CN110879602B - Unmanned aerial vehicle control law parameter adjusting method and system based on deep learning - Google Patents
Unmanned aerial vehicle control law parameter adjusting method and system based on deep learning Download PDFInfo
- Publication number
- CN110879602B CN110879602B CN201911242288.2A CN201911242288A CN110879602B CN 110879602 B CN110879602 B CN 110879602B CN 201911242288 A CN201911242288 A CN 201911242288A CN 110879602 B CN110879602 B CN 110879602B
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- deep learning
- control law
- flight
- 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.)
- Active
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 40
- 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 17
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 230000001133 acceleration Effects 0.000 claims description 12
- 230000001105 regulatory effect Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 5
- 238000005299 abrasion Methods 0.000 abstract description 3
- 230000007774 longterm Effects 0.000 abstract description 2
- 238000013136 deep learning model Methods 0.000 description 5
- 238000013480 data collection Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 230000003750 conditioning effect Effects 0.000 description 3
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 230000036544 posture 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, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/08—Control of attitude, i.e. control of roll, pitch, or yaw
- G05D1/0808—Control of attitude, i.e. control of roll, pitch, or yaw specially adapted for aircraft
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- 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
-
- 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)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
- Feedback Control In General (AREA)
Abstract
The invention provides an unmanned aerial vehicle control law parameter adjusting method and system based on deep learning. The system includes a deep learning training subsystem and a deep learning application subsystem. The deep learning training subsystem comprises an unmanned aerial vehicle data acquisition unit and a deep learning training unit. The deep learning application subsystem comprises a deep learning operation unit and a PID control unit, wherein the deep learning operation unit comprises a deep learning network model trained by the deep learning training unit, the deep learning network model receives unmanned aerial vehicle flight state parameter data, outputs control law parameter offset values of the unmanned aerial vehicle, and serves as input of the PID control unit. The invention ensures that the flight control is more stable and the precision is higher. The unmanned aerial vehicle flight control system and the unmanned aerial vehicle flight control method can quickly enable flight control data of the unmanned aerial vehicle to be compensated to the characteristic value of the optimal control law parameter, and avoid the situation that the actual flight attitude of the unmanned aerial vehicle deviates from the attitude which can be achieved by the control parameter due to abrasion or deformation of the local structure of the unmanned aerial vehicle under long-term service.
Description
Technical Field
The invention relates to the field of automatic control, in particular to an unmanned aerial vehicle control law parameter adjusting method and system based on deep learning.
Background
Along with the wider and wider application of unmanned aerial vehicles, the requirement on the accumulated operation duration of the service period of the unmanned aerial vehicle is higher and higher, the control system of many unmanned aerial vehicles at present adopts a PID (proportion, integration and differentiation) control method, and particularly for the unmanned aerial vehicle with multiple rotor wings, the lifting force generated by each propeller controls the flight state of the unmanned aerial vehicle, and the lifting and steering functions of the unmanned aerial vehicle are achieved through the mutual cooperation of each motor, so that the motor rotation speed regulating function has anti-interference capability, and when the unmanned aerial vehicle is in service for a long time, the corresponding parts of each motor cannot be accurately controlled due to the mechanical abrasion or stress change of the unmanned aerial vehicle, so that the flight performance of the unmanned aerial vehicle is reduced. The PID control method is applied to the unmanned aerial vehicle control system, various parameters can be accurately assigned through a large number of experiments by engineers with abundant experience, and deviation caused by mechanical abrasion or stress change of the unmanned aerial vehicle cannot be timely and accurately adjusted, so that a stable and reliable method and system are required to be designed for adjusting the control parameters in real time, and continuous stability and control precision of unmanned aerial vehicle flight are guaranteed.
Disclosure of Invention
Aiming at the defects existing in the prior art, the invention provides an unmanned aerial vehicle control law parameter adjusting method and system based on deep learning.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
an unmanned aerial vehicle control law parameter adjusting system based on deep learning comprises a deep learning training subsystem and a deep learning application subsystem.
The deep learning training subsystem comprises an unmanned aerial vehicle data acquisition unit and a deep learning training unit. The unmanned aerial vehicle data acquisition unit acquires unmanned aerial vehicle flight state parameter information, wherein the unmanned aerial vehicle flight state parameter information comprises flight attitude information, speed, acceleration, angular velocity, position, airborne equipment, state of a sensor and sensor data.
The deep learning training unit is used for receiving the unmanned aerial vehicle flight state parameter information acquired by the unmanned aerial vehicle data acquisition unit and training to form a deep learning network model.
The deep learning application subsystem comprises a deep learning operation unit and a PID control unit, wherein the deep learning operation unit comprises a deep learning network model trained by the deep learning training unit, the deep learning network model receives unmanned aerial vehicle flight state parameter data, outputs control law parameter offset values of the unmanned aerial vehicle, and serves as input of the PID control unit.
The PID control unit is an unmanned aerial vehicle attitude control unit taking proportional, integral and differential regulation as a framework, and in different stages of the unmanned aerial vehicle flight process, the input parameters of the PID control unit are changed in real time according to the change of the control law parameter offset value output by the deep learning operation unit, and the flight attitude of the unmanned aerial vehicle is adjusted.
The corresponding unmanned aerial vehicle control law parameter adjusting method based on deep learning comprises the following steps:
s1, constructing an unmanned plane deep learning flight database.
The unmanned aerial vehicle deep learning flight database comprises unmanned aerial vehicle flight data corresponding to PID optimal control law parameters and unmanned aerial vehicle flight data obtained by taking the optimal control law parameters as the center and adjusting the control law parameters of a PID control unit.
The flight data of the invention comprises pitch angle, roll angle, yaw angle, each axial angular velocity, each axial acceleration and coordinate position.
S1.1, unmanned aerial vehicle flight data corresponding to PID optimal control law parameters are obtained, and the data is that the unmanned aerial vehicle flight through practice verification can reach an optimal stable state.
S1.2, taking the optimal control law parameter as a center to regulate unmanned aerial vehicle flight data obtained by the control law parameter of the PID control unit.
The method comprises the following steps: s1.2.1, setting a control law parameter adjusting mode of a PID control unit;
the control law parameters of the PID control unit are regulated in three ways, one is that only one of the proportion, the integral and the differential is regulated, and the other two are unchanged; one is to arbitrarily adjust two of the proportion, the integral and the differential, and the other one is unchanged; one is proportional, integral and differential simultaneous adjustment.
The PID control law parameter is adjusted in a mode of taking the optimal control law parameter as a center, and the unmanned aerial vehicle flight data are gradually adjusted and collected towards the two sides of the optimal control law parameter according to the meaning of each parameter.
S1.2.2 the parameters of PID control unit are adjusted one by one according to the regulation mode, the unmanned aerial vehicle flight data taking parameter regulation content as the label is obtained when each time is adjusted, the time for each time of collecting data is based on a classical frame time of unmanned aerial vehicle flight, namely stable flight is carried out for 10 minutes, each time of adjusting control law parameters again carries out unmanned aerial vehicle flight data collection, the unmanned aerial vehicle flight data collection comprises pitch angle, roll angle, yaw angle, each axial angular velocity, each axial acceleration, coordinate position and the like, the label taking parameter regulation content as data is finally obtained, and the unmanned aerial vehicle flight data sample base with the label is finally obtained.
S1.3, combining the step S1.1 and the step S1.2 to obtain the unmanned aerial vehicle deep learning flight database with the labels.
S2, constructing and training a deep learning network model.
S2.1, constructing a deep learning network model;
the deep learning network model is a multi-layer neural network convolution model taking unmanned aerial vehicle flight data with labels as influence factors;
s2.2, training a deep learning network model;
s2.2.1, inputting the unmanned aerial vehicle flight data acquired in the step S1 into a multi-layer neural network convolution model for training;
s2.2.2, minimizing a residual function by adopting a gradient descent method, and extracting parameter characteristics of a multi-layer neural network convolution model;
extracting multi-stage parameter characteristics through a multi-stage convolution layer and a pooling layer;
the gradient descent method minimizes residual functions, namely factors affecting weights in unmanned aerial vehicle flight data are judged, the adjustment direction of parameters in a deep learning network is determined, a result output by the deep learning network is approximated to a direction which enables an unmanned aerial vehicle flight control system to be more stable, and the condition of a stable attitude angle of the unmanned aerial vehicle flight attitude is that a pitch angle, a roll angle and a yaw angle are closest to the unmanned aerial vehicle flight attitude angle corresponding to known optimal control law parameters, and the corresponding deviation is minimum.
S2.2.3, adjusting each weight parameter in the multi-layer neural network convolution model layer by layer according to the step S2.2.2 to obtain a trained deep learning network model;
according to the invention, a Tensorflow frame is selected as a training deep learning model, and data acquired for unmanned aerial vehicle flight control are as follows: and (3) selecting corresponding convolution and pooling factors in the frame for each axial angular speed, acceleration and speed, and performing cyclic iterative training and evaluation by Tensorflow to finally obtain offset values of each control law parameter relative to the optimal control law parameter.
And the offset value comprises the deviation of the mechanical parts, which is caused by the long service time of the unmanned aerial vehicle under the same control law parameters, and the deviation caused by the stress of the unmanned aerial vehicle.
S3, checking a deep learning network model;
inputting the unmanned aerial vehicle flight data without training in the step S1 into the deep learning network model built in the step S2, and learning the offset of the current flight control law parameters of the unmanned aerial vehicle relative to the optimal control law parameters; and comparing the learned offset with the standard deviation, if the confidence coefficient of the learned offset is higher than 90% relative to the standard deviation which is not more than +/-5%, the test is passed, otherwise, repeating the step S2 until the test is passed.
S4, acquiring real-time flight data of the unmanned aerial vehicle. When the unmanned aerial vehicle executes various flight tasks, new unmanned aerial vehicle flight data are collected in real time, wherein the new unmanned aerial vehicle flight data comprise pitch angle, roll angle and yaw angle, and each axial angular speed, acceleration, speed and position.
S5, inputting the acquired real-time flight data of the unmanned aerial vehicle into the tested deep learning network model to obtain the offset relative to the optimal control law parameter;
and S6, the PID control unit adjusts the unmanned aerial vehicle control law parameters according to the offset obtained in the step S5, so that the unmanned aerial vehicle obtains an optimal flight state.
According to the invention, flight control data of the unmanned aerial vehicle under various conditions are collected in the training of the deep learning model, and different flight postures are fused through the constructed deep learning network model, so that the model of the deep learning network is more perfect, the optimal adjustment of the flight parameters of the unmanned aerial vehicle by PID control is improved, the flight system of the unmanned aerial vehicle is more stable, and the control is more accurate.
The beneficial effects of the invention are as follows: according to the invention, by utilizing the good learning ability of the deep learning network and the supervised learning mode, unmanned aerial vehicle flight control data are continuously learned in the deep learning model to be close to the optimal control law parameters of unmanned aerial vehicle flight control, so that the flight control is more stable, the precision is higher, and the unmanned aerial vehicle persistence is better. Compared with the traditional control algorithm, after the deep learning model is trained, the flight control data of the unmanned aerial vehicle can be quickly compensated to the characteristic value of the optimal control law parameter, so that the unmanned aerial vehicle is prevented from being worn or deformed due to the local structure of the unmanned aerial vehicle under long-term service, the actual flight attitude of the unmanned aerial vehicle deviates from the attitude which can be achieved by the control parameter, the flight control law parameter of the unmanned aerial vehicle can be better and quickly adjusted, and the unmanned aerial vehicle can be restored to the optimal control law parameter of the unmanned aerial vehicle, so that the unmanned aerial vehicle flies in a stable attitude.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required in the description of the embodiments will be briefly described below, and it will be apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of training process of deep learning network model in the adjustment method of the present invention.
FIG. 2 is a schematic diagram of an application flow in the conditioning method of the present invention.
FIG. 3 is a schematic diagram of the training system in the conditioning system of the present invention.
FIG. 4 is a schematic diagram of an application system configuration in the conditioning system of the present invention.
Detailed Description
The technical solutions of the present invention will be clearly and perfectly described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1: an unmanned aerial vehicle control law parameter adjusting system based on deep learning comprises a deep learning training subsystem and a deep learning application subsystem.
The deep learning training subsystem, as shown in fig. 3, comprises an unmanned aerial vehicle data acquisition unit and a deep learning training unit. The unmanned aerial vehicle data acquisition unit acquires unmanned aerial vehicle flight state parameter information, wherein the unmanned aerial vehicle flight state parameter information comprises flight attitude information, speed, acceleration, angular velocity, position, airborne equipment, state of a sensor and sensor data.
The deep learning training unit is used for receiving the unmanned aerial vehicle flight state parameter information acquired by the unmanned aerial vehicle data acquisition unit and training to form a deep learning network model.
The deep learning application subsystem, as shown in fig. 4, comprises a deep learning operation unit and a PID control unit, wherein the deep learning operation unit comprises a deep learning network model trained by a deep learning training unit, and the deep learning network model receives unmanned aerial vehicle flight state parameter data, outputs control law parameter offset values of the unmanned aerial vehicle and is used as input of the PID control unit.
The PID control unit is an unmanned aerial vehicle attitude control unit taking proportional, integral and differential regulation as a framework, and in different stages of the unmanned aerial vehicle flight process, the input parameters of the PID control unit are changed in real time according to the change of the control law parameter offset value output by the deep learning operation unit, and the flight attitude of the unmanned aerial vehicle is adjusted.
Example 2: the unmanned aerial vehicle control law parameter adjusting method based on deep learning, as shown in figures 1 and 2, comprises the following steps:
s1, constructing an unmanned plane deep learning flight database.
The unmanned aerial vehicle deep learning flight database comprises unmanned aerial vehicle flight data corresponding to PID optimal control law parameters and unmanned aerial vehicle flight data obtained by taking the optimal control law parameters as the center and adjusting the control law parameters of a PID control unit.
The flight data of the invention comprises pitch angle, roll angle, yaw angle, each axial angular velocity, each axial acceleration and coordinate position.
S1.1, unmanned aerial vehicle flight data corresponding to PID optimal control law parameters are obtained, and the data is that the unmanned aerial vehicle flight through practice verification can reach an optimal stable state.
S1.2, taking the optimal control law parameter as a center to regulate unmanned aerial vehicle flight data obtained by the control law parameter of the PID control unit.
The method comprises the following steps: s1.2.1, setting a control law parameter adjusting mode of a PID control unit;
the control law parameters of the PID control unit are regulated in three ways, one is that only one of the proportion, the integral and the differential is regulated, and the other two are unchanged; one is to arbitrarily adjust two of the proportion, the integral and the differential, and the other one is unchanged; one is proportional, integral and differential simultaneous adjustment.
The PID control law parameter is adjusted in a mode of taking the optimal control law parameter as a center, and the unmanned aerial vehicle flight data are gradually adjusted and collected towards the two sides of the optimal control law parameter according to the meaning of each parameter.
S1.2.2 the parameters of PID control unit are adjusted one by one according to the regulation mode, the unmanned aerial vehicle flight data taking parameter regulation content as the label is obtained when each time is adjusted, the time for each time of collecting data is based on a classical frame time of unmanned aerial vehicle flight, namely stable flight is carried out for 10 minutes, each time of adjusting control law parameters again carries out unmanned aerial vehicle flight data collection, the unmanned aerial vehicle flight data collection comprises pitch angle, roll angle, yaw angle, each axial angular velocity, each axial acceleration, coordinate position and the like, the label taking parameter regulation content as data is finally obtained, and the unmanned aerial vehicle flight data sample base with the label is finally obtained.
S1.3, combining the step S1.1 and the step S1.2 to obtain the unmanned aerial vehicle deep learning flight database with the labels.
S2, constructing and training a deep learning network model.
S2.1, constructing a deep learning network model;
the deep learning network model is a multi-layer neural network convolution model taking unmanned aerial vehicle flight data with labels as influence factors;
s2.2, training a deep learning network model;
s2.2.1, inputting the unmanned aerial vehicle flight data acquired in the step S1 into a multi-layer neural network convolution model for training;
s2.2.2, minimizing a residual function by adopting a gradient descent method, and extracting parameter characteristics of a multi-layer neural network convolution model; the invention specifically extracts multistage parameter characteristics through a multistage convolution layer and a pooling layer.
The gradient descent method minimizes residual functions, namely factors affecting weights in unmanned aerial vehicle flight data are judged, the adjustment direction of parameters in a deep learning network is determined, a result output by the deep learning network is approximated to a direction which enables an unmanned aerial vehicle flight control system to be more stable, and the condition of a stable attitude angle of the unmanned aerial vehicle flight attitude is that a pitch angle, a roll angle and a yaw angle are closest to the unmanned aerial vehicle flight attitude angle corresponding to known optimal control law parameters, and the corresponding deviation is minimum.
S2.2.3, adjusting each weight parameter in the multi-layer neural network convolution model layer by layer according to the step S2.2.2 to obtain a trained deep learning network model;
according to the invention, a Tensorflow frame is selected as a training deep learning model, and data acquired for unmanned aerial vehicle flight control are as follows: and (3) selecting corresponding convolution and pooling factors in the frame for each axial angular speed, acceleration and speed, and performing cyclic iterative training and evaluation by Tensorflow to finally obtain offset values of each control law parameter relative to the optimal control law parameter.
And the offset value comprises the deviation of the mechanical parts, which is caused by the long service time of the unmanned aerial vehicle under the same control law parameters, and the deviation caused by the stress of the unmanned aerial vehicle.
S3, checking a deep learning network model;
inputting the unmanned aerial vehicle flight data without training in the step S1 into the deep learning network model built in the step S2, and learning the offset of the current flight control law parameters of the unmanned aerial vehicle relative to the optimal control law parameters; and comparing the learned offset with the standard deviation, if the confidence coefficient of the learned offset is higher than 90% relative to the standard deviation which is not more than +/-5%, the test is passed, otherwise, repeating the step S2 until the test is passed.
The deep learning network model which passes the inspection can be directly applied to a control system of the unmanned aerial vehicle, and specifically comprises the following steps:
s4, acquiring real-time flight data of the unmanned aerial vehicle. When the unmanned aerial vehicle executes various flight tasks, new unmanned aerial vehicle flight data are collected in real time, wherein the new unmanned aerial vehicle flight data comprise pitch angle, roll angle and yaw angle, and each axial angular speed, acceleration, speed and position.
S5, inputting the acquired real-time flight data of the unmanned aerial vehicle into the tested deep learning network model to obtain the offset relative to the optimal control law parameter;
and S6, the PID control unit adjusts the unmanned aerial vehicle control law parameters according to the offset obtained in the step S5, so that the unmanned aerial vehicle obtains an optimal flight state.
Claims (6)
1. Unmanned aerial vehicle control law parameter adjustment system based on degree of depth study, its characterized in that: comprises a deep learning training subsystem and a deep learning application subsystem; the deep learning training subsystem comprises an unmanned aerial vehicle data acquisition unit and a deep learning training unit; the unmanned aerial vehicle data acquisition unit acquires unmanned aerial vehicle flight state parameter information; the deep learning training unit is used for receiving the unmanned aerial vehicle flight state parameter information acquired by the unmanned aerial vehicle data acquisition unit and training to form a deep learning network model; the deep learning network model is a multi-layer neural network convolution model taking unmanned aerial vehicle flight data with labels as influence factors;
the deep learning application subsystem comprises a deep learning operation unit and a PID control unit, wherein the deep learning operation unit comprises a deep learning network model trained by a deep learning training unit, the deep learning network model receives unmanned aerial vehicle flight state parameter data, outputs unmanned aerial vehicle control law parameter offset values, and is used as input of the PID control unit together with the unmanned aerial vehicle flight data;
the PID control unit is an unmanned aerial vehicle attitude control unit taking proportional, integral and differential regulation as a framework, and in different stages of the unmanned aerial vehicle flight process, the input parameters of the PID control unit are changed in real time according to the change of the control law parameter offset value output by the deep learning operation unit and the current flight data of the unmanned aerial vehicle, and the flight attitude of the unmanned aerial vehicle is adjusted.
2. The unmanned aerial vehicle control law parameter adjustment system based on deep learning of claim 1, wherein: the unmanned aerial vehicle flight state parameter information comprises flight attitude information, speed, acceleration, angular speed, position, airborne equipment and sensor state and sensor data.
3. An unmanned aerial vehicle control law parameter adjusting method based on deep learning is characterized by comprising the following steps of: the method comprises the following steps:
s1, constructing an unmanned plane deep learning flight database; the unmanned aerial vehicle deep learning flight database comprises unmanned aerial vehicle flight data corresponding to PID optimal control law parameters and unmanned aerial vehicle flight data obtained by taking the optimal control law parameters as the center and adjusting the control law parameters of a PID control unit;
s2, constructing and training a deep learning network model; the deep learning network model is a multi-layer neural network convolution model taking unmanned aerial vehicle flight data with labels as influence factors;
s3, checking a deep learning network model;
s4, acquiring real-time flight data of the unmanned aerial vehicle; the flight data comprise pitch angle, roll angle, yaw angle, axial angular velocity, axial acceleration and coordinate position;
s5, inputting the acquired real-time flight data of the unmanned aerial vehicle into the tested deep learning network model to obtain the offset relative to the optimal control law parameter;
and S6, the PID control unit adjusts unmanned aerial vehicle control law parameters according to the offset obtained in the step S5 and current unmanned aerial vehicle flight data, so that the unmanned aerial vehicle obtains an optimal flight attitude.
4. The unmanned aerial vehicle control law parameter adjustment method based on deep learning according to claim 3, wherein in step S1, the specific steps are as follows:
s1.1, unmanned aerial vehicle flight data corresponding to PID optimal control law parameters are obtained;
s1.2, regulating unmanned aerial vehicle flight data obtained by the control law parameters of the PID control unit by taking the optimal control law parameters as the center;
s1.2.1, setting a control law parameter adjusting mode of a PID control unit;
the control law parameters of the PID control unit are regulated in three ways, one is that only one of the proportion, the integral and the differential is regulated, and the other two are unchanged; one is to arbitrarily adjust two of the proportion, the integral and the differential, and the other one is unchanged; one is that the proportion, the integral and the differential are simultaneously regulated;
s1.2.2, adjusting parameters of the PID control unit one by one in an adjusting mode, and obtaining unmanned aerial vehicle flight data taking parameter adjusting content as a label during each adjustment;
s1.3, combining the step S1.1 and the step S1.2 to obtain the unmanned aerial vehicle deep learning flight database with the labels.
5. The unmanned aerial vehicle control law parameter adjustment method based on deep learning according to claim 3, wherein in step S2, the specific steps are as follows: s2.1, constructing a deep learning network model;
the deep learning network model is a multi-layer neural network convolution model taking unmanned aerial vehicle flight data with labels as influence factors;
s2.2, training a deep learning network model;
s2.2.1, inputting the unmanned aerial vehicle flight data acquired in the step S1 into a multi-layer neural network convolution model for training;
s2.2.2, minimizing a residual function by adopting a gradient descent method, and extracting parameter characteristics of a multi-layer neural network convolution model;
s2.2.3, adjusting each weight parameter in the multi-layer neural network convolution model layer by layer according to the step S2.2.2 to obtain a trained deep learning network model.
6. The unmanned aerial vehicle control law parameter adjusting method based on deep learning according to claim 5, wherein in step 3, the unmanned aerial vehicle flight data with the tag which is not trained in step S1 is input into the deep learning network model built in step S2, and the offset of the current unmanned aerial vehicle flight control law parameter relative to the optimal control law parameter is learned; and comparing the learned offset with the standard deviation, if the confidence coefficient of the learned offset is higher than 90% relative to the standard deviation which is not more than +/-5%, the test is passed, otherwise, repeating the step S2 until the test is passed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911242288.2A CN110879602B (en) | 2019-12-06 | 2019-12-06 | Unmanned aerial vehicle control law parameter adjusting method and system based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911242288.2A CN110879602B (en) | 2019-12-06 | 2019-12-06 | Unmanned aerial vehicle control law parameter adjusting method and system based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110879602A CN110879602A (en) | 2020-03-13 |
CN110879602B true CN110879602B (en) | 2023-04-28 |
Family
ID=69730754
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911242288.2A Active CN110879602B (en) | 2019-12-06 | 2019-12-06 | Unmanned aerial vehicle control law parameter adjusting method and system based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110879602B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113485443B (en) * | 2021-08-10 | 2023-12-22 | 北京北航天宇长鹰无人机科技有限公司 | Unmanned aerial vehicle control method based on deep learning, storage medium and equipment |
CN116165880B (en) * | 2023-04-23 | 2023-08-04 | 苏州海豚之星智能科技有限公司 | Method for automatically adjusting PID parameters by AGV speed regulation of convolutional neural network of genetic algorithm |
CN116741019A (en) * | 2023-08-11 | 2023-09-12 | 成都飞航智云科技有限公司 | Flight model training method and training system based on AI |
CN117930881B (en) * | 2024-03-25 | 2024-06-07 | 成都锦城学院 | Unmanned aerial vehicle control method and system based on deep learning |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107817820A (en) * | 2017-10-16 | 2018-03-20 | 复旦大学 | A kind of unmanned plane autonomous flight control method and system based on deep learning |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101813697B1 (en) * | 2015-12-22 | 2017-12-29 | 한국항공대학교산학협력단 | Unmanned aerial vehicle flight control system and method using deep learning |
CN107368091B (en) * | 2017-08-02 | 2019-08-20 | 华南理工大学 | A kind of stabilized flight control method of more rotor unmanned aircrafts based on finite time neurodynamics |
CN109409354B (en) * | 2017-08-18 | 2021-09-21 | 深圳市道通智能航空技术股份有限公司 | Unmanned aerial vehicle intelligent following target determination method, unmanned aerial vehicle and remote controller |
CN109933081A (en) * | 2017-12-15 | 2019-06-25 | 北京京东尚科信息技术有限公司 | Unmanned plane barrier-avoiding method, avoidance unmanned plane and unmanned plane obstacle avoidance apparatus |
CN108803668B (en) * | 2018-06-22 | 2021-08-24 | 中国南方电网有限责任公司超高压输电公司广州局 | Intelligent inspection unmanned aerial vehicle nacelle system for static target monitoring |
-
2019
- 2019-12-06 CN CN201911242288.2A patent/CN110879602B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107817820A (en) * | 2017-10-16 | 2018-03-20 | 复旦大学 | A kind of unmanned plane autonomous flight control method and system based on deep learning |
Non-Patent Citations (1)
Title |
---|
董守田 ; 杨利红 ; 康成吉 ; 黄丹丹 ; 徐娜 ; .固定翼无人机姿态控制及仿真.东北农业大学学报.2015,第46卷(第09期),第92-97页. * |
Also Published As
Publication number | Publication date |
---|---|
CN110879602A (en) | 2020-03-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110879602B (en) | Unmanned aerial vehicle control law parameter adjusting method and system based on deep learning | |
CN106774409B (en) | A kind of semi-autonomous imitative ground flight system and its control method of unmanned plane | |
CN108196532B (en) | Fault detection and separation method for longitudinal flight control system of unmanned aerial vehicle based on nonlinear adaptive observer | |
US20130245879A1 (en) | Method and system for monitoring a structure | |
CN113988202B (en) | Mechanical arm abnormal vibration detection method based on deep learning | |
US20100132446A1 (en) | Wind tunnel testing technique | |
CN102289714B (en) | Method for controlling autonomous take-off and landing of small unmanned rotorcraft based on behavioral model | |
CN114281093B (en) | Defect detection system and method based on unmanned aerial vehicle power inspection | |
CN108100301B (en) | Test flight data processing method for objective test of helicopter simulator | |
CN113190036A (en) | Unmanned aerial vehicle flight trajectory prediction method based on LSTM neural network | |
CN114398049A (en) | Self-adaptive dynamic updating method for digital twin model of discrete manufacturing workshop | |
CN103995529A (en) | On-line diagnosis method for mechanical installation faults of rotor craft | |
CN114004023A (en) | Aircraft pneumatic parameter identification method based on recurrent neural network | |
CN113247302A (en) | Method and system for monitoring condition of VTOL aircraft | |
US20200023994A1 (en) | Method and system for determining a descent profile | |
CN106970632A (en) | A kind of accurate operational method of rotor wing unmanned aerial vehicle based on the canopy vortex stable state of motion | |
CN113110558B (en) | Hybrid propulsion unmanned aerial vehicle demand power prediction method | |
CN113469082B (en) | Satellite actuator fault detection method based on migration component analysis | |
Ahsan et al. | Grey box modeling of lateral-directional dynamics of a uav through system identification | |
EP3683641A1 (en) | Methods for determining usage in fly-by-wire systems | |
KR20220096658A (en) | Method and apparatus for improving the accuracy of 6-dof flight simulation using flight test data | |
CN117891280B (en) | Unmanned aerial vehicle simulated flight maneuvering flight control method | |
CN116700358B (en) | Nonlinear height-fixing compensation control method for unmanned aerial vehicle in turning stage | |
CN115221640B (en) | Intelligent full-automatic aviation rudder wing inspection system based on infrared induction principle | |
CN118189958A (en) | Unmanned aerial vehicle positioning method based on improved matching mechanism |
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 |