CN117540626A - Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network - Google Patents
Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network Download PDFInfo
- Publication number
- CN117540626A CN117540626A CN202311428417.3A CN202311428417A CN117540626A CN 117540626 A CN117540626 A CN 117540626A CN 202311428417 A CN202311428417 A CN 202311428417A CN 117540626 A CN117540626 A CN 117540626A
- Authority
- CN
- China
- Prior art keywords
- aerial vehicle
- unmanned aerial
- data
- enemy
- situation information
- 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.)
- Granted
Links
- 238000013531 bayesian neural network Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims description 31
- 238000013528 artificial neural network Methods 0.000 claims description 29
- 238000012360 testing method Methods 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 4
- 125000004122 cyclic group Chemical group 0.000 claims description 4
- 230000002349 favourable effect Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 12
- 230000015654 memory Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000001351 cycling effect Effects 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a situation prediction method of a fixed wing unmanned aerial vehicle based on a Bayesian neural network, belonging to the technical field of unmanned aerial vehicle situation prediction; the method solves the technical problem that the unmanned aerial vehicle on the my side cannot predict the future situation of the unmanned aerial vehicle on the enemy side under an uncertain environment. The technical proposal is as follows: establishing a Bayesian network suitable for time sequence prediction and collecting limited situation information of the enemy unmanned aerial vehicle; the situation information of the enemy unmanned aerial vehicle is used as input, and the established Bayesian neural network is used for predicting the situation of the enemy unmanned aerial vehicle at the next moment; and predicting again by taking the single-moment predicted value as input to form situation information of the future time period of the enemy unmanned aerial vehicle. The beneficial effects of the invention are as follows: the situation of the enemy unmanned aerial vehicle under a period of time can be predicted by using limited situation information in a battlefield environment, so that the enemy unmanned aerial vehicle can occupy battlefield initiative, the battlefield capability of the unmanned aerial vehicle is improved, and the battlefield loss ratio of the enemy unmanned aerial vehicle is reduced.
Description
Technical Field
The invention relates to the technical field of unmanned aerial vehicle situation prediction, in particular to a fixed wing unmanned aerial vehicle situation prediction method based on a Bayesian neural network.
Background
In recent years, a fixed-wing unmanned aerial vehicle plays a key role in military operations due to the advantages of long endurance, high-speed flight, strong loading capacity and the like, and can perform various tasks such as reconnaissance, striking, air countermeasure and the like. In the combat environment, the situation information of the opposite unmanned aerial vehicle is obtained through prediction, which is important for specifying tactical decisions and implementing striking actions, so that the situation information of the enemy unmanned aerial vehicle is predicted, and not only is a high-precision sensor such as a radar, an infrared sensor, a camera and the like required, but also a mature and reliable prediction algorithm is required to be applied to complete the prediction by using data obtained from the sensor.
In an air combat environment, in order to reduce the combat loss ratio of the my unmanned aerial vehicle, it is often necessary to rely on very reliable algorithms to help the unmanned aerial vehicle predict future situation information of the enemy unmanned aerial vehicle. Some conventional algorithms have excessive confidence in the predictions and cannot evaluate the uncertainty of the predictions, and these defects can have irrecoverable consequences in military operations. For example, in the article Task offloading scheme combining deep reinforcement learning and convolutional neural networks for vehicle trajectory prediction in smart cities, convolutional neural networks, which are neural networks dedicated to time series, are often required to predict vehicle trajectories, and convolutional neural networks, which are black box models, lack interpretation and lack reliability in military operations. The method aims to solve the problem that limited data cannot be utilized to provide reliable prediction for the enemy unmanned aerial vehicle under the uncertain environment of a battlefield.
Disclosure of Invention
The invention aims to provide a situation prediction method of a fixed wing unmanned aerial vehicle based on a Bayesian neural network, which can quickly predict situation information of an enemy unmanned aerial vehicle, gives uncertainty of prediction and is beneficial to reducing over fitting of the neural network.
The invention is characterized in that: firstly, establishing a Bayesian network suitable for time sequence prediction, acquiring the latest limited situation information of an enemy unmanned aerial vehicle through a sensor system of the enemy unmanned aerial vehicle, predicting the situation of the enemy unmanned aerial vehicle at the next moment by using the established Bayesian neural network, inputting data obtained by network prediction into the network again as input, and obtaining the situation information of the enemy unmanned aerial vehicle in a future time period through multiple cycles.
The invention gives the unmanned aerial vehicle the ability to predict the future situation information of the enemy unmanned aerial vehicle and gives the uncertainty of the prediction result.
The invention is realized by the following measures: a situation prediction method of a fixed wing unmanned aerial vehicle based on a Bayesian neural network comprises the following steps:
s1, establishing a Bayesian network suitable for time sequence prediction, and storing trained network parameters and structures, so that real-time prediction can be performed after real-time situation information of an enemy unmanned aerial vehicle is collected in a later period;
s2, acquiring the latest limited situation information of the enemy unmanned aerial vehicle through a sensor system of the unmanned aerial vehicle, and transmitting the situation information of the enemy to a Bayesian neural network after finishing, so that the neural network can predict quickly;
s3, taking situation information of the enemy unmanned aerial vehicle as input, and predicting the situation of the enemy unmanned aerial vehicle at the next moment by using an established Bayesian neural network;
s4, predicting again by taking a single-moment predicted value obtained by network prediction as input, splicing a predicted result with an original data segment to form situation data of a future time segment of the enemy unmanned aerial vehicle, and finally transmitting the predicted situation information of the enemy unmanned aerial vehicle back to the enemy unmanned aerial vehicle, so that the enemy unmanned aerial vehicle can occupy a favorable position in advance in the future time segment.
Further, the first step includes the following steps:
1-1), collecting sufficient situation information data of the enemy unmanned aerial vehicle, arranging the collected data information into a three-dimensional array to form a database, wherein the first dimension is the quantity of the collected situation information data of the enemy unmanned aerial vehicle, the second dimension is the time step of each piece of the collected situation information data, and the third dimension is the characteristic quantity of the situation information of the input neural network;
1-2) randomly selecting training data and test data for training the bayesian neural network from the data set collected in step 1-1) using a dual-loop function and a slicing operation;
1-2-1), selecting an index for the first dimension and the second dimension of the collected data using a double loop;
1-2-2), taking the selected index as the start end of slicing, slicing backwards, and selecting enough data segments with the same data length of each segment through cyclic operation; taking the selected index as the start end of slicing, slicing backwards, and selecting enough data segments with the same data length by cyclic operation;
1-2-3), arranging the data segments collected by slicing to form a three-dimensional array as a data set, wherein the first dimension represents the total number of the collected data segments, the second dimension represents the time step of each data segment, and the third dimension represents the situation information feature quantity of the enemy unmanned plane;
1-2-4), after the collected data sets are disordered, selecting 70% -80% of the first dimension of the data sets as training sets, and the rest as test sets;
1-2-5), for each data segment of the training set and the test set in the data set, cutting the first half of the second dimension as data values using a slicing operation, the remainder as labels, in which case four sub-data sets are obtained in total, respectively the number of inputs for trainingThe data set, the label data set corresponding to the training set, the input data set for testing, and the label data set corresponding to the testing set. The third dimension of the four sub-data sets contains 6 elements, which are situation characteristic information of unmanned aerial vehicles per second and are recorded asWherein: 1-3), the third dimension features_num of the data set obtained in step 1-2-3) being 6, denoted +.>Wherein:
（1）
wherein n is x Is unmanned aerial vehicle tangential overload, n z The normal overload of the unmanned aerial vehicle is achieved, phi is the roll angle of the unmanned aerial vehicle around the speed vector, and g is the gravity acceleration. The output characteristic output_size is 6×1, which indicates that the predicted data is the 6 kinematic characteristics of the enemy unmanned aerial vehicle in the 6 th second;
1-4), establishing a neural network layer with Bayesian characteristics, wherein the neural network layer defines that the weight and bias of each node are random variables obtained by sampling in normal distribution, the mean value and variance of the weight in the neural network layer are two-dimensional matrixes, the first dimension is the input characteristic quantity, the second dimension is the output characteristic quantity, and the weight and variance of the bias are one-dimensional matrixes with the shape of the output characteristic quantity;
1-5), constructing a Bayesian neural network according to the defined neural network layer with Bayesian characteristics, wherein the Bayesian neural network uses the evidence lower bound in variation inference as a loss function, and provides that the log prior distribution, the log posterior distribution and the log likelihood in the neural network layer with Bayesian characteristics are calculated in a forward propagation method, and the calculation mode of the network loss function is defined as the log posterior distribution minus the log prior distribution minus the log likelihood;
1-6), training on the training set by using the network, and storing the trained network model structure and parameters.
In the second step, situation information data of the enemy unmanned aerial vehicle is used as input, and the situation information of the enemy unmanned aerial vehicle at the next moment is predicted through a network.
Further, the second step comprises the following steps: a step of
2-1) arranging the situation information of the enemy unmanned aerial vehicle collected by the unmanned aerial vehicle in real time into an array conforming to an input network, and recording the array as S b ；
2-2) S is to b And inputting the information into the trained Bayesian neural network, and outputting the information by the network to obtain situation information of the enemy unmanned aerial vehicle at the next moment.
Further, in the third step, the predicted value of the single moment is used as input to predict again, so that the predicted value of the single moment of the situation information of the future time period of the enemy unmanned aerial vehicle is used as input to predict again, and the situation information of the future time period of the enemy unmanned aerial vehicle is formed.
Further, the third step comprises the following steps:
3-1) arranging situation information of the enemy unmanned aerial vehicle at a single moment into a two-dimensional array S b One-dimensional array S with same second-dimensional characteristic elements h ；
3-2) one-dimensional array S T0 Splicing to a two-dimensional array S in a first-dimensional order b Two-dimensional array S formed by the extreme end X Two-dimensional array S cut according to a first dimension X Form a new two-dimensional array S T1 ；
3-3) the two-dimensional array S obtained by cyclic operation T1 Inputting a Bayesian neural network to obtain situation information of the enemy unmanned aerial vehicle at the next moment, and circularly performing the operation in the step 2) to obtain a two-dimensional array S T2 ；
3-4) cycling through 3-1) -3-3) above to obtain situation information data S of future time period of enemy unmanned aerial vehicle T The situation data are all composed of situation data of the enemy unmanned aerial vehicle at a single moment predicted by a Bayesian neural network;
3-5) transmitting situation information data of future time periods of enemy to the unmanned aerial vehicle, and utilizing the situation information, the unmanned aerial vehicle seizes the favorable situation in the air combat.
Compared with the prior art, the invention has the beneficial effects that:
(1) According to the invention, uncertainty modeling is introduced, the Bayesian neural network can carry out uncertainty additivity on a predicted result, which is not possessed by the traditional neural network, in unmanned plane situation prediction, factors such as environmental change, sensor noise and the like cause uncertainty, and the Bayesian neural network can provide a more accurate predicted result and better quantify the reliability of prediction.
(2) The method is excellent in limited data, and in unmanned aerial vehicle situation prediction, insufficient data is a common challenge. While conventional neural networks are prone to overfitting in this case, bayesian neural networks reduce the risk of overfitting by introducing a priori probabilities, thus rendering them more robust on small datasets.
(3) The Bayesian neural network provides high interpretability, and is particularly suitable for interpreting situation prediction of the enemy unmanned aerial vehicle. Unlike traditional neural networks, bayesian neural networks more easily interpret their prediction logic, helping decision makers understand the underlying principles of model predictions.
(4) The Bayesian neural network in the invention can show stronger generalization capability, can more effectively process uncertainty, reduce the risk of over fitting, can better adapt to unseen data, can process environment and data instability, and can play an excellent role in complex tasks.
(5) The Bayesian neural network can still provide effective prediction when only limited data of the enemy unmanned aerial vehicle is collected, and can provide relatively reliable situation information of the enemy unmanned aerial vehicle in a future time period.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
Fig. 1 is a flowchart of a situation method of a fixed-wing unmanned aerial vehicle in an uncertain environment based on a bayesian neural network.
Fig. 2 is a flow chart of post-processing of collected situation data of an enemy fixed wing unmanned aerial vehicle.
Fig. 3 is a network training flow chart of a situation prediction method of a fixed wing unmanned aerial vehicle in an uncertain environment based on a bayesian neural network.
Fig. 4 is a histogram of MSE comparison between a predicted trajectory of each step and an actual trajectory in five-step prediction of a method for predicting a situation of a fixed-wing unmanned aerial vehicle in an uncertain environment based on a bayesian neural network; among them, fig. 4 (a) is a first step, fig. 4 (b) is a second step, fig. 4 (c) is a third step, fig. 4 (d) is a fourth step, and fig. 4 (e) is a fifth step.
Fig. 5 is a flowchart of a situation prediction method of a fixed wing unmanned aerial vehicle under an uncertain environment based on a bayesian neural network, when predicting again by using data obtained by network prediction.
Fig. 6 is a graph comparing a predicted track and an actual track obtained by the situation prediction method of the fixed wing unmanned aerial vehicle under the uncertain environment based on the Bayesian neural network.
Fig. 7 is a graph comparing a predicted trajectory and an actual trajectory obtained after replacing a bayesian neural network with a long-short-term memory neural network by the situation prediction method of a fixed-wing unmanned aerial vehicle under an uncertain environment based on the bayesian neural network.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. Of course, the specific embodiments described herein are for purposes of illustration only and are not intended to limit the invention.
Example 1
The embodiment provides a situation prediction method of a fixed wing unmanned aerial vehicle based on a Bayesian neural network, which comprises the following steps:
step 1), establishing a Bayesian network suitable for time sequence prediction, and storing trained network parameters and structures;
step 2) acquiring the latest situation information of the enemy unmanned aerial vehicle through a sensor system of the unmanned aerial vehicle, and transmitting the situation information of the enemy to a Bayesian neural network after finishing;
and 3) predicting the situation of the enemy unmanned aerial vehicle at the next moment by using the established Bayesian neural network, and finally transmitting the predicted situation information of the enemy unmanned aerial vehicle back to the unmanned aerial vehicle.
Step 1), establishing a Bayesian network suitable for time sequence prediction, and storing trained network parameters and structures, wherein the specific steps are as follows:
1-1), collecting situation information data of the enemy unmanned aerial vehicle which are sufficient and are not repeated independently of each other for the processing flow after the collected enemy situation data as shown in fig. 2, wherein the data shape is (tracks_num, time_long, features_num), wherein the tracks_num represents the collected situation information quantity of the enemy unmanned aerial vehicle, the time_long represents the time step of each piece of collected situation information data of the enemy unmanned aerial vehicle, and the features_num represents the situation information feature quantity of the enemy unmanned aerial vehicle input into the network.
1-2) randomly selecting training data and test data for training the bayesian neural network from the data set collected in step 1-1) using a dual-loop function and a slicing operation;
1-2-1), selecting indexes idex for the first dimension tracks_num and the second dimension time_long of the data collected in the step 1-1) by using a double-loop function, wherein the total amount of the indexes is selected as batch_size epsilon (1000, tracks_num multiplied by time_long);
1-2-2), taking the index idex selected in the step 1-2-1) as the start end of a slice, selecting 6 continuous time steps backwards as one piece of data, selecting batch_size pieces of data altogether, wherein each piece of data is 6 in length, and each piece of data is marked as alpha;
1-2-3), the data selected in steps 1-2-1) and 1-2-2) are three-dimensional arrays in the shape of (batch_size, time_step, features_num), wherein batch_size=3000 is the total batch for training and testing, time_step=6 is the time step of single-segment unmanned plane situation information, and features_num represents the feature quantity of situation information of enemy unmanned plane;
1-2-4), selecting 80% of the batch_size as training set three_data after the collected data sets are out of order, wherein the shape is (2400, 6, features_num), and the remaining 20% is as test set test_data, and the shape is (800, 6, features_num);
1-2-5), for each data segment α in the training set train_data and the test set test_data, the first 5 seconds are sliced out as data values, the 6 th second is taken as a label, and 4 data sets are obtained at this time respectively:
input data input_train for training, shape (2400, 5, features_num);
the label data set output_train corresponding to the training data set is in the shape of (2400, 1, features_num);
input data input_test for test, shape (800, 5, features_num);
the label data set output_test corresponding to the test data set is (800, 1, features_num), and the collection and division of the data set is shown in fig. 2.
1-3), the third dimension of the data set obtained in step 1-2-3) having a number of features_num of 6, noted asWherein:
（1）
n x is unmanned aerial vehicle tangential overload, n z The normal overload of the unmanned aerial vehicle is achieved, phi is the roll angle of the unmanned aerial vehicle around the speed vector, and g is the gravity acceleration. The output characteristic output_size is 6×1, which indicates that the predicted data is the 6 kinematic characteristics of the enemy unmanned aerial vehicle in the 6 th second;
1-4), establishing a neural network layer linear_bbb with bayesian characteristics, wherein the neural network layer defines that the weight w and the bias b of each node are random variables obtained by sampling in a normal distribution with a mean value of 0 and a variance of 1, the mean value w_mu of the weights and the variance w_rho of the weights in the neural network layer are matrixes with the shape of (input_size) and the bias weight b_mu and the bias variance b_rho are matrixes with the shape of (output_size);
1-5), constructing a network mlp_bbb, creating a linear_bbb instance defined above in the network, the bayesian neural network using a lower bound of evidence ELBO (Evidence Lower Bound) in variance inference as a loss function, and defining a calculation formula for calculating a log prior distribution log_priority, a log posterior distribution log_post and a log likelihood log_like in a linear_bbb layer in a forward propagation method, and defining a network loss function sample_elambo as follows:
loss＝log_post-log_prior-log_like (2)
in the loss function of the network, log_post-log_priority is the complexity cost, and log_like is the error cost;
1-6), training the constructed Bayesian neural network by using the collected enemy situation information data, setting a training batch epoch_num before training, finishing the network training according to the flow in the second graph, and saving the network parameters and the structure after finishing the training, wherein the change curve of the loss function is shown in figure 4, and the change curve of the complexity cost and the error cost of the loss function is shown in figure 5 when the network and the test network are trained.
Step 2), sensing situation information data S of the enemy unmanned aerial vehicle in real time through a sensor system of the unmanned aerial vehicle b The data should be situation information of the enemy unmanned aerial vehicle for the latest 5 seconds, namely S b The shape of the model is (5, 6), wherein the first dimension is 5, which represents situation information of the last five seconds of the enemy unmanned aerial vehicle, the second dimension is 6, which represents 6 kinematic features of the enemy unmanned aerial vehicle per second, and the perceived data are transmitted to the Bayesian neural network trained in the step 1-6).
Step 3),
3-1) arranging situation information of the enemy unmanned aerial vehicle at a single moment into a three-dimensional array S b Two-dimensional array S with same shape of rear two-dimensional h ；
3-2) two-dimensional array S T0 Splicing to a three-dimensional array S according to a first dimension b Three-dimensional array S formed by the extreme end X Three-dimensional array S cut according to a first dimension X Form a new two-dimensional array S at the initial end of the systemThree-dimensional array S of (2) T1 ；
3-3) the three-dimensional array S obtained by circulation T1 Inputting a Bayesian neural network to obtain situation information of the enemy unmanned aerial vehicle at the next moment, and circularly performing the operation in the step 2) to obtain a three-dimensional array S T2 ；
3-4) cycling through 3-1) -3-3) above to obtain situation information data S of future time period of enemy unmanned aerial vehicle T When the situation data are all composed of situation data of the enemy unmanned aerial vehicle at a single moment predicted by a Bayesian neural network, a program circulation flow chart is shown in fig. 5;
3-5) transmitting situation information data of future time periods of enemy to the unmanned aerial vehicle, and utilizing the situation information, the unmanned aerial vehicle seizes the favorable situation in the air combat.
Example 2
Referring to fig. 7, the technical scheme provided in this embodiment is that a fixed wing situation prediction method based on a long-short-term memory neural network includes the following specific steps:
step 1), a long-short-time memory neural network suitable for time sequence prediction is established, and network parameters and structures are saved.
Step 2) acquiring the latest situation information of the enemy unmanned aerial vehicle through a sensor system of the unmanned aerial vehicle, and transmitting the situation information of the enemy to a Bayesian neural network after finishing;
and 3) predicting the situation of the enemy unmanned aerial vehicle at the next moment by using the established Bayesian neural network, and finally transmitting the predicted situation information of the enemy unmanned aerial vehicle back to the unmanned aerial vehicle.
Step 1), establishing a long-short time memory neural network suitable for a time sequence prediction long-short time memory neural network, and storing trained network parameters and structures, wherein the specific steps are as follows:
steps 1-1) to 1-3) are the same as those of example 1 described above.
1-4) establishing a 3-layer long and short-term memory neural network, using a mean square error MSE as a loss function, prescribing that the calculation of the loss function is carried out in each training, and saving the parameters and the structure of the long and short-term memory neural network after the training is completed.
Step 2) and step 3) are similar to those of the above-described embodiment 1, and a pair of a trajectory chart and an actual trajectory chart of the unmanned aerial vehicle predicted by using the long-short-term memory neural network is shown in fig. 7.
Comparing fig. 6 with fig. 7, the bayesian neural network used in fig. 6 obtains a more accurate prediction result, which is closer to the actual enemy unmanned aerial vehicle track. The bayesian neural network proposed in this embodiment with less training data is also shown to have a stronger generalization ability.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.
Claims (6)
1. A situation prediction method of a fixed wing unmanned aerial vehicle based on a Bayesian neural network is characterized by comprising the following steps:
1) Establishing a database according to the situation information of the enemy fixed wing unmanned aerial vehicle, and constructing a Bayesian neural network for time sequence prediction;
2) The situation information of the enemy unmanned aerial vehicle is taken as input, and the situation information of the enemy unmanned aerial vehicle at the next moment is predicted through a network;
3) And predicting again by taking the single-moment predicted value obtained by network output as input to form situation information of future time periods of the enemy unmanned aerial vehicle, and transmitting back to the unmanned aerial vehicle.
2. The method of claim 1, wherein creating a database based on enemy fixed wing unmanned aerial vehicle situation information comprises the steps of:
1) Collecting sufficient situation information data of the enemy unmanned aerial vehicle, arranging the collected data information into a three-dimensional array to form a database, wherein the first dimension is the quantity of the collected situation information data of the enemy unmanned aerial vehicle, the second dimension is the time step of each piece of the collected situation information data, and the third dimension is the characteristic quantity of the situation information of the input neural network;
2) The collected data is randomly selected and partitioned into training and testing sets using a double loop and slice operation.
3. The prediction method according to claim 2, wherein the randomly selecting and dividing the collected data into the training set and the test set using the double loop and slice operation comprises the steps of: 1) Selecting an index for the first dimension and the second dimension of the collected data using a double loop; 2) Taking the selected index as the start end of slicing, slicing backwards, and selecting enough data segments with the same data length by cyclic operation; 3) The data segments collected through slicing are arranged to form a three-dimensional array as a data set, wherein the first dimension represents the total number of the collected data segments, the second dimension represents the time step of each segment of data segment, and the third dimension represents the situation information feature quantity of the enemy unmanned aerial vehicle; 4) After the collected data sets are disordered, 75 to 90 percent of the first dimension of the data sets are selected as training sets, and the rest of the data sets are selected as test sets; 5) For each data segment of the training set and the test set in the data set, slicing the first half of the second dimension into data values and the rest into labels by using slicing operation, and obtaining four sub-data sets, namely an input data set for training, a label data set corresponding to the training set, an input data set for testing and a label data set corresponding to the test set, wherein the third dimension of the four sub-data sets contains 6 elements, is situation characteristic information of the unmanned aerial vehicle per second and is recorded asWherein (1)>The position increment and the +.A unit time of the unmanned plane in the x, y and z directions are respectively as follows>For the speed increment of unmanned plane per unit time, +.>The yaw angle increment and the pitch angle increment of the unmanned aerial vehicle in unit time are respectively adopted.
4. The prediction method according to claim 2, wherein the construction of the bayesian neural network for the time series comprises the steps of:
1) Establishing a neural network layer with Bayesian characteristics, wherein the neural network layer defines that the weight and bias of each node are random variables obtained by sampling in normal distribution, the mean value and variance of the weight in the neural network layer are two-dimensional matrixes, the first dimension is the number of input features, the second dimension is the number of output features, and the weight and variance of the bias are one-dimensional matrixes with the shape of the number of output features;
2) Constructing a Bayesian neural network according to the defined neural network layer with Bayesian characteristics, wherein the Bayesian neural network uses the evidence lower bound in variation inference as a loss function, and provides that the log prior distribution, the log posterior distribution and the log likelihood in the neural network layer with Bayesian characteristics are calculated in a forward propagation method, and the calculation mode of the network loss function is defined as the log posterior distribution minus the log prior distribution minus the log likelihood;
3) Training on the training set by using the network, and storing the trained network model structure and parameters.
5. The prediction method according to claim 1, wherein the situation information of the enemy unmanned aerial vehicle is input, and the situation information of the enemy at the next moment is predicted through a network, and the method comprises the following steps:
1) The enemy unmanned aerial vehicle situation information collected by the unmanned aerial vehicle in real time is arranged into an array conforming to an input network and recorded as S b ；
2) Will S b And inputting the information into the trained Bayesian neural network, and outputting the information by the network to obtain situation information of the enemy unmanned aerial vehicle at the next moment.
6. The prediction method according to claim 1, wherein the predicting again using the predicted value of the single moment as an input constitutes situation information of a future period of time of the enemy unmanned aerial vehicle, and comprises the steps of:
1) The situation information of the enemy unmanned aerial vehicle at a single moment is arranged into a two-dimensional array S b One-dimensional array S with same shape in rear two dimensions h ；
2) Will be one-dimensional array S T0 Splicing to a two-dimensional array S in order of a first dimension b Two-dimensional array S formed by the extreme end X Two-dimensional array S cut out according to the order of the first dimension X Form a new two-dimensional array S T1 ；
3) The two-dimensional array S is obtained by circulation T1 Inputting a Bayesian neural network to obtain situation information of the enemy unmanned aerial vehicle at the next moment, and circularly performing the operation in the step 2) to obtain a two-dimensional array S T2 ；
4) Circulating the steps 1) to 3) to obtain situation information data S of future time periods of the enemy unmanned aerial vehicle T The situation data are all composed of situation data of the enemy unmanned aerial vehicle at a single moment predicted by a Bayesian neural network;
5) And transmitting situation information data of the enemy future period to the unmanned aerial vehicle, and utilizing the situation information, the unmanned aerial vehicle seizes the favorable situation in the air combat.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311428417.3A CN117540626B (en) | 2023-10-30 | 2023-10-30 | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311428417.3A CN117540626B (en) | 2023-10-30 | 2023-10-30 | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117540626A true CN117540626A (en) | 2024-02-09 |
| CN117540626B CN117540626B (en) | 2024-05-14 |
Family
ID=89794957
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311428417.3A Active CN117540626B (en) | 2023-10-30 | 2023-10-30 | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117540626B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020046213A1 (en) * | 2018-08-31 | 2020-03-05 | Agency For Science, Technology And Research | A method and apparatus for training a neural network to identify cracks |
| CN111240350A (en) * | 2020-02-13 | 2020-06-05 | 西安爱生无人机技术有限公司 | Unmanned aerial vehicle pilot dynamic behavior evaluation system |
| US20200326718A1 (en) * | 2019-04-09 | 2020-10-15 | Robert Bosch Gmbh | Control and monitoring of physical system based on trained bayesian neural network |
| US20200410342A1 (en) * | 2019-06-28 | 2020-12-31 | Robert Bosch Gmbh | Method for training an artificial neural network, artificial neural network, use of an artificial neural network, and corresponding computer program, machine-readable memory medium, and corresponding apparatus |
| CN114510078A (en) * | 2022-02-16 | 2022-05-17 | 南通大学 | Unmanned aerial vehicle maneuver evasion decision-making method based on deep reinforcement learning |
| US20220335297A1 (en) * | 2019-09-17 | 2022-10-20 | Center For Excellence In Molecular Cell Science, Chinese Academy Of Sciences | Anticipatory Learning Method and System Oriented Towards Short-Term Time Series Prediction |
| CN115993835A (en) * | 2022-12-27 | 2023-04-21 | 西北工业大学 | Target maneuver intention prediction-based short-distance air combat maneuver decision method and system |
| CN116069056A (en) * | 2022-12-15 | 2023-05-05 | 南通大学 | Unmanned plane battlefield target tracking control method based on deep reinforcement learning |
| CN116187169A (en) * | 2022-12-30 | 2023-05-30 | 中国人民解放军国防科技大学 | Unmanned aerial vehicle cluster intention inference algorithm and system based on dynamic Bayesian network |
| CN116700079A (en) * | 2023-06-04 | 2023-09-05 | 西北工业大学 | Unmanned aerial vehicle countermeasure occupation maneuver control method based on AC-NFSP |
-
2023
- 2023-10-30 CN CN202311428417.3A patent/CN117540626B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020046213A1 (en) * | 2018-08-31 | 2020-03-05 | Agency For Science, Technology And Research | A method and apparatus for training a neural network to identify cracks |
| US20200326718A1 (en) * | 2019-04-09 | 2020-10-15 | Robert Bosch Gmbh | Control and monitoring of physical system based on trained bayesian neural network |
| US20200410342A1 (en) * | 2019-06-28 | 2020-12-31 | Robert Bosch Gmbh | Method for training an artificial neural network, artificial neural network, use of an artificial neural network, and corresponding computer program, machine-readable memory medium, and corresponding apparatus |
| US20220335297A1 (en) * | 2019-09-17 | 2022-10-20 | Center For Excellence In Molecular Cell Science, Chinese Academy Of Sciences | Anticipatory Learning Method and System Oriented Towards Short-Term Time Series Prediction |
| CN111240350A (en) * | 2020-02-13 | 2020-06-05 | 西安爱生无人机技术有限公司 | Unmanned aerial vehicle pilot dynamic behavior evaluation system |
| CN114510078A (en) * | 2022-02-16 | 2022-05-17 | 南通大学 | Unmanned aerial vehicle maneuver evasion decision-making method based on deep reinforcement learning |
| CN116069056A (en) * | 2022-12-15 | 2023-05-05 | 南通大学 | Unmanned plane battlefield target tracking control method based on deep reinforcement learning |
| CN115993835A (en) * | 2022-12-27 | 2023-04-21 | 西北工业大学 | Target maneuver intention prediction-based short-distance air combat maneuver decision method and system |
| CN116187169A (en) * | 2022-12-30 | 2023-05-30 | 中国人民解放军国防科技大学 | Unmanned aerial vehicle cluster intention inference algorithm and system based on dynamic Bayesian network |
| CN116700079A (en) * | 2023-06-04 | 2023-09-05 | 西北工业大学 | Unmanned aerial vehicle countermeasure occupation maneuver control method based on AC-NFSP |
Non-Patent Citations (2)
| Title |
|---|
| YUN CHENG 等: "Active Disturbance Rejection Generalized Predictive Control of a Quadrotor UAV via Quantitative Feedback Theory", 《DIGITAL OBJECT IDENTIFIER》, 13 April 2022 (2022-04-13), pages 37912 - 37923 * |
| YUSHENG DU 等: "Fault Tolerant Control of a Quadrotor Unmanned Aerial Vehicle Based on Active Disturbance Rejection Control and Two-Stage Kalman Filter", 《DIGITAL OBJECT IDENTIFIER》, 10 July 2023 (2023-07-10), pages 67556 - 67566 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117540626B (en) | 2024-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111104595B (en) | Deep reinforcement learning interactive recommendation method and system based on text information | |
| CN112596515B (en) | Multi-logistics robot movement control method and device | |
| EP3940604A1 (en) | Federated teacher-student machine learning | |
| Zhang et al. | Velc: A new variational autoencoder based model for time series anomaly detection | |
| CN108596327B (en) | Seismic velocity spectrum artificial intelligence picking method based on deep learning | |
| CN112115998B (en) | Method for overcoming catastrophic forgetting based on anti-incremental clustering dynamic routing network | |
| US20220164660A1 (en) | Method for determining a sensor configuration | |
| CN111047078B (en) | Traffic characteristic prediction method, system and storage medium | |
| CN114004383A (en) | Training method of time series prediction model, time series prediction method and device | |
| CN109190787B (en) | Dual particle swarm multi-monitoring point access path planning method for underwater vehicle | |
| CN114819316A (en) | Complex optimization method for multi-agent task planning | |
| CN112527547B (en) | Mechanical intelligent fault prediction method based on automatic convolution neural network | |
| CN117540626B (en) | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network | |
| CN112651499A (en) | Structural model pruning method based on ant colony optimization algorithm and interlayer information | |
| CN117633448A (en) | Cloud-edge collaboration-based equipment cluster health monitoring method | |
| CN116777068A (en) | Causal transducer-based networked data prediction method | |
| CN116993010B (en) | Fixed wing unmanned aerial vehicle situation prediction method based on Bayesian neural network | |
| CN115907120A (en) | Attention mechanism-based VMD-CNN-LSTM short-term wind power prediction method | |
| US20220284261A1 (en) | Training-support-based machine learning classification and regression augmentation | |
| Schitz et al. | Path optimization for autonomous driving using deep learning | |
| CN115204463A (en) | Residual service life uncertainty prediction method based on multi-attention machine mechanism | |
| CN110532660B (en) | Performance prediction method of aircraft brake system and aircraft brake system | |
| CN116802646A (en) | Data processing method and device | |
| CN116198771B (en) | Two-dimensional rectangular strip packaging method based on deep reinforcement learning | |
| US20230034973A1 (en) | Methods and Systems for Predicting Trajectory Data of an Object |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |