CN116432514A - Interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense game - Google Patents
Interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense game Download PDFInfo
- Publication number
- CN116432514A CN116432514A CN202310143437.XA CN202310143437A CN116432514A CN 116432514 A CN116432514 A CN 116432514A CN 202310143437 A CN202310143437 A CN 202310143437A CN 116432514 A CN116432514 A CN 116432514A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- flight data
- enemy
- neural network
- 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
- 238000004088 simulation Methods 0.000 title claims abstract description 89
- 238000000034 method Methods 0.000 title claims abstract description 71
- 230000007123 defense Effects 0.000 title claims abstract description 31
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 238000013528 artificial neural network Methods 0.000 claims description 140
- 230000015654 memory Effects 0.000 claims description 100
- 238000003066 decision tree Methods 0.000 claims description 36
- 230000004927 fusion Effects 0.000 claims description 29
- 238000012545 processing Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 14
- 238000004422 calculation algorithm Methods 0.000 abstract description 9
- 238000013461 design Methods 0.000 abstract description 2
- 238000012795 verification Methods 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 15
- 210000004027 cell Anatomy 0.000 description 12
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 230000009471 action Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000001133 acceleration Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000005094 computer simulation Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000035484 reaction time Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
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
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses an interception intention recognition strategy simulation system and method for an unmanned aerial vehicle attack and defense game, comprising a fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit; the fight simulation data generating unit is used for generating and storing flight performance parameters and flight data of the friend or foe unmanned aerial vehicle; the complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the interception intention of the enemy unmanned aerial vehicle based on the flight data predicted value of the enemy unmanned aerial vehicle and the flight data of the current me unmanned aerial vehicle, and taking corresponding interception countermeasures. According to the invention, through real-time simulation, the algorithm design and verification efficiency is improved, and the test period is shortened.
Description
Technical Field
The invention relates to an unmanned aerial vehicle group attack and defense simulation system and method, in particular to an interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense games.
Background
Currently, with the rapid development of informatization countermeasure, target intention recognition is a key of the countermeasure win or lose as a premise and a basis of intelligent decision. The intention recognition is to obtain the intention of the opposite party by acquiring the state information of the opposite party, processing and analyzing the state information and reasoning the intention. The accuracy of potential risk assessment can be improved through target intention recognition, the degree and type of risk generated by the target of the opposite party can be accurately recognized, more reaction time is provided for taking proper actions, and the overall development trend is mastered. In the process of multi-unmanned plane game, the detected opposite side information mainly comprises the azimuth, distance, flying speed, flight path and the like of the opposite side unmanned plane group, and the characteristics can reflect the action intention of the opposite side unmanned plane to a certain extent, but due to the complex environment of the opposite sides, the situation of the opposite sides instantaneously changes, and how to quickly and accurately identify the intention of the opposite side becomes a research difficulty.
The existing research methods mainly comprise a template matching method, a Bayesian network, a neural network, a deep learning network and the like. The template matching method and the Bayesian network have high dependence on priori knowledge, and the intention of the opposite party is difficult to accurately identify under complex situations. The neural network has the learning capability of potential data characteristics, overcomes the dependence on priori knowledge, improves the accuracy of intention recognition, and has obvious technical advantages in the problem of intention recognition under complex attack and defense situations. However, three problems still need to be solved in performing intent recognition research based on neural networks: (1) how to obtain a learning sample; (2) how to construct the intent space and generate knowledge tags; (3) how to construct an efficient neural network model.
Disclosure of Invention
The invention provides an interception intention recognition strategy simulation system and method for an unmanned aerial vehicle attack and defense game, which aim to solve the technical problems in the prior art.
The invention adopts the technical proposal for solving the technical problems in the prior art that: an interception intention recognition strategy simulation system for an unmanned aerial vehicle attack and defense game comprises a fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit;
the fight simulation data generating unit is used for generating and storing flight data of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group and my base coordinate data; initializing the speeds and coordinates of the enemy unmanned aerial vehicle and the My unmanned aerial vehicle group, and simulating the flight conditions of the two unmanned aerial vehicles to obtain simulated flight data of the two unmanned aerial vehicles;
the complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the steps of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit;
the interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the attack and defense intention of the enemy unmanned aerial vehicle based on the flight data prediction value of the enemy unmanned aerial vehicle, and taking corresponding interception countermeasures based on the flight data of the current unmanned aerial vehicle group.
Further, the complex environment simulation unit includes a gaussian white noise generator.
Further, the interception intention recognition and decision unit comprises: a fusion prediction module, a K neighbor module and a decision tree module which are jointly constructed by a long-short-term memory neural network and a back propagation neural network;
the fusion prediction module is used for predicting flight data of the enemy unmanned aerial vehicle, inputting the flight data of the enemy unmanned aerial vehicle generated through simulation into the long-short-time memory neural network, and taking a plurality of time sequence prediction results output by the long-short-time memory neural network together as the input of the back propagation neural network; outputting predicted flight data of the enemy unmanned aerial vehicle by a back propagation neural network; the flight data comprises a flight heading, a flight coordinate and a flight speed; generating a predicted trajectory from the predicted flight data;
the K nearest neighbor module is used for processing the predicted flight data output by the fusion prediction module and calculating Euclidean distance between the predicted flight data of the enemy unmanned aerial vehicle and the current flight data of each unmanned aerial vehicle; classifying the predicted trajectories with different priorities based on the probability, and outputting the predicted trajectory of the enemy unmanned aerial vehicle with the highest priority as a final predicted trajectory;
the decision tree module is used for a decision interception process and comprises a decision tree, decision classification nodes are set in the decision tree, and corresponding decisions are made according to the predicted value of the flight data of the enemy unmanned aerial vehicle and the current flight data of the unmanned aerial vehicle group.
Further, the fusion prediction module comprises a long-short-time memory neural network module and a back propagation neural network module which are sequentially connected; the long-time and short-time memory neural network module comprises n long-time and short-time memory neural network units which are connected in sequence; let the current time step be i, the first m time steps be i-m, wherein: m=n-1, i is greater than or equal to m+1; let the flight data of enemy unmanned aerial vehicle at the current time step be X i Let the enemy unmanned aerial vehicle flight data predicted by the long-short-term memory neural network module at the current time step be Y i Setting the coordinate data of the current time step my base as Z i The method comprises the steps of carrying out a first treatment on the surface of the The flight data of the enemy unmanned aerial vehicle from the first 1 time step to the first m time steps are sequentially as follows: x is X i-1 ,X i-2 ,…,X i-(m-2) ,X i-(m-1) ,X i-m The enemy unmanned aerial vehicle flight data predicted by the long-short memory neural network module from the first 1 time step to the first m time steps are sequentially as follows: y is Y i-1 ,Y i-2 ,…,Y i-(m-2) ,Y i-(m-1) ,Y i-m The method comprises the steps of carrying out a first treatment on the surface of the The inputs of the 1 st to n th long-short-term memory neural network units correspond to: x is X i-m ,X i-(m-1) ,X i-(m-2) ,…,X i-2 ,X i-1 ,X i The outputs of the 2 nd to nth long and short time memory neural network units correspond to: y is Y i-(m-1) ,Y i-(m-2) ,…,Y i-2 ,Y i-1 ,Y i The method comprises the steps of carrying out a first treatment on the surface of the 1 st to n-1 st long-short-term memory neural network unitThe memory cell state and the memory cell hiding state of the cell are correspondingly output to the next long-short-time memory neural network unit; a back propagation neural network module, the inputs of which include: z is Z i ,X i-1 ,X i-2 ,…,X i-(m-2) ,Y i-1 ,Y i-2 ,…,Y i-(m-1) Its output includes predicted flight data for the enemy drone.
The invention also provides an interception intention recognition strategy simulation method of the unmanned aerial vehicle attack and defense game, which is provided with an fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit;
the fight simulation data generating unit is used for generating and storing flight data of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group and my base coordinate data; initializing speeds and coordinates of the enemy unmanned aerial vehicle and the My unmanned aerial vehicle group, and simulating the flight conditions of the two unmanned aerial vehicles to obtain simulated flight data of the two unmanned aerial vehicles;
the complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the steps of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit;
the interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the attack and defense intention of the enemy unmanned aerial vehicle based on the flight data prediction value of the enemy unmanned aerial vehicle, and taking corresponding interception countermeasures based on the flight data of the current unmanned aerial vehicle group.
Further, the interception intention recognition and decision unit sets up: a fusion prediction module, a K neighbor module and a decision tree module which are jointly constructed by a long-short-term memory neural network and a back propagation neural network;
the fusion prediction module is used for predicting flight data of the enemy unmanned aerial vehicle, inputting the flight data of the enemy unmanned aerial vehicle generated through simulation into the long-short-time memory neural network, and taking a plurality of time sequence prediction results output by the long-short-time memory neural network together as the input of the back propagation neural network; outputting predicted flight data of the enemy unmanned aerial vehicle by a back propagation neural network; the flight data comprises a flight heading, a flight coordinate and a flight speed; generating a predicted trajectory from the predicted flight data;
the K nearest neighbor module is used for processing the predicted flight data output by the fusion prediction module and calculating Euclidean distance between the predicted flight data of the enemy unmanned aerial vehicle and the current flight data of each unmanned aerial vehicle; classifying the predicted trajectories with different priorities based on the probability, and outputting the predicted trajectory of the enemy unmanned aerial vehicle with the highest priority as a final predicted trajectory;
the decision tree module is used for a decision interception process and comprises a decision tree, decision classification nodes are set in the decision tree, and corresponding decisions are made according to the predicted value of the flight data of the enemy unmanned aerial vehicle and the current flight data of the unmanned aerial vehicle group.
Further, a long-short-time memory neural network module and a back propagation neural network module which are sequentially connected are arranged in the fusion prediction module; the long-short-time memory neural network module is provided with n long-short-time memory neural network units which are connected in sequence; let the current time step be i, the first m time steps be i-m, wherein: m=n-1, i is greater than or equal to m+1; let the flight data of enemy unmanned aerial vehicle at the current time step be X i Let the enemy unmanned aerial vehicle flight data predicted by the long-short-term memory neural network module at the current time step be Y i Setting the coordinate data of the current time step my base as Z i The method comprises the steps of carrying out a first treatment on the surface of the The flight data of the enemy unmanned aerial vehicle from the first 1 time step to the first m time steps are sequentially as follows: x is X i-1 ,X i-2 ,…,X i-(m-2) ,X i-(m-1) ,X i-m The enemy unmanned aerial vehicle flight data predicted by the long-short memory neural network module from the first 1 time step to the first m time steps are sequentially as follows: y is Y i-1 ,Y i-2 ,…,Y i-(m-2) ,Y i-(m-1) ,Y i-m The method comprises the steps of carrying out a first treatment on the surface of the X is to be i-m ,X i-(m-1) ,X i-(m-2) ,…,X i-2 ,X i-1 ,X i Sequentially and correspondingly inputting the data to the 1 st to nth long-short-time memory neural network units; the 2 nd to n th long and short time memory neural network units sequentially and correspondingly output: y is Y i-(m-1) ,Y i-(m-2) ,…,Y i-2 ,Y i-1 ,Y i The method comprises the steps of carrying out a first treatment on the surface of the The memory cell states and the memory cell hiding states of the 1 st to n-1 st long-short-time memory neural network units are correspondingly output to the next long-short-time memory neural network unit, and Z is calculated i ,X i-1 ,X i-2 ,…,X i-(m-2) ,Y i-1 ,Y i-2 ,…,Y i-(m-1) And inputting the predicted flight data to a back propagation neural network module, and outputting predicted flight data of the enemy unmanned aerial vehicle by the back propagation neural network module.
Further, the method comprises the following steps:
step 3, inputting current and historical flight data of each unmanned aerial vehicle of the enemy into a long-short-time memory neural network to conduct multi-step time sequence prediction, inputting multi-step time sequence prediction results into a counter-propagation neural network, and predicting next flight data of the unmanned aerial vehicle of the enemy;
step 4, inputting the predicted value of the flight data of the enemy unmanned aerial vehicle and the flight data of each unmanned aerial vehicle in the current My unmanned aerial vehicle group into a K nearest neighbor module to obtain the predicted value of the flight data of the enemy unmanned aerial vehicle with the highest probability;
step 5, inputting predicted values of flight data of the enemy unmanned aerial vehicles and the flight data of each unmanned aerial vehicle in the current my unmanned aerial vehicle group into a decision tree for decision making so as to give a decision whether the current iteration step is successfully intercepted and whether the interception is continued;
and 6, repeating the steps 3-5 until the iteration ending condition is met.
Further, when the complex environment simulation unit performs noise adding processing on the generated enemy unmanned aerial vehicle flight data, gaussian white noise with the same signal-to-noise ratio is added to all the data based on different assignments of different characteristic signals so as to simulate noise of the acquired signals under actual conditions; the different signal-to-noise ratios are then varied to test differences in prediction accuracy under noise interference of different strengths.
The invention has the advantages and positive effects that: the invention adopts the fight simulation data generating unit to generate and store the flight performance parameters and the flight data of the friend or foe unmanned aerial vehicle; the flight data of the simulated unmanned aerial vehicles are obtained by constructing an unmanned aerial vehicle dynamic simulation model, initializing the speeds and coordinates of the enemy unmanned aerial vehicle and the unmanned aerial vehicle group, simulating the flight conditions of the unmanned aerial vehicles, and simulating a real environment countermeasure scene by adopting the simulated flight data. The complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the step of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit. And predicting the flight data of the enemy unmanned aerial vehicle by adopting an interception intention recognition and decision unit, and determining corresponding interception countermeasures based on the flight data of the enemy unmanned aerial vehicle. Through real-time simulation, the algorithm design and verification efficiency is improved, and the test period is shortened.
According to the method, an unmanned aerial vehicle attack and defense simulation scene is constructed in a preset scene mode; and providing decision information and corresponding decisions in the attack and defense interception process through decision tree construction, and deciding the interception process. The unmanned aerial vehicle countermeasure process is performed by fusing a fusion prediction module, a K neighbor module and a decision tree module which are jointly constructed by the long-short-term memory neural network and the back propagation neural network, the fusion prediction module predicts the respective running tracks of the enemy unmanned aerial vehicle aiming at the complex prediction scene of the unmanned aerial vehicle group, the K neighbor module processes the predicted flight data of the enemy unmanned aerial vehicle, and the decision tree module makes a corresponding decision.
Drawings
Fig. 1 is a schematic structural diagram of an interception intention recognition strategy simulation system for an unmanned aerial vehicle attack and defense game.
FIG. 2 is a schematic diagram of a complex environment simulation unit applying different signal-to-noise ratios.
FIG. 3 is a schematic diagram of the structure of a decision tree in the intercept intent recognition and decision unit.
Fig. 4 is a schematic diagram of a simulated hostile unmanned aerial vehicle attack and defense scenario.
Fig. 5 is a workflow diagram of the K-nearest neighbor module in the intercept intent recognition and decision unit.
Fig. 6 is a schematic diagram of a fusion module structure constructed by combining a long-short-term memory neural network and a back propagation neural network in an interception intention recognition and decision unit.
In fig. 4: r denotes the effective intercept radius.
In fig. 6: x is X i-(m-2) 、X i-(m-1) 、X i-m 、X i The flight data of the enemy unmanned aerial vehicle are sequentially corresponding to the first m-2 time steps, the first m-1 time steps and the first m time steps; y is Y i-(m-2) ,Y i-(m-1) ,Y i Sequentially storing enemy unmanned aerial vehicle flight data predicted by the neural network module for long and short time in the first m-2 time steps, the first m-1 time steps and the current time step; z is Z i The method comprises the steps of providing my base coordinate data for the current time step; LSTM (Linear drive TM) 1 ,LSTM 2 ,...,LSTM n The 1 st to nth long and short time memory neural network units are sequentially and correspondingly arranged; c 1 ,c 2 ,c 3 ,…c n-1 The memory cell states of the neural network units are correspondingly memorized for the 1 st to n-1 st long and short time; h is a 1 ,h 2 ,h 3 ,…h n-1 The hidden states of the memory cells of the neural network units are correspondingly memorized for the 1 st to n-1 st long and short time.
Detailed Description
For a further understanding of the invention, its features and advantages, reference is now made to the following examples, which are illustrated in the accompanying drawings in which:
referring to fig. 1 to 6, an interception intention recognition strategy simulation system for an unmanned aerial vehicle attack and defense game comprises a fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit;
the fight simulation data generating unit is used for generating and storing flight data of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group and my base coordinate data; and initializing the speeds and coordinates of the enemy unmanned aerial vehicle and the My unmanned aerial vehicle group, and simulating the flight conditions of the two unmanned aerial vehicles to obtain the flight data of the simulated two unmanned aerial vehicles.
The complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the step of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit.
The interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the attack and defense intention of the enemy unmanned aerial vehicle based on the flight data prediction value of the enemy unmanned aerial vehicle, and taking corresponding interception countermeasures based on the flight data of the current unmanned aerial vehicle group.
The fight simulation data generating unit can simulate an air fight by using an unmanned aerial vehicle dynamic model. By constructing an unmanned aerial vehicle dynamic model, the speeds and coordinates of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group are initialized, and various complex actual unmanned aerial vehicle group attack and defense conditions can be simulated through programming and parameter setting. And setting a lower limit on the initial value of the speed of the two sides, and judging that the interception is successful when the distance between the unmanned aerial vehicles of the two sides is smaller than the effective interception range. In the simulated flight process of the dynamic simulation model, the flight coordinates and the flight heading of the enemy unmanned aerial vehicle, and the flight speeds and the flight heading of each unmanned aerial vehicle of the unmanned aerial vehicle group are continuously changed to simulate the complex and changeable conditions of the real air combat. After the unmanned aerial vehicle dynamic model is operated, the flight speed and the flight heading which are simulated by the unmanned aerial vehicle of the two parties are selected and recorded, the sampling frequency of data is 100Hz, and the combination of the sampling frequency and the multi-dimension which are high enough provides a plurality of and sufficient data set bases for the subsequent decision.
Simplifying an unmanned aerial vehicle dynamic model into a particle model, wherein the unmanned aerial vehicle dynamic model comprises the following components:
wherein t is the start of the interception activityAny time later; m represents the number of unmanned aerial vehicles in the unmanned aerial vehicle group;represents p j (t) derivative; p is p j (t) represents the position of the jth unmanned aerial vehicle in the my unmanned aerial vehicle group at the moment t; />Representing v j (t) derivative; v j (t) represents the flight speed of the jth unmanned aerial vehicle in the my unmanned aerial vehicle group at the time t; a, a j (t) represents the acceleration control amount of the jth unmanned aerial vehicle in the My unmanned aerial vehicle group at the time t; mu (mu) j Representing the flight resistance constant of the jth unmanned aerial vehicle in the My unmanned aerial vehicle group;
simplifying an unmanned aerial vehicle dynamic model into a particle model, wherein the enemy unmanned aerial vehicle dynamic model comprises the following components:
wherein t is any time after the interception activity starts;represents p e (t) derivative; p is p e (t) represents the position of the enemy unmanned aerial vehicle at the time t; v e (t) represents the flight speed of the enemy unmanned aerial vehicle at the time t; a, a e (t) represents the acceleration control quantity of the enemy unmanned aerial vehicle at the moment t; mu (mu) e And represents the flight resistance constant of the enemy unmanned aerial vehicle.
According to the real environment, the flight data of each unmanned aerial vehicle of the unmanned aerial vehicle group can be detected through the sensor of the unmanned aerial vehicle group, and the flight data of each unmanned aerial vehicle of the unmanned aerial vehicle group can be obtained without being interfered by enemy unmanned aerial vehicles and other environment interference signals. The flight data of the enemy unmanned aerial vehicle needs to be detected by a radar detection device, so that the enemy unmanned aerial vehicle is easy to be interfered by other environmental interference signals.
Preferably, the combat simulation data generation unit may comprise a flight database; the flight database unit is used for storing flight data of the enemy unmanned aerial vehicle and the unmanned aerial vehicle group.
Preferably, the complex environment simulation unit may include a gaussian white noise generator. The complex environment simulation unit can simulate noise interference such as natural environment and electromagnetic signals of enemy when the actual combat situation obtains signals, and Gaussian white noise is added to flight data generated by unmanned aerial vehicle dynamic model simulation flight.
Preferably, the interception intention recognition and decision unit may comprise: the fusion prediction module, the K neighbor module and the decision tree module; the fusion prediction module is constructed by combining a long-short-time memory neural network and a back propagation neural network.
The fusion prediction module which is jointly constructed by the long-short-time memory neural network and the back propagation neural network can be used for predicting flight data of the enemy unmanned aerial vehicle, the flight data of the enemy unmanned aerial vehicle which is generated by simulation of the fight simulation data generation unit is input into the long-short-time memory neural network, and a plurality of time sequence prediction results output by the long-short-time memory neural network are jointly used as input of the back propagation neural network; outputting predicted flight data of the enemy unmanned aerial vehicle by a back propagation neural network; the flight data comprises a flight heading, a flight coordinate and a flight speed; a predicted trajectory is generated from the predicted flight data.
Preferably, the fusion prediction module structure may be as shown in fig. 6, and the fusion prediction module may include a long-short-time memory neural network module and a back propagation neural network module connected in sequence; the long-short-time memory neural network module can comprise n long-short-time memory neural network units which are connected in sequence; the current time step may be set as i, the first m time steps as i-m, where: m=n-1, i is greater than or equal to m+1; can set the flight data of the enemy unmanned aerial vehicle at the current time step as X i The enemy unmanned aerial vehicle flight data predicted by the long-short-term memory neural network module in the current time step can be set as Y i The coordinate data of the current time step my base can be set as Z i The method comprises the steps of carrying out a first treatment on the surface of the The flight data of the enemy unmanned aerial vehicle from the first 1 time step to the first m time steps can be sequentially: x is X i-1 ,X i-2 ,…,X i-(m-2) ,X i-(m-1) ,X i-m The enemy unmanned aerial vehicle flight data predicted by the long-short memory neural network module from the first 1 time step to the first m time steps can be sequentially: y is Y i-1 ,Y i-2 ,…,Y i-(m-2) ,Y i-(m-1) ,Y i-m The method comprises the steps of carrying out a first treatment on the surface of the The inputs of the 1 st to nth long and short term memory neural network units may correspond in order to: x is X i-m ,X i-(m-1) ,X i-(m-2) ,…,X i-2 ,X i-1 ,X i The outputs of the 2 nd to nth long-short-term memory neural network units may correspond in order to: y is Y i-(m-1) ,Y i-(m-2) ,…,Y i-2 ,Y i-1 ,Y i The method comprises the steps of carrying out a first treatment on the surface of the The memory cell states of the 1 st to n-1 st long-short-time memory neural network units and the memory cell hiding states can be correspondingly output to the next long-short-time memory neural network unit; the back propagation neural network module, the inputs of which may include: z is Z i ,X i-1 ,X i-2 ,…,X i-(m-2) ,Y i-1 ,Y i-2 ,…,Y i-(m-1) The output may include predicted flight data for the enemy drone.
The flight data of a plurality of time steps of the enemy unmanned aerial vehicle generated by simulation can be input into the long-short-time memory neural network module, the output of the long-short-time memory neural network module and the coordinate data of the base of the enemy are input into the back propagation neural network module together, and the back propagation neural network module outputs the predicted flight data of the enemy unmanned aerial vehicle.
The long-short-term memory neural network unit sequentially comprises an input layer, an LSTM layer, a full-connection layer, a softmax layer and a classification layer. The LSTM layer is provided with 100 hidden units, and each unit cell structure in the LSTM layer is mainly divided into three links of a forgetting gate, an input gate and an output gate. The long-short-time memory neural network module is used for processing and predicting important events spaced in the time sequence, and predicting the next-step enemy unmanned aerial vehicle flight data according to the current and historical flight data of the existing enemy unmanned aerial vehicle.
The back propagation neural network module trains according to an error back propagation algorithm, input data is gradually processed through the action of an implicit layer until the input data is output, and each neuron parameter and a threshold value are reversely regulated according to the output and the expected error, so that the output is more and more close to the expected value. The training steps include network initialization, forward transfer, reverse transfer, modifying weights, generating a network. The back propagation neural network module is provided with 30 hidden layers, the training algorithm is a Bayesian algorithm, the learning rate is 0.02, and the activation function of the hidden layers is a hyperbolic tangent function:
in the formula, x is E (-1, +1) and is E (-infinity, + -infinity).
The output layer activation function adopts a sigmoid function:
in the formula, x is E (- ≡, ++ infinity a) of the above-mentioned components, S (x) ε (0, 1).
The back propagation neural network module can be used for correcting and optimizing the operation process of the long-short-term memory neural network module on interception target track prediction, and the back propagation neural network module enters BP network operation through returning errors of a plurality of time step flight data and flight data predicted values of the enemy unmanned aerial vehicle in a simulation environment so as to realize parameter updating of weights and thresholds. In the method, the long-short-time memory neural network is combined with the back propagation neural network, the long-short-time memory neural network performs preliminary flight prediction aiming at the multiple possibility of the flight data of the enemy unmanned aerial vehicle, and the output is transmitted to the back propagation neural network, so that the prediction accuracy and stability are improved, and the method is suitable for the current simulation scene.
The K nearest neighbor module can be used for processing the predicted flight data output by the fusion prediction module and calculating Euclidean distance between the predicted flight data of the enemy unmanned aerial vehicle and the current flight data of each unmanned aerial vehicle; the method comprises the steps of classifying predicted tracks according to different priorities based on probability, and outputting the predicted track of the enemy unmanned aerial vehicle with the highest priority as a final predicted track.
In the K nearest neighbor module, a feature space taking the flight coordinates as a label is established, and the predicted value of the flight coordinates of the enemy unmanned aerial vehicle and the Euclidean distance of the current flight coordinates of each unmanned aerial vehicle of the unmanned aerial vehicle group in the feature space are sequentially calculated; and selecting the predicted value of flight data of k (k is more than 0 and less than or equal to 20) enemy unmanned aerial vehicles with the shortest Euclidean distance, wherein the enemy unmanned aerial vehicle with the largest occurrence number is selected.
The K nearest neighbor module adopts a K nearest neighbor algorithm, which is also called a KNN algorithm, and is a learning algorithm commonly used in data mining and machine learning, and before the KNN algorithm is carried out, data is normalized at first, so that influence of dimension on calculation distance is avoided; calculating the distance between the data to be classified and each sample in the training set; finding k samples closest to the sample to be classified; observing classification conditions of the k samples; and taking the category with the largest occurrence number as the category of the data to be classified. The distance calculation mode in the K neighbor module is Euclidean distance calculation method, the number of adjacent points is 1, the distance weights are equal, and the distance of the adjacent points in the N-dimensional space is d (x, y):
in the method, in the process of the invention,
x i (i=1..n.) is the i-th dimensional coordinate of sample x in the N-dimensional space;
y i (i=1..n) is the i-th dimensional coordinate of sample y in the N-dimensional space.
The K nearest neighbor module is used for calculating Euclidean distance between each predicted value of the flight track of the enemy unmanned aerial vehicle and the current coordinate of the unmanned aerial vehicle, classifying the predicted track with different priorities based on the probability, and outputting predicted flight data with the highest priority.
The probability calculation formula is as follows:
wherein P is j Representing the probability corresponding to the j-th enemy unmanned aerial vehicle flight data predicted value; k represents the selected enemy droneNumber of predicted values of flight data; h represents the number of times that the j-th enemy unmanned aerial vehicle flight data predicted value appears in the selected predicted values.
The decision tree module is used for a decision interception process, and can comprise a decision tree, wherein decision classification nodes can be set in the decision tree, and corresponding decisions can be made according to the predicted value of the flight data of the enemy unmanned aerial vehicle and the current flight data of the unmanned aerial vehicle group.
The method comprises the steps that a base course of the enemy, a course of the unmanned aerial vehicle of the enemy, an effective interception distance of the unmanned aerial vehicle of the enemy and a speed of the unmanned aerial vehicle of the enemy are selected as characteristic labels, so that corresponding decisions are made according to a predicted value of flight data of the unmanned aerial vehicle of the enemy and a predicted value of Euclidean distance between the unmanned aerial vehicles of the two sides, and the method comprises an unmanned aerial vehicle group interception resource allocation scheme with highest interception probability and a next interception maneuvering scheme of the unmanned aerial vehicle of the me; taking the correlation of the base course of the friend or foe value target course as the characteristic, making a binary decision, if the correlation is the correlation, making a next decision, otherwise, keeping the current action; taking the course correlation of the unmanned aerial vehicle of the friend or foe as a characteristic, making a binary decision, if the binary decision is correlated, making a next decision, otherwise, keeping the current action; taking the effective interception distance of the unmanned aerial vehicle or the effectiveness of the interception distance of the unmanned aerial vehicle as a characteristic, making a binary decision, if the unmanned aerial vehicle is in the effective interception range of the unmanned aerial vehicle, making a next decision, otherwise, making a formation flight, and executing a combined maneuver; and when the speed of the unmanned aerial vehicle is taken as the characteristic, making a binary decision, if the speed is matched with the speed of the unmanned aerial vehicle, executing interception, and otherwise, making a decision again.
The decision classification ending conditions are: when one enemy unmanned aerial vehicle approaches to the my base, the enemy unmanned aerial vehicle is at the maximum endurance time t max The method is characterized in that the method is in at least one effective interception range of the my unmanned aerial vehicle, and the interception task is successful when the my base is always out of the attack range of the enemy unmanned aerial vehicle after the interception activity is started, and the formula is expressed as follows:
|p n (t end )-p e (t end )|≤r cap ,1≤n≤M;
|p base -p e (t)|≥r att ,0<t≤t end ;
0<t end ≤t max ;
wherein p is n Representing the position of an nth unmanned aerial vehicle which can effectively intercept an enemy unmanned aerial vehicle in the my unmanned aerial vehicle group; p is p e Representing the location of the enemy drone; p is p base Representing the location of my base; r is (r) cap Representing an effective interception range of the unmanned aerial vehicle; r is (r) att Representing an attack range of the enemy unmanned aerial vehicle; m represents the number of unmanned aerial vehicles in the unmanned aerial vehicle group; t represents any time after the start of the interception activity; t is t end Representing an interception activity ending time; t is t max Representing the maximum endurance time of the enemy unmanned aerial vehicle; p is p n (t end ) Representing the position of an nth unmanned aerial vehicle which can be effectively intercepted in the unmanned aerial vehicle group when the interception activity is finished; p is p e And (t) represents the position of the enemy unmanned aerial vehicle at any time after the interception activity starts.
The invention also provides an embodiment of an interception intention recognition strategy simulation method for the unmanned aerial vehicle attack and defense game, which is provided with a fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit.
The fight simulation data generating unit is used for generating and storing flight data of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group and my base coordinate data; and initializing the speeds and coordinates of the enemy unmanned aerial vehicle and the My unmanned aerial vehicle group, and simulating the flight conditions of the two unmanned aerial vehicles to obtain the flight data of the simulated two unmanned aerial vehicles.
The complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the step of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit.
The interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the attack and defense intention of the enemy unmanned aerial vehicle based on the flight data prediction value of the enemy unmanned aerial vehicle, and taking corresponding interception countermeasures based on the flight data of the current unmanned aerial vehicle group.
Preferably, the fight simulation data generating unit may set a flight database; the flight database unit can be used for storing performance parameters of the unmanned aerial vehicle and flight data of the unmanned aerial vehicle of both the enemy and the person. By collecting and analyzing performance data of various unmanned aerial vehicles in the prior art, the maximum speed, the maximum acceleration, the maximum flight distance and other parameters related to the maneuverability of the unmanned aerial vehicle are set for the unmanned aerial vehicle type, the label data are preset and stored in a flight database, and the flight database provides the corresponding relation between the performance parameters of the unmanned aerial vehicle and the unmanned aerial vehicle type. Analyzing the relationship between the characteristic parameters of the unmanned aerial vehicle and the threat degree of the unmanned aerial vehicle to the object, constructing a related mathematical model for characterization, including a friend or foe maneuvering angle, a friend or foe flying height, a friend or foe flying speed and a friend or foe distance, finishing mapping from the object characteristics to threat degree evaluation grades, developing a real-time query function, and providing reference for evaluation of indexes such as further interception intention recognition and the like.
Preferably, the interception intention recognition and decision unit may set: and the fusion prediction module, the K neighbor module and the decision tree module are jointly constructed by the long-short-term memory neural network and the back propagation neural network.
The fusion prediction module can be used for predicting flight data of the enemy unmanned aerial vehicle, the flight data of the enemy unmanned aerial vehicle generated through simulation can be input into a long-short-time memory neural network, and a plurality of time sequence prediction results output by the long-short-time memory neural network are jointly used as input of a back propagation neural network; outputting predicted flight data of the enemy unmanned aerial vehicle by a back propagation neural network; the flight data comprises a flight heading, a flight coordinate and a flight speed; a predicted trajectory is generated from the predicted flight data.
The K nearest neighbor module can be used for processing the predicted flight data output by the fusion prediction module, and can calculate Euclidean distance between the predicted flight data of the enemy unmanned aerial vehicle and the current flight data of each unmanned aerial vehicle; the method comprises the steps of classifying predicted tracks according to different priorities based on probability, and outputting the predicted track of the enemy unmanned aerial vehicle with the highest priority as a final predicted track.
The decision tree module is used for a decision interception process, and can comprise a decision tree, decision classification nodes can be set in the decision tree, and corresponding decisions can be made according to the predicted value of the flight data of the enemy unmanned aerial vehicle and the current flight data of the unmanned aerial vehicle group.
Preferably, a long-short-time memory neural network module and a back propagation neural network module which are sequentially connected can be arranged in the fusion prediction module; the long-short-time memory neural network module can be provided with n long-short-time memory neural network units which are connected in sequence; the current time step may be set as i, the first m time steps as i-m, where: m=n-1, i is greater than or equal to m+1; can set the flight data of the enemy unmanned aerial vehicle at the current time step as X i The enemy unmanned aerial vehicle flight data predicted by the long-short-term memory neural network module in the current time step can be set as Y i The coordinate data of the current time step my base can be set as Z i The method comprises the steps of carrying out a first treatment on the surface of the The flight data of the enemy unmanned aerial vehicle from the first 1 time step to the first m time steps can be sequentially: x is X i-1 ,X i-2 ,…,X i-(m-2) ,X i-(m-1) ,X i-m The enemy unmanned aerial vehicle flight data predicted by the long-short memory neural network module from the first 1 time step to the first m time steps can be sequentially: y is Y i-1 ,Y i-2 ,…,Y i-(m-2) ,Y i-(m-1) ,Y i-m The method comprises the steps of carrying out a first treatment on the surface of the Can be used for X i-m ,X i-(m-1) ,X i-(m-2) ,…,X i-2 ,X i-1 ,X i Sequentially and correspondingly inputting the data to the 1 st to nth long-short-time memory neural network units; the 2 nd to n th long and short time memory neural network units can output correspondingly in sequence: y is Y i-(m-1) ,Y i-(m-2) ,…,Y i-2 ,Y i-1 ,Y i The method comprises the steps of carrying out a first treatment on the surface of the The memory cell states and the memory cell hiding states of the 1 st to n-1 st long-short-time memory neural network units can be correspondingly output to the next long-short-time memory neural network unit, and Z can be calculated i ,X i-1 ,X i-2 ,…,X i-(m-2) ,Y i-1 ,Y i-2 ,…,Y i-(m-1) The predicted flight data of the enemy unmanned aerial vehicle can be output by the back propagation neural network module.
Preferably, the method for setting the decision tree may comprise: classifying the flight data of the unmanned aerial vehicle, and taking tag data with larger information gain as decision classification nodes.
Preferably, the method may comprise the steps of:
And 2, selecting a flight heading, a flight coordinate and a flight speed as characteristic labels, compiling data generated by the fight simulation data generating unit into a prediction data set, and establishing a decision tree.
And 3, inputting the current and historical flight data of each unmanned aerial vehicle of the enemy into a long-short-time memory neural network to conduct multi-step time sequence prediction, inputting a multi-step time sequence prediction result into a counter-propagation neural network, and predicting the next flight data of the unmanned aerial vehicle of the enemy.
And 4, inputting the predicted value of the flight data of the enemy unmanned aerial vehicle and the flight data of each unmanned aerial vehicle in the current My unmanned aerial vehicle group into a K-nearest neighbor module to obtain the predicted value of the flight data of the enemy unmanned aerial vehicle with the highest probability.
And 5, inputting predicted values of flight data of the enemy unmanned aerial vehicles and the flight data of each unmanned aerial vehicle in the current my unmanned aerial vehicle group into a decision tree to make a decision so as to give a decision whether the current iteration step is successfully intercepted and whether the interception is continued.
And 6, repeating the steps 3-5 until the iteration ending condition is met.
Preferably, when the complex environment simulation unit performs noise adding processing on the generated enemy unmanned aerial vehicle flight data, gaussian white noise with the same signal-to-noise ratio is added to all the data based on different assignments of different characteristic signals so as to simulate noise of the acquired signals under actual conditions; the different signal-to-noise ratios are then varied to test differences in prediction accuracy under noise interference of different strengths.
The above-mentioned fight simulation data generating unit, complex environment simulation unit, interception intention recognition and decision unit, flight database, long and short time memory neural network, back propagation neural network module, K neighbor module, decision tree and other functional units, modules and components can be functional units, modules and components in the prior art, or can be constructed by adopting functional units, modules and components in the prior art and adopting conventional technical means.
The above-described embodiments are only for illustrating the technical spirit and features of the present invention, and it is intended to enable those skilled in the art to understand the content of the present invention and to implement it accordingly, and the scope of the present invention is not limited to the embodiments, i.e. equivalent changes or modifications to the spirit of the present invention are still within the scope of the present invention.
Claims (9)
1. The interception intention recognition strategy simulation system for the unmanned aerial vehicle attack and defense game is characterized by comprising a fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit;
the fight simulation data generating unit is used for generating and storing flight data of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group and my base coordinate data; initializing the speeds and coordinates of the enemy unmanned aerial vehicle and the My unmanned aerial vehicle group, and simulating the flight conditions of the two unmanned aerial vehicles to obtain simulated flight data of the two unmanned aerial vehicles;
the complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the steps of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit;
the interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the attack and defense intention of the enemy unmanned aerial vehicle based on the flight data prediction value of the enemy unmanned aerial vehicle, and taking corresponding interception countermeasures based on the flight data of the current unmanned aerial vehicle group.
2. The intercept intent recognition strategy simulation system of unmanned aerial vehicle attack and defense gaming of claim 1 wherein the complex environment simulation unit comprises a gaussian white noise generator.
3. The intercept intent recognition strategy simulation system of unmanned aerial vehicle attack and defense gaming of claim 1 wherein the intercept intent recognition and decision unit comprises: a fusion prediction module, a K neighbor module and a decision tree module which are jointly constructed by a long-short-term memory neural network and a back propagation neural network;
the fusion prediction module is used for predicting flight data of the enemy unmanned aerial vehicle, inputting the flight data of the enemy unmanned aerial vehicle generated through simulation into the long-short-time memory neural network, and taking a plurality of time sequence prediction results output by the long-short-time memory neural network together as the input of the back propagation neural network; outputting predicted flight data of the enemy unmanned aerial vehicle by a back propagation neural network; the flight data comprises a flight heading, a flight coordinate and a flight speed; generating a predicted trajectory from the predicted flight data;
the K nearest neighbor module is used for processing the predicted flight data output by the fusion prediction module and calculating Euclidean distance between the predicted flight data of the enemy unmanned aerial vehicle and the current flight data of each unmanned aerial vehicle; classifying the predicted trajectories with different priorities based on the probability, and outputting the predicted trajectory of the enemy unmanned aerial vehicle with the highest priority as a final predicted trajectory;
the decision tree module is used for a decision interception process and comprises a decision tree, decision classification nodes are set in the decision tree, and corresponding decisions are made according to the predicted value of the flight data of the enemy unmanned aerial vehicle and the current flight data of the unmanned aerial vehicle group.
4. The interception intention recognition strategy simulation system of the unmanned aerial vehicle attack and defense game according to claim 3, wherein the fusion prediction module comprises a long-short-term memory neural network module and a back propagation neural network module which are sequentially connected; the long-time and short-time memory neural network module comprises n long-time and short-time memory neural network units which are connected in sequence; let the current time step be i, the first m time steps be i-m, wherein: m=n-1, i is greater than or equal to m+1; let the flight data of enemy unmanned aerial vehicle at the current time step be X i Let the enemy unmanned aerial vehicle flight data predicted by the long-short-term memory neural network module at the current time step be Y i Setting the current time stepThe coordinate data of the my base is Z i The method comprises the steps of carrying out a first treatment on the surface of the The flight data of the enemy unmanned aerial vehicle from the first 1 time step to the first m time steps are sequentially as follows: x is X i-1 ,X i-2 ,…,X i-(m-2) ,X i-(m-1) ,X i-m The enemy unmanned aerial vehicle flight data predicted by the long-short memory neural network module from the first 1 time step to the first m time steps are sequentially as follows: y is Y i-1 ,Y i-2 ,…,Y i-(m-2) ,Y i-(m-1) ,Y i-m The method comprises the steps of carrying out a first treatment on the surface of the The inputs of the 1 st to n th long-short-term memory neural network units correspond to: x is X i-m ,X i-(m-1) ,X i-(m-2) ,…,X i-2 ,X i-1 ,X i The outputs of the 2 nd to nth long and short time memory neural network units correspond to: y is Y i-(m-1) ,Y i-(m-2) ,…,Y i-2 ,Y i-1 ,Y i The method comprises the steps of carrying out a first treatment on the surface of the The memory cell states of the 1 st to n-1 st long-short-time memory neural network units and the memory cell hiding states are correspondingly output to the next long-short-time memory neural network unit; a back propagation neural network module, the inputs of which include: z is Z i ,X i-1 ,X i-2 ,…,X i-(m-2) ,Y i-1 ,Y i-2 ,…,Y i-(m-1) Its output includes predicted flight data for the enemy drone.
5. An interception intention recognition strategy simulation method for an unmanned aerial vehicle attack and defense game is characterized in that an fight simulation data generation unit, a complex environment simulation unit and an interception intention recognition and decision unit are arranged in the method;
the fight simulation data generating unit is used for generating and storing flight data of the enemy unmanned aerial vehicle and the my unmanned aerial vehicle group and my base coordinate data; initializing speeds and coordinates of the enemy unmanned aerial vehicle and the My unmanned aerial vehicle group, and simulating the flight conditions of the two unmanned aerial vehicles to obtain simulated flight data of the two unmanned aerial vehicles;
the complex environment simulation unit is used for simulating the natural environment and the electromagnetic interference condition applied by the enemy when the flight data of the enemy unmanned aerial vehicle are acquired; the method comprises the steps of carrying out noise adding processing on the enemy unmanned aerial vehicle flight data generated by the fight simulation data generation unit;
the interception intention recognition and decision unit is used for predicting the flight data of the enemy unmanned aerial vehicle, recognizing the attack and defense intention of the enemy unmanned aerial vehicle based on the flight data prediction value of the enemy unmanned aerial vehicle, and taking corresponding interception countermeasures based on the flight data of the current unmanned aerial vehicle group.
6. The method for simulating the interception intention recognition strategy of the unmanned aerial vehicle attack and defense game according to claim 5, wherein the interception intention recognition and decision unit is configured to: a fusion prediction module, a K neighbor module and a decision tree module which are jointly constructed by a long-short-term memory neural network and a back propagation neural network;
the fusion prediction module is used for predicting flight data of the enemy unmanned aerial vehicle, inputting the flight data of the enemy unmanned aerial vehicle generated through simulation into the long-short-time memory neural network, and taking a plurality of time sequence prediction results output by the long-short-time memory neural network together as the input of the back propagation neural network; outputting predicted flight data of the enemy unmanned aerial vehicle by a back propagation neural network; the flight data comprises a flight heading, a flight coordinate and a flight speed; generating a predicted trajectory from the predicted flight data;
the K nearest neighbor module is used for processing the predicted flight data output by the fusion prediction module and calculating Euclidean distance between the predicted flight data of the enemy unmanned aerial vehicle and the current flight data of each unmanned aerial vehicle; classifying the predicted trajectories with different priorities based on the probability, and outputting the predicted trajectory of the enemy unmanned aerial vehicle with the highest priority as a final predicted trajectory;
the decision tree module is used for a decision interception process and comprises a decision tree, decision classification nodes are set in the decision tree, and corresponding decisions are made according to the predicted value of the flight data of the enemy unmanned aerial vehicle and the current flight data of the unmanned aerial vehicle group.
7. The method for simulating interception intention recognition strategy of unmanned aerial vehicle attack and defense game according to claim 6, wherein the fusion prediction module is provided with the following components connected in sequenceA long-short-term memory neural network module and a back propagation neural network module; the long-short-time memory neural network module is provided with n long-short-time memory neural network units which are connected in sequence; let the current time step be i, the first m time steps be i-m, wherein: m=n-1, i is greater than or equal to m+1; let the flight data of enemy unmanned aerial vehicle at the current time step be X i Let the enemy unmanned aerial vehicle flight data predicted by the long-short-term memory neural network module at the current time step be Y i Setting the coordinate data of the current time step my base as Z i The method comprises the steps of carrying out a first treatment on the surface of the The flight data of the enemy unmanned aerial vehicle from the first 1 time step to the first m time steps are sequentially as follows: x is X i-1 ,X i-2 ,…,X i-(m-2) ,X i-(m-1) ,X i-m The enemy unmanned aerial vehicle flight data predicted by the long-short memory neural network module from the first 1 time step to the first m time steps are sequentially as follows: y is Y i-1 ,Y i-2 ,…,Y i-(m-2) ,Y i-(m-1) ,Y i-m The method comprises the steps of carrying out a first treatment on the surface of the X is to be i-m ,X i-(m-1) ,X i-(m-2) ,…,X i-2 ,X i-1 ,X i Sequentially and correspondingly inputting the data to the 1 st to nth long-short-time memory neural network units; the 2 nd to n th long and short time memory neural network units sequentially and correspondingly output: y is Y i-(m-1) ,Y i-(m-2) ,…,Y i-2 ,Y i-1 ,Y i The method comprises the steps of carrying out a first treatment on the surface of the The memory cell states and the memory cell hiding states of the 1 st to n-1 st long-short-time memory neural network units are correspondingly output to the next long-short-time memory neural network unit, and Z is calculated i ,X i-1 ,X i-2 ,…,X i-(m-2) ,Y i-1 ,Y i-2 ,…,Y i-(m-1) And inputting the predicted flight data to a back propagation neural network module, and outputting predicted flight data of the enemy unmanned aerial vehicle by the back propagation neural network module.
8. The method for simulating the interception intention recognition strategy of the unmanned aerial vehicle attack and defense game according to claim 6, wherein the method comprises the following steps:
step 1, generating and storing flight data of an enemy unmanned aerial vehicle and a group of my unmanned aerial vehicles and coordinate data of a my base by a fight simulation data generating unit; the complex environment simulation unit performs noise adding processing on the generated enemy unmanned aerial vehicle flight data;
step 2, selecting a flight heading, a flight coordinate and a flight speed as characteristic labels, compiling data generated by the fight simulation data generating unit into a prediction data set, and establishing a decision tree;
step 3, inputting current and historical flight data of each unmanned aerial vehicle of the enemy into a long-short-time memory neural network to conduct multi-step time sequence prediction, inputting multi-step time sequence prediction results into a counter-propagation neural network, and predicting next flight data of the unmanned aerial vehicle of the enemy;
step 4, inputting the predicted value of the flight data of the enemy unmanned aerial vehicle and the flight data of each unmanned aerial vehicle in the current My unmanned aerial vehicle group into a K nearest neighbor module to obtain the predicted value of the flight data of the enemy unmanned aerial vehicle with the highest probability;
step 5, inputting predicted values of flight data of the enemy unmanned aerial vehicles and the flight data of each unmanned aerial vehicle in the current my unmanned aerial vehicle group into a decision tree for decision making so as to give a decision whether the current iteration step is successfully intercepted and whether the interception is continued;
and 6, repeating the steps 3-5 until the iteration ending condition is met.
9. The method for simulating the interception intention recognition strategy of the unmanned aerial vehicle attack and defense game according to claim 5, wherein when the complex environment simulation unit performs noise adding processing on generated enemy unmanned aerial vehicle flight data, gaussian white noise with the same signal-to-noise ratio is added to all data based on different assignments of different characteristic signals so as to simulate noise of acquired signals under actual conditions; the different signal-to-noise ratios are then varied to test differences in prediction accuracy under noise interference of different strengths.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310143437.XA CN116432514A (en) | 2023-02-21 | 2023-02-21 | Interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense game |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310143437.XA CN116432514A (en) | 2023-02-21 | 2023-02-21 | Interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense game |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116432514A true CN116432514A (en) | 2023-07-14 |
Family
ID=87083804
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310143437.XA Pending CN116432514A (en) | 2023-02-21 | 2023-02-21 | Interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense game |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116432514A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116993010A (en) * | 2023-07-28 | 2023-11-03 | 南通大学 | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network |
-
2023
- 2023-02-21 CN CN202310143437.XA patent/CN116432514A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116993010A (en) * | 2023-07-28 | 2023-11-03 | 南通大学 | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network |
CN116993010B (en) * | 2023-07-28 | 2024-02-06 | 南通大学 | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11794898B2 (en) | Air combat maneuvering method based on parallel self-play | |
CN111753881B (en) | Concept sensitivity-based quantitative recognition defending method against attacks | |
CN110348708B (en) | Ground target dynamic threat assessment method based on extreme learning machine | |
US7324979B2 (en) | Genetically adaptive neural network classification systems and methods | |
CN113255936A (en) | Deep reinforcement learning strategy protection defense method and device based on simulation learning and attention mechanism | |
CN108320051B (en) | Mobile robot dynamic collision avoidance planning method based on GRU network model | |
CN113467508A (en) | Multi-unmanned aerial vehicle intelligent cooperative decision-making method for trapping task | |
Bai et al. | Adversarial examples construction towards white-box Q table variation in DQN pathfinding training | |
CN112036556B (en) | Target intention inversion method based on LSTM neural network | |
CN115578876A (en) | Automatic driving method, system, equipment and storage medium of vehicle | |
CN116432514A (en) | Interception intention recognition strategy simulation system and method for unmanned aerial vehicle attack and defense game | |
CN113743509A (en) | Incomplete information-based online combat intention identification method and device | |
CN116679719A (en) | Unmanned vehicle self-adaptive path planning method based on dynamic window method and near-end strategy | |
CN115480582A (en) | Maneuver prediction method for LSTM-based target, electronic device and storage medium | |
CN112305913A (en) | Multi-UUV collaborative dynamic maneuver decision method based on intuitive fuzzy game | |
CN116027198A (en) | Lithium battery health state estimation method based on combined weighted domain countermeasure network | |
CN115909027B (en) | Situation estimation method and device | |
CN116360503A (en) | Unmanned plane game countermeasure strategy generation method and system and electronic equipment | |
CN115204286A (en) | Target tactical intention online identification method based on deep learning in simulation environment | |
CN113919425B (en) | Autonomous aerial target allocation method and system | |
CN114445456B (en) | Data-driven intelligent maneuvering target tracking method and device based on partial model | |
CN114296067A (en) | Pulse Doppler radar low-slow small target identification method based on LSTM model | |
CN115015908A (en) | Radar target data association method based on graph neural network | |
CN115422995A (en) | Intrusion detection method for improving social network and neural network | |
CN115935773A (en) | Layered identification method for target tactical intentions in air combat simulation environment |
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 |