CN112699603A - Machine learning-oriented ultra-fast air combat simulation method and system - Google Patents
Machine learning-oriented ultra-fast air combat simulation method and system Download PDFInfo
- Publication number
- CN112699603A CN112699603A CN202011594401.6A CN202011594401A CN112699603A CN 112699603 A CN112699603 A CN 112699603A CN 202011594401 A CN202011594401 A CN 202011594401A CN 112699603 A CN112699603 A CN 112699603A
- Authority
- CN
- China
- Prior art keywords
- simulation
- airplane
- data
- control
- oriented
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The application belongs to the technical field of simulation model design, and particularly relates to an ultra-fast air combat simulation method and system for machine learning. The system comprises a plurality of nodes, the plurality of nodes collect data uniformly through a cloud platform, and each node comprises: the intelligent decision module is used for carrying out AI algorithm learning, generating an airplane control instruction and giving a planned command and an initialization command; the simulation control system is used for carrying out simulation process control and data synchronous management on a plurality of simulation operation frameworks connected with the simulation control system; and the simulation operation frameworks are used for simulating the countermeasure maneuver according to the airplane control instruction. According to the method and the device, data interaction among multiple processes is carried out in a memory sharing mode, and the problem that deployment cannot be achieved due to network configuration is solved.
Description
Technical Field
The application belongs to the technical field of simulation model design, and particularly relates to a machine learning-oriented ultra-fast air combat simulation system construction method.
Background
With 2017, AlphaGo wins the world in man-machine war of go, and a new era of intelligent algorithm is opened. A new era of intelligent systems for machine learning. In the field of aircraft design, the application of machine learning algorithms is still in the search phase. Because the machine learning algorithm extremely depends on the simulation operation environment, the existing simulation environment can not support machine learning, and the main reason is that:
a) the system operation efficiency is not high. Taking AlphaGo Zero as an example, it has been self-paced 490 ten thousand times before challenge, which requires ultra-fast computing capability of the simulation system. The existing aircraft simulation system is complex, one-time air combat confrontation is operated for about ten minutes, and the machine learning is difficult to converge in a short time due to the operation efficiency, so that the period of intelligent design is prolonged.
b) Large-scale deployment cannot be supported. The training amount of AlphaGo in self-playing is about 3000 thousands of disks, and the AlphaGo is deployed on 16000 simulation nodes. The 16000 nodes must be adopted, the network configuration of the current simulation system is complicated, and large-scale node deployment cannot be supported mostly.
In sum, the defects of the simulation system are an important factor for restricting intelligent air combat, the problem of the simulation system is solved, and the method plays an important role in the application of the machine learning method in the aircraft design.
Most of the existing simulation systems are based on real-time running environments, and the interaction of countermeasure information is usually realized in a network mode, so that the design has the following defects:
firstly, the operating efficiency of the system cannot be guaranteed, and secondly, the deployment of a plurality of nodes cannot be guaranteed and the machine learning and use cannot be well supported due to the complex configuration of the network.
Disclosure of Invention
The application provides an ultra-fast air combat simulation method and system for machine learning, which mainly aim at the problems of machine learning in the field of air combat design and solve the problems of machine learning in the application of air combat design. The method comprises the following specific steps:
a) the problem of system operation efficiency is solved;
b) the problem of connection and control of an intelligent algorithm and a simulation system is solved;
c) the problem of large-scale deployment of the system is solved.
The application provides an ultrafast air combat simulation method facing machine learning in a first aspect, which comprises the following steps:
s1, pushing a preset command to the simulation control system by the intelligent decision module, wherein the preset command is used for initializing a simulation environment;
step S2, the simulation control system distributes the planned command and the initialization information to each simulation operation frame in a shared memory mode, wherein each simulation operation frame is respectively used for simulating one airplane to solve the countermeasure data;
step S3, the intelligent decision module sends an operation instruction, and the simulation control system respectively performs simulation operation control on each simulation operation frame after receiving the operation instruction;
step S4, the simulation control system receives the airplane simulation situation data sent by the simulation operation framework and uploads the airplane simulation situation data to the intelligent decision module;
step S5, after receiving the airplane simulation situation data, the intelligent decision module learns an AI algorithm and generates an airplane control instruction;
and step S6, each simulation operation frame carries out simulation countermeasure maneuver according to the airplane control command.
Preferably, in step S1, the initializing simulation environment includes initializing a battlefield area and initializing an aircraft state, the initializing of the battlefield area includes setting a start longitude, a start latitude, an end longitude and an end latitude, and the initializing of the aircraft state includes setting an aircraft position, a heading, a speed, an oil amount and a number of guided missiles to be carried.
Preferably, in step S2, the countermeasure data solution includes simulating aircraft platform flight control, airframe characteristics, mission management, electronic reconnaissance, electronic countermeasure, fire control management, and weapons management.
Preferably, in step S4, the simulation control system controls the simulation operation framework by data synchronization pushing.
Preferably, in step S6, the simulated countermeasure maneuver includes airplane maneuver control, missile launching, radar switching, and electronic warfare jamming.
The application second aspect provides a machine learning-oriented ultrafast air combat simulation system, including a plurality of nodes, a plurality of nodes carry out data and collect in unison through the cloud platform, and every node includes:
the intelligent decision module is used for carrying out AI algorithm learning, generating an airplane control instruction and giving a planned command and an initialization command;
the simulation control system is used for carrying out simulation process control and data synchronous management on a plurality of simulation operation frameworks connected with the simulation control system;
and the simulation operation frameworks are used for simulating the countermeasure maneuver according to the airplane control instruction.
Preferably, a data recording system is also included for recording process data generated during the simulation.
Preferably, each simulation operation framework is operated in a concurrent mode of shared memory and multiple threads.
Preferably, the simulation control system controls the simulation operation framework in a data synchronous pushing mode.
According to the method and the device, data interaction among multiple processes is carried out in a memory sharing mode, and the problem that deployment cannot be achieved due to network configuration is solved.
Drawings
Fig. 1 is a diagram of an intelligent training structure of a preferred embodiment of the machine learning oriented ultra-fast air combat simulation method of the present application.
Fig. 2 is an architecture diagram of the ultra-fast air combat simulation system for machine learning according to the present application.
FIG. 3 is a schematic diagram of parallelization processing according to the present application.
Detailed Description
In order to make the implementation objects, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be described in more detail below with reference to the accompanying drawings in the embodiments of the present application. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all embodiments of the present application. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application, and should not be construed as limiting the present application. All other embodiments obtained by a person of ordinary skill in the art without any inventive work based on the embodiments in the present application are within the scope of protection of the present application. Embodiments of the present application will be described in detail below with reference to the drawings.
The application provides an ultrafast air combat simulation method facing machine learning in a first aspect, which comprises the following steps:
s1, pushing a preset command to the simulation control system by the intelligent decision module, wherein the preset command is used for initializing a simulation environment;
step S2, the simulation control system distributes the planned command and the initialization information to each simulation operation frame in a shared memory mode, wherein each simulation operation frame is respectively used for simulating one airplane to solve the countermeasure data;
step S3, the intelligent decision module sends an operation instruction, and the simulation control system respectively performs simulation operation control on each simulation operation frame after receiving the operation instruction;
step S4, the simulation control system receives the airplane simulation situation data sent by the simulation operation framework and uploads the airplane simulation situation data to the intelligent decision module;
step S5, after receiving the airplane simulation situation data, the intelligent decision module learns an AI algorithm and generates an airplane control instruction;
and step S6, each simulation operation frame carries out simulation countermeasure maneuver according to the airplane control command.
Correspondingly, this application second aspect provides a machine learning-oriented ultrafast air combat simulation system, including a plurality of nodes, a plurality of nodes carry out data and collect in unison through the cloud platform, and every node includes:
the intelligent decision module is used for carrying out AI algorithm learning, generating an airplane control instruction and giving a planned command and an initialization command;
the simulation control system is used for carrying out simulation process control and data synchronous management on a plurality of simulation operation frameworks connected with the simulation control system;
and the simulation operation frameworks are used for simulating the countermeasure maneuver according to the airplane control instruction.
In this embodiment, the multi-node deployment is a deployment mode of a multi-node intelligent training environment facing a cloud platform, in this case, a deployment schematic diagram of an ultrafast air combat simulation system facing machine learning is shown in fig. 1, it is required that the intelligent system and the simulation system are in the same node, and finally, the confrontation results are collected by the cloud platform in a unified manner.
The intelligent system mainly refers to an intelligent decision-making module, the simulation system adopts an ultra-fast air combat simulation system facing machine learning, and as shown in figure 2, the simulation has three components: an ultrafast simulation control system, an ultrafast countermeasure simulation framework, and a data recording system.
And the simulation control system is responsible for performing simulation process control and data synchronous management on the framework. The system comprises functions of a planning management part, an operation control management part, a model management part, a data management part, a communication management part and a configuration management part. The simulation framework is used as a core unit in the air combat simulation countermeasure simulation system and is responsible for simulating functional characteristics of aircraft platform flight control, organism characteristics, task management, electronic reconnaissance, electronic countermeasure, fire control management, weapon management and the like. The data recording system is used for recording process data generated in the simulation process and can be used for analyzing results.
In some optional embodiments, a high concurrent process scheduling mechanism is adopted, and in order to improve the operation efficiency in the process of carrying out air combat countermeasure by the simulation system, the simulation control system stops a clock propulsion mode and changes a data synchronous propulsion mode. When the method is adopted, after the simulation framework is operated, the state is returned to the simulation control system, the simulation control system can immediately push the simulation framework to continue to operate next beat, and the time delay caused by clock waiting is reduced, so that the operation efficiency is improved.
In some optional embodiments, a high-concurrency data interaction mechanism is adopted, the intelligent decision module receives situation information of the simulation system in each simulation period to carry out resolving, and then the decision information and the control information are sent to the simulation system, so that action of the warplane is controlled. After the configuration file is determined, the simulation control system allocates a memory address to each simulation frame, each simulation frame uses a shared memory according to the identifier in the simulation, each simulation frame processes data by adopting multiple processes or multiple threads, concurrent execution is adopted between the processes and the threads, and the execution time and the response time of the program are reduced by adopting a mode that a CPU scheduling program executes in turn.
In some optional embodiments, the machine learning oriented ultra-fast air combat simulation method comprises the following steps:
1. and starting the intelligent decision module and the ultra-fast simulation system, and performing network connection (TCP) between the systems after the intelligent decision module and the ultra-fast simulation system are started.
2. After the system network is connected, the intelligent decision module pushes a default command to the ultra-fast simulation control system, and the default command is used for an initial simulation environment, including a battlefield area (initial longitude, initial latitude, end longitude and end latitude), and aircraft initialization information (including position, course, speed, oil quantity, guided missile carrying quantity and the like).
3. After receiving the set information instruction, the ultra-fast simulation control system distributes the information to the ultra-fast simulation operation frames (the airplane 1 and the airplane 2) in a shared memory mode, and then returns a set instruction response message to the intelligent decision module.
4. The intelligent decision module sends an initialization instruction, the ultra-fast simulation control system respectively initializes the ultra-fast simulation operation frames after receiving the initialization instruction, and then returns an initialization instruction response message to the intelligent decision module.
5. And the intelligent decision module sends an operation instruction, the ultra-fast simulation control system respectively carries out simulation operation control on the ultra-fast simulation operation framework after receiving the operation instruction, and then returns an operation instruction response message to the intelligent decision module.
6. And in the running process of the ultra-fast simulation system, the ultra-fast simulation control system controls the step length propulsion of the ultra-fast simulation running frame according to a synchronous propulsion mechanism, and in the simulation running resolving process, the ultra-fast simulation running frame sends the aircraft simulation situation data to the intelligent decision module in each step.
7. And after receiving the airplane simulation situation data, the intelligent decision module performs AI algorithm learning and generates airplane control instructions.
8. The ultra-fast simulation operation framework carries out simulation countermeasure maneuver (including airplane maneuver control, missile launching, radar switch, electronic warfare interference and the like) through an intelligent decision control instruction sent by the intelligent decision module.
9. And when the simulation training is finished, the intelligent decision module sends a stop command to finish the confrontation simulation training.
According to the method, the machine learning-oriented ultra-fast air combat simulation method is adopted, data interaction among multiple processes is carried out in a memory sharing mode, and the problem that deployment cannot be achieved due to network configuration is solved. The method is put into use, and each simulation beat takes 80 microseconds by means of a shared memory and multithreading concurrency. Under the single node 4 core 8G memory, 4000 simulation nodes can be supported for machine learning at the same time.
Having thus described the present application in connection with the preferred embodiments illustrated in the accompanying drawings, it will be understood by those skilled in the art that the scope of the present application is not limited to those specific embodiments, and that equivalent modifications or substitutions of related technical features may be made by those skilled in the art without departing from the principle of the present application, and those modifications or substitutions will fall within the scope of the present application.
Claims (9)
1. A machine learning-oriented ultra-fast air combat simulation method is characterized by comprising the following steps:
s1, pushing a preset command to the simulation control system by the intelligent decision module, wherein the preset command is used for initializing a simulation environment;
step S2, the simulation control system distributes the planned command and the initialization information to each simulation operation frame in a shared memory mode, wherein each simulation operation frame is respectively used for simulating one airplane to solve the countermeasure data;
step S3, the intelligent decision module sends an operation instruction, and the simulation control system respectively performs simulation operation control on each simulation operation frame after receiving the operation instruction;
step S4, the simulation control system receives the airplane simulation situation data sent by the simulation operation framework and uploads the airplane simulation situation data to the intelligent decision module;
step S5, after receiving the airplane simulation situation data, the intelligent decision module learns an AI algorithm and generates an airplane control instruction;
and step S6, each simulation operation frame carries out simulation countermeasure maneuver according to the airplane control command.
2. The machine learning-oriented ultrafast air combat simulation method of claim 1, wherein in step S1, the initialization simulation environment includes a battlefield area initialization and an aircraft state initialization, the battlefield area initialization includes setting a start longitude, a start latitude, an end longitude, and an end latitude, and the aircraft state initialization includes setting an aircraft position, a heading, a speed, an oil amount, and a number of guided missiles to be carried.
3. The machine-learning-oriented ultrafast air war simulation method of claim 1, wherein in step S2, the solution of countermeasure data includes simulation of aircraft platform flight control, airframe characteristics, mission management, electronic reconnaissance, electronic countermeasure, fire control management, and weapons management.
4. The machine-learning-oriented ultrafast air war simulation method of claim 1, wherein in step S4, the simulation control system controls the simulation operation framework by data synchronous propulsion.
5. The machine learning-oriented ultrafast air combat simulation method of claim 1, wherein in step S6, the simulated countermeasure maneuver includes airplane maneuver control, missile launching, radar switch, and electronic combat disturbance.
6. The utility model provides a machine learning oriented ultrafast air combat simulation system which characterized in that, includes a plurality of nodes, and a plurality of nodes carry out data and collect in unison through the cloud platform, and every node includes:
the intelligent decision module is used for carrying out AI algorithm learning, generating an airplane control instruction and giving a planned command and an initialization command;
the simulation control system is used for carrying out simulation process control and data synchronous management on a plurality of simulation operation frameworks connected with the simulation control system;
and the simulation operation frameworks are used for simulating the countermeasure maneuver according to the airplane control instruction.
7. The machine-learning-oriented ultrafast air combat simulation system of claim 6, further comprising a data recording system for recording process data generated during the simulation process.
8. The machine-learning-oriented ultrafast air combat simulation system of claim 6, wherein each simulation running framework runs in a shared memory and multi-thread concurrent manner.
9. The machine-learning-oriented ultrafast air combat simulation system of claim 6, wherein the simulation control system controls the simulation operation framework by data synchronous propulsion.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011594401.6A CN112699603A (en) | 2020-12-29 | 2020-12-29 | Machine learning-oriented ultra-fast air combat simulation method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011594401.6A CN112699603A (en) | 2020-12-29 | 2020-12-29 | Machine learning-oriented ultra-fast air combat simulation method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112699603A true CN112699603A (en) | 2021-04-23 |
Family
ID=75511884
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011594401.6A Pending CN112699603A (en) | 2020-12-29 | 2020-12-29 | Machine learning-oriented ultra-fast air combat simulation method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112699603A (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060142903A1 (en) * | 2002-12-05 | 2006-06-29 | Nir Padan | Dynamic guidance for close-in maneuvering air combat |
CN108021754A (en) * | 2017-12-06 | 2018-05-11 | 北京航空航天大学 | A kind of unmanned plane Autonomous Air Combat Decision frame and method |
CN109656147A (en) * | 2018-11-23 | 2019-04-19 | 中国航空工业集团公司沈阳飞机设计研究所 | Air-combat simulation system |
-
2020
- 2020-12-29 CN CN202011594401.6A patent/CN112699603A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060142903A1 (en) * | 2002-12-05 | 2006-06-29 | Nir Padan | Dynamic guidance for close-in maneuvering air combat |
CN108021754A (en) * | 2017-12-06 | 2018-05-11 | 北京航空航天大学 | A kind of unmanned plane Autonomous Air Combat Decision frame and method |
CN109656147A (en) * | 2018-11-23 | 2019-04-19 | 中国航空工业集团公司沈阳飞机设计研究所 | Air-combat simulation system |
Non-Patent Citations (3)
Title |
---|
刘超等: "面向战术级空战智能决策的仿真软件系统设计", 2019中国系统仿真与虚拟现实技术高层论坛, pages 24 - 27 * |
张博;高晓光;: "并行空战仿真系统框架的设计", 微型电脑应用, no. 04, pages 32 - 35 * |
董彦非;汪凯;: "对抗性智能空战目标机设计与实现", 飞行力学, no. 02, pages 13 - 17 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107193639B (en) | Multi-core parallel simulation engine system supporting combined combat | |
CN112295229B (en) | Intelligent game confrontation platform | |
CN101382968B (en) | Advanced synthesis type aviation electronic simulation system and simulation method thereof | |
CN105205271B (en) | A kind of aircraft mission reliability modeling method based on dynamic characteristic | |
CN108108329B (en) | Multi-characteristic analysis method for dynamic reconfiguration strategy of IMA system | |
CN108287959A (en) | Artificial intelligence program person writes digital aircraft source code specifications decision and executes method | |
Paunicka et al. | The OCP-an open middleware solution for embedded systems | |
CN111400895B (en) | Multi-level and multi-granularity cross-domain joint simulation event scheduling method and system | |
CN102122136B (en) | Simulation method for synchronously controlling a plurality of airplanes | |
CN107220107B (en) | Multi-clock multi-task parallel real-time simulation system and method | |
CN112699603A (en) | Machine learning-oriented ultra-fast air combat simulation method and system | |
CN109656147A (en) | Air-combat simulation system | |
Ruan et al. | Task assignment under constraint of timing sequential for cooperative air combat | |
Wu et al. | Mission reliability modeling and evaluation of multi-mission phased mission system based on extended object-oriented petri net | |
CN113205311A (en) | Auxiliary support device and method for carrying equipment for maneuvering transition of aviation soldier troops | |
CN115756789A (en) | GPU scheduling optimization method for deep learning inference service system | |
CN114757057B (en) | Multithreading parallel combat simulation method and system based on hybrid propulsion | |
CN112486036A (en) | Real-time visual cooperative anti-interference simulation platform based on Unity3D, RTX and VMIC technologies | |
Pan et al. | Design of a Distributed Multi-missile Simulation System Based on Optical Fiber Reflective Memory Network | |
RU2623799C2 (en) | Method of task execution sequence assignment in multi-core processor | |
Jing et al. | The Research on Simulation Technology for multi-aircraft cooperative guidance | |
Alver et al. | Assessing the robustness of UAV assignments | |
Bu et al. | Design of Cooperative Combat Task Architecture for Unknown Wide Area Cluster Reconnaissance and Strike | |
Li et al. | Software architecture of C2 mission system based on MAS and SOA for manned/unmanned aerial vehicle formation | |
Agrawal et al. | Real-time adaptive resource management for multi-model control |
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 |