CN113534837A - Unmanned aerial vehicle flight state fusion control system and control method based on generation of countermeasure network - Google Patents
Unmanned aerial vehicle flight state fusion control system and control method based on generation of countermeasure network Download PDFInfo
- Publication number
- CN113534837A CN113534837A CN202110763783.9A CN202110763783A CN113534837A CN 113534837 A CN113534837 A CN 113534837A CN 202110763783 A CN202110763783 A CN 202110763783A CN 113534837 A CN113534837 A CN 113534837A
- Authority
- CN
- China
- Prior art keywords
- loss function
- unmanned aerial
- aerial vehicle
- input signal
- vgg
- 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
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000004927 fusion Effects 0.000 title claims abstract description 23
- RZVHIXYEVGDQDX-UHFFFAOYSA-N 9,10-anthraquinone Chemical compound C1=CC=C2C(=O)C3=CC=CC=C3C(=O)C2=C1 RZVHIXYEVGDQDX-UHFFFAOYSA-N 0.000 claims abstract description 36
- 238000005070 sampling Methods 0.000 claims description 7
- 230000003042 antagnostic effect Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000001174 ascending effect Effects 0.000 claims description 3
- 230000006872 improvement Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012384 transportation and delivery Methods 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000005405 multipole Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/10—Simultaneous control of position or course in three dimensions
- G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft
- G05D1/106—Change initiated in response to external conditions, e.g. avoidance of elevated terrain or of no-fly zones
-
- 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 unmanned aerial vehicle flight state fusion control system and method based on a generation countermeasure network. The unmanned aerial vehicle comprises a plurality of rotors, a fixed wing, a power system and a flight control system, wherein the flight control system controls the output power of the power system so as to fuse and adjust the rotating speed of the rotors and the attack angle of the fixed wing, and an original flight state parameter output module reads rated working parameters of the rotors and the fixed wing in real time to generate a first input signal image; the actual flight state parameter acquisition module acquires actual working parameters of the multiple rotors and the fixed wings in real time to generate a second input signal image; the generator network module generates an enhanced picture according to the second input signal image, and the discriminator network module outputs a discrimination result according to the first input signal image and the enhanced picture. The unmanned aerial vehicle control system provided by the invention realizes the deep fusion of the multiple rotors and the fixed wings.
Description
[ technical field ] A method for producing a semiconductor device
The invention relates to the technical field of unmanned aerial vehicles, in particular to a control system and a control method for fusing flight states of an unmanned aerial vehicle with multiple rotors and fixed wings on the basis of an antagonistic network model.
[ background of the invention ]
In recent years, with the rapid development of the unmanned aerial vehicle industry, the demand of the unmanned aerial vehicle in the field of short-distance transportation delivery is more urgent. Especially, the unmanned aerial vehicle is operated at a small load capacity of ten kilograms to fifty kilograms in a distance range of five kilometers to fifty kilometers and in a severe environment, so that the unmanned aerial vehicle is more widely applied to the production and living fields.
According to the differentiation to the flight in-process, that provides the mode difference of power, unmanned aerial vehicle is divided into many rotor unmanned aerial vehicle and fixed wing unmanned aerial vehicle. Wherein, many rotor unmanned aerial vehicle is because small and exquisite portable, and it is convenient to take off and land, expandes rapidly, low cost, and environmental suitability is good, becomes the first-selected type of short distance operation and switching. However, the existing multi-rotor unmanned aerial vehicle has two technical short boards: short endurance time and small load capacity.
In order to solve the technical problems:
the prior art combines stationary vane and many rotors, forms "hang down stationary vane unmanned aerial vehicle" system from this, and this technique uses stationary vane unmanned aerial vehicle as leading, mainly utilizes many rotors to solve the problem in stationary vane unmanned aerial vehicle take-off and land place, has also kept the performance that the stationary vane load capacity is strong, the voyage is far away simultaneously. Wherein the working states of the fixed wing and the multiple rotors are as follows:
in the stage of taking off and landing, the fixed wing does not work, by many rotors work, and control unmanned aerial vehicle's taking off and landing.
And in the cruising stage, the multiple rotors stop working, and the fixed wings work to control the navigation of the unmanned aerial vehicle. However, the 'suspended fixed wing unmanned aerial vehicle' cannot realize hovering and accurate launching and is not suitable for transportation and delivery in the existing market. Its leading cause is at whole flight in-process, this kind of "fixed wing unmanned aerial vehicle hangs down" only implements simple fixed wing and many rotor mode switching at the stage of taking off and landing, and the two independent work separately does not carry out the degree of depth with fixed wing and many rotors and fuses, leads to in the course of sailing, when needing accurate location and input, many rotors are in the out-of-service condition, can't realize accurate location and input.
On the other hand, when switching to fixed wing operating condition from many rotors operating condition, perhaps when unmanned aerial vehicle's flight state changes, in view of unmanned aerial vehicle self motion state's inertial property, can have the violent change of flight attitude or the serious problem of power loss in the moment that leads to switching.
[ summary of the invention ]
In order to solve among the prior art many rotors and stationary vane fuse not fully, lead to many rotors and the unmanned aerial vehicle that the stationary vane combines to can't realize long-time accurate hovering and put in, can not steadily smoothly switch, switch the serious technical problem of in-process power loss between different flight states.
In view of the above, the invention provides an unmanned aerial vehicle flight state fusion control system and a control method based on a generation countermeasure network.
An unmanned aerial vehicle flight state fusion control system based on generation of an countermeasure network is disclosed, wherein the unmanned aerial vehicle comprises ascending, cruising, fixed-point hovering and landing states, the unmanned aerial vehicle comprises a plurality of rotors, a fixed wing, a power system and a flight control system, the flight control system controls the output power of the power system to further fuse and adjust the rotating speed of the rotors and the attack angle of the fixed wing, the flight control system comprises an original flight state parameter output module, an actual flight state parameter acquisition module and a generation countermeasure network model, and the original flight state parameter output module reads rated working parameters of the rotors and the fixed wing in real time to generate a first input signal image; the actual flight state parameter acquisition module acquires actual working parameters of the multiple rotors and the fixed wings in real time to generate a second input signal image; the generation countermeasure network model comprises a generator network module, a discriminator network module and a loss function, the generator network module generates an enhanced picture according to the second input signal image, and the discriminator network module outputs a discrimination result according to the first input signal image and the enhanced picture.
As a further improvement of the present invention, the generator network module employs an Attention net, wherein the Attention net network includes a down-sampling, an up-sampling and an Attention gate, the down-sampling is performed on the second input signal image, a deep feature of the second input signal image is learned, the deep feature is deconvoluted and up-sampled, and finally the enhanced image is output.
As a further improvement of the present invention, the discriminator network module uses a patch gan, the enhanced image and the first input signal image output from the generator network module are input into the discriminator network module to obtain an n × n matrix, and finally, a discrimination result is output, and the discrimination result takes a mean value of the matrix as a true/false result.
As a further improvement of the invention, the loss functions include a penalty loss function, a mean square error loss function, and a VGG loss function.
As a further improvement of the invention, the penalty function is:
wherein D represents the discriminator network module, G represents the generator network module, x represents the second input signal image, G (x) represents the enhanced image, y represents the first input signal image, k is the gain coefficient, and v is the aircraft airspeed.
As a further improvement of the present invention, the mean square error loss function is:
where W represents the width of the input image and H represents the height of the input image.
As a further improvement of the present invention, the VGG loss function is:
wherein phiiThe characteristic output of the i-th layer convolution of the VGG model is obtained, and the definition of a generated image can be improved by the VGG loss function.
As a further improvement of the invention, the total loss function is:
Loss＝LG+λMSELMSE+λVGGLVGG
wherein L isGIs a function of the penalty of fighting, LMSEIs a function of the mean square error loss, LVGGIs a VGG loss function, λMSEIs a penalty factor, λ, of a mean square error loss functionVGGIs the penalty factor for the VGG loss function. Setting a loss function, and then using image data to generate a countermeasure network model for continuous alternate iterative training to optimize model parameters of the generated countermeasure network model.
An unmanned aerial vehicle flight state fusion control method based on an countermeasure network, wherein the unmanned aerial vehicle comprises a plurality of rotors, a fixed wing, a power system and a flight control system, the flight control system comprises an original flight state parameter output module, an actual flight state parameter acquisition module and a generated countermeasure network model, the generated countermeasure network model comprises a generator network module, a discriminator network module and a loss function, and the method comprises the following steps:
step S01, providing an original flight state parameter output module, reading rated working parameters of the multi-rotor wing and the fixed wing, and generating a first input signal image;
step S02, providing an actual flight state parameter acquisition module, acquiring actual working parameters of the multiple rotors and the fixed wings in real time, and generating a second input signal image;
step S03, inputting the second input signal image to the generator network module to obtain an enhanced image;
step S04, inputting the enhanced image and the first input signal image into the discriminator network module, and outputting a discrimination result after discrimination;
and step S05, the generation countermeasure network model correspondingly feeds back the fused enhanced image as a flight control signal according to the output results of the step S03 and the step S04 to drive the power system to adjust the working states of the multiple rotors and the fixed wings.
As a further improvement of the invention, the loss function comprises a penalty loss function, a mean square error loss function and a VGG loss function, wherein:
the penalty function is:
wherein D represents a discriminator network module, G represents a generator network module, x represents a second input signal image, G (x) represents an enhanced image, y represents a first input signal image, k is a gain coefficient, and v is an aircraft airspeed;
the mean square error loss function is:
wherein W represents the width of the input image and H represents the height of the input image;
the VGG loss function is:
wherein phiiThe characteristic output of the i-th layer convolution of the VGG model is obtained, and the definition of a generated image can be improved by the VGG loss function;
the total loss function is:
Loss＝LG+λMSELMSE+λVGGLVGG
wherein L isGIs to fight against the damageLoss function, LMSEIs a function of the mean square error loss, LVGGIs a VGG loss function, λMSEIs a penalty factor, λ, of a mean square error loss functionVGGIs the penalty factor for the VGG loss function. Setting the loss function, then using the image data to generate a countermeasure network model for continuous alternate iterative training, and optimizing the model parameters of the generated countermeasure network model.
Compared with the prior art, the unmanned aerial vehicle flight state fusion control method based on the countermeasure network is provided by the invention. Through hovering and coordinating the turn in-process to unmanned aerial vehicle taking off, descending, cruising, fixed point, flight control system simultaneously passes through many rotors reach the stationary vane cooperation collaborative work controls unmanned aerial vehicle's horizontal position, vertical height, preceding speed, lateral velocity and aircraft course angle make unmanned aerial vehicle takes off, descends, cruises, fixed point hovers and coordinates the turn in-process, many rotors with the stationary vane simultaneous working. And adjusting the attack angle of the fixed wing according to the flight state so as to control the unmanned aerial vehicle to fly in the state of the maximum ratio of the lift force to the resistance. Improve unmanned aerial vehicle is at the aerodynamic lift under the high-speed flight state, reduce system's consumption.
Secondly, the loss function in the generation of the countermeasure network model can better restore the actual input parameter information, the defects that the generated picture has excessive top-end texture characteristics and lacks high-frequency information are overcome, and the VGG loss enables the enhanced image generated by the countermeasure network model again to be closer to the effect of the first input signal image no matter the bottom layer characteristic pixel value or the high-layer abstract characteristics.
More importantly, in the flight control process of the unmanned aerial vehicle, a feedback control mechanism is built between the actual flight state of the unmanned aerial vehicle and a predicted flight control plan under the control of a flight control system through an antagonistic network model, so that the flight precision control of the unmanned aerial vehicle is enhanced, and the loss, the flight parameter error caused by unsmooth switching between different flight states and the power loss are reduced.
[ description of the drawings ]
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without inventive efforts, wherein:
fig. 1 is a schematic perspective view of the unmanned aerial vehicle disclosed in the present invention;
fig. 2 is a block diagram of the structure of the drone shown in fig. 1;
FIG. 3 is a block diagram of the flight control system of FIG. 2;
FIG. 4 is a block diagram of the structure of the generator network module shown in FIG. 3;
fig. 5 is a flowchart of a method for controlling the flight state fusion of an unmanned aerial vehicle based on a generation countermeasure network model disclosed in the present invention.
[ detailed description ] embodiments
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Please refer to fig. 1 and fig. 2 in combination, wherein fig. 1 is a schematic perspective view of an unmanned aerial vehicle according to the present invention, and fig. 2 is a block diagram of the unmanned aerial vehicle shown in fig. 1. Unmanned aerial vehicle 10 includes fuselage 11, a plurality of rotor 13, stationary vane 15, driving system 17 and flight control system 19. The plurality of rotary wings 13 and the fixed wing 15 are fixed to the airframe 11. The power system 17 outputs power to adjust the operating states of the rotors 13 and the fixed wing 15. The flight control system 19 controls the output power of the power system 17 to adjust the rotation speed of the multiple rotors 13 and the attack angle of the fixed wing 15 in a fusion manner.
Fuselage 11 supports fixedly a plurality of rotors 13, stationary vane 15 and driving system 17.
The plurality of rotors 13 are fixed to the fuselage 11 and are evenly distributed around the fuselage 11. Each rotor 13 drives the operating state of the propeller itself by means of a rotating electrical machine, i.e. the propeller speed and the direction of rotation of each rotor 13 are driven by said rotating electrical machine. The plurality of rotating motors are respectively electrically connected with the flight control system 19 and receive the control signal of the power system 17, namely, the power system 17 receives the control signal of the flight control system 19 to drive the rotating motors to rotate, so that the adjustment of the working state of the rotor wing 13 is realized. The rotary motor drives the propeller to operate in a set rotation direction and rotation speed range according to a control signal from the flight control system 19, that is, the power system 17 adjusts the operating state of the multiple rotors 13 according to the output power.
The fixed wing 15 is fixed to the top end of the fuselage 11 and is electrically connected with the flight control system 19 through a support motor. The support motor drives the inclination angle of the fixed wing 15 relative to the horizontal direction during flight, namely: angle of attack. The change of the attack angle value of the fixed wing 15 correspondingly adjusts the flight resistance of the unmanned aerial vehicle 10 in the flight process.
The power system 17 includes a multi-pole output driving device with multiple output branches, which includes a rotating motor for driving the multiple rotors 13 to work and a supporting motor for adjusting the attack angle of the fixed wing 15. Each rotating motor and the supporting motor correspond to one output branch of the power system, and the rotating speed and the rotating direction of the multiple output branches are coordinated; the support motor corresponds to the other output branch, and the attack angle of the fixed wing is dynamically adjusted, so that the unmanned aerial vehicle 10 can change in space flight state in real time, and smoothly switch between different states.
Specifically, the unmanned aerial vehicle 10 includes an ascending state, a cruising state, and a landing state during an actual flight, and the dynamic flight of the unmanned aerial vehicle 10 is arbitrarily switched between the three operating states. Simultaneously, unmanned aerial vehicle 10 is in different operating condition, the rotation speed and the direction of rotation of rotor, the angle of attack of stationary vane fuses the effect jointly.
Referring to fig. 2 and fig. 3 in combination, fig. 3 is a block diagram of the flight control system shown in fig. 2. The flight control system 19 is electrically connected to the input shaft of the power system 17. The flight control system 19 collects the flight resistance value, the speed value, the height value and the like of the unmanned aerial vehicle 10 in real time, namely the actual flight state parameter value of the unmanned aerial vehicle 10. The flight control system 19 includes an original flight state parameter output module 191, an actual flight state parameter collection module 193, and a generation countermeasure network model 195.
The original flight state parameter output module 191 reads rated working parameters of the multiple rotors 13 and rated working parameters of the fixed wings 15 from the unmanned aerial vehicle 10 in real time, and generates an original working state flight curve of the unmanned aerial vehicle 10 as a first input signal image 1910.
Actual flight state parameter gathers module 193 and gathers in real time unmanned aerial vehicle 10 the actual working parameter of many rotors 13 the actual working parameter of stationary vane 15 generates unmanned aerial vehicle 10's actual working state flight curve is as second input signal image 1930. Specifically, the actual flight state parameter acquisition module 193 includes the rotation speed inductor of rotating electrical machines, fixed wing and horizontal direction included angle detector, altitude sensor, flight speed inductor and resistance inductor etc. and real-time induction respectively the rotation speed, angle of attack, flight altitude parameter, flight speed parameter and the flight resistance value of the rotating electrical machines of unmanned aerial vehicle 10 are drawn into the flight curve, and serve as the second input signal image 1930.
The generative countermeasure network model 195 includes a generator network module 1951, a discriminator network module 1953, and a loss function 1955.
Please refer to fig. 4, which is a block diagram of the generator network module shown in fig. 3. The generator network module 1951 employs an Attention Unet. Wherein the Attention Unet network includes downsampling, upsampling, and Attention gate. The second input signal image 1930 is downsampled, deep features of the second input signal image 1930 are learned, the deep features are deconvoluted and upsampled, and finally an enhanced image 19510 is output.
The Attention Unet network has the advantages that features of different layers can be captured and integrated in a feature superposition mode, more information of the second input signal image 1930 can be carried, and the enhanced image 19510 can be generated better.
The discriminator network module 1953 receives the first input signal image 1910 outputted from the original flight status parameter output module 191 and the enhanced image 19510 generated by the generator network module 1951, and provides a discrimination result 19530, i.e. the authenticity between the enhanced image 19510 and the first input signal image 1910 is obtained by continuous iterative training, so as to obtain the enhanced image 19510 closer to the first input signal image 1910.
The discriminator network module 1953 inputs the enhanced image 19510 and the first input signal image 1910 output from the generator network module 1951 into the discriminator network module 1953 using a PatchGAN, resulting in an n x n matrix, and finally outputs the discrimination result 19530, which takes the mean value of the matrix as a true/false result.
The loss function 1955 calculates image loss, optimizes model parameters that generate the opposing network model 195, the loss function 1955 includes an opposing loss function, a mean square error loss function, and a VGG loss function, wherein:
the penalty function is:
wherein D represents the discriminator network module, G represents the generator network module, x represents the second input signal image, G (x) represents the enhanced image, y represents the first input signal image, k is the gain coefficient, and v is the aircraft airspeed.
The mean square error loss function is:
where W represents the width of the input image and H represents the height of the input image.
The VGG loss function is:
wherein phiiThe characteristic output of the i-th layer convolution of the VGG model is obtained, and the definition of a generated image can be improved by the VGG loss function.
The total loss function is:
Loss＝LG+λMSELMSE+λVGGLVGG
wherein L isGIs a function of the penalty of fighting, LMSEIs a function of the mean square error loss, LVGGIs a VGG loss function, λMSEIs a penalty factor, λ, of a mean square error loss functionVGGIs the penalty factor for the VGG loss function. Setting the loss function, then continuously and alternately training the generated confrontation network model by using image data, optimizing the model parameters of the generated confrontation network model, wherein the trained generated confrontation network model can have better enhancement effect on the second input signal image.
Compared with the prior art, in the unmanned aerial vehicle flight state fusion control system 10 based on the generation of the countermeasure network, the discriminator network module 1953 adopts PatchGAN to ensure that all smooth transitions of the enhanced image 19510 are closer to the rated first input signal image.
Meanwhile, the loss function in the generation countermeasure network model 195 can better restore the actual input parameter information, and overcomes the defects of excessive sharp texture features and lack of high-frequency information of the generated picture, and the VGG loss makes the enhanced image generated by the countermeasure network model again more approximate to the effect of the first input signal image no matter the pixel value of the bottom layer feature or the high-level abstract feature.
The loss function senses flight loss according to the second input signal image, adjusts the enhanced image, and further feeds back the determination result 19530, so that the enhanced image 19510 is closer to the first input signal image 1910, and the flight enhancement effect is improved.
Please refer to fig. 5, which is a flowchart illustrating a method for controlling the flight state fusion of an unmanned aerial vehicle based on a generation countermeasure network model according to the present invention. The unmanned aerial vehicle flight state fusion control system based on the generation countermeasure network model is applied to the flight state control of the unmanned aerial vehicle 10, and comprises the following steps:
step S01, providing an original flight state parameter output module 191, reading rated operating parameters of the multiple rotors 13 and the fixed wing 15, and generating a first input signal image 1910;
step S02, providing an actual flight state parameter acquisition module 193, acquiring actual working parameters of the multiple rotors 13 and the fixed wings 15 in real time, and generating a second input signal image 1930;
step S03, inputting the second input signal image 1930 to the generator network module 195, obtaining an enhanced image 19510;
step S04, inputting the enhanced image 19510 and the first input image 1910 into the discriminator network module 1953, and outputting the discrimination result 19530 after discrimination;
in step S05, the generated countermeasure network model 195 responds to the feedback fused enhanced image 19510 according to the output results of steps S03 and S04 and drives the power system 17 to adjust the operating states of the multiple rotors 13 and the fixed wings 15 as flight control signals.
When the unmanned aerial vehicle 10 works, the multiple rotary wings 13 and the fixed wings 15 work synchronously. An actual flight state parameter acquisition module 193 of the flight control system 19 acquires actual working parameters of the multiple rotors 13 and the fixed wings 15, and transmits the actual working parameters as a second input signal image 1930 to the pair of anti-network generation models 195; meanwhile, the original flight state parameter collecting module 191 reads rated working parameters of the multiple rotors 13 and the fixed wing 15 to generate a first input signal image 1910. In the actual flight process, the first input signal image 1910 and the second input signal image 1930 have a difference, and when the difference is large, the phenomena of jamming and drifting of the unmanned aerial vehicle 10 can be caused, and the flight state of the controller cannot be accurately controlled.
The discriminator network module 1953 fuses the first input signal image 1910 and the second input signal image 1930 to output and synthesize the flight control signal, and outputs enhanced flight parameters, dynamically modifies the dynamic control of the flight control system 19 on the output power of the power system 17, and meets the actual flight requirement; the loss function senses flight loss according to the actual working parameters, adjusts the enhanced flight parameter enhancement, and performs feedback control on the output power of the power system 17 to improve the flight enhancement effect.
The unmanned aerial vehicle 10 is in the actual flight process, including the working condition of flying over, the working condition of cruising and descending flight condition. According to actual flight needs, constantly adjust unmanned aerial vehicle 10's actual flight state for unmanned aerial vehicle 10 smoothly switches between different flight states, in the moment of switching, generate to resist network model 19 basis first input signal image 1910, second input signal image 1930, corresponding feedback flight control signal is in order to strengthen control unmanned aerial vehicle 10's smooth switch reduces the loss, improves cruise ability and accurate control unmanned aerial vehicle 10's flight state.
More importantly, in the process, a feedback control mechanism is built through an antagonistic network model between the actual flight state of the unmanned aerial vehicle 10 and a predicted flight control plan under the control of a flight control system, so that the flight precision control of the unmanned aerial vehicle 10 is enhanced, and the loss and the flight parameter error and the power loss caused by unsmooth switching among different flight states are reduced.
Compared with the related art, the flight control method with the fusion of the multiple rotors 13 and the fixed wings 15 is provided. Through cruising the in-process to unmanned aerial vehicle 10, control system simultaneously through many rotors reach the stationary vane cooperation collaborative work controls unmanned aerial vehicle 10's position makes unmanned aerial vehicle is cruising the in-process, many rotors with the stationary vane simultaneous work. And adjusting the attack angle of the fixed wing according to the flight state so as to control the unmanned aerial vehicle to fly in the state of the maximum ratio of the lift force to the resistance. Improve unmanned aerial vehicle is at the aerodynamic lift under the high-speed flight state, reduce system's consumption.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. An unmanned aerial vehicle flight state fusion control system based on generation of an countermeasure network is disclosed, wherein the unmanned aerial vehicle comprises ascending, cruising and descending states, the unmanned aerial vehicle comprises a plurality of rotors, a fixed wing, a power system and a flight control system, the flight control system controls the output power of the power system so as to fuse and adjust the rotating speed of the rotors and the attack angle of the fixed wing, and the unmanned aerial vehicle flight state fusion control system is characterized by comprising an original flight state parameter output module, an actual flight state parameter acquisition module and a generation countermeasure network model, wherein the original flight state parameter output module reads rated working parameters of the rotors and the fixed wing in real time to generate a first input signal image; the actual flight state parameter acquisition module acquires actual working parameters of the multiple rotors and the fixed wings in real time to generate a second input signal image; the generation countermeasure network model comprises a generator network module, a discriminator network module and a loss function, the generator network module generates an enhanced picture according to the second input signal image, and the discriminator network module outputs a discrimination result according to the first input signal image and the enhanced picture.
2. The countermeasure-network-based unmanned aerial vehicle flight state fusion control system of claim 1, wherein the generator network module employs an Attention Unet, wherein the Attention Unet network includes down-sampling, up-sampling, and Attention gate, the second input signal image is subjected to the down-sampling, deep features of the second input signal image are learned, the deep features are subjected to deconvolution up-sampling, and the enhanced image is finally output.
3. The unmanned aerial vehicle flight state fusion control system based on the countermeasure network of claim 2, wherein the discriminator network module employs PatchGAN, the enhanced image and the first input signal image output from the generator network module are input into the discriminator network module to obtain an n x n matrix, and finally a discrimination result is output, and the discrimination result takes a mean value of the matrix as a true/false result.
4. The system of claim 3, wherein the loss functions include a countermeasure loss function, a mean square error loss function, and a VGG loss function.
5. The unmanned aerial vehicle flight state fusion control system based on countermeasure network of claim 4, wherein the countermeasure loss function is:
wherein D represents the discriminator network module, G represents the generator network module, x represents the second input signal image G (x) represents the enhanced image, y represents the first input signal image, k is the gain coefficient, and v is the aircraft airspeed.
6. The countermeasure-network-based unmanned aerial vehicle flight state fusion control system of claim 5, wherein the mean square error loss function is:
where W represents the width of the input image and H represents the height of the input image.
7. The countermeasure network-based unmanned aerial vehicle flight state fusion control system of claim 6, wherein the VGG loss function is:
wherein phiiThe characteristic output of the i-th layer convolution of the VGG model is obtained, and the definition of a generated image can be improved by the VGG loss function.
8. The countermeasure network-based unmanned aerial vehicle flight state fusion control system of claim 7, wherein the total loss function is:
Loss＝LG+λMSELMSE+λVGGLVGG
wherein L isGIs a function of the penalty of fighting, LMSEIs a function of the mean square error loss, LVGGIs a VGG loss function, λMSEIs a penalty factor, λ, of a mean square error loss functionVGGIs the penalty factor for the VGG loss function. Setting the loss function, and then continuously and alternately training the generated confrontation network model by using image data to optimize the model parameters of the generated confrontation network model.
9. An unmanned aerial vehicle flight state fusion control method based on an antagonistic network is disclosed, wherein the unmanned aerial vehicle comprises a plurality of rotors, a fixed wing, a power system and a flight control system, the flight control system comprises an original flight state parameter output module, an actual flight state parameter acquisition module and an antagonistic network generation model, and the antagonistic network generation model comprises a generator network module, a discriminator network module and a loss function, and is characterized by comprising the following steps:
step S01, providing an original flight state parameter output module, reading rated working parameters of the multi-rotor wing and the fixed wing, and generating a first input signal image;
step S02, providing an actual flight state parameter acquisition module, acquiring actual working parameters of the multiple rotors and the fixed wings in real time, and generating a second input signal image;
step S03, inputting the second input signal image to the generator network module to obtain an enhanced image;
step S04, inputting the enhanced image and the first input signal image into the discriminator network module, and outputting a discrimination result after discrimination;
and step S05, the generation countermeasure network model correspondingly feeds back the fused enhanced image as a flight control signal according to the output results of the step S03 and the step S04 to drive the power system to adjust the working states of the multiple rotors and the fixed wings.
10. The method of claim 9, wherein the loss function comprises a countermeasure loss function, a mean square error loss function, and a VGG loss function, and wherein:
the penalty function is:
wherein D represents a discriminator network module, G represents a generator network module, x represents a second input signal image, G (x) represents an enhanced image, y represents a first input signal image, k is a gain coefficient, and v is an aircraft airspeed;
the mean square error loss function is:
wherein W represents the width of the input image and H represents the height of the input image;
the VGG loss function is:
wherein phiiThe characteristic output of the i-th layer convolution of the VGG model is obtained, and the definition of a generated image can be improved by the VGG loss function;
the total loss function is:
Loss＝LG+λMSELMSE+λVGGLVGG
wherein L isGIs a function of the penalty of fighting, LMSEIs a function of the mean square error loss, LVGGIs a VGG loss function, λMSEIs a penalty factor, λ, of a mean square error loss functionVGGIs the penalty factor for the VGG loss function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110763783.9A CN113534837B (en) | 2021-07-06 | 2021-07-06 | Unmanned aerial vehicle flight state fusion control system and control method based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110763783.9A CN113534837B (en) | 2021-07-06 | 2021-07-06 | Unmanned aerial vehicle flight state fusion control system and control method based on generation countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113534837A true CN113534837A (en) | 2021-10-22 |
| CN113534837B CN113534837B (en) | 2024-03-22 |
Family
ID=78126938
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110763783.9A Active CN113534837B (en) | 2021-07-06 | 2021-07-06 | Unmanned aerial vehicle flight state fusion control system and control method based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113534837B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10373026B1 (en) * | 2019-01-28 | 2019-08-06 | StradVision, Inc. | Learning method and learning device for generation of virtual feature maps whose characteristics are same as or similar to those of real feature maps by using GAN capable of being applied to domain adaptation to be used in virtual driving environments |
| CN110866472A (en) * | 2019-11-04 | 2020-03-06 | 西北工业大学 | Unmanned aerial vehicle ground moving target identification and image enhancement system and method |
| CN111783545A (en) * | 2020-06-02 | 2020-10-16 | 山西潞安环保能源开发股份有限公司五阳煤矿 | Coal mine unmanned aerial vehicle image enhancement method based on generation of countermeasure network |
| CN112508929A (en) * | 2020-12-16 | 2021-03-16 | 奥比中光科技集团股份有限公司 | Method and device for training generation of confrontation network |
| WO2021088101A1 (en) * | 2019-11-04 | 2021-05-14 | 中国科学院自动化研究所 | Insulator segmentation method based on improved conditional generative adversarial network |
-
2021
- 2021-07-06 CN CN202110763783.9A patent/CN113534837B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10373026B1 (en) * | 2019-01-28 | 2019-08-06 | StradVision, Inc. | Learning method and learning device for generation of virtual feature maps whose characteristics are same as or similar to those of real feature maps by using GAN capable of being applied to domain adaptation to be used in virtual driving environments |
| CN110866472A (en) * | 2019-11-04 | 2020-03-06 | 西北工业大学 | Unmanned aerial vehicle ground moving target identification and image enhancement system and method |
| WO2021088101A1 (en) * | 2019-11-04 | 2021-05-14 | 中国科学院自动化研究所 | Insulator segmentation method based on improved conditional generative adversarial network |
| CN111783545A (en) * | 2020-06-02 | 2020-10-16 | 山西潞安环保能源开发股份有限公司五阳煤矿 | Coal mine unmanned aerial vehicle image enhancement method based on generation of countermeasure network |
| CN112508929A (en) * | 2020-12-16 | 2021-03-16 | 奥比中光科技集团股份有限公司 | Method and device for training generation of confrontation network |
Non-Patent Citations (1)
| Title |
|---|
| 毕晓君;潘梦迪;: "基于生成对抗网络的机载遥感图像超分辨率重建", 智能系统学报, no. 01 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113534837B (en) | 2024-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA2571372C (en) | Method and apparatus for flight control of tiltrotor aircraft | |
| CA2673642C (en) | Dual rotor vertical takeoff and landing rotorcraft | |
| US8561937B2 (en) | Unmanned aerial vehicle | |
| WO2017184438A1 (en) | Combined pitch and forward thrust control for unmanned aircraft systems | |
| JP2003527272A (en) | Aircraft flight control system | |
| CN116529159A (en) | System and method for a VTOL aircraft flight control manipulator | |
| CN108594839A (en) | Control method, aircraft based on more vectoring technologies and storage medium | |
| CN105000175A (en) | Multi-rotor aircraft and control method thereof | |
| AU2018303294A1 (en) | Hybrid multi-rotor unmanned aerial vehicle with adjustable wings | |
| JP2009143268A (en) | Flight control system for aircraft and aircraft with the flight control system | |
| CN204776021U (en) | Multi -rotor aircraft | |
| US20220350348A1 (en) | Aircraft with a multi-fan propulsion system for controlling flight orientation transitions | |
| JP6952389B1 (en) | Aircraft | |
| CN113534837A (en) | Unmanned aerial vehicle flight state fusion control system and control method based on generation of countermeasure network | |
| US20220402605A1 (en) | Flight Control Apparatus | |
| CN116443246A (en) | Short-distance take-off and landing control device for tiltrotor aircraft | |
| WO2022145045A1 (en) | Flying object control method | |
| JPH07132893A (en) | Rotary-wing aircraft | |
| EP4212430A1 (en) | Flying vehicle | |
| CN107284658A (en) | A kind of compound vertical/STOL aircraft | |
| CN114180046A (en) | Electric vertical take-off and landing unmanned aerial vehicle | |
| CN113879051A (en) | Vertical take-off and landing and fixed wing aerocar | |
| CN220948591U (en) | Flying body | |
| CN209649973U (en) | Manned air-ground amphibious aircraft | |
| JP2022150020A (en) | Unmanned aircraft and control method thereof |
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 |