CN109740283A - Autonomous multiple agent confronting simulation method and system - Google Patents
Autonomous multiple agent confronting simulation method and system Download PDFInfo
- Publication number
- CN109740283A CN109740283A CN201910045079.2A CN201910045079A CN109740283A CN 109740283 A CN109740283 A CN 109740283A CN 201910045079 A CN201910045079 A CN 201910045079A CN 109740283 A CN109740283 A CN 109740283A
- Authority
- CN
- China
- Prior art keywords
- artificial intelligence
- simulation
- scene
- information
- algorithm
- 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
Landscapes
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a kind of autonomous multiple agent confronting simulation method and system, wherein this method comprises: in simulated environment, according to analogue system setting and simulation system parameters, building sea, day one real-time simulation scene and artificial intelligence human-computer interaction model;In simulated environment, according to the parameter information of different artificial intelligence bodies, artificial intelligence body Model is constructed, different artificial intelligence bodies load different confrontation tasks and algorithm, construct autonomous Multi-Agent simulation frame;Each artificial intelligence body is iterated deduction in simulated environment to complete confrontation task and algorithm;Judge whether each artificial intelligence body completes confrontation task and algorithm, if completing, the simulation process information of each artificial intelligence body and simulation result information are exported and saved, if not completing, continues iteration deduction.This method establishes a multi-dimensional complicated environment man-machine coordination Counter Simulation System, can the more true and comprehensive simulation to the Scene realization really fought.
Description
Technical field
The present invention relates to computer simulation technique field, in particular to a kind of autonomous multiple agent confronting simulation method and it is
System.
Background technique
Current true Antagonistic Environment is increasingly sophisticated, and multiple complex targets form one to major embodiment before on the one hand
The shared collective of a situation information, each target between another aspect sea, land and air also will form the confrontation unit of collaboration one,
The confrontation under round-the-clock, all the period of time multi-dimensional complicated environment also becomes new demand simultaneously.Due under actual scene, intelligent body
Between confrontation mode have developed into a comprehensive general space solid confrontation, it is man-machine to establish a multi-dimensional complicated environment
Cooperate with Counter Simulation System, come to true confrontation Scene realization it is more true and comprehensively simulate it is particularly important.
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.
For this purpose, this method is established an object of the present invention is to provide a kind of autonomous multiple agent confronting simulation method
One multi-dimensional complicated environment man-machine coordination Counter Simulation System, can be more true and comprehensive to the Scene realization really fought
Simulation.
It is another object of the present invention to propose a kind of autonomous multiple agent Counter Simulation System.
In order to achieve the above objectives, one aspect of the present invention embodiment proposes a kind of autonomous multiple agent confronting simulation method,
The following steps are included: S1, in simulated environment, according to analogue system setting and simulation system parameters, building sea, day are integrally real-time
Simulating scenes and artificial intelligence human-computer interaction model;S2 believes in the simulated environment according to the parameter of different artificial intelligence bodies
Breath constructs artificial intelligence body Model, and different artificial intelligence bodies load different confrontation tasks and algorithm, and building is autonomous mostly intelligent
Body simulation frame;S3, each artificial intelligence body are iterated deduction in the simulated environment that step S1 and S2 are configured with complete
At confrontation task and algorithm;S4, judges whether each artificial intelligence body completes confrontation task and algorithm, will if completing
The simulation process information and simulation result information of each artificial intelligence body are exported and are saved, if do not complete, return S3 after
It is continuous to carry out the iteration deduction.
The autonomous multiple agent confronting simulation method of the embodiment of the present invention, by establishing the man-machine association of multi-dimensional complicated environment
Same Counter Simulation System more true to the Scene realization really fought can be simulated with comprehensive, can also be to simulated environment
Visualization is carried out convenient for user's adjusting parameter.
In addition, autonomous multiple agent confronting simulation method according to the above embodiment of the present invention can also have following add
Technical characteristic:
Further, in one embodiment of the invention, S3 further comprises: S31, updates in the simulated environment
Ambient condition;S32, each artificial intelligence body update itself algorithm according to updated ambient condition, according to confrontation task
It is deduced to obtain the algorithm that each artificial intelligence body completes confrontation task;S33, each artificial intelligence body is according to pushing away
The algorithm drilled takes response action to complete confrontation task.
Further, in one embodiment of the invention, the sea, day one real-time simulation scene include: that three-dimensional can
Light-exposed scene and three-dimensional IR Scene;The three-dimensional visible light scene be the sea of Real-time modeling set and rendering, sky one three
Tie up simulating scenes;The three-dimensional IR Scene is the three-dimensional artificial field using black body radiation model and atmospheric radiation model construction
Scape.
Further, in one embodiment of the invention, the autonomous Multi-Agent simulation frame includes: visual perception
Model and cooperative awareness model;Wherein, the human perceptual model for realizing each artificial intelligence body obtain itself week
The visible light and infrared image image-forming information enclosed;The cooperative awareness model for each artificial intelligence body obtain itself week
The information of other each artificial intelligence bodies in sensing range is enclosed, and carries out information friendship with each artificial intelligence body
It changes.
Further, in one embodiment of the invention, there are human-computer interaction interfaces for visual in simulation process
Change the parameter information for showing the simulated environment, simulation model and each artificial intelligence body so that user carry out observation and
Modification;Control in simulation process there are interactive voice interface for realizing user to each artificial intelligence body simulation process
System.
In order to achieve the above objectives, another aspect of the present invention embodiment proposes a kind of autonomous multiple agent confronting simulation system
System, comprising: the first configuration module, in simulated environment, according to analogue system setting and simulation system parameters, building sea,
Its integrated real-time simulation scene and artificial intelligence human-computer interaction model;Second configuration module is used for the root in the simulated environment
According to the parameter information of different artificial intelligence bodies, artificial intelligence body Model is constructed, and different artificial intelligence bodies load different confrontation
Task and algorithm construct autonomous Multi-Agent simulation frame;Iteration deduces module, for each artificial intelligence body in the emulation
Deduction is iterated in environment to complete confrontation task and algorithm;Judgment module, for judging that each artificial intelligence body is
No completion confrontation task and algorithm;Output module, after completing confrontation task and algorithm for each artificial intelligence body, by institute
The simulation process information and simulation result information for stating each artificial intelligence body are exported and are saved.
The autonomous multiple agent Counter Simulation System of the embodiment of the present invention, by establishing the man-machine association of multi-dimensional complicated environment
Same Counter Simulation System more true to the Scene realization really fought can be simulated with comprehensive, can also be to simulated environment
Visualization is carried out convenient for user's adjusting parameter.In addition, autonomous multiple agent confronting simulation system according to the above embodiment of the present invention
System can also have following additional technical characteristic:
Further, in one embodiment of the invention, the iteration is deduced mould and is further used for, and updates the emulation
Ambient condition in environment;Each artificial intelligence body updates itself algorithm according to updated ambient condition, according to confrontation
Task is deduced to obtain the algorithm that each artificial intelligence body completes confrontation task;Each artificial intelligence body is according to pushing away
The algorithm drilled takes response action to complete confrontation task.
Further, in one embodiment of the invention, the sea, day one real-time simulation scene include: that three-dimensional can
Light-exposed scene and three-dimensional IR Scene;The three-dimensional visible light scene be the sea of Real-time modeling set and rendering, sky one three
Tie up simulating scenes;The three-dimensional IR Scene is the three-dimensional artificial field using black body radiation model and atmospheric radiation model construction
Scape.
Further, in one embodiment of the invention, the autonomous Multi-Agent simulation frame includes: visual perception
Model and cooperative awareness model;The human perceptual model for realizing each artificial intelligence body obtain around itself can
Light-exposed and infrared image image-forming information;The cooperative awareness model is obtained for each artificial intelligence body and is perceived around itself
The information of other each artificial intelligence bodies in range, and each true intelligent body progress information exchange is imitated with described.
Further, in one embodiment of the invention, further includes: artificial intelligence human-computer interaction module;The emulation
Intelligent human-machine interaction module includes: human-computer interaction interface and interactive voice interface;The human-computer interaction interface is aobvious for visualizing
The parameter information for showing the simulated environment, simulation model and each artificial intelligence body, so that user observes and modifies;
Control of the interactive voice interface for realizing user to each artificial intelligence body simulation process.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, in which:
Fig. 1 is the autonomous multiple agent confronting simulation overall model figure according to one embodiment of the invention;
Fig. 2 is the flow chart according to the autonomous multiple agent confronting simulation method of one embodiment of the invention;
Fig. 3 is to model effect according to the high dynamic three-dimensional visible light of one embodiment of the invention and the rendering of IR Scene
Figure;
Fig. 4 is the autonomous multiple agent architecture diagram according to one embodiment of the invention;
Fig. 5 is the data path schematic diagram according to more autonomous agent parameters of one embodiment of the invention:
Fig. 6 is the intelligent coordinated perception core information center architecture diagram according to one embodiment of the invention;
Fig. 7 is the design frame chart according to the visual perception subsystem of one embodiment of the invention;
Fig. 8 is the autonomous multiple agent confronting simulation overall flow block diagram according to one embodiment of the invention;
Fig. 9 is to deduce subsystem flow chart according to the core of the analogue system of one embodiment of the invention;
Figure 10 is according to the confrontation flow of task in the complex multi-dimensional environment combined simulation system of one embodiment of the invention
Flow chart;
Figure 11 is the autonomous multiple agent Counter Simulation System structural schematic diagram according to one embodiment of the invention.
Specific embodiment
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to is used to explain the present invention, and is not considered as limiting the invention.
The autonomous multiple agent confronting simulation method and system proposed according to embodiments of the present invention are described with reference to the accompanying drawings,
The autonomous multiple agent confronting simulation method proposed according to embodiments of the present invention is described with reference to the accompanying drawings first.
In an embodiment of the present invention, Holistic modeling is carried out to the autonomous multiple agent confronting simulation under complex scene, pressed
Division according to function respectively constructs it, is broadly divided into: main Simulation Model, autonomous Multi-Agent simulation model and
Associative simulation kernel model.Fig. 1 is the autonomous multiple agent confronting simulation overall model figure according to one embodiment of the invention.Under
Face is introduced the modeling of each model therein according to the process in the autonomous multiple agent confronting simulation method of proposition.
Fig. 2 is the flow chart according to the autonomous multiple agent confronting simulation method of one embodiment of the invention.
As shown in Fig. 2, the autonomous multiple agent confronting simulation method the following steps are included:
In step sl, in simulated environment, according to analogue system setting and simulation system parameters, building sea, day one
Real-time simulation scene and artificial intelligence human-computer interaction model.
Further, in one embodiment of the invention, sea, day one real-time simulation scene include: three-dimensional visible light
Scene and three-dimensional IR Scene;Three-dimensional visible light scene is the three-dimensional artificial field on the sea of Real-time modeling set and rendering, sky one
Scape;Three-dimensional IR Scene is the three-dimensional artificial scene using black body radiation model and atmospheric radiation model construction.
In particular it is required that modeled to the dynamic environment of Complex Sea, day one in simulated environment, including visible light with
And infrared scene, while also human-computer interaction model is constructed.
As shown in figure 3, in order to real simulation environment, needing simulation building three-dimensional in autonomous Multi-Agent simulation
Visible light scene and IR Scene, since artificial intelligence body needs to carry out environment autonomous visual perception, three-dimensional visible light
Scene and IR Scene need authenticity with higher, and need to repeat to render multiple.OpenGL is used in the present embodiment
(Open Graphics Library, open graphic library or " open GL) with CUDA (Compute Unified
Device Architecture unifiedly calculates equipment framework) method that combines, utilize GPU (Graphics Processing
Unit, graphics processor) high degree of parallelism, the rendering of high dynamic real-time model three-dimensional visible light scene, IR Scene and
Smog jamming signal type.Three-dimensional visible light scene and IR Scene construct the running environment for emulating autonomous multiple agent together, real
The deduction of confronting simulation process visualization is showed.
Three-dimensional visible light scene is located at the model Visible Scene in the three-dimensional scene models of emulation.In Visible
The rendering to the three-dimensional visible scene of entire simulation model is realized in Scene.It is rendered carries out essentially according to following procedure:
First be setting illumination, set light source position (position) and each component size.Here illumination
Model uses Phong Feng Shi illumination model, principle are as follows: the light for being issued light source is divided into three components, environment light
(ambient), it diffuses (diffuse) and specular light (specular), it is corresponding for each model, for light
Three kinds of components have corresponding three kinds of reflection coefficients, and reflection of ambient light coefficient (ambient), diffuse reflection coefficient
(diffuse) and specular light reflections coefficient (specular).Therefore for observer, the color for the light seen is
The component of three kinds of light of light source accumulated result again after three corresponding coefficient weightings on model.Below to three kinds of light
According to explaining:
Ambient lighting (Ambient Lighting): it even if in the case where dark, has usually remained in the world
Bright the light of distant place (moon), so object is never almost complete darkness.In order to simulate this illumination, will use
One ambient lighting constant, always give object some colors.
Diffuse photograph (Diffuse Lighting): beam-shaping effect (Directional of the analog light source to object
Impact).It is visually most significant component in Phong Feng Shi illumination model.Certain a part of object faces light
Source, can be brighter.
Specular light shines (Specular Lighting): simulating the bright spot that glossy object occurs above, the face that specular light is shone
Color can be more likely to the color of light compared to the color of object.
The superposition of three kinds of light just constitutes the color for the object that observer can see.
Next the color of sky is drawn according to temporal information, and sets the fog information in environment.Subsequent root
The position at visual angle, adjustment perspective and zoom factor are set according to the position of target, are set using the method for perspective projection transformation
It sets.Finally, start the drafting of scene, the obj file of reading model and the letter such as parse vertex therein, normal, face element first
Breath, is then plotted in the three-dimensional scenic of emulation again, then starts the Sea Model of real-time rendering high dynamic.For
The target of interference has been cast, dynamic rendering has been carried out to the smog that it is interfered, render the form and colouring information of smog, just complete
The process of scene rendering.
The rendering modeled segments of environment on sea, an independent module being individually pulled out into analogue system
OceanModule reserves tri- function interfaces of init, update and display_func.Wherein init interface is for module
Initialization, for distributing vertex data and texture cache on GPU, update is used to handle the update of data and display_
The rendering that func interface is then responsible for sea is shown.In init interface, vertex cache to be used in OpenGL is carried out just
Beginningization simultaneously opens up memory headroom, while generating vertical array.Then setting texture information, the texture classes carried using Qt
QOpenGLTexture is loaded into texture, is arranged size and allocation space.After being loaded into texture, generate
The tinter Shader model of OpenGL, finally binds the space and tinter of texture and caching, completes sea mould
The operation of initialization block.And be then the calling realized to CUDA C code in update interface, main function is to texture
Data carry out dynamic realtime update.The display on sea is realized in display_func interface: being to set sight first
Position, direction and the visual field parameter for examining visual angle, then set the information such as projection matrix, the light source position at visual angle, last basis
Each vertex carries out color applying drawing in sequence can be achieved with the Dynamically Announce on sea, in the caching calculation resources that will have been bound
It is discharged with the memory of texture cell.
The render process of sky and the render process on sea be it is similar, difference therein is that the rendering of sky is not used
To the acceleration of CUDA, but the texture mapping for the sky being loaded directly into.Sky has been used in the rendering of sky in simulating scenes
The technology of box (Sky Sphere), the core of day sylphon technology be the scene of emulation to be abstracted into a cuboid box, so
The textures of a sky are attached on six faces of box afterwards, the processing such as some splicings, smooth, filtering are carried out on boundary, just
As soon as a celestial environment true to nature can be simulated, a complete sky can be seen from the visual angle in box.In the present embodiment
A sky textures next life true to nature sylphon all day long is used, while this puts up figure can also generate the inverted image of sky across the sea,
After having added the environment of sea sky, simulating scenes are more true, and there is an inverted image of sky on sea, and illumination ripple etc. can be with
Accomplish that the random variation of dynamic realtime, the introducing of new environment can preferably simulate actual ambient conditions.
Further, three-dimensional IR Scene rendering is located in the three-dimensional scene models of emulation in model InfraredScene.
IR Scene InfraredScene and three-dimensional visible light scene VisibleScene have similar structure and process, uniquely not
Same is: for sea, sky and other each models, no longer by the way of textures, but according to its temperature information
Black body radiation is calculated, the color of object is calculated further according to radiant illumination information, to realize the simulation to infrared visual angle.
Further, in one embodiment of the invention, there are human-computer interaction interfaces for visual in simulation process
The parameter information for changing display simulated environment, simulation model and each artificial intelligence body, so that user observes and modifies;Imitative
Control during true there are interactive voice interface for realizing user to each artificial intelligence body simulation process.
Intelligent human-machine interaction interface is modeled, graphical interface of user is built using Qt GUI frame, utilizes the graphical user
Interface may be implemented to carry out visualization input to the parameters of parameters and artificial intelligence body in emulation, read and deposit
Storage, ease for use with higher have the characteristics that high efficiency and visual.For parameter setting circle of each artificial intelligence body
Face includes update_ui interface and update_param interface, the former intuitively shows all data in Parameters data structure
Show on interface, after user makes modification, the updated data on interface are passed to Parameters data structure by the latter, just
User is realized to update data manually and be transmitted to the process in simulation model.
It emulates in set interface, inner in tabs " artificial intelligence body ", user oneself can select the intelligence for needing to emulate
Body, can oneself selection type and quantity.Left mouse button clicks some or point in one column intelligent body Target Photo of left side
Hit the tabs of different intelligent body type, so that it may be switched to the introduction page of corresponding intelligent body, click " addition XX mesh
Mark " button can add corresponding intelligent body target, and the intelligent body of addition can be shown in the list of lower section.And adding
On the ListItem for adding each intelligent body in target, it is arranged there are two button and deletes, it respectively can be to current after click
The parameter setting and the current intelligent body target of deletion of intelligent body target.The setting for clicking the ListItem of some intelligent body is pressed
Button just can enter secondary parameter setting modification interface, can modify to every simulation parameter of the intelligent body.
Intelligent human-machine interaction interface is modeled, the man-machine interactive interface of autonomous multiple agent opposed decision-making analogue system can be with
Realize the insertion and control of external complex decision making algorithm.The distributed fortune of analogue system may be implemented using the man-machine interactive interface
Row, i.e., analogue system and external algorithm end can carry out data exchange by way of network, while the interface realizes emulation
The isomerization of system, the different programming language of different platforms can realize fusion.Man-machine interactive interface calculates external decision
Method and intelligent body binding, extend the scope of application of emulation, provide the foundation for the emulation study verifying of increasingly complex algorithm,
Embody the strong antagonism of game, the rapidity of response and the uncertainty on boundary.
Man-machine coordination Decision Control interface realizes from the external control for reading trainer and inputs information, such as keyboard, mouse
The input equipments such as mark, rocking bar, according to different mission requirements and training method, by the input signal for controlling equipment carry out coding and
Recombination, is translated into the instruction that artificial intelligence body is understood that, further artificial intelligence body root according to certain control protocol
Corresponding behavior and decision are made according to command content, realize intervention and interaction of the external staff to intelligent body behavior.
In autonomous multiple agent joint Counter Simulation System, emulation is realized using the communication mode of Server-Client
Human-computer interaction and control in the process.Wherein, main emulation platform establishes network monitoring service, according to certain authentication mode, prison
It listens from external connection request.After receiving external request, determined to need according to the content of external request to be given
Simulated environment data and intelligent volume data, the corresponding control instruction returned later further according to outside are come to each intelligence in emulation
Energy body is updated iteration, to realize external artificial control and intervention and control of the intelligent algorithm to simulation process.External people
It can be any type and type for control and data algorithm, it is only necessary to according to certain specification protocol.Specifically, imitative
True human-computer interactive control will be carried out according to following process:
(1) after autonomous Multi-Agent simulation system brings into operation, using in the QNetwork module of Qt
QNetworkSocket establishes socket server, monitors outside port connection.
(2) Client client sends connection request, connects the designated port of Server server-side.
(3) Server termination is by the end Client connection request, and sends primary data information.Primary data information contains
In current simulation parameter environment, the essential information of each autonomous multiple agent and the essential information of simulated environment.
(4) end Client receives initial data information, and according to mission requirements or decision, selection needs the autonomous intelligence controlled
The ID number of energy body, and the end Server is sent back, the end Server sends confirmation message, and ensures the end Client of each connection
It has been ready.
(5) end Server sends emulation commencing signal to each end Client.During iteration of simulations each time,
The information of each autonomous agent and real-time status are sent to the end Client of its corresponding control, Client root by the end Server
Policy-maker is carried out according to the information of intelligent body, and the control signal of intelligent body is sent to the end Server.The end Server is according to receipts
To the control signal of Client each autonomous agent controlled, to complete the loop iteration once emulated.
In step s 2, in simulated environment, according to the parameter information of different artificial intelligence bodies, artificial intelligence body mould is constructed
Type, and different artificial intelligence bodies load different confrontation tasks and algorithm, construct autonomous Multi-Agent simulation frame.
Further, in one embodiment of the invention, autonomous Multi-Agent simulation frame includes: human perceptual model
And cooperative awareness model;Wherein, human perceptual model for realizing each artificial intelligence body obtain visible light around itself and
Infrared image image-forming information;Cooperative awareness model obtains other each in sensing range around itself for each artificial intelligence body
The information of a artificial intelligence body, and information exchange is carried out with each artificial intelligence body.
It emulates autonomous multiple agent and represents each independent simulation unit in simulation process, each artificial intelligence body
There are oneself threedimensional model and parameter model, while different task or algorithm can be carried on each intelligent body.It is followed in emulation
During ring iterative is run, intelligent body can obtain information from the environment around itself, or from other intelligence around itself
Body obtains information, to go to complete specific task and algorithm.In autonomous multiple agent confronting simulation, as shown in figure 4, from
The framework of main multiple agent are as follows:
(1) artificial intelligence body: referring to the side attacked in antagonistic process in emulation platform, needs to pass through release in simulations
Smog interference come influence attack the infrared detector of tracker realize confrontation, can also be evaded by flexible mode
It threatens.Artificial intelligence body has a naval vessel, helicopter, fighter plane, transporter these four, contain each ring from sea to sky
Border.For each intelligent body, they have different a model and characteristic, but some basic parameter types be it is similar, such as
The basic informations such as speed, position, temperature.There can be the jamming equipment of different number to cope with prestige on each intelligent body simultaneously
The side of body, can detect threaten after in suitable distance release jamming bomb, while each intelligent body can carry different type and
The sensor of quantity, such as camera, radar, sensor module are primarily referred to as building for intelligent body visual perception in analogue system
Mould realizes perception of the intelligent body to simulated environment around for simulating the practical function of camera under real Antagonistic Environment.
(2) jamming equipment: refer to the device threatened on each intelligent body to simulate interference, it contains multiple interference moulds
Type, and each interference also has different types, such as smoke screen, the interference of face source etc..Each interference has different mist forms, can
To simulate different actual conditions.Jamming equipment will also record the type of each interference, whether emitted, transmitting angle simultaneously
The information such as degree and time, and be updated when iteration of simulations.The quantity of jamming equipment on each intelligent body is different
, jamming bomb type nor identical.
(3) it attacks tracker: referring to side's intelligent body of the threat attack in simulation process.Due to there is sea in the scene
The intelligent body of the different levels of face sky threatens tracker also to have different types, carries out in the type of parameter and numerical value
Difference, for example, size, speed, maximum speed and acceleration, the radius of damage, the available machine time and distance etc..Simultaneously in three-dimensional scenic
In can be also distinguish to attack tracker using different threedimensional models to be different types of.It is also being had on tracker to attack
Detector module, to simulate the infrared detector in real equipment.Core on detector module is the infrared of its operation
Track algorithm, the infrared image for threatening intelligent body to obtain are transmitted to detector, and the track algorithm on detector is obtained according to input picture
To displacement information, intelligent body is threatened to update oneself the speed of service and traffic direction according to the feedback of track algorithm, to realize pair
The tracking of intelligent body.The ginsengs such as different imaging sizes, field angle and detection wavelength can be set to attack the detector of tracker
Number, above- mentioned information can be very good the operating status and function of different situations under simulation truth.
(4) policy module: policy module refers to detecting threat, release interference, escape evasion tactics to intelligent body
A kind of modeling.After detecting interference, artificial intelligence body according to itself Situation Awareness to ambient enviroment, in conjunction with itself certainly
Plan algorithm, it is each in order to guarantee the success of confrontation after releasing jamming bomb according to certain strategy release smog interference
Target can generally be fled from according to maximum velocity and acceleration.Tactics module can modify according to the demand of emulation,
Each intelligent body can be selected to formulate different strategies, to realize optimal opposed decision-making.
(5) noise module: noise module is a kind of simulation of a kind of pair of actual environment bring interference.Environment in reality
It inevitably influences whether the imaging of tracker infrared detector and introduces noise, therefore is in simulations that noise section is independent
At a module modeling, the Gaussian noise of varying strength can be loaded on the infrared image of tracker, to reality
Carry out more true simulation.Mainly after the completion of IR Scene rendering, the infrared image of generation is passed to for the introducing of noise
Before tracking module, Gaussian noise can be added according to the parameter of setting to infrared image among this.
In order to make above-mentioned module have unified interface, each submodule can all be inherited from one when realizing and be called
The virtual base class of all models of the representative of Model.In this virtual base class, it must be realized there are two interface:
1) Initialize: initialization interface.It can be updated in agent model in initialization interface intelligent body parameter,
Including information such as position, speed, direction, interference.Simultaneously for each by attack artificial intelligence body, it is also necessary to its jamming equipment
It is initialized.Initialization operation typically occurs in when analogue system operation for the first time renders interface and each emulation terminates preparation
When starting to emulate next time.
2) Update: more new interface.Update interface simulation intelligent body carry out the update to oneself parameter, including oneself
The information such as position, the direction of motion, speed, interference.And for each tracker, jamming equipment can also be updated to determine
Whether the strategy of release is interfered and is interfered in release, updates operation generation in simulation cycles iteration.
The two interfaces can be simulated nucleus module Simulation emulation and call, and therefore, ensure that respective characteristic
In the case where, each intelligent module in emulation platform can accomplish modularization to greatest extent, realize main program and intelligent body
The coupling divided in discrete data in modular structure.
To autonomous multiple agent three-dimensional modeling, the threedimensional model for creating each autonomous agent needs to create their geometry
Then model is reused the technology of OpenGL and is shown to this model in program in a manner of as true as possible.With aircraft intelligence
The geometrical model of building intelligent body, the first method by being scanned to true aircraft are illustrated how for energy body, it can
To obtain the geometrical model data of the aircraft intelligent body, there are many file memory formats for obtained data, wherein using to be
Obj file.The letter such as apex coordinate, vertex scheme vector, vertex order, texture coordinate of object module is stored in obj file
Breath.Next it seeks to using OpenGL show the obj file of target to carry out rendering, a three-dimensional is created in OpenGL
Object needs the method using triangle surface, needs using the coordinate and normal information for arriving each face of target and vertex, because
This, obj file is loaded into program platform, reads out the vertex information of target, and mainly data type is GLdouble's
Array vertices [] three big, normal [], faces [], have respectively represented coordinate, normal vector direction and the vertex on vertex
Sequentially, then each triangle surface of target is rendered using the technology of OpenGL, establishes and has more the three of the sense of reality
Dimension module.
There is the information of basic structure and each plane by aforesaid operations agent model, model does not have
Three-dimensional sense and the sense of reality.In order to solve this problem, the lighting function inside using OpenGL is needed.In real life, people
The color of seen object depend on the frequency of electromagnetic wave that the surface of the object is radiated.For transparent substance, light irradiation
When to its surface, the color of object seen by person is codetermined by the reflected light and transmitted light of object;For opaque object
Body, color seen by person are only determined by the reflected light of object.Therefore, it is necessary to the mathematical models by establishing illumination come analogies
Body surface face is calculated the color of object reflected light by the phenomenon after illumination.Object under true environment can thus be simulated
Luminous situation so that the threedimensional model of target is more true.
Have in OpenGL to model carry out illumination processing mode, wherein the illumination of object be divided into diffuse, mirror surface
Reflected light and environment light.It is using the method for illumination system rendering triangle surface in OpenGL: setting Lighting information first,
Determine the position of light source, light source ambient light component, the component that diffuses, Specular light components, object material and decaying
Then the parameters such as the factor open illumination system using sentence glEnable (GL_LIGHT), illumination system is by default
It closes, specifies the color of triangle surface, each vertex of a triangle is then specified by glNormal3f () when drawing
Coordinate specifies the normal vector direction on each vertex by function glVertex3f (), is one necessary to rendering illumination
Important parameter finally closes illumination system using sentence glDisable (GL_LIGHT), completes drafting.
After introducing illumination, the surface transition of object module is more naturally, also seem more true.It is each in platform
The visible light model of a intelligent body can appear in the position of setting according to the setting of parameter, and being capable of adjustment direction and angle.
In OpenGL, the model of target is put into the position specified in scene using glTranslate () function, is then used
GlRotate () function carrys out rolling target model to reach the direction of setting.
Autonomous multiple agent parameter is modeled, autonomous multiple agent is the main body simulation object in emulation platform.On
It states independently, refers to that each intelligent body is independent from each other, each other the mutual independent artificial physical in structure, and in number
According to can above have mutual association and transmitting, multiple agent refers to there are multiple targets in simulating scenes, multiple fingers
It is in quantitative and type, including naval vessel, helicopter, fighter plane, transporter target, artificial intelligence body constitute an intelligence
Energy body cluster, is a spatially associated combat unit.
In autonomous Multi-Agent simulation system, in order to track clarification of objective and spy in real simulation reality
Point needs to take out the critical parameter information of target, can be derived that the characteristic of target by parameter information, can judge that target exists
The action or manner of execution that possibility in various situations is taken devise two class target components: base with actual conditions according to demand
This parameter and Monte Carlo parameter and current simulation parameter.
(1) essential information parameter: the basic parameter of artificial intelligence body refers to the ginseng of description target build-in attribute and feature
Number, it is the own attribute of intelligent body.Basic parameter has just been set before emulation is at the beginning, during emulation
It will not change.The basic parameter of target is illustrated by taking naval vessel intelligent body as an example, the basic parameter on naval vessel include: X,
Y, Z coordinate, i.e., initial position in the scene;Maximum speed, peak acceleration and maximum angular rate;Sea X, Y, Z tri-
Information is rocked on direction;Length, the i.e. size on naval vessel;The value volume and range of product of jamming equipment;The temperature of hull and chimney;
Camera number;Model size angle position of camera etc..Above-mentioned parameter is more the essential characteristic and category for describing naval vessel
Property.For helicopter, fighter plane, transporter model, their basic parameter is similar, only in certain parameters class
It type and numerically will be different.
(2) Monte Carlo parameter: if all parameters are all constant in simulation process, it is believed that each time imitative
Very result is all constant, because not introducing Random Effect.Monte Carlo parameter refers to having in simulation process random
Property that a kind of parameter, these parameters have a maximum value and minimum value range, all can be according to equal when emulation each time
Even distribution can be good at the situation under simulating actual conditions after by Multi simulation running from value in this range.Together
When, the presence of this kind of parameters can introduce certain uncertainty to emulation, and therefore, the mode of Multi simulation running can also eliminate it
Randomness.For each intelligent body, their Monte Carlo parameter is similar, for example horizontal driving direction, is attacked
Tracker horizontal direction, interference cast etc..And some parameters are then that certain targets are exclusive, such as aircraft have it is vertical
The direction of motion, it is corresponding to have vertical direction to attack tracker.And threatening the initial distance between target is also to belong to Meng Te
Carlow parameter a, because key factor in emulating can also be become from different distance to attack tracker.Similar goes back
Have to attack the vertical incident direction of the level of tracker and horizontal vertical incidence angle, influences whether the track algorithm and intelligence of tracking equipment
The jamming exposure area of energy body.
In emulation platform, when the setting of parameter occurs mainly in the load of main program frame, graphical interfaces rendering;Parameter
Transmitting occur mainly in platform initialized, each intelligent body start initialization when;And parameter update occurs mainly in emulation
When iterative cycles are iterated each time.As shown in figure 5, therefore various parameters have different data paths in the different stages:
1) it brings into operation in platform, when main program starts to render graphic user interface, main program is read in local file
Parameter.The comprehensive parameters comprising all intelligent body information are read first in a file, contain each intelligent body
The whole information such as quantity, the quantity of tracker, the corresponding relationship of each intelligent body and tracker and title.Then according to intelligence
Energy body number of species information, main program read parameter in the corresponding file of these intelligent bodies.Main program reads each emulation intelligence
Can body basic parameter (BasicParam) and Monte Carlo parameter (MonteCarloParam), complete parameter from local to
The transmitting of main program.Then the value volume and range of product that main program obtains intelligent body before creates the intelligent body of corresponding number,
And the dll file of intelligent body is loaded into memory, then the parameter of each intelligent body is transmitted to intelligent module from main program
On, the final transmitting of simulation parameter is completed, each intelligent body has been obtained for the parameter information of oneself.
2) then the emulation nucleus module Simulation of emulation platform starts to carry out initial work.In emulation core
In initialization procedure, main program can to each agent model call Initialize interface, to each intelligent body into
Row initialization.It, will be in both the basic parameter that obtained before it and Monte Carlo parameter in each intelligent body initialization section
A part is updated in its current simulation parameter (CurrentParam).Master when current simulation parameter is iteration of simulations circulation
The parameter to be updated, and the source of last data when saving simulation result.
3) when carrying out simulation cycles, the update of parameter is occurred mainly on current simulation parameter (CurrentParam).
In intelligent body module initialization before, current simulation parameter has passed through basic parameter and Monte Carlo gain of parameter
Initial value, in each iteration of simulations, current simulation parameter will be updated according to other parameters, updated parameter
The state for illustrating intelligent body after current iteration, at the end of emulation, current simulation parameter be can serve as in emulation
Key parameters are taken as simulation result to extract.By taking naval vessel as an example, when being updated, it will be updated sea first and rock
Phase, frequency and amplitude, the position on naval vessel is then updated according to the rate on this and naval vessel, while also determining whether to detect
To tracker is attacked, if detected it is also predefined attacking direction and discharges interference, the naval vessel after releasing interference
It will be evaded according to escape mechanism, after emulation, the information of emulation last moment is had recorded in current simulation parameter
And whether confrontation succeeds, finally, basic parameter, Monte Carlo parameter and current simulation parameter are used as one of simulation result
Divide and is recorded.
Further, a variety of multiple artificial intelligence body-sensings-are established and know collaboration confrontation model, core is to establish core
Acquisition of information and perception approach.For " sense ", each artificial intelligence body can obtain perception in a manner of visual pattern sequence
The variation of ambient enviroment situation information;For " knowing ", a high-level center for information management is established in simulations, is responsible for collecting week
The information of local environment situation information and each artificial intelligence body is enclosed, the mode that each target can then simulate realistic communication is logical
Center for information management is crossed to obtain " the knowing " of Incomplete information and itself situation, the input for itself opposed decision-making algorithm.
Information centre can determine intelligent body according to the situation of local environment locating for intelligent body and intelligent body self character
The range and content for the information that can be obtained.Therefore, the core of Multi-Agent simulation independent, artificial evolution alone by multiple target
It is shared for multiple-object information, the Coordination Decision emulation of information Perception, information transmitting.
During emulating deduction, confrontation target is enemy to attack tracker target, and intelligent body is obtained according to from information centre
The local situation information taken and the status information of itself, according to corresponding opposed decision-making algorithm and the letter of combination ambient intelligence body
Breath realizes that typical " attack-keep " process embodies more to attacking the interference of target and hiding to make corresponding motor-driven and behavior
Confrontation and game between intelligent body and threat target.
Specifically, in confronting simulation before, the antagonistic process of emulation emphasizes the independent confrontation of single intelligent body
Process, therefore, single intelligent body intelligently cope with the threat (such as one piece is attacked a naval vessel to attack tracker) of single intelligent body, such as
There are multiple intelligent bodies in fruit, then can interfere to come the single intelligent body attack strategies for attacking tracker, can not determine target of attack;
If there is multiple threats, the jamming exposure area under the single threat of intelligent body can also be impacted.Such emulation is without real
Now collaboration confrontation, that is, the ability of information sharing substantially, being embodied in each artificial intelligence body can only obtain locating for oneself
The situation information of environment, and each artificial intelligence body is not known the information of other intelligent bodies in current environment or is tracked to attack
The information of device.The thinking for solving the problems, such as this is to establish an artificial intelligence center module, as shown in fig. 6, the information center module
Different from other artificial intelligence bodies, it has higher permission, and each emulation under current simulated environment can be collected by being embodied in it
The information of intelligent body, and each artificial intelligence body can obtain the situation information in its sensing range from information center module,
Information center module constructs the approach of information sharing, becomes the core of intelligent coordinated system.
Therefore, in analogue system, the core of global information center module (Global) and other artificial intelligence bodies in emulation
There is the different stages in heart iterative cycles, in the different stages in global information center module and artificial intelligence cognition make
Different behaviors, and realize mutual information transmitting.Mainly include following sections:
1) artificial intelligence body initializes: each artificial intelligence body carries out initial according to the parameter that main platform framework passes over
Change, each artificial intelligence body can be placed on corresponding position in simulating scenes according to the location information in parameter, directional information
It sets and direction, and is configured and updates according to every essential information of the parameter to artificial intelligence body.It is tracked simultaneously for attack
Device target loads corresponding track algorithm plug-in unit according to its parameter information and is initialized.In each artificial intelligence of this step
On your marks for body, completes the transmitting and effect of this simulation parameter, starts to get ready for emulation in next step.
2) global information center initializes: after the initialization for completing each artificial intelligence body, global information center
From whole global angle, the information of each intelligent body in emulation Antagonistic Environment at present is collected, and according to each intelligent body
Independent unduplicated ID to carry out categorised collection to the information of intelligent body.Realize sense of the Global module to current Antagonistic Environment
Know, obtain the situation information of each intelligent body of current Antagonistic Environment, and is initialized the state of each artificial intelligence body.It is above-mentioned
The state of artificial intelligence body refers to whether emulation terminates (DefenceOver) and whether confrontation succeeds
(DefenceSuccessful), it is True that the former, which is False the latter, under original state.
3) artificial intelligence body iteration updates: artificial intelligence body is made according to the opposed decision-making algorithm of its operation to be updated, packet
Include motor-driven variation, interference release and information sharing etc..This stage artificial intelligence body can be current to Global module polls
It whether there is possible threat intelligent body outside with the presence or absence of other intelligent bodies and itself sensing range in itself sensing range.
According to the limited local information acquired, intelligent body can make increasingly complex decision and motor-driven.
4) global information center iteration updates: after completing artificial intelligence body iteration and updating, Global module is to mesh
Preceding updated emulation Antagonistic Environment is perceived, collect and update confrontation database in each intelligent body information, and according to
Surrounding's situation information needed for the demand transmission intelligent body of intelligent body attacks tracker information with potential.In last Global mould
Block the relationship (relative position, radius of damage etc.) of tracker and each artificial intelligence body can be attacked according to each come judge it is each come attack tracking
The intelligent body whether device hits, thus to judge whether the antagonistic process of intelligent body terminates and whether succeed.
Perception and acquisition of the global information center module to Antagonistic Environment are realized according to aforementioned four step, is built simultaneously
The bridge communicated between each artificial intelligence body, the information of itself available sensing range of artificial intelligence body realize
The situation information of part is shared and transmits, and realizes the collaborative and intelligence of confrontation.
Further, it to realize perception of the artificial intelligence body to environment, needs to design visual perception subsystem, utilizes vision
Subsystem is perceived, artificial intelligence body can be from angle, according to specific imaging mode --- including visible light, infrared, depth
Deng --- the image information of the situation of ambient enviroment is obtained, to assist carrying out decision using these image informations and appoint accordingly
Business.As shown in fig. 7, the perceptive mode that the subsystem can greatly extend artificial intelligence body is believed from simple situation is passively obtained
Breath is changed into the situation information for initiatively obtaining perception ambient enviroment, and the emulation of visual perception subsystem is artificial intelligence body
" eye ".
The core of the design of visual perception subsystem is that each intelligent body is designed multiple cameras for it and perceives mould
Block Camera, each Camera can simulate corresponding imaging function, and parameter includes the size of imaging, the type of imaging
And view directions of imaging etc..By the way that different imaging parameters are carried out with the building and modeling of image, can be realized from three-dimensional
Object to two-dimension picture conversion, to realize to the visual perception function of environment in visual angle.
Realize the modeling to the autonomous visual perception of intelligent body with the following method in analogue system:
Firstly for each intelligent body target, it is arranged corresponding camera parameter, i.e., each intelligent body can be taken
With any number of Camera module, and each Camera module has parameter setting alone, as camera is taken relative to it
The position of the intelligent body of load, relative to the angle of the intelligent body carried, the size of imaging surface, the type of imaging and imaging
The information such as field angle.
Secondly it in simulation process to the rendering composition of scene, using the mechanism of OpenGL, needs to each
Scene under the visual angle Camera carries out imaging building, and using the picture in present viewing field as picture acquired in Camera present frame
Face, so that it may the perception as intelligent body to the scene under the current visual angle Camera.For from the three-dimensional scenic to two-dimensional imaging
Building, using following formula:
Wherein right side coordinate is the three-dimensional coordinate of threedimensional model and scene, using its own as coordinate system;Left side is to be transformed into
The coordinate with depth information after two dimension indicates the coordinate on two-dimensional imaging face.The two coordinates are next coordinate
Form, and three transformation matrixs therein are the homogeneous transform matrix of 4x4, finally realize three-dimensional coordinate to two-dimensional coordinate
Conversion, is below introduced each matrix.
MmodelIt is the matrix of a linear transformation of standard, including translation (Transform), rotates (Rotate) and scale
(Scale) matrix.The effect of the matrix is to be transformed into the point on each three-dimension object entirely from the coordinate system of three-dimension object itself
In the middle of the three-dimensional system of coordinate of office, therefore, which can unify all three-dimension objects to a global whole seat
Mark system, realizes the unification of different coordinate systems, the transformation and rendering convenient for the later period to each three-dimension object.
MviewIt is viewing field of camera transformation matrix, is defined as:
Wherein p indicates the position of observer (camera) in three-dimensional world coordinate system.Defining c is center in camera fields of view
Pointed position, then vector r illustrates that the vector of standard dextrad under present viewing field, vector u illustrate present viewing field subscript
Quasi- upward vector.By the transformation of the matrix, the coordinate points in three-dimensional world have been switched to the imaging under camera perspective, right
It can be carried out rejecting without rendering in the point outside camera imaging face, wash with watercolours can greatly be promoted by only rendering visible point
The efficiency of dye.
MprofectionIt is perspective projection transformation matrix, is defined as:
Wherein θ illustrates the vertical field of view angle of camera imaging, and the parameter is bigger, and the scene that can be seen is also wider.
Aspect is the ratio of width to height that picture is imaged, and f indicates the position of imaging far plane, and n indicates the position of imaging hither plane.Perspective projection
Transformation matrix can convert coordinate points according to the imaging law of reality, so that transformed point is with " close big remote
It is small " transparent effect, better authenticity.
By the transformation of three matrixes, the coordinate points of a threedimensional model can be transformed into two-dimensional camera imaging plane
On, to realize that the rendering to three-dimensional scenic is modeled and obtained.The image is the presentation of the three-dimensional scenic under camera perspective,
The information source of intelligent body Coordination Decision perception will be become as visual perception information, to realize more complicated decision.
The embodiment of the present invention specifically includes that main simulation universal framework, Multi-Agent simulation frame and associative simulation core
Module.The basic unit of main simulation universal framework building collaboration confronting simulation, composition includes: man-machine interactive system and three-dimensional scenic
System, wherein three-dimensional scenic system constructs complicated high dynamic sea, day one simulated environment in real time, and man-machine interactive system handles people
Machine interaction and control, Multi-Agent simulation frame includes: a variety of, multiple artificial intelligence bodies, visual perception system and collaborative perception
System, intelligent body carries out decision and motor-driven according to algorithms of different task, in conjunction with the local situation information of itself and perception, thus complete
At target, associative simulation nucleus module constructs and coordinates each simulation unit, realizes that emulation promotes and develops.
The present invention has taken out each element in simulation process, the basic framework function of main simulation universal framework building emulation
Can, can be realized it is man-machine can interactive controlling simulation process deduction, and construct high dynamic complex three-dimensional sea, day integrally emulation ring
Border, Real Time Dynamic Modeling has simultaneously rendered sea with true effect, day one environment, and builds expansible human-computer interaction and connect
Mouthful.
In an embodiment of the present invention, Multi-Agent simulation frame unit further comprises: infrared target radiation detection mould
Block, for generating the infrared imaging in confrontation scene and being applied to a variety of, multiple intelligent bodies;Algoritic module instructs intelligent body to exist
Corresponding task and target are completed according to itself perception information in analogue system.
In an embodiment of the present invention, associative simulation nucleus module further comprises: simulation result module, for recording simultaneously
The information of each intelligent body and final simulation result in simulation process are generated, and is analyzed for subsequent algorithm;Dynamic data mould
Block obtains the image that there is high dynamic smog to block and interfere and corresponding Pixel-level mark in real time in simulation process, from
And construct complicated shielded image database.
As shown in figure 8, in step s3, each artificial intelligence body changes in the simulated environment that step S1 and S2 are configured
In generation, deduces to complete confrontation task and algorithm.
Before emulation starts, the parameter of the parameter of each artificial intelligence body, emulation platform system parameter and simulated environment,
It is read from outside to analogue system in the way of graphical interface of user, and for constructing and initializing each artificial intelligence
Body.Artificial intelligence body loads corresponding assignment algorithm according to mission requirements, or is bound with external control decision algorithm,
To realize the loading to Decision-making of Agent module.Complex environment model required for emulation is established simultaneously, includes high dynamic
Sea and celestial environment model.After everything in readiness, simulation process starts under artificial triggering, intracardiac into simulated core
In the iteration cycle process in portion, start to deduce emulation.After to simulation process, generated according to the information emulated at this time
This emulation as a result, on the one hand be each intelligent body algorithm deduction result, can be carried out with true value matching and comparison;
On the other hand the image sequence that emulation generates constitutes the image sequence data library under complex environment, subsequent offline to assist
Supervised decision making algorithm research.
Further, in one embodiment of the invention, S3 further comprises: S31, updates the environment in simulated environment
State;S32, each artificial intelligence body update itself algorithm according to updated ambient condition, are deduced according to confrontation task
Obtain the algorithm that each artificial intelligence body completes confrontation task;S33, the algorithm that each artificial intelligence body is obtained according to deduction, is adopted
Response action is taken to complete confrontation task.
Wherein, simulated environment model is updated, comprising: wave of the sea, wind speed and direction, the interference of dynamic smog.Due to intelligent body
Type is different, and Decision-making of Agent behavior includes but is not limited to: release interference, tracks and identifies adjustment posture.
The core iteration deduction system of emulation platform, is deduced by the way of loop iteration.Therefore emulation platform base
In step-length period regular time, the modules in analogue system are updated repeatedly according to certain sequence within the period
In generation, deduces emulation according to the sequence of time to realize.It is first right before simulation process integrally brings into operation among these
The whole visualization interface of emulation platform is initialized, and reads parameters, to subsystems and artificial intelligence body
Model is constructed, when simulation control signal starts emulation, as shown in figure 9, the core iterative cycles of emulation are according to following suitable
Sequence carries out:
1) it updates ambient condition: ambient condition being updated according to environmental evolution rule, it is empty such as waving for wave of the sea
In the information such as wind speed and direction, while acquiring the infrared smoke shielded image of scene, and generate mark.
2) intelligent body algorithm is updated: carried decision making algorithm or external binding decision making algorithm on intelligent body, according to environmental state
The variation of gesture is deduced according to different mission requirements, obtains the algorithm output under current time.
3) update intelligent body state: intelligent body takes different response actions according to the output of decision making algorithm, such as carries out
Evasive maneuvering, release smog jamming bomb etc..
4) judge whether antagonistic process terminates, if being not finished, return to 1;Otherwise current iteration of simulations circulation is exited.
Specifically, in the beginning of simulation cycles iteration, be first emulation update is carried out to scene, wherein it is most important just
It is to be updated to environment.For information such as wind speed, wind directions in environment, belong to Monte Carlo parameter, according to the original of randomness
Then, one group of new value can be generated at random to simulate truth according to the range of parameter setting.It updates after completing, it can be by environment
For information function on each intelligent body, each intelligent body can obtain updated environmental element.It is followed by update intelligent body
Maneuvering characteristics, according to information such as the inceptive direction of intelligent body and speed, accelerations, intelligence when can calculate next iteration
Position and direction at body.Since the iterative cycles of emulation are timed calling, calculating each time is all benefit
With regular time step-length.By taking naval vessel intelligent body as an example, intelligent computing agent it is motor-driven when, need to consider sea X, Y,
Brought influence is rocked on tri- directions Z, so needing after calculating the position where intelligent body next time plus one
It is a that introduced thin tail sheep information is rocked by sea.It should be noted that the motor-driven of each intelligent body is updated before,
In contain all artificial intelligence bodies such as naval vessel, aircraft, detector, intelligent body can be to Global global information center die later
Block obtains to attack the information of tracker, and intelligent body can accurately be known to attack the exact position of tracker.If there is intelligence
Body has released interference, to be also updated to interference, and what is mainly updated is the motion change of the smoke particle of jamming bomb
And temperature change, so that the visible light of jamming bomb smog and infrared form change.
Intelligent body carries out motor-driven variation or release cigarette according to its own algorithm carried or the decision making algorithm of external binding
The behaviors such as mist interference, to realize to come the interference of attacking tracker.Later, due to each intelligent body have been completed it is motor-driven, from
Each visual field to attack the acquisition of the tracker direction of motion can also change.In this part, mainly by each detector
On the obtained infrared image of tracker be updated.Each piece will obtain its movement come the tracking equipment attacked on tracker
Infrared view on direction, wherein may including the infrared view of intelligent body and interference.If threatening tracker at this time
Reach booting distance or just start to track, then the detector on tracker will be run, using the infrared view as defeated
Enter.Track algorithm is also to exist as independent module in emulation platform, they are all inherited from this module of Tracker,
With unified interface format.
After the track algorithm in tracking equipment has been run, available new motion information.What track algorithm obtained
Data are stored in the structural body of entitled DLL_MISSILE_DECISION, contain the information of track algorithm, such as mould
The horizontal adjusting value and vertical adjustment amount of the version size and location of window, the size and location of search window and tracking center.
Intelligent body before can feeding back to this result after the operation result for obtaining track algorithm, tracker obtain tracking knot
After fruit, it will adjust the position of oneself direction according to the horizontal adjusting value of tracking center and vertical adjustment amount, while each
Come attack tracker also can using track algorithm result adjust oneself movement velocity and direction, come so that intelligent body goes out always
The centre at present detector visual angle, thus will not be with losing target.
The iteration of primary emulation core is completed by the above process.When first time iteration, input parameter is each
The initial parameter of target;In iteration each time, there would not be data input again, each subparameter update is all the ginseng before
On the basis of number;After emulation, the picture and mark of the image perception of simulation result and tracker and artificial intelligence body
It is generated and exports, to realize the process once emulated.
It as shown in Figure 10, is the confrontation flow of task flow chart in complex multi-dimensional environment combined simulation system.Below to every
One stage is described in detail.
1) stage started in program, main program frame will create the display interface view of program first, generate on interface
Each control that can be shown, including detector home court scape view, infrared view, polar coordinates view, Simulation Control button etc..Together
When, main program can also read the simulation parameter information of preservation in local file, can read one first and contain population parameter
The file of ModelParam, in the platform recorded in parameter presently, there are intelligent body type, number and each intelligent body
Task type and the requirement of association and each target between their corresponding detectors, go again according to this related information
The parameter information of each target is read in corresponding Parameter File.After obtaining the intelligent body information in current emulation,
Main program will start to create Simulation data structure, and Simulation represents the core of emulation, in its building
During, mainly the dll file of each intelligent module is loaded into platform, according to jamming exposure area, jamming equipment, is imitated
True intelligent body, track algorithm, tracker sequence be loaded into.
2) user can open the simulation parameter set interface in program, to modify to emulation.It is imitated in face of different
True demand, user can choose addition or delete each intelligent body, can modify to the parameters of each intelligent body.
After confirmation is over the setting to artificial intelligence body, for removed target, DLL module can be released from memory
The DLL module for the intelligent body put, and be newly added can then be loaded into memory.Complete the modification to intelligent body and parameter it
Afterwards, emulation core Simulation will be by the parameter information of artificial intelligence body (source is that local file is read and interface inputs)
Each module is passed to, and each module is initialized.The primary operational of initialization is exactly by each intelligent body
Parameter be configured.During this time, the links of emulation are ready, and main program interface has generated, each
Intelligent body has been loaded, has obtained parameter and be initialised, and waits the beginning of emulation.
3) after emulating start button when the user clicks, the timer of control emulation core brings into operation, analogue system
Center iterative cycles start to be called according to certain frequency.By taking infrared counteraction tracing task as an example: in this emulation major cycle
In the middle, main operation is updated in parameter, the motor-driven and task to each intelligent body: it is more new environment first, including
Wind direction of ocean surface wind speed temperature etc., after the completion of environmental renewal, each intelligent body can obtain current environmental information, then update
The jamming bomb (if intelligent body has discharged) of the release of each intelligent body, mainly update the movement of its each particle so that
The temperature of entire jamming bomb puff profile changes, by the body directly in the infrared view of home court scape visible light view detector
Reveal and.Starting to update the motor-driven of each intelligent body again, each intelligent body is moved according to the speed and the direction of motion of oneself,
Motor-driven due to intelligent body, the infrared view that each detector obtains also just is changed, then operates in detector tracker
On track algorithm new output will be also generated according to this, this output in contain the motion excursion amount of tracking center.It
Detector afterwards will correct oneself movement velocity and direction according to the data that track algorithm obtains, to realize the mesh of tracking
's.This completes the updates of detector intelligent body.
4) simulation cycles are just iterated according to mode as above, recycling each time finally, all can be to each mesh
Target task is determined whether end.After determining that the task to each target has finished on, emulation is followed at this time
Ring will stop iteration, and simulation run terminates.After emulation, emulation core can make some important parameters of each intelligent body
For simulation result generation, the simulation result of generation can be then written in local file according to certain format, in order to later
Analysis to the algorithm in emulation, while the sensor being arranged on artificial intelligence body perceptual image generated is labeled life
At, be finally saved in together with simulation result this.It, will be again to main program if being provided with whole Multi simulation running
It is initialized with each intelligent body, is ready to emulate next time.
In emulation overall flow, the beginning, pause and end of emulation can be controlled by clicking the button on interface.
Before emulation starts, parameter setting interface can be entered by click " parameter set button ", to modify the intelligent body of emulation
Type and number and the parameter of they and detector.It, can due to there are multiple intelligent bodies simultaneously during emulation
Main view angle of field and infrared window are shown that content switches between each intelligent body by realizing by Ctrl key, thus
It can be realized the observation to the antagonistic process of each intelligent body.
In step s 4, judge whether each artificial intelligence body completes confrontation task and algorithm, it, will be each imitative if completing
The simulation process information and simulation result information of true intelligent body are exported and are saved, if not completing, are returned to S3 and are continued iteration
It deduces.
In emulation core iterative cycles module, most important function is that modules are updated with simultaneously Transfer Parameters,
But in each iterative cycles finally, will judge whether that emulation should terminate.And the end of a simulation process is meant that
The confrontation of all intelligent bodies is in current platform, therefore how to determine that the end of each intelligent body antagonistic process is this
The key of problem.Judge whether the confrontation of an intelligent body terminates to depend primarily on two o'clock:
If 1) in simulation cycles final stage, the small Mr. Yu of the distance between some intelligent body tracker corresponding with its
A threshold value, it is possible to determine that the confrontation of intelligent body terminates, and intelligent body confrontation failure, intelligent body is hit to attack tracker.
2) when the condition in 1) is unsatisfactory for, the distance between intelligent computing agent and corresponding detector, if detecting this
Distance both when carving the distance of intelligent body and detector greater than last iteration, determines that the confrontation of intelligent body terminates, intelligent body
It fights successfully, does not hit intelligent body to attack tracker.Above-mentioned implied condition is that detector is to beginning to target is all directed towards eventually
Flight, then distance between the two must just successively decrease.If the distance of both a certain moment discoveries becomes larger,
Mean that the distance of last moment between the two has had reached minimum value, following detector will be far from intelligent body flight.
Tracker does not track intelligent body at this time, brushes past with intelligent body, therefore, it is possible to determine that confrontation terminates, and detector is not hit
Middle intelligent body.
Assuming that have N number of intelligent body in current platform, phase whether end whether end with the confrontation of this N number of intelligent body of emulation
It closes.Only when the antagonistic process of each intelligent body terminates, it is believed that emulation terminates.For being over the intelligence of confrontation
Body, will no longer to its parameter and it is motor-driven be updated, also no longer rendering is thirdly the infrared model of peacekeeping, to reduce calculation amount.
The autonomous multiple agent confronting simulation method proposed according to embodiments of the present invention, by establishing a multi-dimensional complicated ring
Man-machine coordination Counter Simulation System in border more true to the Scene realization really fought can be simulated with comprehensive, can also be right
Simulated environment carries out visualization convenient for user's adjusting parameter.
The autonomous multiple agent Counter Simulation System proposed according to embodiments of the present invention is described referring next to attached drawing.
Figure 11 is the autonomous multiple agent Counter Simulation System structural schematic diagram according to one embodiment of the invention.
As shown in figure 11, which includes: the configuration of the first configuration module 100, second
Module 200, iteration deduce module 300, judgment module 400 and output module 500.
Wherein, the first configuration module 100 be used in simulated environment, according to analogue system setting and simulation system parameters,
Building sea, day one real-time simulation scene and artificial intelligence human-computer interaction model.
Second configuration module 200 is used in simulated environment, according to the parameter information of different artificial intelligence bodies, building emulation
Agent model, and different artificial intelligence bodies load different confrontation tasks and algorithm, construct autonomous Multi-Agent simulation frame.
Iteration deduces module 300 and is iterated deduction in simulated environment for each artificial intelligence body to complete confrontation times
Business and algorithm.
Judgment module 400 is for judging whether each artificial intelligence body completes confrontation task and algorithm.
After output module 500 completes confrontation task and algorithm for each artificial intelligence body, by each artificial intelligence body
Simulation process information and simulation result information are exported and are saved.
The system 10, can be to really fighting by establishing a multi-dimensional complicated environment man-machine coordination Counter Simulation System
The more true and comprehensive simulation of Scene realization.
Further, in one embodiment of the invention, iteration is deduced mould and is further used for, and updates in simulated environment
Ambient condition;Each artificial intelligence body updates itself algorithm according to updated ambient condition, is deduced according to confrontation task
Obtain the algorithm that each artificial intelligence body completes confrontation task;The algorithm that each artificial intelligence body is obtained according to deduction, takes sound
It should act to complete confrontation task.
Further, in one embodiment of the invention, sea, day one real-time simulation scene include: three-dimensional visible light
Scene and three-dimensional IR Scene;Three-dimensional visible light scene is the three-dimensional artificial field on the sea of Real-time modeling set and rendering, sky one
Scape;Three-dimensional IR Scene is the three-dimensional artificial scene using black body radiation model and atmospheric radiation model construction.
Further, in one embodiment of the invention, autonomous Multi-Agent simulation frame includes: human perceptual model
And cooperative awareness model;Human perceptual model obtains the visible light and infrared figure around itself for realizing each artificial intelligence body
As image-forming information;Cooperative awareness model obtains other each emulation around itself in sensing range for each artificial intelligence body
The information of intelligent body, and information exchange is carried out with each true intelligent body is imitated.
Further, in one embodiment of the invention, further includes: artificial intelligence human-computer interaction module;Artificial intelligence
Human-computer interaction module includes: human-computer interaction interface and interactive voice interface;Human-computer interaction interface emulates ring for visualization display
Border, simulation model and each artificial intelligence body parameter information so that user observes and modifies;Interactive voice interface is used for
Realize control of the user to each artificial intelligence body simulation process.
It should be noted that the aforementioned explanation to autonomous multiple agent confronting simulation embodiment of the method is also applied for this
The system of embodiment, details are not described herein again.
The autonomous multiple agent Counter Simulation System proposed according to embodiments of the present invention, by establishing a multi-dimensional complicated ring
Man-machine coordination Counter Simulation System in border more true to the Scene realization really fought can be simulated with comprehensive, can also be right
Simulated environment carries out visualization convenient for user's adjusting parameter.
In addition, term " first ", " second " are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of autonomous multiple agent confronting simulation method, which comprises the following steps:
S1, in simulated environment, according to analogue system setting and simulation system parameters, building sea, day one real-time simulation scene
With artificial intelligence human-computer interaction model;
S2, according to the parameter information of different artificial intelligence bodies, constructs artificial intelligence body Model in the simulated environment, and
Different artificial intelligence bodies load different confrontation tasks and algorithm, construct autonomous Multi-Agent simulation frame;
S3, each artificial intelligence body are iterated deduction in the simulated environment that step S1 and S2 are configured and are appointed with completing confrontation
Business and algorithm;
S4, judges whether each artificial intelligence body completes confrontation task and algorithm, if completing, by each emulation intelligence
The simulation process information and simulation result information of energy body are exported and are saved, if not completing, are returned to S3 and are continued the iteration
It deduces.
2. autonomous multiple agent confronting simulation method according to claim 1, which is characterized in that S3 further comprises:
S31 updates the ambient condition in the simulated environment;
S32, each artificial intelligence body update itself algorithm according to updated ambient condition, are pushed away according to confrontation task
It drills to obtain the algorithm that each artificial intelligence body completes confrontation task;
S33, the algorithm that each artificial intelligence body is obtained according to deduction, takes response action to complete confrontation task.
3. autonomous multiple agent confronting simulation method according to claim 1, which is characterized in that
The sea, day one real-time simulation scene include: three-dimensional visible light scene and three-dimensional IR Scene;
The three-dimensional visible light scene is the three-dimensional artificial scene on the sea of Real-time modeling set and rendering, sky one;
The three-dimensional IR Scene is the three-dimensional artificial scene using black body radiation model and atmospheric radiation model construction.
4. autonomous multiple agent confronting simulation method according to claim 1, which is characterized in that
The autonomous Multi-Agent simulation frame includes: human perceptual model and cooperative awareness model;
Wherein, the human perceptual model obtains visible light around itself and infrared for realizing each artificial intelligence body
Image image-forming information;
The cooperative awareness model obtains other described each in sensing range around itself for each artificial intelligence body
The information of a artificial intelligence body, and information exchange is carried out with each artificial intelligence body.
5. autonomous multiple agent confronting simulation method according to claim 1, which is characterized in that further include:
There are human-computer interaction interfaces for simulated environment, simulation model described in visualization display and described each in simulation process
The parameter information of artificial intelligence body, so that user observes and modifies;
Control in simulation process there are interactive voice interface for realizing user to each artificial intelligence body simulation process
System.
6. a kind of autonomous multiple agent Counter Simulation System characterized by comprising
First configuration module, for according to analogue system setting and simulation system parameters, constructing sea, day one in simulated environment
Body real-time simulation scene and artificial intelligence human-computer interaction model;
Second configuration module, in the simulated environment, according to the parameter information of different artificial intelligence bodies, building to emulate intelligence
Energy body Model, and different artificial intelligence bodies load different confrontation tasks and algorithm, construct autonomous Multi-Agent simulation frame;
Iteration deduces module, is iterated deduction in the simulated environment for each artificial intelligence body to complete confrontation task
With algorithm;
Judgment module, for judging whether each artificial intelligence body completes confrontation task and algorithm;
Output module, after completing confrontation task and algorithm for each artificial intelligence body, by each artificial intelligence body
Simulation process information and simulation result information export and save.
7. autonomous multiple agent Counter Simulation System according to claim 6, which is characterized in that the iteration deduce mould into
One step is used for,
Update the ambient condition in the simulated environment;
Each artificial intelligence body updates itself algorithm according to updated ambient condition, deduce according to confrontation task
The algorithm of confrontation task is completed to each artificial intelligence body;
The algorithm that each artificial intelligence body is obtained according to deduction, takes response action to complete confrontation task.
8. autonomous multiple agent Counter Simulation System according to claim 6, which is characterized in that
The sea, day one real-time simulation scene include: three-dimensional visible light scene and three-dimensional IR Scene;
The three-dimensional visible light scene is the three-dimensional artificial scene on the sea of Real-time modeling set and rendering, sky one;
The three-dimensional IR Scene is the three-dimensional artificial scene using black body radiation model and atmospheric radiation model construction.
9. autonomous multiple agent Counter Simulation System according to claim 6, which is characterized in that
The autonomous Multi-Agent simulation frame includes: human perceptual model and cooperative awareness model;
The human perceptual model obtains visible light and infrared image around itself for realizing each artificial intelligence body
Image-forming information;
The cooperative awareness model obtains other described each in sensing range around itself for each artificial intelligence body
The information of a artificial intelligence body, and each true intelligent body progress information exchange is imitated with described.
10. autonomous multiple agent Counter Simulation System according to claim 6, which is characterized in that further include: artificial intelligence
Human-computer interaction module;
The artificial intelligence human-computer interaction module includes: human-computer interaction interface and interactive voice interface;
The human-computer interaction interface is used for simulated environment described in visualization display, simulation model and each artificial intelligence body
Parameter information, so that user observes and modifies;
Control of the interactive voice interface for realizing user to each artificial intelligence body simulation process.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910045079.2A CN109740283A (en) | 2019-01-17 | 2019-01-17 | Autonomous multiple agent confronting simulation method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910045079.2A CN109740283A (en) | 2019-01-17 | 2019-01-17 | Autonomous multiple agent confronting simulation method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109740283A true CN109740283A (en) | 2019-05-10 |
Family
ID=66365219
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910045079.2A Pending CN109740283A (en) | 2019-01-17 | 2019-01-17 | Autonomous multiple agent confronting simulation method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109740283A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110147883A (en) * | 2019-05-28 | 2019-08-20 | 航天科工系统仿真科技(北京)有限公司 | Training method, device, equipment and the storage medium of model for emulation of fighting |
CN110309861A (en) * | 2019-06-10 | 2019-10-08 | 浙江大学 | A kind of multi-modal mankind's activity recognition methods based on generation confrontation network |
CN110428057A (en) * | 2019-05-06 | 2019-11-08 | 南京大学 | A kind of intelligent game playing system based on multiple agent deeply learning algorithm |
CN110728636A (en) * | 2019-09-17 | 2020-01-24 | 杭州群核信息技术有限公司 | Monte Carlo rendering image denoising model, method and device based on generative confrontation network |
JP2020193885A (en) * | 2019-05-29 | 2020-12-03 | 川崎重工業株式会社 | Underwater object search project review supporting system, and underwater object search project review supporting method |
CN112131786A (en) * | 2020-09-14 | 2020-12-25 | 中国人民解放军军事科学院评估论证研究中心 | Target detection and distribution method and device based on multi-agent reinforcement learning |
CN112257875A (en) * | 2020-11-13 | 2021-01-22 | 北京理工大学 | Task understanding method of multiple intelligent agents based on extreme learning machine |
CN112381454A (en) * | 2020-12-04 | 2021-02-19 | 中国科学院自动化研究所 | Role-based multi-agent task cooperative system |
CN113468735A (en) * | 2021-06-24 | 2021-10-01 | 国汽(北京)智能网联汽车研究院有限公司 | Laser radar simulation method, device and system and storage medium |
CN113656963A (en) * | 2021-08-16 | 2021-11-16 | 中国电子科技集团公司第五十四研究所 | Distributed combat deduction simulation system capable of realizing real-time interactive control |
CN113705102A (en) * | 2021-08-31 | 2021-11-26 | 湖南苍树航天科技有限公司 | Deduction simulation system, method, equipment and storage medium for sea-air cluster confrontation |
CN113836754A (en) * | 2021-11-26 | 2021-12-24 | 湖南高至科技有限公司 | Multi-agent simulation modeling oriented simulation method, device, equipment and medium |
CN114247144A (en) * | 2021-12-21 | 2022-03-29 | 清华大学 | Multi-agent confrontation simulation method and device, electronic equipment and storage medium |
CN112131786B (en) * | 2020-09-14 | 2024-05-31 | 中国人民解放军军事科学院评估论证研究中心 | Target detection and distribution method and device based on multi-agent reinforcement learning |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101964019A (en) * | 2010-09-10 | 2011-02-02 | 北京航空航天大学 | Against behavior modeling simulation platform and method based on Agent technology |
WO2012023725A2 (en) * | 2010-08-16 | 2012-02-23 | Lee Jeong-Gab | Method and server for providing advertisements |
CN103279587A (en) * | 2013-04-02 | 2013-09-04 | 北京航空航天大学 | Fighting simulation system based on imaginary description |
CN104135659A (en) * | 2014-08-01 | 2014-11-05 | 西安电子科技大学 | Visible light and infrared target scene simulation and loading system, and application method |
CN104732019A (en) * | 2015-03-12 | 2015-06-24 | 清华大学 | Sea-surface infrared countermeasure simulation system |
CN106682351A (en) * | 2017-01-10 | 2017-05-17 | 北京捷安申谋军工科技有限公司 | Fight simulation system generating military strength based on computer and simulation method |
CN108733897A (en) * | 2018-04-28 | 2018-11-02 | 上海烜翊科技有限公司 | Attack-defense Confrontation Simulation System based on architectural framework model and emulation mode |
-
2019
- 2019-01-17 CN CN201910045079.2A patent/CN109740283A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012023725A2 (en) * | 2010-08-16 | 2012-02-23 | Lee Jeong-Gab | Method and server for providing advertisements |
CN101964019A (en) * | 2010-09-10 | 2011-02-02 | 北京航空航天大学 | Against behavior modeling simulation platform and method based on Agent technology |
CN103279587A (en) * | 2013-04-02 | 2013-09-04 | 北京航空航天大学 | Fighting simulation system based on imaginary description |
CN104135659A (en) * | 2014-08-01 | 2014-11-05 | 西安电子科技大学 | Visible light and infrared target scene simulation and loading system, and application method |
CN104732019A (en) * | 2015-03-12 | 2015-06-24 | 清华大学 | Sea-surface infrared countermeasure simulation system |
CN106682351A (en) * | 2017-01-10 | 2017-05-17 | 北京捷安申谋军工科技有限公司 | Fight simulation system generating military strength based on computer and simulation method |
CN108733897A (en) * | 2018-04-28 | 2018-11-02 | 上海烜翊科技有限公司 | Attack-defense Confrontation Simulation System based on architectural framework model and emulation mode |
Non-Patent Citations (3)
Title |
---|
王三喜 等: "基于复杂适应系统理论多智能体模型技术的分队对抗仿真模型建立", 《中国运筹学会第八届学术交流会议集》 * |
鄢超波 等: "多智能体并行仿真框架", 《系统仿真学报》 * |
陈艳彪 等: "基于多智能体的坦克分队对抗仿真模型研究", 《系统仿真学报》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110428057A (en) * | 2019-05-06 | 2019-11-08 | 南京大学 | A kind of intelligent game playing system based on multiple agent deeply learning algorithm |
CN110147883A (en) * | 2019-05-28 | 2019-08-20 | 航天科工系统仿真科技(北京)有限公司 | Training method, device, equipment and the storage medium of model for emulation of fighting |
JP7240256B2 (en) | 2019-05-29 | 2023-03-15 | 川崎重工業株式会社 | Underwater object search plan study support system and underwater object search plan study support method |
JP2020193885A (en) * | 2019-05-29 | 2020-12-03 | 川崎重工業株式会社 | Underwater object search project review supporting system, and underwater object search project review supporting method |
CN110309861A (en) * | 2019-06-10 | 2019-10-08 | 浙江大学 | A kind of multi-modal mankind's activity recognition methods based on generation confrontation network |
CN110728636A (en) * | 2019-09-17 | 2020-01-24 | 杭州群核信息技术有限公司 | Monte Carlo rendering image denoising model, method and device based on generative confrontation network |
CN112131786A (en) * | 2020-09-14 | 2020-12-25 | 中国人民解放军军事科学院评估论证研究中心 | Target detection and distribution method and device based on multi-agent reinforcement learning |
CN112131786B (en) * | 2020-09-14 | 2024-05-31 | 中国人民解放军军事科学院评估论证研究中心 | Target detection and distribution method and device based on multi-agent reinforcement learning |
CN112257875A (en) * | 2020-11-13 | 2021-01-22 | 北京理工大学 | Task understanding method of multiple intelligent agents based on extreme learning machine |
CN112381454A (en) * | 2020-12-04 | 2021-02-19 | 中国科学院自动化研究所 | Role-based multi-agent task cooperative system |
CN112381454B (en) * | 2020-12-04 | 2023-10-31 | 中国科学院自动化研究所 | Role-based multi-agent task collaboration system |
CN113468735A (en) * | 2021-06-24 | 2021-10-01 | 国汽(北京)智能网联汽车研究院有限公司 | Laser radar simulation method, device and system and storage medium |
CN113468735B (en) * | 2021-06-24 | 2024-03-22 | 国汽(北京)智能网联汽车研究院有限公司 | Laser radar simulation method, device, system and storage medium |
CN113656963A (en) * | 2021-08-16 | 2021-11-16 | 中国电子科技集团公司第五十四研究所 | Distributed combat deduction simulation system capable of realizing real-time interactive control |
CN113656963B (en) * | 2021-08-16 | 2024-06-11 | 中国电子科技集团公司第五十四研究所 | Distributed combat deduction simulation system capable of being interactively controlled in real time |
CN113705102A (en) * | 2021-08-31 | 2021-11-26 | 湖南苍树航天科技有限公司 | Deduction simulation system, method, equipment and storage medium for sea-air cluster confrontation |
CN113705102B (en) * | 2021-08-31 | 2024-05-10 | 湖南苍树航天科技有限公司 | Deduction simulation system, deduction simulation method, deduction simulation equipment and deduction simulation storage medium for sea-air cluster countermeasure |
CN113836754A (en) * | 2021-11-26 | 2021-12-24 | 湖南高至科技有限公司 | Multi-agent simulation modeling oriented simulation method, device, equipment and medium |
CN114247144A (en) * | 2021-12-21 | 2022-03-29 | 清华大学 | Multi-agent confrontation simulation method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109740283A (en) | Autonomous multiple agent confronting simulation method and system | |
US11461958B2 (en) | Scene data obtaining method and model training method, apparatus and computer readable storage medium using the same | |
Vince | Virtual reality systems | |
AU2017361061A1 (en) | Deep learning system for cuboid detection | |
CN111192354A (en) | Three-dimensional simulation method and system based on virtual reality | |
CN112076473B (en) | Control method and device of virtual prop, electronic equipment and storage medium | |
CN102129709A (en) | Visualizing depth | |
Zhang et al. | Collaborative virtual laboratory environments with hardware in the loop | |
CN115082607A (en) | Virtual character hair rendering method and device, electronic equipment and storage medium | |
CN112802091B (en) | DQN-based agent countermeasure behavior realization method under augmented reality condition | |
Thalmann et al. | Crowd and group animation | |
Cavazza et al. | Intelligent virtual environments for virtual reality art | |
WO2023142609A1 (en) | Object processing method and apparatus in virtual scene, device, storage medium and program product | |
Nahrstedt et al. | Computational Model of Human Creativity in Dance Choreography. | |
Buro et al. | On the development of a free RTS game engine | |
Lan | Simulation of Animation Character High Precision Design Model Based on 3D Image | |
KR102655962B1 (en) | Remote Drone Tactical Training System | |
CN112870694B (en) | Picture display method and device of virtual scene, electronic equipment and storage medium | |
Brown et al. | Quake ii as a robotic and multi-agent platform | |
Schmalstieg et al. | Constructing a highly immersive virtual environment: a case study | |
Meloni | Learning from Video Streams: Virtual Environments and Parallel Computation | |
Ho et al. | Fame, soft flock formation control for collective behavior studies and rapid games development | |
Jędrasiak et al. | Interactive application using augmented reality and photogrammetric scanning | |
Smith et al. | Augmented reality sandpit simulating ant colonies | |
Dimitriu et al. | A Reinforcement Learning Approach to Military Simulations in Command: Modern Operations |
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 |