CN110542913A - Satellite coordinate estimation method and device - Google Patents
Satellite coordinate estimation method and device Download PDFInfo
- Publication number
- CN110542913A CN110542913A CN201910743877.2A CN201910743877A CN110542913A CN 110542913 A CN110542913 A CN 110542913A CN 201910743877 A CN201910743877 A CN 201910743877A CN 110542913 A CN110542913 A CN 110542913A
- Authority
- CN
- China
- Prior art keywords
- satellite
- error
- neural network
- weight
- abc
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/40—Correcting position, velocity or attitude
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/42—Determining position
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides a satellite coordinate estimation method and a device, wherein the method comprises the following steps: obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite; obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm; inputting the collected historical data into an ABC-DR model to obtain an optimal error compensation value; and adding the rough satellite position and the optimal error compensation value to obtain an accurate satellite position.
Description
Technical Field
The invention belongs to the field of GPS satellite navigation processing, and particularly relates to a satellite coordinate estimation method and device.
Background
in recent years, the GPS industry has been rapidly developed and widely used in various fields, and a GPS receiver needs to know the spatial position of each visible satellite at an arbitrary time during positioning, and the spatial position of the satellite that changes with time is called the orbit of the satellite. The orbit error directly affects the positioning accuracy of the receiver and therefore the accuracy of the satellite orbit is a critical task in GPS positioning. The coordinate position of the satellite is obtained through ephemeris, the precision of the broadcast ephemeris is low, and the current broadcast ephemeris is based on orbital elements. Different parameter ephemeris models need to be designed according to the orbit characteristics. And due to the characteristics of the complexity of the space perturbation factor, the high dynamic of the model parameters and the like, the complexity of the mathematical modeling is seen to be general. In many papers on simulated satellite orbits, it can be found that the broadcast ephemeris orbital error obviously has periodicity, so that besides establishing a more precise ephemeris parameter model, the orbit error can be used for trying to correct the positioning coordinates of the satellite.
the neural network technology is rapidly developed in recent years, the basic structure of the artificial neural network can imitate the human brain, reflect a plurality of basic characteristics of the human brain function, and can adapt to the environment, summarize the rule and complete a certain operation, identification or control process. The method has strong self-learning, self-adaptability and nonlinearity, so that the error prediction has outstanding functions in many fields.
although the neural network has strong nonlinear mapping capability and fault tolerance capability, the error feedback algorithm is substantially a gradient descent algorithm, so that the model is sensitive to the weight and the threshold and is easy to fall into a local optimal solution.
Disclosure of Invention
in view of the above disadvantages of the prior art, an object of the present invention is to provide a method and an apparatus for estimating satellite coordinates, which fuse a broadcast ephemeris and a neural network algorithm and well avoid the complexity and adaptability of mathematical modeling.
to achieve the above and other related objects, the present invention provides a satellite coordinate estimation method, including:
Obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite;
obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm;
inputting the collected historical data into an ABC-DR model to obtain an optimal error compensation value;
and adding the rough satellite position and the optimal error compensation value to obtain an accurate satellite position.
optionally, the obtaining the ABC-DR model through the DR neural network and the artificial bee colony algorithm includes:
S21 learning the non-linear error of the ephemeris orbit by using a DR neural network;
s22, using an ABC algorithm to optimize parameters of the DR neural network;
S23 retraining the optimized parameters as initial values of the DR algorithm;
s24 repeating the steps S21-S23 until the optimal solution is obtained.
alternatively, a unipolar type Sigmoid function is used as an activation function of the DR neural network,
where ρ (·) is the activation function of the hidden neuron and ε is the jitter parameter.
optionally, a BP learning algorithm is used for the DR neural network, and the error function is:
where j (k) represents the error function and d (k) represents the original value of the sample, representing the actual output of the sample.
optionally, the obtaining of the ABC-DR model through the DR neural network and the artificial bee colony algorithm specifically includes:
Step1, selecting training sample data, using 16 parameters related to satellite orbit information as network input, using position error compensation value as network output, randomly generating connection weight of network input layer and middle layer, middle layer and output layer, feedback weight wij of middle layer,
Step2 calculating the actual output of the sample
step3, calculating a sample error value J;
step4, if J < epsilon meets the error requirement, ending the training, turning to Step11, otherwise, entering Step 5;
step5, calculating the weight variation and the threshold variation of each layer;
step6 recalculating the connection weights wij,
Step7, recalculating the actual output and error according to the new weight and the sample data;
Step8, if J < epsilon meets the error requirement, ending the training and turning to Step 11;
Step9, taking the threshold and the weight of each layer as the initial solution of the ABC algorithm, setting parameters MCN and Limit, and taking an error value J as a target function of the quality fit of the solution corresponding to the nectar amount of the food source;
Step10, taking the weight and the threshold generated according to the ABC algorithm as the initial weight and the threshold of the next training of the DR neural network, and turning to Step 5;
step11, after training, outputting the weight wij meeting the training precision,
and Step12, obtaining a final result according to the finally determined weight and the input value selected in Step 1.
the present invention also provides a satellite coordinate estimation apparatus, including:
The first position estimation value is used for obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite;
the model creating module is used for obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm;
the compensation value calculation module is used for inputting the collected historical data into the ABC-DR model to obtain an optimal error compensation value;
and the second satellite position estimation value is used for adding the rough satellite position and the optimal error compensation value to obtain an accurate satellite position.
optionally, the obtaining the ABC-DR model through the DR neural network and the artificial bee colony algorithm includes:
S21 learning the non-linear error of the ephemeris orbit by using a DR neural network;
s22, using an ABC algorithm to optimize parameters of the DR neural network;
s23 retraining the optimized parameters as initial values of the DR algorithm;
s24 repeating the steps S21-S23 until the optimal solution is obtained.
alternatively, a unipolar type Sigmoid function is used as an activation function of the DR neural network,
where ρ (·) is the activation function of the hidden neuron and ε is the jitter parameter.
Optionally, a BP learning algorithm is used for the DR neural network, and the error function is:
where j (k) represents the error function and d (k) represents the original value of the sample, representing the actual output of the sample.
optionally, the obtaining of the ABC-DR model through the DR neural network and the artificial bee colony algorithm specifically includes:
Step1, selecting training sample data, using 16 parameters related to satellite orbit information as network input, using position error compensation value as network output, randomly generating connection weight of network input layer and middle layer, middle layer and output layer, feedback weight wij of middle layer,
step2 calculating the actual output of the sample
Step3, calculating a sample error value J;
step4, if J < epsilon meets the error requirement, ending the training, turning to Step11, otherwise, entering Step 5;
step5, calculating the weight variation and the threshold variation of each layer;
step6 recalculating the connection weights wij,
step7, recalculating the actual output and error according to the new weight and the sample data;
step8, if J < epsilon meets the error requirement, ending the training and turning to Step 11;
step9, taking the threshold and the weight of each layer as the initial solution of the ABC algorithm, setting parameters MCN and Limit, and taking an error value J as a target function of the quality fit of the solution corresponding to the nectar amount of the food source;
Step10, taking the weight and the threshold generated according to the ABC algorithm as the initial weight and the threshold of the next training of the DR neural network, and turning to Step 5;
Step11, after training, outputting the weight wij meeting the training precision,
And Step12, obtaining a final result according to the finally determined weight and the input value selected in Step 1.
As described above, the method and apparatus for estimating satellite coordinates according to the present invention have the following advantages:
according to the invention, on the basis of the rule of the periodicity of the broadcast ephemeris orbit error, various difficulties of mathematical modeling are avoided, and good orbit precision can be obtained through strong self-learning adaptivity of a neural network.
Drawings
To further illustrate the description of the present invention, the following detailed description of the embodiments of the present invention is provided with reference to the accompanying drawings. It is appreciated that these drawings are merely exemplary and are not to be considered limiting of the scope of the invention.
FIG. 1 is a schematic diagram of a satellite coordinate estimation method according to the present invention;
fig. 2 is a diagram showing a structure of a diagonal neural network according to the present invention.
Detailed Description
the embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
it should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention, and the components related to the present invention are only shown in the drawings rather than drawn according to the number, shape and size of the components in actual implementation, and the type, quantity and proportion of the components in actual implementation may be changed freely, and the layout of the components may be more complicated.
As shown in fig. 1, the present invention provides a method for estimating satellite coordinates, comprising:
S1, obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite;
S2, obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm;
S3, inputting the collected historical data into an ABC-DR model to obtain an optimal error compensation value;
s4 adds the coarse satellite position to the optimal error compensation value to obtain an accurate satellite position.
The satellite coordinate estimation method provided by the invention applies a DR neural network (diagonal recurrent neural network) algorithm to satellite orbit coordinate calculation, combines the DR neural network with an ABC algorithm (artificial bee colony algorithm), and discusses an unconventional method for increasing the prediction precision.
In step S1, the coarse satellite positions are obtained by the following steps.
step1: calculating a normalized time tk, wherein tk is t-toe;
step2: calculating the average angular velocity n of the satellite; wherein n is n0+ Δ n;
step3: calculating a mean approximate point angle Mk of the signal emission time, wherein Mk is M0+ ntk;
Step4: calculating a near point angle Ek of a signal transmitting moment, wherein the Ek is obtained through Ej being M + essin (Ej-1) according to es and Mk;
Step5: calculating a true near point angle vk of the signal emission moment;
step6: calculating the rising intersection angular distance phi k of the signal transmitting moment, wherein phi k is vk + omega;
Step7: calculating perturbation correction terms delta uk, delta rk, delta ik of signal emission time, wherein
δu＝Csin(2Φ)+Ccos(2Φ)
δr＝Csin(2Φ)+Ccos(2Φ)
δi＝Csin(2Φ)+Ccos(2Φ)
Step8: calculating the perturbation corrected lift-off point uk, satellite radial length rk and orbit inclination ik, wherein
u＝Φ+δu
r＝a(1-ecosE)+δr
step9: calculating the position (x 'k, y' k) of the satellite in the orbital plane at the moment of signal transmission, wherein x 'k is rkcosuk, and y' k is rksinuk;
Step10: calculating the rising point right ascension omega k of the signal emission moment, wherein
step11: calculating coordinates (xk, yk, zk) of the satellite in a WGS-84 Earth-centered-Earth-fixed rectangular coordinate system (XT, YT, ZT), wherein
x＝x′cosΩ-y′cosisinΩ
y＝x′sinΩ+y′cosicosΩ
z＝y′sini
the above unknown parameters are obtained from the broadcast ephemeris.
In various recurrent neural networks, the diagonal recurrent neural network can embody the system dynamics, the dynamic neural network can process the problem which can not be overcome by the static neural network, the diagonal recurrent neural network can be obtained by improving the common multilayer forward neural network, the structure is simple, and the network can process the dynamic information by itself only increasing the internal feedback channel. Its simple connectivity and the general nature of recursive networks. So that the gradient descent algorithm can be used directly to train different parameters in the network. A DR neural network is used to learn such non-linear errors of ephemeris orbits.
The diagonal neural network can be simplified to an output neuron comprising an input layer, a hidden layer and an output layer. The structure is shown in fig. 2.
the neurons in the hidden layer are delayed to feed back, so the network has dynamic characteristics, and the input of the network is x (k) ═ u1(k), u2(k), …, um (k), y (k) ] T; h neurons are in the hidden layer, the input of the hidden layer neuron i is si (k), and the output of the hidden layer neuron i is the output of the network at the moment k + 1; the connection weight of the jth input neuron and the ith hidden layer neuron component is a threshold value of the hidden layer neuron i, a feedback connection weight of the hidden layer neuron i and a connection weight between the hidden layer neuron i and the output neuron.
the input and output relationships are:
h(k)＝ρ(s(k)) (2)
in the formula: rho (·) is an activation function of a hidden neuron, and a unipolar Sigmoid function is taken as the activation function: where epsilon is the jitter parameter.
the diagonal recurrent neural network uses a BP learning algorithm, and the error function is as follows:
The network parameters are adjusted as follows:
(1) Adjustment of join weights for input and hidden neurons using a gradient descent algorithm
in the formula: i 1,2, h, j 1,2, q 0, α is a momentum factor, η is a step size and represents an adjustment value of the weight at the k-th time, q j (k-1) represents a partial derivative of the hi (k-1) pair, and ρ' (si (k)) represents a derivative of the activation function with si (k) as an argument.
(2) the threshold value adjusting algorithm comprises the following steps:
in the formula: i ═ 1,2, ·, h, ri (0) · 0, ri (k-1) denotes the partial derivatives of the hi (k-1) pairs.
(3) feedback weight value adjusting algorithm
in the formula: i ═ 1,2, ·, h, pi (0) · 0, pi (k-1) denotes the partial derivatives of the hi (k-1) pairs.
(4) output weight value adjusting algorithm
In the formula: i-1, 2, h.
because the neural network algorithm is sensitive to the weight and is easy to fall into a local optimal solution, the ABC algorithm is adopted to complete the optimization of the DR neural network.
in the ABC algorithm, bees are led to search for food sources, the following bees select the food sources according to information, and the detecting bees randomly search for the food sources. The amount of nectar in the food source corresponds to the mass fit of the corresponding solution, which is calculated according to the following equation:
fi is the error function J, abs (fi) represents the absolute value of fi.
The method comprises the following steps:
step1, initializing all parameters of the algorithm;
Step2, the leading bees are in one-to-one correspondence with the honey sources, the honey source information is updated according to a formula, and meanwhile, the nectar amount of the honey sources is determined, xij represents the jth optimization parameter, the ith honey source solution, k is any honey source which is not equal to i and is any random number between [ -1,1 ];
step3: selecting a honey source by the observation bee according to information provided by the leading bee by adopting a certain strategy, updating honey source information according to a formula of Step2, and simultaneously determining the nectar amount of the honey source;
Step4, determining scout bees, and according to the search for a new honey source, rand (0,1) is an arbitrary random number between (0,1), which represents a new solution substituted by the scout bees when the local optimal solution is trapped in Step2, and max and min represent the maximum value and the minimum value of the residual honey source after the ith honey source is removed;
Step5, memorizing the best honey source so far;
step6, judging whether the termination condition is satisfied, if not, repeating Step2-Step 5;
the ABC algorithm can accelerate the convergence of the algorithm, reduce the oscillation of the algorithm and jump out the local optimal solution. Therefore, the ABC algorithm is combined with the DR neural network algorithm, namely the ABC-DR algorithm is used for learning data.
the basic idea of the ABC-DR algorithm is as follows: when the DR algorithm has a low convergence speed or falls into a local optimal solution, the ABC algorithm is started to optimize parameters of the DR algorithm, and the optimized result is used as an initial value of the DR algorithm for retraining, so that the two algorithms are alternately used until a global optimal solution is calculated. The specific algorithm steps are as follows:
Step1, selecting training sample data, wherein 7 parameters (1 reference time t, 3 position x, y, z and velocity vector vx, vy, vz information) related to satellite orbit information and 9 relatively sensitive perturbation parameters (earth aspheric attraction, solar attraction, lunar attraction, solid tide, ocean tide, atmospheric tide, relativistic effect, gravity gradient moment and earth magnetic field moment) are used as network output, and the position error parameters are used as network output. Randomly generating the connection weight of the network input layer and the middle layer, the middle layer and the output layer, the feedback weight wij of the middle layer,
step2 calculating the actual output of the sample according to equations (1) to (3)
step3, calculating a cost function J of the square sum of the sample errors according to the formula (4);
step4, if J < epsilon meets the error requirement, ending the training, turning to Step11, otherwise, entering Step 5;
Step5, calculating the weight variation and the threshold variation of each layer according to the (5) to (8);
step6 recalculating the connection weights wij,
step7, calculating the actual input and error again according to the formula according to the new weight and the sample data;
step8, if J < epsilon meets the error requirement, ending the training and turning to Step 11;
step9, taking the threshold and each layer weight as the initial solution of the ABC algorithm, setting parameters MCN (maximum iteration number) and Limit (abandon search threshold), and taking the error J as the target function of the formula (9);
And Step10, calling an ABC algorithm to solve the optimal solution. Taking the weight and the threshold generated by the ABC algorithm as the initial weight and the threshold of the next training of the DR neural network, and turning to Step 5;
step11, after training, outputting the weight wij meeting the training precision,
And Step12, obtaining a final result according to the finally determined weight and the input value selected in Step 1.
The invention provides satellite coordinate estimation of a DR neural network based on a broadcast ephemeris and an artificial bee colony, and the algorithm can find the optimal error value in an iterative mode without complex mathematical modeling. The satellite positioning precision is greatly improved. Provides a new idea for further development of the GPS navigation field.
the present invention also provides a satellite coordinate estimation apparatus, including:
The first position estimation value is used for obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite;
The model creating module is used for obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm;
The compensation value calculation module is used for inputting the collected historical data into the ABC-DR model to obtain an optimal error compensation value;
and the second satellite position estimated value is used for adding the rough satellite position and the optimal error compensation value to obtain an accurate satellite position. Wherein, a rough satellite position is obtained according to the broadcast ephemeris of the GPS satellite, comprising the following steps:
Step1: calculating a normalized time tk, wherein tk is t-toe, wherein toe is ephemeris reference time;
step2: calculating the average angular velocity n of the satellite; where n is n0+ Δ n, where n0 is the (average) angular velocity of a hypothetical satellite travelling in a circumferential orbit and Δ n is the average angular velocity correction value;
Step3: calculating a mean-near-point angle Mk of the signal emission time, wherein Mk is M0+ ntk, and M0 is an initial value of the mean-near-point angle;
step4: calculating a near point angle Ek of a signal transmitting moment, wherein the Ek is obtained through Ej being M + essin (Ej-1) according to es and Mk, wherein the es is the eccentricity of an elliptical orbit, because the process is an iteration process, k and k-1 respectively represent the current and last near point angle values, and M is Mk;
step5: calculating a true near point angle vk of the signal emission moment;
step6: calculating the rising intersection point angular distance phi k of the signal transmitting moment, wherein phi k is vk + omega, and omega is the orbit geodesic angular distance;
Step7: calculating perturbation correction terms delta uk, delta rk, delta ik of signal emission time, wherein
δu＝Csin(2Φ)+Ccos(2Φ)
δr＝Csin(2Φ)+Ccos(2Φ)
δi＝Csin(2Φ)+Ccos(2Φ)
Cus is a rising point angular distance sine harmonic correction amplitude, a Crs track radius sine harmonic correction amplitude, and Cis is a track inclination angle sine harmonic correction amplitude;
Step8: calculating the perturbation corrected lift-off point uk, satellite radial length rk and orbit inclination ik, wherein
u＝Φ+δu
r＝a(1-ecosE)+δr
Wherein as is the satellite orbit major semi-axis, i0 is the orbit inclination when toe, which is the change rate of the orbit inclination to time;
Step9: calculating the position (x 'k, y' k) of the satellite in the orbital plane at the moment of signal transmission, wherein x 'k is rkcosuk, and y' k is rksinuk;
Step10: calculating rising intersection point right ascension Ω k at the time of signal emission, wherein Ω 0 is orbit rising intersection point right ascension when the time in the week is equal to 0, is the change rate of the orbit rising intersection point right ascension to time, and is the value of the earth rotation angular velocity constant;
Step10: calculating coordinates (xk, yk, zk) of the satellite in a WGS-84 Earth-centered-Earth-fixed rectangular coordinate system (XT, YT, ZT), wherein
x＝x′cosΩ-y′cosisinΩ
y＝x′sinΩ+y′cosicosΩ
z＝y′sini
wherein ik is the rotation angle of the X 'axis, and Ω k is the rotation angle of the Z' axis;
the above unknown parameters are obtained from the broadcast ephemeris.
In some embodiments, the obtaining the ABC-DR model through the DR neural network and the artificial bee colony algorithm includes:
s21 learning the non-linear error of the ephemeris orbit by using a DR neural network;
s22, using an ABC algorithm to optimize parameters of the DR neural network;
s23 retraining the optimized parameters as initial values of the DR algorithm;
s24 repeating the steps S21-S23 until the optimal solution is obtained.
In some embodiments, a unipolar Sigmoid function is used as the activation function for the DR neural network,
Where ρ (·) is the activation function of the hidden neuron and ε is the jitter parameter.
In some embodiments, a BP learning algorithm is used for the DR neural network, and the error function is:
where j (k) represents the error function and d (k) represents the original value of the sample, representing the actual output of the sample.
In some embodiments, the obtaining the ABC-DR model by the DR neural network and the artificial bee colony algorithm specifically includes:
Step1, selecting training sample data, using 16 parameters (1 reference time t, 3 position x, y, z and velocity vector vx, vy, vz information, earth aspheric attraction, solar attraction, moon attraction, solid tide, ocean tide, atmospheric tide, relativistic effect, gravity gradient moment, earth magnetic field moment) of satellite orbit information as network input and output, using position error compensation value as network output, randomly generating connection weight of network input layer and middle layer, middle layer and output layer, feedback weight wij of middle layer, and calculating the position error compensation value,
Step2 calculating the actual output of the sample
step3, calculating a sample error value J;
Step4, if J < epsilon meets the error requirement, ending the training, turning to Step11, otherwise, entering Step 5;
step5, calculating the weight variation and the threshold variation of each layer;
Step6 recalculating the connection weights wij,
Step7, recalculating the actual input and error according to the new weight and the sample data;
Step8, if J < epsilon meets the error requirement, ending the training and turning to Step 11;
step9, taking the threshold and each layer weight as the initial solution of the ABC algorithm, setting parameters MCN (maximum iteration number) and Limit (abandon search threshold), and taking an error value J as a target function of the quality fit of the solution corresponding to the nectar amount of the food source;
step10, taking the weight and the threshold generated according to the ABC algorithm as the initial weight and the threshold of the next training of the DR neural network, and turning to Step 5;
Step11, after training, outputting the weight wij meeting the training precision,
step12, finding the final result according to the final determined weight and the input value selected in Step1
It should be noted that, because the embodiment of the apparatus portion and the embodiment of the method portion correspond to each other, please refer to the description of the embodiment of the method portion for the content of the embodiment of the apparatus portion, which is not repeated here.
the invention also provides a storage medium storing a computer program which, when executed by a processor, performs the method as described above.
The present invention also provides an electronic terminal, comprising:
a memory for storing a computer program;
a processor for executing the computer program stored by the memory to cause the apparatus to perform the aforementioned method.
the computer program comprises computer program code which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may comprise any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a Random Access Memory (RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, etc.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
the memory may be an internal storage unit or an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital Card (SD), a Flash memory Card (Flash Card), and the like. Further, the memory may also include both an internal storage unit and an external storage device. The memory is used for storing the computer program and other programs and data. The memory may also be used to temporarily store data that has been or will be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
in the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
in the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal device and method may be implemented in other ways. For example, the above-described embodiments of the apparatus/terminal device are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (10)
1. a method for estimating satellite coordinates, the method comprising:
Obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite;
obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm;
Inputting the collected historical data into an ABC-DR model to obtain an optimal error compensation value;
And adding the rough satellite position and the optimal error compensation value to obtain an accurate satellite position.
2. the satellite coordinate estimation method according to claim 1, wherein the obtaining of the ABC-DR model through the DR neural network and the artificial bee colony algorithm comprises:
S21 learning the non-linear error of the ephemeris orbit by using a DR neural network;
s22, using an ABC algorithm to optimize parameters of the DR neural network;
S23 retraining the optimized parameters as initial values of the DR algorithm;
s24 repeating the steps S21-S23 until the optimal solution is obtained.
3. The satellite coordinate estimation method of claim 2, wherein a unipolar type Sigmoid function is used as an activation function of the DR neural network,
where ρ (·) is the activation function of the hidden neuron and ε is the jitter parameter.
4. the satellite coordinate estimation method of claim 3, wherein a BP learning algorithm is used for the DR neural network, and the error function is:
where j (k) represents the error function and d (k) represents the original value of the sample, representing the actual output of the sample.
5. the satellite coordinate estimation method according to claim 3, wherein the obtaining of the ABC-DR model through the DR neural network and the artificial bee colony algorithm specifically comprises:
step1, selecting training sample data, using 16 parameters related to satellite orbit information as network input, using position error compensation value as network output, randomly generating connection weight of network input layer and middle layer, middle layer and output layer, feedback weight wij of middle layer,
Step2 calculating the actual output of the sample
step3, calculating a sample error value J;
Step4, if J < epsilon meets the error requirement, ending the training, turning to Step11, otherwise, entering Step 5;
Step5, calculating the weight variation and the threshold variation of each layer;
step6 recalculating the connection weights wij,
step7, recalculating the actual output and error according to the new weight and the sample data;
step8, if J < epsilon meets the error requirement, ending the training and turning to Step 11;
step9, taking the threshold and the weight of each layer as the initial solution of the ABC algorithm, setting parameters MCN and Limit, and taking an error value J as a target function of the quality fit of the solution corresponding to the nectar amount of the food source;
Step10, taking the weight and the threshold generated according to the ABC algorithm as the initial weight and the threshold of the next training of the DR neural network, and turning to Step 5;
step11, after training, outputting the weight wij meeting the training precision,
And Step12, obtaining a final result according to the finally determined weight and the input value selected in Step 1.
6. A satellite coordinate estimation apparatus, characterized by comprising:
the first position estimation value is used for obtaining a rough satellite position according to the broadcast ephemeris of the GPS satellite;
the model creating module is used for obtaining an ABC-DR model through a DR neural network and an artificial bee colony algorithm;
the compensation value calculation module is used for inputting the collected historical data into the ABC-DR model to obtain an optimal error compensation value;
and the second satellite position estimation value is used for adding the rough satellite position and the optimal error compensation value to obtain an accurate satellite position.
7. The satellite coordinate estimation apparatus according to claim 6, wherein the obtaining of the ABC-DR model through the DR neural network and the artificial bee colony algorithm includes:
S21 learning the non-linear error of the ephemeris orbit by using a DR neural network;
s22, using an ABC algorithm to optimize parameters of the DR neural network;
s23 retraining the optimized parameters as initial values of the DR algorithm;
s24 repeating the steps S21-S23 until the optimal solution is obtained.
8. The satellite coordinate estimation apparatus according to claim 7, wherein a unipolar type Sigmoid function is used as an activation function of the DR neural network,
where ρ (·) is the activation function of the hidden neuron and ε is the jitter parameter.
9. The satellite coordinate estimation apparatus according to claim 8, wherein a BP learning algorithm is used for the DR neural network, and the error function is:
where j (k) represents the error function and d (k) represents the original value of the sample, representing the actual output of the sample.
10. The satellite coordinate estimation apparatus according to claim 8, wherein the obtaining of the ABC-DR model by the DR neural network and the artificial bee colony algorithm specifically includes:
Step1, selecting training sample data, using 16 parameters related to satellite orbit information as network input, using position error compensation value as network output, randomly generating connection weight of network input layer and middle layer, middle layer and output layer, feedback weight wij of middle layer,
step2 calculating the actual output of the sample
Step3, calculating a sample error value J;
Step4, if J < epsilon meets the error requirement, ending the training, turning to Step11, otherwise, entering Step 5;
Step5, calculating the weight variation and the threshold variation of each layer;
step6 recalculating the connection weights wij,
Step7, recalculating the actual output and error according to the new weight and the sample data;
step8, if J < epsilon meets the error requirement, ending the training and turning to Step 11;
Step9, taking the threshold and the weight of each layer as the initial solution of the ABC algorithm, setting parameters MCN and Limit, and taking an error value J as a target function of the quality fit of the solution corresponding to the nectar amount of the food source;
step10, taking the weight and the threshold generated according to the ABC algorithm as the initial weight and the threshold of the next training of the DR neural network, and turning to Step 5;
Step11, after training, outputting the weight wij meeting the training precision,
and Step12, obtaining a final result according to the finally determined weight and the input value selected in Step 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910743877.2A CN110542913A (en) | 2019-08-13 | 2019-08-13 | Satellite coordinate estimation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910743877.2A CN110542913A (en) | 2019-08-13 | 2019-08-13 | Satellite coordinate estimation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110542913A true CN110542913A (en) | 2019-12-06 |
Family
ID=68711478
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910743877.2A Pending CN110542913A (en) | 2019-08-13 | 2019-08-13 | Satellite coordinate estimation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110542913A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101535832A (en) * | 2006-11-10 | 2009-09-16 | 高通股份有限公司 | Method and apparatus for position determination with extended SPS orbit information |
| CN103886368A (en) * | 2014-03-26 | 2014-06-25 | 南京航空航天大学 | Satellite accurate orbit prediction method |
| CN104834285A (en) * | 2015-03-30 | 2015-08-12 | 福州大学 | Realization method of diagonal recurrent neural network controller in multiple platforms |
| EP3467662A1 (en) * | 2017-10-09 | 2019-04-10 | Accenture Global Solutions Limited | Verification of applications that utilize artificial intelligence |
| CN109977871A (en) * | 2019-03-27 | 2019-07-05 | 中国人民解放军战略支援部队航天工程大学 | A kind of Satellite Targets recognition methods based on wideband radar data and GRU neural network |
| US20190245564A1 (en) * | 2018-02-08 | 2019-08-08 | Raytheon Company | Preprocessor for device navigation |
-
2019
- 2019-08-13 CN CN201910743877.2A patent/CN110542913A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101535832A (en) * | 2006-11-10 | 2009-09-16 | 高通股份有限公司 | Method and apparatus for position determination with extended SPS orbit information |
| CN103886368A (en) * | 2014-03-26 | 2014-06-25 | 南京航空航天大学 | Satellite accurate orbit prediction method |
| CN104834285A (en) * | 2015-03-30 | 2015-08-12 | 福州大学 | Realization method of diagonal recurrent neural network controller in multiple platforms |
| EP3467662A1 (en) * | 2017-10-09 | 2019-04-10 | Accenture Global Solutions Limited | Verification of applications that utilize artificial intelligence |
| US20190245564A1 (en) * | 2018-02-08 | 2019-08-08 | Raytheon Company | Preprocessor for device navigation |
| CN109977871A (en) * | 2019-03-27 | 2019-07-05 | 中国人民解放军战略支援部队航天工程大学 | A kind of Satellite Targets recognition methods based on wideband radar data and GRU neural network |
Non-Patent Citations (3)
| Title |
|---|
| DENG XIAOLONG: ""Nonlinear Identification Based on Diagonal Recurrent Neural Network and Particle Filter"", 《2009 FIFTH INTERNATIONAL CONFERENCE ON NATURAL COMPUTATION》 * |
| 张欣: ""基于DRNN的传感器的建模研究"", 《电脑开发与应用》 * |
| 李卫华 等: ""基于人工蜂群的BP神经网络算法"", 《计算机系统应用》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106803271B (en) | Camera calibration method and device for visual navigation unmanned aerial vehicle | |
| CN111156987A (en) | Inertia/astronomical combined navigation method based on residual compensation multi-rate CKF | |
| Jouaber et al. | Nnakf: A neural network adapted kalman filter for target tracking | |
| JPH10124658A (en) | Method for correcting image distortion of camera by utilizing neural network | |
| CN116245015A (en) | Data change trend prediction method and system based on deep learning | |
| CN114612556A (en) | Training method of visual inertial odometer model, pose estimation method and pose estimation device | |
| Sommer et al. | Continuous-time estimation of attitude using b-splines on lie groups | |
| Yousuf et al. | Information fusion of GPS, INS and odometer sensors for improving localization accuracy of mobile robots in indoor and outdoor applications | |
| CN109443355B (en) | Visual-inertial tight coupling combined navigation method based on self-adaptive Gaussian PF | |
| Mainenti-Lopes et al. | Design of a nonlinear controller for a rigid-flexible satellite using multi-objective Generalized Extremal Optimization with real codification | |
| Praczyk | Using evolutionary neural networks to predict spatial orientation of a ship | |
| CN112648999B (en) | Brain-like navigation method based on multi-scale grid cell path integral | |
| CN110542913A (en) | Satellite coordinate estimation method and device | |
| Zheng et al. | An rnn-based learnable extended kalman filter design and application | |
| WO2020247721A1 (en) | Systems and methods to improve geo-referencing using a combination of magnetic field models and in situ measurements | |
| Malleswaran et al. | A hybrid approach for GPS/INS integration using Kalman filter and IDNN | |
| CN115102868A (en) | Web service QoS prediction method based on SOM clustering and depth self-encoder | |
| CN114357872A (en) | Ship motion black box identification modeling and motion prediction method based on stacking model fusion | |
| Agogino et al. | Efficient credit assignment through evaluation function decomposition | |
| Liu et al. | A camera calibration method based on neural network optimized by genetic algorithm | |
| Chiang et al. | An ANN–RTS smoother scheme for accurate INS/GPS integrated attitude determination | |
| CN110782033A (en) | AGV positioning method based on fuzzy neural network | |
| Deng et al. | Measurement model and observability analysis for optical flow-aided inertial navigation | |
| Alndiwee et al. | A Novel Bundle Adjustment Approach Based on Guess-Aided and Angle Quantization Multiobjective Particle Swarm Optimization (GAMOPSO) for 3D Reconstruction Applications | |
| CN114812565B (en) | Dynamic navigation method based on artificial intelligence network |
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 |