CN115285380A - Microsatellite magnetic torquer attitude control method based on neural network - Google Patents
Microsatellite magnetic torquer attitude control method based on neural network Download PDFInfo
- Publication number
- CN115285380A CN115285380A CN202211055986.3A CN202211055986A CN115285380A CN 115285380 A CN115285380 A CN 115285380A CN 202211055986 A CN202211055986 A CN 202211055986A CN 115285380 A CN115285380 A CN 115285380A
- Authority
- CN
- China
- Prior art keywords
- attitude
- control
- satellite
- neural network
- model
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 title claims abstract description 39
- 108091092878 Microsatellite Proteins 0.000 title claims abstract description 25
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 21
- 238000005457 optimization Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000004088 simulation Methods 0.000 claims abstract description 7
- 230000007246 mechanism Effects 0.000 claims description 15
- 210000002569 neuron Anatomy 0.000 claims description 14
- 235000019580 granularity Nutrition 0.000 claims description 13
- 230000005484 gravity Effects 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 5
- 230000005358 geomagnetic field Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 230000005855 radiation Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 2
- 230000002068 genetic effect Effects 0.000 claims description 2
- 239000002245 particle Substances 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 230000036544 posture Effects 0.000 abstract description 8
- 230000007613 environmental effect Effects 0.000 abstract description 4
- 230000008878 coupling Effects 0.000 abstract description 3
- 238000010168 coupling process Methods 0.000 abstract description 3
- 238000005859 coupling reaction Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 12
- 239000011159 matrix material Substances 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000005389 magnetism Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000002310 reflectometry Methods 0.000 description 2
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 206010034719 Personality change Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000005034 decoration Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012938 design process Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B64—AIRCRAFT; AVIATION; COSMONAUTICS
- B64G—COSMONAUTICS; VEHICLES OR EQUIPMENT THEREFOR
- B64G1/00—Cosmonautic vehicles
- B64G1/22—Parts of, or equipment specially adapted for fitting in or to, cosmonautic vehicles
- B64G1/24—Guiding or controlling apparatus, e.g. for attitude control
- B64G1/244—Spacecraft control systems
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B64—AIRCRAFT; AVIATION; COSMONAUTICS
- B64G—COSMONAUTICS; VEHICLES OR EQUIPMENT THEREFOR
- B64G1/00—Cosmonautic vehicles
- B64G1/22—Parts of, or equipment specially adapted for fitting in or to, cosmonautic vehicles
- B64G1/24—Guiding or controlling apparatus, e.g. for attitude control
- B64G1/244—Spacecraft control systems
- B64G1/245—Attitude control algorithms for spacecraft attitude control
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B64—AIRCRAFT; AVIATION; COSMONAUTICS
- B64G—COSMONAUTICS; VEHICLES OR EQUIPMENT THEREFOR
- B64G1/00—Cosmonautic vehicles
- B64G1/22—Parts of, or equipment specially adapted for fitting in or to, cosmonautic vehicles
- B64G1/24—Guiding or controlling apparatus, e.g. for attitude control
- B64G1/244—Spacecraft control systems
- B64G1/247—Advanced control concepts for autonomous, robotic spacecraft, e.g. by using artificial intelligence, neural networks or autonomous agents
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Automation & Control Theory (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Combustion & Propulsion (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Robotics (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention provides a microsatellite magnetic torquer attitude control method based on a neural network. The method comprises the following steps: establishing a satellite orbit attitude dynamic model and an environment model; establishing a satellite attitude control subsystem model with proper granularity; optimizing the initial attitude angle and the angular speed of the satellite in one orbit period at different times to ensure that the satellite does not actively control to return to the initial attitude in one orbit period; taking the postures of the satellites at different time and orbit positions generated by the simulation of the optimization result as a sample training network and carrying out network optimization; and tracking the target posture by utilizing the target posture generated by the neural network and selecting a posture control algorithm. The attitude control method of the microsatellite magnetic torquer based on the neural network can effectively improve the magnetic control efficiency of the microsatellite, reduces the coupling of each shaft in the active control of the magnetic torquer to a certain extent, fully utilizes the environmental moment to reduce the consumption of the active magnetic control, and has certain robustness.
Description
Technical Field
The application relates to the technical field of aircraft attitude control, in particular to a microsatellite magnetic torquer attitude control method based on a neural network.
Background
The high-performance low-cost microsatellite platform is one of the directions of future aerospace development. The attitude control system is a core part of the microsatellite, is the most important part of a general commercial satellite platform in development, and is also the part with the most faults of the satellite. Therefore, it is important to improve the reliability of the attitude and orbit control subsystem as much as possible under the limit of the small satellite volume, mass, power and cost.
The magnetic torquer is a commonly used attitude control actuating mechanism, electric energy on a satellite is utilized to generate dipole magnetic moments through coils on a magnetic bar, and a geomagnetic field and the generated magnetic moments interact to generate magnetic control torque to control the attitude of the satellite. The satellite has the advantages of no need of working medium consumption, low power consumption, low cost and small volume and mass, and is applied to about two thirds of satellites at present. The satellite is acted by various different moments on the orbit, and the moments can influence the attitude of the satellite, so that the normal operation of the satellite is further influenced. Because the magnetic torquer needs to generate control torque depending on geomagnetism, the satellite three-axis can not be completely controlled in real time only by adopting the magnetic torquer for active control, and therefore, the control results of different control algorithms are very different. How to design a good control algorithm is a main problem facing the active magnetic control of the satellite.
The requirement of a large part of microsatellites on the attitude pointing accuracy is not very high, and if the attitude can be adjusted in a small amplitude to enable interference moments to be mutually counteracted, the consumption of active control can be reduced, and the coupling among magnetic control shafts is reduced. The magnetic field of the satellite is a time-varying system, and the direction of the magnetic moment generated by the magnetic torquer is constrained by the magnetic field, so that a good control effect is difficult to obtain only by considering the current state to control in real time, and a good control effect can be obtained only by considering the model change rule in a period of time according to the rule of environment change. The neural network is more and more widely applied to the field of aircraft control, and the neural network can learn the rule and fully consider the model change rule in a period of time.
Therefore, how to provide a method for controlling the attitude of a microsatellite magnetic torquer based on a neural network is a problem to be solved urgently by those skilled in the art.
Disclosure of Invention
The purpose of the application is to provide a microsatellite magnetic torquer attitude control method based on a neural network, aiming at the problems.
The application provides a method for controlling the attitude of a microsatellite magnetic torquer based on a neural network, which comprises the following steps:
s1, establishing a satellite orbit attitude dynamic model and an environment model;
s2, establishing a satellite attitude control subsystem model with a proper granularity;
s3, optimizing the initial attitude angle and the angular speed of the satellite in one orbit period at different times by using the models in the step S1 and the step S2, so that the satellite does not actively control to return to the initial attitude in one orbit period;
s4, taking the postures of the satellite at different time and orbit positions, which are generated by simulating the optimization result in the step S3, as a sample training network and carrying out network optimization;
and S5, utilizing the optimal neural network obtained in the step S4 to generate the target attitude of the satellite at each moment and selecting a magnetic torquer attitude control algorithm to track the attitude.
According to the technical solution provided by the embodiment of the present application, in step S1, the established environment model includes: a gravity field model, an atmosphere and wind field model, a solar radiation model, a geomagnetic field model and an unknown interference model.
According to the technical solution provided in the embodiment of the present application, in step S2, the components of the satellite attitude control subsystem model with appropriate granularity include: modeling of the attitude sensor, the attitude control computer, the attitude control actuating mechanism and corresponding algorithms thereof; wherein the attitude sensor comprises:
an angular velocity sensor: a top;
an angle sensor: sun sensors, star sensors;
a magnetic field sensor: a magnetometer;
the position sensor is a GPS;
attitude control actuating mechanism: magnetic torquer, thruster, momentum wheel.
According to the technical scheme provided by the embodiment of the present application, in step S2, the granularity of the satellite attitude control subsystem model includes: the control computer directly obtains attitude information under the track system from the environment and directly generates control torque under the system;
respectively establishing a sensor instrument coordinate system and an actuating mechanism instrument coordinate system according to the part specification document;
the sensor model acquires attitude information under the instrument coordinate system and transmits the attitude information to the control computer;
the control computer generates a control instruction and transmits the control instruction to the executing mechanism, and calculates a control moment according to the executing mechanism model;
adding uncertainty into the sensor and the execution mechanism, and processing data in the control computer;
the control computer, the sensor and the executing mechanism are connected through a bus and unpacked according to a component communication protocol package to establish a data transmission analysis model;
and selecting one granularity from the granularities of the satellite attitude control subsystem models for modeling.
According to the technical scheme provided by the embodiment of the application, in the step S3, the method for optimizing the initial attitude angle and the angular velocity of the satellite in one orbit period at different times by using the models in the steps S1 and S2 comprises the following steps: interior point method, exterior point method, genetic algorithm and particle swarm algorithm.
According to the technical scheme provided by the embodiment of the application, in the step S4, the trained network is a back propagation neural network, and the influencing factors to be preferably considered for the network include the number of layers of units, the number of neurons in each layer, the activation function of each layer and the termination condition.
According to the technical solution provided by the embodiment of the present application, in the step S4, the method for optimizing the neural network includes:
setting a proper step length to traverse to generate a large number of working conditions;
and training the network in batch by using a neural network optimization tool and automatically screening the network configuration with the minimum mean square error.
According to the technical scheme provided by the embodiment of the application, in the step S5, the selected attitude control algorithm of the magnetic torquer comprises: the method comprises PD control, LQR control, sliding mode control and fuzzy control.
According to the technical scheme provided by the embodiment of the application, in the step S5, the attitude control algorithm of the magnetic torquer is selected to track the attitude, parameters in the control algorithm need to be designed, and the parameters are optimized by utilizing the optimization algorithm.
Compared with the prior art, the beneficial effect of this application: the invention provides a method for controlling the attitude of a magnetic torquer of a microsatellite based on a neural network, which takes full consideration of the in-orbit environment of a satellite, trains and preferably selects a target attitude which enables environmental moments to be mutually counteracted as much as possible in a certain time generated by the network, and tracks the set of attitude sequences by using the magnetic torquer. Coupling among all shafts in the magnetic control is reduced, three-shaft stable attitude control is completed only by using a magnetic torquer, and a better attitude control effect can be obtained with lower energy consumption compared with a traditional method.
Drawings
Fig. 1 is a schematic overall flow chart of a microsatellite magnetic torquer attitude control method based on a neural network according to an embodiment of the present application;
fig. 2 is a control algorithm framework diagram of the method for controlling the attitude of a magnetorquer of a microsatellite based on a neural network provided by the embodiment, which is applied to a certain satellite.
Detailed Description
The following detailed description of the present application is given for the purpose of enabling those skilled in the art to better understand the technical solutions of the present application, and the description in this section is only exemplary and explanatory, and should not be taken as limiting the scope of the present application in any way.
Referring to fig. 1-2, the present embodiment provides a method for controlling an attitude of a microsatellite magnetic torquer based on a neural network, including:
s1, establishing a satellite orbit attitude dynamic model and an environment model;
specifically, a dynamic model of the satellite is established, wherein the dynamic model of the satellite comprises an orbit dynamic model and an attitude dynamic model;
establishing an orbit dynamics equation of the satellite:
the satellite orbit dynamics equation can be expressed in the geocentric inertial system in the following differential form:
wherein m is the total mass of the satellite, r is the position vector of the satellite in the geocentric inertial system, and F is the resultant force received by the satellite.
F＝F E +F S +F M +F A +F SP +F U
Wherein F E Is the gravity of the earth, F, received by the satellite S Is the gravitational force of the sun, F M Is the gravity of the moon, F A Is atmospheric resistance, F SP Is sunlight pressure, F U Other unknown forces.
Attitude dynamics equation:
by quaternions [ q ] 0 q 1 q 2 q 3 ]Representing the attitude of the satellite system in other reference systems by ω b ＝[ω xb ω yb ω zb ]Representing the component form of the angular velocity of the satellite relative to the reference frame under the satellite's own system.
The above equation is the attitude change of the satellite described by the angular velocity, i.e. the attitude kinematics equation.
Wherein M is b Is the component form of the satellite's resultant external moment in the system, I b Is a rotational inertia matrix under the system.
Wherein, I x 、I y 、I z Is the main inertia, I xy 、I yz 、I zx Is the product of inertia.
When the axis of the satellite body is the principal axis of inertia, that is, the product of inertia is 0, the attitude kinetic equation of the satellite is:
the environment model building method comprises the following steps:
gravity gradient moment:
wherein M is g In the embodiment, a JGM gravity field model, r, is used as the gravity field model, and g is a gravity field strength vector and is obtained by calculation through the gravity field model in the form of a component of a gravity gradient moment borne by a satellite in the system b The component form of the position of the satellite in the geocentric inertial system after the coordinate transformation from the inertial system to the main system.
Aerodynamic moment:
V a ＝V-V W
wherein V a The velocity of the satellite centroid relative to the atmosphere, V is the absolute velocity of the satellite under the geocentric inertial system, V W The rotation speed of the atmosphere is generally the same as the rotation direction of the earth, and is 1 to 1.5 times the rotation speed of the earth.
The aerodynamic drag experienced by the jth surface of the satellite is:
wherein n is j Unit vector in the direction of surface normal, C D Is a coefficient of resistance, S j.eff ρ is the atmospheric density, which is the effective windward area.
Atmospheric resistance to the satellite:
aerodynamic moment to which the satellite is subjected:
M A ＝r A ×F A
sunlight pressure moment:
wherein lambda is the reflectivity of the surface material of the satellite, sigma is the diffuse reflectivity of the surface material of the satellite, A is the component form of the surface area of each axis direction under the satellite system, A is the sunning area, r sb ＝[r sbx r sby r sbz ] T In the form of the component of the sun position vector in the system, p is the solar radiation pressure.
M sp ＝r P ×F sP
Wherein M is sp Is the sunlight pressure moment r received by the satellite P Is the vector from the satellite centroid to the satellite centroid.
Remanence moment:
M rm ＝m rm ×B b
wherein M is rm Is the residual magnetic moment vector, m, to which the satellite is subjected rm Is the satellite remanent magnetic moment vector, B b Is the component form of the geomagnetic induction intensity under the system.
And establishing a geomagnetic field model, wherein an international reference geomagnetic field (IGRF) is adopted to obtain spherical harmonic coefficients for constructing a magnetic potential Laplace function in the example, and the components of the magnetic induction intensity in a geographic horizontal coordinate system are obtained through recursion solution.
S2, establishing a satellite attitude control subsystem model with a proper granularity;
specifically, the attitude and orbit control subsystem of a satellite according to this embodiment is composed of an ADCS computer, an MEMS gyroscope, a star sensor, a sun sensor, a GPS, a thruster, and a magnetic torquer. In this embodiment, the simulation granularity of the component is selected according to the task requirement, and the method for selecting the simulation granularity of the component further includes that the method may be experience of a designer or iteration in a design process; and the component model sets related parameters and errors according to the product related description and unpacks according to the communication protocol package. The relevant parameters and errors of the components and the communication protocol are extremely different and can be obtained through product description documents, the principle models of the components are also extremely different, a designer needs to build a model according to own experience and by referring to relevant technical documents, and only a star sensor model used by a certain satellite is specifically given as an example in the embodiment.
The star sensor is connected with an attitude and orbit control computer through an RS422 bus;
each control cycle of the attitude and orbit control computer sends a data brief packet request instruction to the star sensor;
the star sensor returns a data short packet containing quaternion information after receiving the instruction;
and establishing a star sensor simulation model, taking the satellite attitude quaternion under the geocentric inertial system as model input, and outputting the attitude quaternion under the coordinate system of the star sensor instrument.
The instrument coordinate system o of the star sensor m X m Y m Z m The definition is as follows:
o m the origin of the coordinate system is the intersection point of the optical axis and the photosensitive surface of the image sensor, and can be generally regarded as the center of the photosensitive surface of the image sensor. Z m Directed in the direction of the optical axis, X m Directed to an external electrical connector, Y m Determined by the right hand rule. X m Y m The axis is parallel to the horizontal and vertical arrangement direction of the pixels of the image sensor.
The star sensor mounting matrix under the satellite body system is R sb ；
Q＝q 0 +q 1 i+q 2 j+q 3 k is the satellite body system oX b Y b Z b Relative to the earth's center inertial system OX i Y i Z i The attitude quaternion of (3), the earth-centered inertial system OX i Y i Z i To the body system oX b Y b Z b The coordinate transformation matrix of (a) is:
the earth's center inertia system OX i Y i Z i To the instrument coordinate system o of the star sensor m X m Y m Z m The coordinate transformation matrix of (a) is:
R si ＝R sb R bi
obtaining the instrument coordinate system o from the coordinate transformation matrix m X m Y m Z m Relative to the earth's center inertial system OX i Y i Z i The attitude quaternion of (a):
Q s ＝q s0 +q s1 i+q s2 j+q s3 k
further solve out Q s The corresponding euler angles are:
describing the measurement error of the star sensor by using the random error which obeys normal distribution, and then the star sensitivity measurement value is as follows:
wherein
Theta and psi are real Euler angles,
θ d 、ψ d in order to measure the error in the euler angle,
is the mean value of the errors measured at the euler angles,
the standard deviation of the euler angle measurement error is.
Transforming euler angle measurements into attitude quaternions:
unitizing the attitude quaternion to obtain an attitude quaternion measured value of the star sensor:
processing the quaternion data according to a star sensor communication protocol to obtain quaternion data sent to an attitude and orbit control computer:
Q so ＝Q U ·2 30
s3, optimizing the initial attitude angle and the angular speed of the satellite in one orbit period at different times by using the models in the step S1 and the step S2, so that the satellite does not actively control to return to the initial attitude in one orbit period;
specifically, if the pitch angle of the satellite can be periodically changed only under the action of the environmental moment in a period of time, the interference moments received by the satellite in the period of time can be mutually offset, and the attitude control is not influenced. Therefore, the target attitude is changed to make the target attitude angle not 0, but have different target attitude angles and angular velocities at different positions in the orbit cycle, so that if the satellite is in the desired attitude at the current moment, the satellite can be in the desired attitude corresponding to the next moment without control at the next moment, and the control requirement of the magnetic torquer on the axis can be reduced. The target track can be obtained by learning the rule of environmental change through a neural network, and the track is tracked by using simple PD control.
At an initial angular velocity delta omega other than the orbital angular velocity 0 And as an optimization variable, simulating an orbit period without controlling the attitude, and taking the pitch angle error epsilon (theta) of the satellite at the end time as an objective function value.
The objective function is:
ε(θ)＝|θ final |＝|f(Δω 0 )|
wherein f () is the process of the satellite simulation program not controlling and simulating the attitude for one orbit period, delta omega 0 To optimize the variables, | θ final And | is a module value of the attitude angle at the end of the simulation.
Optimizing by using an interior point method to obtain an optimized variable delta omega corresponding to the minimum objective function value epsilon (theta) 0 The value of (d) is the optimization result.
S4, taking the postures of the satellite at different time and orbit positions, which are generated by simulating the optimization result in the step S3, as a sample training network and carrying out network optimization;
specifically, the optimization result of the previous step is brought into a simulation model to simulate for a period to obtain target pitch angles and angular velocities at different track positions, and the target pitch angles and the target angular velocities are used as sample data.
And training by adopting a back propagation neural network.
The back propagation neural network consists of an input layer, an output layer and a plurality of hidden unit layers, wherein n of the neural network of the input layer O For each component of the input vector of the network, an individual node:
the input of the hidden unit layer is determined by the output of the previous layer (possibly the network input or the output of the previous hidden unit layer) and the weight value from the previous layer to the current layer:
wherein HI j Is the input of the jth neuron of the current hidden unit layer,
is the output of the ith neuron in the n neurons of the previous layer, omega ij Is the weight between these two neurons.
The output of a hidden unit layer is determined by the input of the layer and the activation function of the layer:
HO j ＝f(HI j )
wherein HO is j For the output of the jth neuron of the current hidden unit layer, f () represents the activation function of that layer.
The input of the output layer is determined by the output of the last hidden unit layer and the weight from the previous layer to the current layer:
wherein YI k For the kth neuron of the output layerThe input of the input data is carried out,
is the output of the jth neuron in the n neurons in the previous layer, omega jk Is the weight between these two neurons.
The output of the output layer is determined by the input of the layer and the activation function of the layer:
YO k ＝f(YI k )
[θ n ω yn ] T ＝YO
wherein YO k For the output of the kth neuron of the current hidden unit layer, f () represents the activation function of that layer.
The maximum characteristic of the back propagation neural network is that the weight of each layer is corrected according to the back propagation of network errors, and the error is calculated to obtain:
wherein d is k Is the desired output in the sample.
And if the calculated error meets the training termination condition, finishing network training, if the error does not meet the training termination condition, calculating the derivative of the error function to the weight layer by layer, and updating the weight between layers, namely performing back propagation.
Network input: the ascension point right ascension omega and the true paraxial angle f of the satellite (characterizing the orbital position of the satellite);
and (3) network output: target attitude angle theta of pitch axis n And a target angular velocity ω yn (representing a target track which enables the pitch axis environment moments to be offset mutually as much as possible);
traversing the number of hidden unit layers, the number of neurons in each layer and the activation function of each layer, processing the training network in batch, and taking the mean square error as an evaluation index. The obtained optimal network is a three-layer hidden unit layer neural network with activation functions of purelin, radbas and radbas respectively, and the number of neurons in each layer is 140, 140 and 60 respectively.
And S5, obtaining the target attitude of the optimal neural network generated satellite at each moment by utilizing the step S4, and selecting a magnetic torquer attitude control algorithm to track the attitude.
The neural network magnetic torquer controller consists of two parts, one part is a posture control link, and the neural network outputs the expected posture of the pitching axis according to the current track information. And tracking the expected attitude of the pitch axis by using PD control, and generating a magnetic moment instruction for controlling the attitude. The other part is a remanence compensation link, and a magnetic moment instruction for compensating remanence is generated according to magnetic field information and satellite remanence. The two parts of magnetic moment instructions jointly form a magnetic moment instruction which is sent to the magnetic torquer by the controller.
Taking a pitch axis as an example, tracking the optimal track calculated by the neural network, wherein the expected moment in the direction is as follows:
wherein
Is the neural network with the ascension angle and the true periapical angle of the ascension point as inputs and the target pitch angle and the angular velocity as outputs, K py 、K dy The PD control parameters are selected and verified by the designer based on experience.
Calibrating m by utilizing remanence rmb The result calculates the remanence moment:
M rm ＝m rmb ×B b
and (3) calculating the residual magnetism compensation magnetic moment which should be output by the magnetic torquer according to the calculation result of the residual magnetism moment:
further obtaining a magnetic torquer instruction magnetic moment containing a residual magnetism compensation magnetic moment and a network control magnetic moment
m c ＝m N +m con
Fig. 2 is a block diagram of a satellite neural network control system according to this embodiment.
The principles and embodiments of the present application are explained herein using specific examples, which are provided only to help understand the method and the core idea of the present application. The foregoing is only a preferred embodiment of the present application, and it should be noted that there are no specific structures which are objectively limitless due to the limited character expressions, and it will be apparent to those skilled in the art that a plurality of modifications, decorations or changes can be made without departing from the principle of the present invention, and the technical features mentioned above can be combined in a suitable manner; such modifications, variations, combinations, or adaptations of the invention in other instances, which may or may not be practiced, are intended to be within the scope of the present application.
Claims (9)
1. A microsatellite magnetic torquer attitude control method based on a neural network is characterized by comprising the following steps:
s1, establishing a satellite orbit attitude dynamic model and an environment model;
s2, establishing a satellite attitude control subsystem model with a proper granularity;
s3, optimizing the initial attitude angle and the angular speed of the satellite in one orbit period at different times by using the models in the step S1 and the step S2, so that the satellite does not actively control to return to the initial attitude in one orbit period;
s4, taking attitudes of the satellite at different time and orbit positions, generated by simulation of the optimization result in the step S3, as a sample training network and carrying out network optimization;
and S5, utilizing the optimal neural network obtained in the step S4 to generate the target attitude of the satellite at each moment and selecting a magnetic torquer attitude control algorithm to track the attitude.
2. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein the environment model established in the step S1 includes: a gravity field model, an atmospheric and wind field model, a solar radiation model, a geomagnetic field model, and an unknown interference model.
3. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in the step S2, the components of the satellite attitude control subsystem model with proper granularity comprise: modeling of the attitude sensor, the attitude control computer, the attitude control actuating mechanism and corresponding algorithms thereof; wherein the attitude sensor comprises:
an angular velocity sensor: a top;
an angle sensor: sun sensors, star sensors;
a magnetic field sensor: a magnetometer;
the position sensor is a GPS;
attitude control actuating mechanism: magnetic torquer, thruster, momentum wheel.
4. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in the step S2, the granularity of the satellite attitude control subsystem model comprises: the control computer directly obtains attitude information under the track system from the environment and directly generates control torque under the system;
establishing a sensor instrument coordinate system according to the part specification document;
the sensor model acquires attitude information under the sensor instrument coordinate system and transmits the attitude information to the control computer;
the control computer generates a control instruction and transmits the control instruction to the execution mechanism, and calculates a control moment according to the execution mechanism model;
adding uncertainty into the sensor and the actuating mechanism, and processing data in the control computer;
establishing a fine-grained component model based on the physical characteristics of the sensor and the actuating mechanism;
the control computer, the sensor and the executing mechanism are connected through a bus and unpacked according to a component communication protocol package to establish a data transmission analysis model;
and selecting one granularity from the granularities of the satellite attitude control subsystem models for modeling.
5. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in step S3, the model of step S1 and step S2 is used to optimize the initial attitude angle and angular velocity of the satellite in one orbit period at different times, and the method comprises: interior point method, exterior point method, genetic algorithm and particle swarm algorithm.
6. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in step S4, the trained network is a back propagation neural network, and the influencing factors to be preferably considered for the network include the number of layers of units, the number of neurons in each layer, the activation function of each layer and termination conditions.
7. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in the step S4, the method of optimizing the neural network comprises the following steps:
setting a proper step length to traverse to generate a large number of working conditions;
and training the network in batch by using a neural network optimization tool and automatically screening the network configuration with the minimum mean square error.
8. The attitude control method of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in the step S5, the selected attitude control algorithm of the magnetic torquer comprises: the method comprises PD control, LQR control, sliding mode control and fuzzy control.
9. The method for controlling the attitude of a microsatellite magnetic torquer based on a neural network as claimed in claim 1, wherein in the step S5, the parameters in the control algorithm are designed according to the requirement of selecting the attitude control algorithm of the magnetic torquer to track the attitude, and the parameters are optimized by utilizing the optimization algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211055986.3A CN115285380B (en) | 2022-08-30 | 2022-08-30 | Micro-satellite magnetic torquer attitude control method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211055986.3A CN115285380B (en) | 2022-08-30 | 2022-08-30 | Micro-satellite magnetic torquer attitude control method based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115285380A true CN115285380A (en) | 2022-11-04 |
| CN115285380B CN115285380B (en) | 2024-08-02 |
Family
ID=83831444
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211055986.3A Active CN115285380B (en) | 2022-08-30 | 2022-08-30 | Micro-satellite magnetic torquer attitude control method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115285380B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117674977A (en) * | 2024-01-29 | 2024-03-08 | 中国卫通集团股份有限公司 | End-to-end simulation method and system for satellite communication physical layer |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106019950A (en) * | 2016-08-09 | 2016-10-12 | 中国科学院软件研究所 | Mobile phone satellite self-adaptive attitude control method |
| RU2669481C1 (en) * | 2017-10-02 | 2018-10-11 | Акционерное общество "Научно-исследовательский институт электромеханики" | Method and device for controlling motion of a spacecraft with controlled orientation |
| CN112254723A (en) * | 2020-10-13 | 2021-01-22 | 天津津航计算技术研究所 | Small unmanned aerial vehicle MARG attitude estimation method based on adaptive EKF algorithm |
| CN113184222A (en) * | 2021-05-13 | 2021-07-30 | 上海卫星工程研究所 | Magnetic torquer signal processing method and system of satellite attitude and orbit control comprehensive test equipment |
-
2022
- 2022-08-30 CN CN202211055986.3A patent/CN115285380B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106019950A (en) * | 2016-08-09 | 2016-10-12 | 中国科学院软件研究所 | Mobile phone satellite self-adaptive attitude control method |
| RU2669481C1 (en) * | 2017-10-02 | 2018-10-11 | Акционерное общество "Научно-исследовательский институт электромеханики" | Method and device for controlling motion of a spacecraft with controlled orientation |
| CN112254723A (en) * | 2020-10-13 | 2021-01-22 | 天津津航计算技术研究所 | Small unmanned aerial vehicle MARG attitude estimation method based on adaptive EKF algorithm |
| CN113184222A (en) * | 2021-05-13 | 2021-07-30 | 上海卫星工程研究所 | Magnetic torquer signal processing method and system of satellite attitude and orbit control comprehensive test equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117674977A (en) * | 2024-01-29 | 2024-03-08 | 中国卫通集团股份有限公司 | End-to-end simulation method and system for satellite communication physical layer |
| CN117674977B (en) * | 2024-01-29 | 2024-04-16 | 中国卫通集团股份有限公司 | End-to-end simulation method and system for satellite communication physical layer |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115285380B (en) | 2024-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104898642B (en) | A kind of integration testing analogue system for Spacecraft Attitude Control algorithm | |
| CN104527994B (en) | Multi-polar cross-over becomes the track set time soon and holds position sensing tracking and controlling method | |
| CN104133479B (en) | A kind of employing single-axle air bearing table simulates the method for testing of flexible satellite three-axis attitude coupled motions | |
| Lu et al. | Real-time simulation system for UAV based on Matlab/Simulink | |
| CN112082574A (en) | Star sensor correction method and system | |
| CN102354123A (en) | Cross-platform extendible satellite dynamic simulation test system | |
| CN114580224B (en) | Distributed pneumatic fusion track coupling attitude perturbation analysis method | |
| CN104570736B (en) | A kind of kinetic parameter in-orbit identification method and apparatus of star arm coupled system | |
| Liu et al. | Attitude control for astronaut assisted robot in the space station | |
| CN113359444B (en) | Flexible spacecraft rigid-flexible coupling characteristic intelligent identification method based on neural network | |
| CN111753464B (en) | Unmanned helicopter model online learning method and system | |
| Bello et al. | Experimental verification and comparison of fuzzy and PID controllers for attitude control of nanosatellites | |
| CN115285380B (en) | Micro-satellite magnetic torquer attitude control method based on neural network | |
| Srivastava et al. | Attitude determination and control system for a leo debris chaser small satellite | |
| Ticona et al. | Attitude determination and control system for nadir pointing and detumbling using magnetorquer for 1u bolivian cubesat | |
| CN113312028A (en) | Dynamics software architecture method and system for satellite attitude and orbit control universal test | |
| CN116252969B (en) | Magnetic disk satellite off-orbit control method, device and computer storage medium | |
| Prieto et al. | A drag free control based on model predictive techniques | |
| Maghami et al. | Laser interferometer space antenna dynamics and controls model | |
| Dracopoulos et al. | Neuro-genetic adaptive attitude control | |
| Lovera | Control-oriented modelling and simulation of spacecraft attitude and orbit dynamics | |
| CN110703595B (en) | Master satellite attitude forecasting method and system of satellite-arm coupling system | |
| Wilder Smith et al. | Dynamic and Robust CubeSat Control with Machine Learning | |
| Wu et al. | Nonlinear dynamic modeling and simulation of an atmospheric re-entry spacecraft | |
| Santiago et al. | Low cost magnetic simulator for attitude control |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |