 separation points on each coordinate axis of the normalized solution spaceReferring to the existing simplex lattice method of Das and Dennis, the number of individuals shown in FIG. 10 (b) was generated asA set of reference points Λ.
If only the population Sorting selection mechanism is considered, the computation complexity mainly depends on the scaling-based Sorting Method (SSM), the determination of Ideal point and Nadir point in the standardization process, and the polar coordinate space Γ of each dimension i The computational complexity of the segmentation operation of (1) is O (MN), respectively 2 O (MN/2) and O (MN), wherein M is solution space dimension, and N represents the former population P' g+1 Is determined (assumed to be 2 times the number of reference points). Therefore, the reference point self-adaptive generation method based on the population distribution does not increase the computational complexity in the population sorting selection stage. Furthermore, thanks to the scaled ranking method adopted, the computational complexity of the population ranking selection phase is even in the worst case smaller than the computational complexity O (MN) of a non-dominated ranking based multi-objective optimization algorithm 2 )。
As shown in fig. 11, with the normalized space of the initial population in fig. 7Direction for example, individuals in the normalized solution set distributed in the first quadrantAnd e 3 Angle of (2)Defined polar coordinate space Γ 3 In the range of [0, π/2]. Assuming n =2, the polar coordinate space Γ 3 Can be respectively according to the spacingAndpartitioned into H subspacesMake a space of a zygoteAndthe number of the solution sets in (1) is the same.
It should be noted that although the reference point is constructed by using the conventional simplex method, on one hand, due to the uneven division of the coordinate axis, and on the other hand, due to the different intercepts of the linear hyperplane on each axis of the standard solution space, the sums of the target values of the dimensions of the generated reference point are not the same, and are distributed on the nonlinear hyperplane.
As shown in fig. 12, due to the uneven distribution on the curved surface, compared with the reference points uniformly distributed on the linear hyperplane in fig. 9, the regions where the reference points in fig. 12 are individually gathered in the solution space are more densely distributed, so that the solution set has a better mapping effect in the case of irregular distribution of the solution set. Each coordinate axis in fig. 12 represents an objective function value of the planning model, and is a dimensionless number.
In some embodiments, after case simulation is performed by using the method provided by the application, the result shows that the method provides an initial solution set generation operation and a local optimization operation on the population update iteration mechanism level, and the rapid planning effect of the training task of the efficient and practical amphibious helicopter is fully achieved. In practical engineering application, aiming at the optimization contents of target selection, time sequence arrangement and the like of the helicopter training tasks, an amphibious helicopter training task planning model is constructed and expressed as a multi-objective optimization model; in order to effectively evaluate the promotion effect of the helicopter training mission on the integral situation development of the landing action, a simulation-based landing action efficiency evaluation model is constructed by using the related ideas of system efficiency evaluation, and the objective function of an amphibious helicopter training mission planning model is calculated through the model; innovations are respectively carried out on the aspects of a population updating iteration mechanism and a sequencing selection mechanism, the provided algorithm has excellent performances in convergence speed and diversity, and reliable reference is provided for training task planning of a helicopter in landing operation.
It should be understood that although the various steps in the flow diagrams of fig. 1 and 3 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps of fig. 1 and 3 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least some of the sub-steps or stages of other steps.
As shown in fig. 13, in an embodiment, there is further provided an adaptive multi-objective task planning system 100 based on reference points, which includes a parameter obtaining module 11, a model calling module 12, a population building module 13, a child generation module 14, a merging calculation module 15, a set generation module 16, a new population module 17, and an iteration control module 18. Wherein:
the parameter obtaining module 11 is configured to obtain known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, the solution space dimensionality, the simplex segmentation parameters and the maximum iteration times. The model calling module 12 is used for calling the constructed amphibious helicopter training task planning model; the constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and the optimization objective functions of the amphibious helicopter training mission planning model comprise a time-consuming objective function of a minimum landing phase, a minimum manpower loss objective function and a ground threat objective function of a minimum helicopter.
The population construction module 13 is used for calculating the population scale and constructing an initial population for the amphibious helicopter training mission planning model based on the mission parameters. The offspring generation module 14 is configured to perform a crossover operation, a mutation operation, and a local optimization search operation on the parent population to generate an offspring population. The merging calculation module 15 is configured to merge the generated child population and parent population to obtain a merged population, update the IDEAL point and the NADIR point according to elements in the solution set corresponding to the merged population, and calculate a normalized solution set according to the IDEAL point and the NADIR point. The set generating module 16 is configured to generate a set of reference points and a set of corresponding solutions of interest based on the normalized solution set by using a reference point adaptive generating method based on population distribution. The new population module 17 is configured to assign each individual in the standardized solution set to a littoral solution set by using a clustering method based on an interior angle measurement method, perform non-dominated sorting operation in each littoral solution set based on a scaled sorting method, and select an individual corresponding to an element ranked at the top to form a new generation population. The iteration control module 18 is used for outputting the training mission planning scheme data of the amphibious helicopter when the population iteration times reach the maximum iteration times.
The self-adaptive multi-target task planning system 100 based on the reference point extracts and analyzes the main task mode and characteristics of the helicopter in the amphibious landing operation, provides a training task planning model of the amphibious helicopter on the basis of reasonably assuming a real landing training scene, and expresses the training task planning model as a multi-target optimization model, and aims to minimize the time consumption of the landing operation, personal casualties of own parties and the threat of the air defense fire of the own helicopter to the blue party. The training mission planning model of the amphibious helicopter considers the specific practical constraints of two aspects of training resources, training and guaranteeing the full-link mission stage time sequence, which can be used by multiple platforms. The reference point self-adaptive generation method based on population distribution is applied to the population sorting selection mechanism, the optimization capability of irregular populations is effectively enhanced on the premise of not increasing the computational complexity, a new initial solution generation, crossing, variation and local optimization method is provided on the population updating iteration mechanism, the search efficiency of a code definition space is better improved, and the optimization efficiency and the population diversity maintenance are well expressed, so that the aim of quickly planning training tasks of the amphibious helicopter is fulfilled.
In one embodiment, the modules of the reference point-based adaptive multi-objective task planning system 100 may be further configured to implement further processing steps in embodiments of the reference point-based adaptive multi-objective task planning method.
For specific limitations of the reference point-based adaptive multi-objective task planning system 100, reference may be made to the corresponding limitations of the reference point-based adaptive multi-objective task planning method described above, and details thereof are not repeated here. The various modules of the reference point-based adaptive multi-objective task planning system 100 described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules may be embedded in a hardware form or a device independent of a specific data processing function, or may be stored in a memory of the device in a software form, so that a processor can call and execute operations corresponding to the modules, where the device may be, but is not limited to, various computer devices existing in the art.
In one embodiment, there is also provided a computer device comprising a memory and a processor, the memory storing a computer program, the processor when executing the computer program implementing the steps of: acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, solution space dimensionality, simplex segmentation parameters and the maximum iteration times; calling the constructed amphibious helicopter training task planning model; constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and an optimization objective function of the amphibious helicopter training mission planning model comprises a time-consuming objective function of a minimized landing phase, a minimized manpower loss objective function and a target function of ground threat of a minimized helicopter;
calculating the population scale of the amphibious helicopter training task planning model based on the task parameters and constructing an initial population; performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population; combining the generated child population and the parent population to obtain a combined population, updating IDEAL points and NADIR points according to elements in a solution set corresponding to the combined population, and calculating to obtain a standardized solution set according to the IDEAL points and the NADIR points; generating a reference point set and a corresponding set of a liberal solution set on the standardized solution set by using a reference point self-adaptive generation method based on population distribution; distributing each individual in the standardized solution set to a lyre solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each lyre solution set based on a scaling sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation of population; and returning to the step of performing cross operation, mutation operation and local optimization search operation on the parent population to generate the offspring population, and outputting training task planning scheme data of the amphibious helicopter until the iteration number of the population reaches the maximum iteration number.
It is understood that the computer device includes, in addition to the memory and the processor, other software and hardware components not listed in this specification, which may be determined according to the model of the specific computer device in different application scenarios, and detailed descriptions are not listed in this specification.
In one embodiment, the processor when executing the computer program may also implement the additional steps or sub-steps in the embodiments of the reference point-based adaptive multi-objective task planning method described above.
In one embodiment, there is also provided a computer readable storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of: acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, solution space dimensionality, simplex segmentation parameters and the maximum iteration times; calling the constructed amphibious helicopter training task planning model; constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and an optimization objective function of the amphibious helicopter training mission planning model comprises a time-consuming objective function of a minimum landing phase, a minimum manpower loss objective function and a ground threat objective function of the minimum helicopter;
calculating the population scale of the amphibious helicopter training task planning model based on the task parameters and constructing an initial population; performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population; combining the generated child population and the parent population to obtain a combined population, updating an IDEAL point and an NADIR point according to elements in a solution set corresponding to the combined population, and calculating to obtain a standardized solution set according to the IDEAL point and the NADIR point; generating a reference point set and a corresponding set of a Liji solution set on the standardized solution set by using a reference point self-adaptive generation method based on population distribution; distributing each individual in the standardized solution set to a lyre solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each lyre solution set based on a scaling sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation of population; and returning to the step of performing cross operation, mutation operation and local optimization search operation on the parent population to generate the offspring population, and outputting training task planning scheme data of the amphibious helicopter until the iteration number of the population reaches the maximum iteration number.
In one embodiment, the computer program, when executed by the processor, may further implement the additional steps or sub-steps of the embodiments of the reference point-based adaptive multi-objective task planning method described above.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware related to instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), rambus DRAM (RDRAM), and interface DRAM (DRDRAM).
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for those skilled in the art, various changes and modifications can be made without departing from the spirit of the present application, and all of them fall within the scope of the present application. Therefore, the protection scope of the present patent should be subject to the appended claims.
Claims (11)
1. A self-adaptive multi-objective task planning method based on a reference point is characterized by comprising the following steps:
acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, the solution space dimensionality, the simplex segmentation parameters and the maximum iteration times;
calling the constructed amphibious helicopter training task planning model; the constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and the optimization objective function of the amphibious helicopter training mission planning model comprises a time-consuming objective function of a minimum landing phase, a minimum manpower loss objective function and a ground threat objective function of the minimum helicopter;
calculating the population scale of the amphibious helicopter training task planning model based on the task parameters and constructing an initial population;
performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population;
combining the generated child population and the parent population to obtain a combined population, updating an IDEAL point and an NADIR point according to elements in a solution set corresponding to the combined population, and calculating according to the IDEAL point and the NADIR point to obtain a standardized solution set;
generating a reference point set and a corresponding set of a Liji solution set on the standardized solution set by using a reference point self-adaptive generation method based on population distribution;
distributing each individual in the standardized solution set to the edge solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each edge solution set based on a scaling sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation of population;
and returning to the step of executing the cross operation, the mutation operation and the local optimization searching operation on the parent population to generate the offspring population, and outputting training task planning scheme data of the amphibious helicopter until the iteration number of the population reaches the maximum iteration number.
2. The reference point-based adaptive multi-objective task planning method according to claim 1, wherein a chromosome coding mode adopted in a population iteration process is as follows:
decomposing all training tasks meeting the condition of setting the number of helicopter marshalls into a single task list with the number of helicopter marshalls equal to 1 and adopting discrete double-row chromosome coding;
adding a gap gene in the chromosome coding of the task list;
the chromosome decoding mode adopted in the population iteration process is as follows:
converting chromosome codes of the task list into an action list representing the action plan of each helicopter;
obtaining a list of initiating moments of vertical login tasks and close range support tasks of each target through the going-out list;
according to the action list, the initiation time list and a predetermined training plan of a plane landing team, an improved Lankaster model is used for calculating the change of the force of each target node of a landing field along with time, a landing field node state vector is obtained, the progress direction of a landing action is deduced through a landing field state transfer model, and the achievement condition of a training purpose and a training task efficiency index set are determined.
3. The method of claim 2, wherein the process of converting to the action list comprises:
calculating the number of all vertical landing sub-queues delivered by each vertical landing task in the chromosome code of the task list and the number of various helicopters required by each vertical landing task and the close distance support task;
selecting 1 class training tasks meeting training resource constraint in the task list according to the number of all vertical landing teams and the number of all classes of helicopters, assigning values to elements in the action list according to a set assignment formula, and adding the executed tasks to an executed task list according to the original list sequence; the class 1 training task is the vertical login task;
after all class-1 training tasks meeting training resource constraints are inserted into the action list, selecting class-2 training tasks meeting the training resource constraints in the task list according to the number of all vertical landing squads and the number of all classes of helicopters, and correspondingly updating the action list and the executed task list; the class 2 training task is the close range support task;
the solving process of the initiation time list comprises the following steps:
according to the task targets of the training tasks in the action list and the distribution condition of target air defense nodes in a landing field, planning a route set with the minimum risk from the helicopter to each task target by utilizing Dijkstra algorithm;
and solving the initiation moment list of the task execution stage of each training task by using the run list and the airline set.
A process for determining the set of training task performance indicators, comprising:
setting the value of a corresponding landing field node state vector to be zero at the moment of initiating a landing action;
solving the bearing risk of each helicopter in the marshalling task flight process based on the setting;
according to the initiation time list and a preset training plan of a plane landing team, finding out landing site nodes which are confronted at each simulation time, calculating the labor loss and time consumption of each landing site node confronting both parties, updating corresponding landing site node state vectors according to landing site state transfer conditions, and performing numerical integration operation on the actual force vectors of the manpower of both parties at each target node by setting a simulation step length until the landing site node state vectors meet a preset target of a landing action opening landing site stage or a training target which cannot be achieved is determined;
and (4) counting the total consumption time of the stage of opening the landing field, the accumulated manpower loss of the vertical landing sub-team and the plane landing team and the bearing risk of all helicopters executing the training tasks to form the training task efficiency index set.
4. The reference point-based adaptive multi-objective task planning method according to any one of claims 1 to 3, wherein the process of constructing the initial population comprises:
searching all 1-class training task combinations meeting the vertical login squad number constraint to form a 1-class training task set;
determining the total number of tasks contained in the target chromosome according to the helicopter number constraint; the target chromosome is a chromosome comprising elements in the class 1 training task set;
constructing an arrangement set comprising all arrangement forms of the corresponding elements in the class 1 training task set in the chromosome with the length being the total task number based on the total task number and the task number contained in the corresponding elements in the class 1 training task set;
for each element in the permutation set, taking the residual space in the chromosome corresponding to the element as a 2-type neighborhood and randomly assigning values to obtain a permutation complete code; the 2-type neighborhood is a neighborhood where a gene corresponding to the 2-type training task is located;
carrying out chromosome decoding on the permutation integral codes to obtain corresponding permutation target values;
if the permutation target value is a feasible solution, combining the permutation complete code and the corresponding permutation target value as an individual in the initial population;
and returning to the step of determining the total number of tasks contained in the target chromosome according to the number constraint of the helicopters until each task contained in each element in the arrangement set is traversed, and outputting the initial population containing an initial coding set and an initial solution set.
5. The reference point-based adaptive multi-objective task planning method according to any one of claims 1 to 3, wherein the process of performing the crossover operation on the parent population comprises:
carrying out length transformation on the two chromosome codes to be crossed, and determining the position of a gene point which can be crossed and can carry out double-point crossing operation in the two chromosome codes to be crossed after transformation;
selecting two optional points in the crossed gene point positions in one chromosome code as selected gene segments, and combining the rest unselected gene segments with the other chromosome code to form a candidate gene library;
selecting genes from the candidate gene library and the selected gene segments to form a complete code, restoring the code length to the original length, and decoding to obtain an initial solution;
if the initial solution is a feasible solution, ending the cross operation and outputting a new individual in the filial generation population, otherwise, repeating the selection and gene recombination processing of the selected gene segment until the initial solution is the feasible solution.
6. The reference point-based adaptive multi-objective mission planning method according to any one of claims 1 to 3, wherein the course of the mutation operation includes:
generating a random number through 0-1 binomial distribution with the probability of the set mutation probability, if the random number is 1, performing mutation, and otherwise, outputting the input initial chromosome code and the initial solution as a new individual;
randomly selecting two genes in the input initial chromosome codes to carry out position exchange, and carrying out decoding operation on newly generated chromosome codes after exchange to obtain a new solution;
if the new solution is feasible, outputting the offspring individuals, otherwise, performing position exchange operation again until the corresponding new solution is feasible.
7. The method for reference point-based adaptive multi-objective task planning according to any one of claims 1 to 3, wherein the process of the local optimal search operation comprises:
assigning a current chromosome code as an input chromosome code, and then assigning an optimal code and an optimal solution corresponding to an optimal individual as an input chromosome code and a solution individual corresponding to the input chromosome code respectively;
setting a tabu list as an empty set and calculating the number of 2 types of neighborhood genes in the input chromosome code;
randomly selecting a class 2 neighborhood gene in the complete permutation code, randomly selecting a target sequence number to assign a value to a target sequence number segment of the class 2 neighborhood gene, and decoding the updated current complete permutation code to obtain a current solution;
if the current solution is not feasible, repeating randomly selecting and arranging 2 types of neighborhood genes in the complete codes and decoding operation until the corresponding current solution is feasible;
judging the quality relation between the current solution and the optimal solution, if the pareto dominance of the first solution individual is determined, assigning a Boolean variable with an initial value of 0 to be 1, and respectively updating the optimal code and the optimal solution into the completely arranged code and the current solution;
if the Boolean variable value is 1, assigning the positions of 2 types of neighborhood genes in the chromosome codes and the target sequence number segments and inserting the assignment into the tabu list;
and re-assigning the initial value of the Boolean variable to be 0, returning to the process of randomly selecting one class 2 neighborhood gene in the complete permutation code and randomly selecting a target sequence number to assign a value to the target sequence number segment of the class 2 neighborhood gene, decoding the updated current permutation complete code to obtain a current solution, and outputting a new offspring population until the iteration number reaches the algorithm iteration number.
8. The reference point-based adaptive multi-objective task planning method according to claim 1, wherein the step of generating a reference point set by using a reference point adaptive generation method based on population distribution on the standardized solution set comprises:
constructing a central vector of the standardized solution set based on median values of included angles between all individuals in the standardized solution set and the direction vectors of all dimensions;
constructing a pseudo linear hyperplane which is perpendicular to the central vector and passes through the maximum individual sum of all dimension target values in the standardized solution set to obtain the intercept of the pseudo linear hyperplane on each coordinate axis of the standardized solution space;
in each dimension direction of a standardized solution space, dividing a polar coordinate space defined by an included angle into H subspaces according to n different intervals according to the distribution condition of the included angle between an individual in the standardized solution set and a corresponding dimension direction vector; n is a partition coefficient, and H is a simplex coefficient;
projecting the boundary angle between the subspaces to a coordinate axis of the direction of the dimension direction vector in the standardized solution space according to the selected length to obtain H-1 separation points in a target interval on the coordinate axis;
and generating the reference point set based on H-1 separation points on each coordinate axis of the standardized solution space.
9. An adaptive multi-objective mission planning system based on reference points, comprising:
the parameter acquisition module is used for acquiring known task parameters; the task parameters comprise the total number of the platforms, the total wave times of helicopter operation, the solution space dimensionality, the simplex segmentation parameters and the maximum iteration times;
the model calling module is used for calling the constructed amphibious helicopter training task planning model; the constraint conditions of the amphibious helicopter training mission planning model comprise vertical landing squad quantity constraint, helicopter quantity constraint, platform guarantee resource constraint, mission phase time sequence constraint based on execution sequence and mission phase time sequence constraint based on platform deck scheduling, and the optimization objective function of the amphibious helicopter training mission planning model comprises a time-consuming objective function of a minimum landing phase, a minimum manpower loss objective function and a ground threat objective function of the minimum helicopter;
the population construction module is used for calculating the population scale of the amphibious helicopter training task planning model based on the task parameters and constructing an initial population;
the child generation module is used for performing cross operation, mutation operation and local optimization search operation on the parent population to generate a child population;
a merging calculation module, configured to merge the generated child population and the parent population to obtain a merged population, update an IDEAL point and an NADIR point according to elements in a solution set corresponding to the merged population, and calculate a normalized solution set according to the IDEAL point and the NADIR point;
a set generating module, configured to generate a reference point set and a set of corresponding solution sets of interest bases on the standardized solution set by using a reference point adaptive generation method based on population distribution;
the new population module is used for allocating each individual in the standardized solution set to the edge solution set by using a clustering method based on an interior angle measurement method, performing non-dominated sorting operation in each edge solution set based on a scaling sorting method, and selecting the individual corresponding to the element with the top rank to form a new generation population;
and the iteration control module is used for outputting the training task planning scheme data of the amphibious helicopter when the population iteration times reach the maximum iteration times.
10. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor when executing the computer program implements the steps of the reference point based adaptive multi-objective mission planning method of any one of claims 1 to 7.
11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the reference point-based adaptive multi-objective task planning method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211201299.8A CN115730700A (en) | 2022-09-29 | 2022-09-29 | Self-adaptive multi-target task planning method, system and equipment based on reference point |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211201299.8A CN115730700A (en) | 2022-09-29 | 2022-09-29 | Self-adaptive multi-target task planning method, system and equipment based on reference point |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115730700A true CN115730700A (en) | 2023-03-03 |
Family
ID=85293365
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211201299.8A Pending CN115730700A (en) | 2022-09-29 | 2022-09-29 | Self-adaptive multi-target task planning method, system and equipment based on reference point |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115730700A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116362652A (en) * | 2023-06-01 | 2023-06-30 | 上海仙工智能科技有限公司 | Transport allocation task scheduling method and system and storage medium |
CN116542468A (en) * | 2023-05-06 | 2023-08-04 | 中国人民解放军32370部队 | Unmanned aerial vehicle cluster task planning method |
CN117130391A (en) * | 2023-10-19 | 2023-11-28 | 中国人民解放军96901部队 | Unmanned aerial vehicle task planning method and equipment based on software definition |
-
2022
- 2022-09-29 CN CN202211201299.8A patent/CN115730700A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116542468A (en) * | 2023-05-06 | 2023-08-04 | 中国人民解放军32370部队 | Unmanned aerial vehicle cluster task planning method |
CN116542468B (en) * | 2023-05-06 | 2023-10-20 | 中国人民解放军32370部队 | Unmanned aerial vehicle cluster task planning method |
CN116362652A (en) * | 2023-06-01 | 2023-06-30 | 上海仙工智能科技有限公司 | Transport allocation task scheduling method and system and storage medium |
CN116362652B (en) * | 2023-06-01 | 2023-10-31 | 上海仙工智能科技有限公司 | Transport allocation task scheduling method and system and storage medium |
CN117130391A (en) * | 2023-10-19 | 2023-11-28 | 中国人民解放军96901部队 | Unmanned aerial vehicle task planning method and equipment based on software definition |
CN117130391B (en) * | 2023-10-19 | 2024-01-12 | 中国人民解放军96901部队 | Unmanned aerial vehicle task planning method and equipment based on software definition |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN115730700A (en) | Self-adaptive multi-target task planning method, system and equipment based on reference point | |
CN112817330B (en) | Multi-unmanned aerial vehicle four-dimensional track collaborative planning method and system | |
CN112733421B (en) | Task planning method for cooperation of unmanned aerial vehicle with ground fight | |
CN102880186B (en) | flight path planning method based on sparse A* algorithm and genetic algorithm | |
Fu et al. | Route planning for unmanned aerial vehicle (UAV) on the sea using hybrid differential evolution and quantum-behaved particle swarm optimization | |
Shima et al. | UAV cooperative multiple task assignments using genetic algorithms | |
Wang et al. | A hybrid metaheuristic DE/CS algorithm for UCAV three-dimension path planning | |
Rasmussen et al. | Tree search algorithm for assigning cooperating UAVs to multiple tasks | |
CN109357678B (en) | Multi-unmanned aerial vehicle path planning method based on heterogeneous pigeon swarm optimization algorithm | |
CN112016812A (en) | Multi-unmanned aerial vehicle task scheduling method, system and storage medium | |
CN105225003A (en) | A kind of cuckoo searching algorithm solves the method for UAV multitask investigation decision problem | |
CN111813144B (en) | Multi-unmanned aerial vehicle collaborative route planning method based on improved flocks of sheep algorithm | |
CN115755963B (en) | Unmanned aerial vehicle group collaborative mission planning method considering carrier delivery mode | |
CN112733251A (en) | Multi-unmanned aerial vehicle collaborative track planning method | |
CN115494873A (en) | Heterogeneous multi-unmanned aerial vehicle cooperative task allocation method based on Monte Carlo tree search architecture under time sequence constraint | |
CN115759328A (en) | Helicopter task planning method, system and equipment based on multi-objective optimization | |
CN106840164A (en) | A kind of many fragments actively remove online weight planning algorithm | |
CN116451934A (en) | Multi-unmanned aerial vehicle edge calculation path optimization and dependent task scheduling optimization method and system | |
Rasmussen et al. | Branch and bound tree search for assigning cooperating UAVs to multiple tasks | |
Peng et al. | Modeling and solving the dynamic task allocation problem of heterogeneous UAV swarm in unknown environment | |
CN113625767A (en) | Fixed-wing unmanned aerial vehicle cluster collaborative path planning method based on preferred pheromone gray wolf algorithm | |
Jing et al. | Cooperative task assignment for heterogeneous multi-UAVs based on differential evolution algorithm | |
CN113608546A (en) | Quantum sea lion mechanism unmanned aerial vehicle group task allocation method | |
CN116088586B (en) | Method for planning on-line tasks in unmanned aerial vehicle combat process | |
Zhou et al. | A novel mission planning method for UAVs’ course of action |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |