CN109459586A - A kind of unmanned plane accelerometer scaling method based on LM algorithm - Google Patents
A kind of unmanned plane accelerometer scaling method based on LM algorithm Download PDFInfo
- Publication number
- CN109459586A CN109459586A CN201811477562.XA CN201811477562A CN109459586A CN 109459586 A CN109459586 A CN 109459586A CN 201811477562 A CN201811477562 A CN 201811477562A CN 109459586 A CN109459586 A CN 109459586A
- Authority
- CN
- China
- Prior art keywords
- accelerometer
- formula
- algorithm
- coefficient
- unmanned plane
- 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
- G01P—MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION, OR SHOCK; INDICATING PRESENCE, ABSENCE, OR DIRECTION, OF MOVEMENT
- G01P21/00—Testing or calibrating of apparatus or devices covered by the preceding groups
Abstract
The invention belongs to air vehicle technique fields, disclose a kind of unmanned plane accelerometer scaling method based on LM algorithm, and selecting three calibration factors all is 1, other 9 unknowm coefficients are all 0, ε 10‑6, λ 0.1, μ 10；It calculatesConstruct Equation With Damping；The formula acquired solves increment δ.The unmanned plane accelerometer scaling method that the present invention designs is not needed first using high-precision turntable and GPS information, outdoor indoors, whether there is or not the occasions of turntable can demarcate accelerometer at any time.Secondly, this method should fully consider the error source of accelerometer, a comprehensive error model is constructed, to guarantee stated accuracy.Data are filtered out using filtering algorithm during the calibration process again, avoid the interference of random error.Calculation amount finally is reduced using the iterative algorithm of optimization when solving calibration coefficient, reduces prover time.
Description
Technical field
The invention belongs to air vehicle technique field more particularly to a kind of unmanned plane accelerometer calibration sides based on LM algorithm
Method.
Background technique
Currently, the prior art commonly used in the trade is such that
3 axis MEMS accelerometer is small in size, light weight and cheap.Overwhelming majority Small and micro-satellite is all adopted at present
With it as attitude-measuring sensor.The measurement accuracy of mems accelerometer has UAV Attitude resolving bigger
It influences.However accelerometer precision is lower, needs to improve its measurement accuracy using certain scaling method.Traditional acceleration
Degree meter calibration generally utilizes high-precision three-axle table, a kind of entitled MEMS inertial sensor automatic batch calibration of existing application
The patent of method and system is exactly to demarcate accelerometer using three-axle table.But this method is needed using high-precision turn
Platform cannot demarcate at any time accelerometer in the occasion of no turntable, it is therefore desirable to a kind of calibration method independent of turntable.It is existing
A kind of entitled patent based on the acceleration planned immunization turntable scaling method of genetic algorithm of application can be demarcated not against turntable
Accelerometer, but this method use genetic algorithm construction cost function be multidimensional nonlinear function, which is asked
There are the problems such as computationally intensive, prover time is long for solution minimum, it is difficult in the unmanned plane of application.A kind of entitled base of existing application
Accelerometer is demarcated using GPS velocity information in the patent of the accelerometer scaling method of GPS velocity information, but this method
It needs to use GPS information, the occasion for interior etc. without GPS signal, this method is not just available, and can not just be realized indoors yet
The accelerometer calibration of unmanned plane.The patent of the scaling method of the entitled accelerometer of existing application, device and electronic equipment is not
Accelerometer can be also demarcated dependent on turntable and GPS information, but the error model of this method only considered three reference axis
Scale factor errors and zero point offset error do not account for the non-orthogonal error between three reference axis, establish in this way
Error model it is fairly simple.In addition, the data of this method measurement are the data of six different predeterminated positions, need artificial
Accelerometer is placed in preset position, otherwise can not be demarcated, it is clear that this method is more demanding.There are model letters for this method in a word
List and the not high problem of stated accuracy.
In conclusion problem of the existing technology is:
(1) it needs in the prior art using high-precision turntable, accelerometer cannot be demarcated at any time in the occasion of no turntable；
Using genetic algorithm, accelerometer can be demarcated not against turntable, but genetic algorithm is computationally intensive, prover time is long.
(2) it needs to use GPS information in the prior art, the occasion for interior etc. without GPS signal is not just available, nothing
Method realizes the accelerometer calibration of unmanned plane indoors.
(3) error model in the prior art is fairly simple and stated accuracy is not high.
Solve the difficulty and meaning of above-mentioned technical problem:
It can not just be demarcated first not against high precision turntable for accelerometer and reference standard is provided, lacked GPS information and then lose
Part effective information is lost.Although increasing error model in addition, existing error model is simple but small using the model calculation amount
Complexity may also bring along computationally intensive problem while increasing precision.
For the difficult point of the above prior art problem, of the invention can be in the condition independent of high-precision turntable
Under, accelerometer is demarcated by multi collect measured value, and do not need using to GPS, the equipment needed in this way
Few, indoor and outdoor occasion can use.In addition, this method is in the complexity appropriate for increasing error model, while using
Method, which can be reduced, calculates the time, avoids while calculation amount increases and improves stated accuracy.
Summary of the invention
In view of the problems of the existing technology, the unmanned plane accelerometer calibration based on LM algorithm that the present invention provides a kind of
Method.
The invention is realized in this way a kind of unmanned plane accelerometer scaling method based on LM algorithm, specifically include with
Lower step:
Step 1: parameter needed for initialization iterative process.Selecting three calibration factors all is 1, other 9 unknowm coefficients are all
It is 0, terminating control constant ε is 10-6, damped coefficient λ be 0.1 and damping scaling multiple μ be 10；Then it calculates
Step 2: the formula for asking local derviation to obtain 12 unknown numbers to be asked by measured value and respectively finds out Jacobian matrix,
Construct Equation With Damping；
Step 3: increment δ is solved by the formula that step 2 acquires.
Further, in step 3, if | f (βk+ δ) | < εk, then β is enabledk+1=βk+δ；If | δ | < ε stops iteration,
Otherwise output is as a result, enable λk+1=λk/ μ, goes to step 2.
Further, in step 3, if | f (βk+1)|≥εk, then λ is enabledk+1=λk*μ；Again solution Equation With Damping obtains increment
δ, return step 1.
Further, LM algorithm specifically:
Do not consider that temperature drift and the installation error of accelerometer, the calibrating patterns of accelerometer can indicate are as follows:
In formula, amx、amyAnd amzFor the measured value of three axis of accelerometer, arx、aryAnd arzIt is accelerometer three
The true value of axis, a0x、aoyAnd a0zFor the zero drift error of three axis of accelerometer, kx、kyAnd kzFor accelerometer three
The calibration factor k of a axisxy、kyx、kxz、kzx、kyz、kzyFor the non-orthogonal error coefficient of three between centers of accelerometer；The calibration of acceleration
Exactly find out above 12 unknown numbers；Acceleration evaluation after formula (1) expansion can must be calibrated
When the static placement of accelerometer, because there are relationships as follows between the accelerometer output valve after correction
In formula, g indicates local gravitational acceleration.Wushu (1) substitution formula (3) can obtain the equation about 12 unknown numbers；
This 12 unknown numbers can be found out using the equation and corresponding measured value, to complete to calibrate；
LM algorithm is mainly used in least square curve fitting problem, that is: given m group data are to (xi,yi), it solves
The parameter beta of curve f (x, β) model, so that data are minimum to the quadratic sum of the range deviation to curve, i.e.,
LM algorithm is an iterative algorithm；.Initial value is assigned firstly the need of to parameter beta；In each step iterative step, parameter beta quilt
New estimation β+δ is replaced；.Functional value f (xi, β+δ) and available linearization is approximately
f(xi,β+δ)≈f(xi,β)+Jiδ \*MERGEFORMAT(5)
In formula, JiGradient for function f relative to β, expression formula are
The J when minimum variance and S (β) are minimizediIt is zero, S (xi, β+δ) and it can be expressed as with first approximation
Above formula expansion can be obtained
S (β+δ)=| | y-f (β)-J δ | |2
=[y-f (β)-J δ]T[y-f(β)-Jδ]
=[y-f (β)]T[y-f(β)]-[y-f(β)]TJδ
-(Jδ)T[y-f(β)]+δTJTJδ
=[y-f (β)]T[y-f(β)]-2[y-f(β)]TJδ+δTJTJδ\*MERGEFORMAT(8)
It is that zero can obtain that S (β+δ), which seeks local derviation relative to δ and enables result,
(JTJ) δ=JT[y-f(β)]\*MERGEFORMAT(9)
In formula, J is the Jacobian matrix of function；The formula is system of linear equations, can be with direct solution δ；I.e.
δ=(JTJ)-1JT[y-f(β)] \*MERGEFORMAT(10)
Levenberg proposes an Equation With Damping on this basis, i.e.,
δ=(JTJ+λI)-1JT[y-f(β)] \*MERGEFORMAT(11)
In formula, λ is known damped coefficient, and I is unit matrix；
Abnormal data elimination is carried out using Schottky photodetectors；If the absolute value of the difference of certain measured value and average value is greater than mark
The product of quasi- difference and Xiao Weile coefficient, then reject the measured value, be expressed as follows with formula:
In formula, xiFor measured value to be tested,For the average value of all measured values, SxFor the standard deviation of all measured values,
ωnFor Xiao Weile coefficient, specific value can be obtained by inquiring Xiao Weile coefficient table, can be with when requiring is not very stringent
It is calculated according to following formula:
ωn=1+0.4ln (n) * MERGEFORMAT (13)
Smothing filtering algorithm can reduce the random noise of acquisition data and eliminate the interference of abnormal sample signal；Using
5 points of weighting moving average methods are filtered sampled data, and weight coefficient is determined according to the principle of least square, really
It protects smoothed out data and initial data is approached with minimum variance, i.e.,
In formula, p (ti) it is polynomial fitting, the polynomial fitting of 5 moving weighted averages are as follows:
In formula, ajFor multinomial coefficient, there are following relationships between the coefficient and minimum variance Q:
5 moving weighted average filtering iteration formula according to the principle of least square, after fitting are as follows:
Smothing filtering can be carried out to acquisition data by formula (17).
By formula (3) it is found that the nonlinear model of system can be expressed as
Above formula asks local derviation that can obtain 12 unknown numbers to be asked respectively
In formula, intermediate variable R is defined as follows:
The matrix of formula (19) composition is exactly the Jacobian matrix of function, and choosing suitable damped coefficient λ can be by formula (11)
Find out the increment δ of iterative process.
In conclusion advantages of the present invention and good effect are as follows:
The unmanned plane accelerometer scaling method that the present invention designs does not need to believe using high-precision turntable and GPS first
Breath, it is outdoor indoors, whether there is or not the occasions of turntable can demarcate accelerometer at any time.Add secondly, this method should fully consider
The error source of speedometer constructs a comprehensive error model, to guarantee stated accuracy.It is using what this method calibrated
Number is as shown in table 1, and 6 of the extra existing patent as mentioned above of calibration coefficient.It is big that Fig. 5 and Fig. 6 demonstrates the data calibrated
Measurement accuracy is improved greatly.Data are filtered out using filtering algorithm during the calibration process again, avoid the interference of random error.Fig. 3
It is exactly to filter out data using filtering algorithm with Fig. 4, as seen from the figure, is obviously become using filtering algorithm post-acceleration measurement magnitude
Steadily, the interference of random error is avoided.Calculation amount finally is reduced using the iterative algorithm of optimization when solving calibration coefficient,
Reduce prover time.
Detailed description of the invention
Fig. 1 is the unmanned plane accelerometer scaling method flow chart provided in an embodiment of the present invention based on LM algorithm.
Fig. 2 is navigation module schematic diagram provided in an embodiment of the present invention.
Fig. 3 is X-axis provided in an embodiment of the present invention smothing filtering Contrast on effect schematic diagram upward vertically.
Fig. 4 is Z axis provided in an embodiment of the present invention smothing filtering Contrast on effect schematic diagram upward vertically.
Fig. 5 is calibrated each axle acceleration component schematic diagram provided in an embodiment of the present invention.
Fig. 6 is calibrated each axle acceleration component schematic diagram provided in an embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
Application principle of the invention is described in detail with reference to the accompanying drawing.
As shown in Figure 1, the unmanned plane accelerometer scaling method provided in an embodiment of the present invention based on LM algorithm, specific to wrap
Include following steps:
S101: selecting three calibration factors all is 1, other 9 unknowm coefficients are all 0, and terminating control constant ε is 10-6, resistance
Buddhist nun's coefficient lambda is 0.1 and damping scaling multiple μ is 10；Then it calculates
S102: the formula for asking local derviation to obtain 12 unknown numbers to be asked by measured value and respectively finds out Jacobian matrix, structure
Produce Equation With Damping；
S103: increment δ is solved by the formula that step 2 acquires.
In step S103, if provided in an embodiment of the present invention | f (βk+ δ) | < εk, then β is enabledk+1=βk+δ；If | δ | <
ε then stops iteration, exports as a result, otherwise enabling λk+1=λk/ μ, goes to step 2.
In step S103, if provided in an embodiment of the present invention | f (βk+1)|≥εk, then λ is enabledk+1=λk*μ；Again solution damping
Equation obtains increment δ, return step 1.
LM algorithm provided in an embodiment of the present invention specifically:
The error source of three axis accelerometer can be mainly divided into three classes, be fixed error, random error and temperature respectively
Degree drift.Wherein fixed error mainly includes zero migration, calibration factor and the non-orthogonal error of three between centers.If not considering to add
The calibrating patterns of the temperature drift of speedometer and installation error, accelerometer can indicate are as follows:
In formula, amx、amyAnd amzFor the measured value of three axis of accelerometer, arx、aryAnd arzIt is accelerometer three
The true value of axis, a0x、aoyAnd a0zFor the zero drift error of three axis of accelerometer, kx、kyAnd kzFor accelerometer three
The calibration factor k of a axisxy、kyx、kxz、kzx、kyz、kzyFor the non-orthogonal error coefficient of three between centers of accelerometer；The calibration of acceleration
Exactly find out above 12 unknown numbers；Acceleration evaluation after formula (1) expansion can must be calibrated
When the static placement of accelerometer, because there are relationships as follows between the accelerometer output valve after correction
In formula, g indicates local gravitational acceleration.Wushu (1) substitution formula (3) can obtain the equation about 12 unknown numbers；
This 12 unknown numbers can be found out using the equation and corresponding measured value, to complete to calibrate；
LM algorithm is mainly used in least square curve fitting problem, that is: given m group data are to (xi,yi), it solves
The parameter beta of curve f (x, β) model, so that data are minimum to the quadratic sum of the range deviation to curve, i.e.,
LM algorithm is an iterative algorithm；Initial value is assigned firstly the need of to parameter beta；In each step iterative step, parameter beta is new
Estimation β+δ replace；Functional value f (xi, β+δ) and available linearization is approximately
f(xi,β+δ)≈f(xi,β)+Jiδ \*MERGEFORMAT(5)
In formula, JiGradient for function f relative to β, expression formula are
The J when minimum variance and S (β) are minimizediIt is zero, S (xi, β+δ) and it can be expressed as with first approximation
Above formula expansion can be obtained
S (β+δ)=| | y-f (β)-J δ | |2
=[y-f (β)-J δ]T[y-f(β)-Jδ]
=[y-f (β)]T[y-f(β)]-[y-f(β)]TJδ
-(Jδ)T[y-f(β)]+δTJTJδ
=[y-f (β)]T[y-f(β)]-2[y-f(β)]TJδ+δTJTJδ\*MERGEFORMAT(8)
It is that zero can obtain that S (β+δ), which seeks local derviation relative to δ and enables result,
(JTJ) δ=JT[y-f(β)]\*MERGEFORMAT(9)
In formula, J is the Jacobian matrix of function；The formula is system of linear equations, can be with direct solution δ；I.e.
δ=(JTJ)-1JT[y-f(β)] \*MERGEFORMAT(10)
Levenberg proposes an Equation With Damping on this basis, i.e.,
δ=(JTJ+λI)-1JT[y-f(β)] \*MERGEFORMAT(11)
In formula, λ is known damped coefficient, and I is unit matrix；
Damped coefficient is adjusted when each iteration；If variance reduction is too fast, lesser damped coefficient, algorithm can be used
Just close to Gauss-Newton Methods, and if the residual error that provides of iteration is insufficient, damped coefficient can be increased, algorithm just close to
Gradient descent method；So each iteration of LM algorithm can be regarded as finding a suitable damped coefficient to improve algorithm effect
Rate；It can be said that compared with gauss-newton method and gradient descent method；LM algorithm avoids while guaranteeing calibration accuracy
The problem of gauss-newton method Singular Value, arithmetic speed is faster than gradient descent method again；So LM algorithm better than gauss-newton method and
Gradient descent method；When unmanned aerial vehicle control system has certain operational capability, LM algorithm can be accelerated applied to unmanned plane
Spend the calibration of meter.
Calibrating accolerometer needs to acquire the output valve of accelerometer under different conditions, in order to reduce the dry of random error
It disturbs, it is necessary first to which acquisition data are pre-processed；Preprocessing process is broadly divided into two steps, and the first step is excluding outlier, the
Two steps are smoothing processing；Commonly judge that mainly there are the quasi- side of 3 ∑s, the quasi- side Xiao Weile and Grubbs test method in the quasi- side of exceptional value
Deng；These three methods respectively have advantage and disadvantage, and the present invention selects Schottky photodetectors to carry out abnormal data elimination；If certain measured value and flat
The absolute value of the difference of mean value is greater than the product of standard deviation and Xiao Weile coefficient, then rejects the measured value, be expressed as follows with formula:
In formula, xiFor measured value to be tested,For the average value of all measured values, SxFor the standard deviation of all measured values,
ωnFor Xiao Weile coefficient, specific value can be obtained by inquiring Xiao Weile coefficient table, can be with when requiring is not very stringent
It is calculated according to following formula:
ωn=1+0.4ln (n) * MERGEFORMAT (13)
Smothing filtering algorithm can reduce the random noise of acquisition data and eliminate the interference of abnormal sample signal；Using
5 points of weighting moving average methods are filtered sampled data, and weight coefficient is determined according to the principle of least square, really
It protects smoothed out data and initial data is approached with minimum variance, i.e.,
In formula, p (ti) it is polynomial fitting, the polynomial fitting of 5 moving weighted averages are as follows:
In formula, ajFor multinomial coefficient, there are following relationships between the coefficient and minimum variance Q:
5 moving weighted average filtering iteration formula according to the principle of least square, after fitting are as follows:
Smothing filtering can be carried out to acquisition data by formula (17).
By formula (3) it is found that the nonlinear model of system can be expressed as
Above formula asks local derviation that can obtain 12 unknown numbers to be asked respectively
In formula, intermediate variable R is defined as follows:
The matrix of formula (19) composition is exactly the Jacobian matrix of function, and choosing suitable damped coefficient λ can be by formula (11)
Find out the increment δ of iterative process.
It elaborates below with reference to having to test to application principle of the invention；
Experiment 1；
In order to which the accelerometer error model and LM algorithm of verifying above-mentioned are suitable for actual unmanned plane accelerometer calibration.
As shown in Fig. 2, experiment uses Navigation of Pilotless Aircraft module diagram.The navigation module includes MPU9250 chip, the core
Piece is integrated with the sensors such as three axis accelerometer, three-axis gyroscope and three axle magnetometer.This experiment is added to three axis therein
Speedometer is calibrated.In order to guarantee the precision of LM algorithm, at least need to acquire six kinds of different accelerometer posture informations, and
And need this six kinds of postures that should cover the positive negative direction of each reference axis substantially.Compare for the ease of analysis, once makes this
Six times acquisition posture informations be X-axis upward, X-axis downward, Y-axis upward, Y-axis downward, Z axis upward and Z axis downward, every kind of appearance
State down-sampling number is 100 times.
In experimentation, acceleration under the different gesture modes that STM32 chip passes through the transmission of IIC interface navigation module
Each axis original measurement value is counted, then STM32 chip is filtered initial data first with 5 weighted moving averages,
Then accelerometer is calibrated using LM algorithm, finally by serial ports by calibration result and the original measurement of accelerometer
Value after the completion of value and calibration is sent to host computer.
X-axis is had chosen respectively to be shown for upward with Z axis upward.First by X-axis vertically upward for, to acceleration
The output valve of meter samples 100 times, and the output of three axis of smothing filtering fore-aft acceleration meter is as shown in Figure 3.
As shown in figure 3, X-axis smothing filtering Contrast on effect schematic diagram upward vertically.
Again by Z axis vertically upward for, the output valve of accelerometer is sampled 100 times, three axis of fore-aft acceleration meter is filtered
Output it is as shown in Figure 4.
As shown in figure 4, vertically upward with X-axis, Z axis downwards when, filtering algorithm equally has good effect,
Pretreated data can be used directly.
It is calculated after the completion of data prediction using LM iterative algorithm.The calibration result of accelerometer is as shown in table 1.
Table 1: the calibration result of accelerometer
First by the X-axis of accelerometer straight up for, respectively to calibration before and calibration after accelerometer output valve
100 samplings are carried out, the 3-axis acceleration measured value for calibrating front and back is as shown in Figure 5
As shown in figure 5, the precision compared with before calibration of the acceleration evaluation after calibration is significantly improved, with six position methods
It compares, the precision of LM algorithm is also higher.
Again by the Z axis of accelerometer straight up for, respectively to calibration before and calibration after accelerometer output valve into
Row 100 times samplings, the 3-axis acceleration measured value for calibrating front and back are as shown in Figure 6
As shown in fig. 6, on X axis, accelerometer output valve after calibration is either with original value compared to still
Compared with six position methods, precision is all significantly improved.The algorithm has both gauss-newton method and gradient descent method, and both are non-
The advantages of linear algorithm.It avoids the problem of gauss-newton method matrix can not invert, and can be by adjusting damped coefficient λ
To increase arithmetic speed.So LM algorithm is suitable for the calibration of unmanned plane accelerometer, and accelerometer can be significantly improved
Measurement accuracy.The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within spirit and principle.
Claims (6)
1. a kind of unmanned plane accelerometer scaling method based on LM algorithm, which is characterized in that nobody based on LM algorithm
Machine accelerometer scaling method the following steps are included:
Step 1: selecting three calibration factors all is 1, other 9 unknowm coefficients are all 0, and terminating control constant ε is 10-6, damping
Coefficient lambda is 0.1 and damping scaling multiple μ is 10；Then it calculates
Step 2: the formula for asking local derviation to obtain 12 unknown numbers to be asked by measured value and respectively finds out Jacobian matrix, constructs
Equation With Damping out；
Step 3: increment δ is solved by the formula that step 2 acquires.
2. as described in claim 1 based on the unmanned plane accelerometer scaling method of LM algorithm, which is characterized in that the step
In three, if | f (βk+ δ) | < εk, then β is enabledk+1=βk+δ；If | δ | < ε stops iteration, exports as a result, otherwise enabling λk+1
=λk/ μ, goes to step 2.
3. as described in claim 1 based on the unmanned plane accelerometer scaling method of LM algorithm, which is characterized in that the step
In three, if | f (βk+1)|≥εk, then λ is enabledk+1=λk*μ；Again solution Equation With Damping obtains increment δ, return step one.
4. as described in claim 1 based on the unmanned plane accelerometer scaling method of LM algorithm, which is characterized in that the LM is calculated
Method specifically:
Do not consider that temperature drift and the installation error of accelerometer, the calibrating patterns of accelerometer indicate are as follows:
In formula, amx、amyAnd amzFor the measured value of three axis of accelerometer, arx、aryAnd arzFor three axis of accelerometer
True value, a0x、aoyAnd a0zFor the zero drift error of three axis of accelerometer, kx、kyAnd kzFor three axis of accelerometer
Calibration factor kxy、kyx、kxz、kzx、kyz、kzyFor the non-orthogonal error coefficient of three between centers of accelerometer；The calibration of acceleration is exactly
Find out above 12 unknown numbers；Acceleration evaluation after must calibrating:
When the static placement of accelerometer, because there are relationships as follows between the accelerometer output valve after correction:
In formula, g indicates local gravitational acceleration, obtains the equation about 12 unknown numbers；Utilize the equation and corresponding measurement
Value can find out this 12 unknown numbers, complete calibration.
5. as described in claim 1 based on the unmanned plane accelerometer scaling method of LM algorithm, which is characterized in that calculated based on LM
The unmanned plane accelerometer scaling method of method assigns initial value to parameter beta；In each step iterative step, parameter beta is by new estimation β+δ generation
It replaces；Functional value f (xi, β+δ) and available linearization is approximately:
f(xi,β+δ)≈f(xi,β)+Jiδ；
In formula, JiGradient for function f relative to β, expression formula are as follows:
The J when minimum variance and S (β) are minimizediIt is zero, S (xi, β+δ) and first approximation is expressed as；
Above formula expansion can be obtained:
S (β+δ)=| | y-f (β)-J δ | |2
=[y-f (β)-J δ]T[y-f(β)-Jδ]
=[y-f (β)]T[y-f(β)]-[y-f(β)]TJδ
-(Jδ)T[y-f(β)]+δTJTJδ
=[y-f (β)]T[y-f(β)]-2[y-f(β)]TJδ+δTJTJδ；
It is that zero can obtain that S (β+δ), which seeks local derviation relative to δ and enables result:
(JTJ) δ=JT[y-f(β)]；
In formula, J is the Jacobian matrix of function；The formula be system of linear equations, direct solution δ:
δ=(JTJ)-1JT[y-f(β)]；
Levenberg proposes an Equation With Damping on this basis:
δ=(JTJ+λI)-1JT[y-f(β)]；
In formula, λ is known damped coefficient, and I is unit matrix；
Abnormal data elimination is carried out using Schottky photodetectors；If the absolute value of the difference of certain measured value and average value is greater than standard deviation
With the product of Xiao Weile coefficient, then the measured value is rejected, is expressed as follows with formula:
In formula, xiFor measured value to be tested,For the average value of all measured values, SxFor the standard deviation of all measured values, ωnFor
Xiao Weile coefficient, specific value can be obtained by inquiring Xiao Weile coefficient table, when requiring is not very stringent, according to following public affairs
Formula calculates:
ωn=1+0.4ln (n)；
Sampled data is filtered using 5 points of weighting moving average methods, weight coefficient is according to the principle of least square
It determines, it is ensured that smoothed out data approach initial data with minimum variance:
In formula, p (ti) it is polynomial fitting, the polynomial fitting of 5 moving weighted averages are as follows:
p(ti)=a0+a1ti+a2ti 2；
In formula, ajFor multinomial coefficient, there are following relationships between the coefficient and minimum variance Q:
5 moving weighted average filtering iteration formula according to the principle of least square, after fitting are as follows:
Nonlinear model indicates are as follows:
Above formula asks local derviation that can obtain 12 unknown numbers to be asked respectively:
In formula, intermediate variable R is defined as follows:
Choose the increment δ that damped coefficient λ finds out iterative process.
6. a kind of nothing using the unmanned plane accelerometer scaling method described in Claims 1 to 5 any one based on LM algorithm
It is man-machine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811477562.XA CN109459586A (en) | 2018-12-05 | 2018-12-05 | A kind of unmanned plane accelerometer scaling method based on LM algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811477562.XA CN109459586A (en) | 2018-12-05 | 2018-12-05 | A kind of unmanned plane accelerometer scaling method based on LM algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109459586A true CN109459586A (en) | 2019-03-12 |
Family
ID=65612376
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811477562.XA Pending CN109459586A (en) | 2018-12-05 | 2018-12-05 | A kind of unmanned plane accelerometer scaling method based on LM algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109459586A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110390697A (en) * | 2019-07-11 | 2019-10-29 | 浙江大学 | A kind of millimetre-wave radar based on LM algorithm and camera combined calibrating method |
| CN111913440A (en) * | 2020-07-29 | 2020-11-10 | 深圳市汇川技术股份有限公司 | Machining speed optimization method for machining equipment, machining equipment and storage medium |
| CN112180122A (en) * | 2020-09-28 | 2021-01-05 | 中国人民解放军海军航空大学 | MEMS accelerometer turntable-free calibration method based on improved drosophila optimization algorithm |
| CN112710470A (en) * | 2020-12-10 | 2021-04-27 | 沈阳航空航天大学 | Self-adaptive wavelet threshold MEMS gyroscope noise reduction method based on maneuver identification |
| CN112781613A (en) * | 2020-12-02 | 2021-05-11 | 普宙飞行器科技（深圳）有限公司 | Calibration method of unmanned aerial vehicle sensor |
| CN113157728A (en) * | 2021-02-23 | 2021-07-23 | 北京科技大学 | Method for identifying circulation working condition of underground diesel carry scraper |
| CN114105056A (en) * | 2021-11-02 | 2022-03-01 | 杭州远视智能科技有限公司 | High-cargo-level forklift safety operation control system and method |
| CN114526756A (en) * | 2022-01-04 | 2022-05-24 | 华南理工大学 | Unmanned aerial vehicle airborne multi-sensor correction method and device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101025824A (en) * | 2007-03-30 | 2007-08-29 | 东南大学 | Calibrating method based on fixed parameters and variable parameters for three-dimensional scanning system |
| CN101286234A (en) * | 2007-04-12 | 2008-10-15 | 佳能株式会社 | Information processing apparatus and information processing method |
| CN101950426A (en) * | 2010-09-29 | 2011-01-19 | 北京航空航天大学 | Vehicle relay tracking method in multi-camera scene |
| US20140274016A1 (en) * | 2013-03-15 | 2014-09-18 | General Motors Llc | Wirelessly provisioning a vehicle telematics unit |
| WO2014182246A1 (en) * | 2013-05-07 | 2014-11-13 | Singapore University Of Technology And Design | A method and/ or system for magnetic localization |
-
2018
- 2018-12-05 CN CN201811477562.XA patent/CN109459586A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101025824A (en) * | 2007-03-30 | 2007-08-29 | 东南大学 | Calibrating method based on fixed parameters and variable parameters for three-dimensional scanning system |
| CN101286234A (en) * | 2007-04-12 | 2008-10-15 | 佳能株式会社 | Information processing apparatus and information processing method |
| CN101950426A (en) * | 2010-09-29 | 2011-01-19 | 北京航空航天大学 | Vehicle relay tracking method in multi-camera scene |
| US20140274016A1 (en) * | 2013-03-15 | 2014-09-18 | General Motors Llc | Wirelessly provisioning a vehicle telematics unit |
| WO2014182246A1 (en) * | 2013-05-07 | 2014-11-13 | Singapore University Of Technology And Design | A method and/ or system for magnetic localization |
Non-Patent Citations (1)
| Title |
|---|
| YUAN ZHONG 等: "A New Drone Accelerometer Calibration Method", 《第37届中国控制会议论文集》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110390697A (en) * | 2019-07-11 | 2019-10-29 | 浙江大学 | A kind of millimetre-wave radar based on LM algorithm and camera combined calibrating method |
| CN110390697B (en) * | 2019-07-11 | 2021-11-05 | 浙江大学 | Millimeter wave radar and camera combined calibration method based on LM algorithm |
| CN111913440A (en) * | 2020-07-29 | 2020-11-10 | 深圳市汇川技术股份有限公司 | Machining speed optimization method for machining equipment, machining equipment and storage medium |
| CN112180122A (en) * | 2020-09-28 | 2021-01-05 | 中国人民解放军海军航空大学 | MEMS accelerometer turntable-free calibration method based on improved drosophila optimization algorithm |
| CN112781613A (en) * | 2020-12-02 | 2021-05-11 | 普宙飞行器科技（深圳）有限公司 | Calibration method of unmanned aerial vehicle sensor |
| CN112710470A (en) * | 2020-12-10 | 2021-04-27 | 沈阳航空航天大学 | Self-adaptive wavelet threshold MEMS gyroscope noise reduction method based on maneuver identification |
| CN112710470B (en) * | 2020-12-10 | 2022-12-27 | 沈阳航空航天大学 | Self-adaptive wavelet threshold MEMS gyroscope noise reduction method based on maneuver identification |
| CN113157728A (en) * | 2021-02-23 | 2021-07-23 | 北京科技大学 | Method for identifying circulation working condition of underground diesel carry scraper |
| CN113157728B (en) * | 2021-02-23 | 2024-03-19 | 北京科技大学 | Method for identifying circulation working conditions of underground diesel scraper |
| CN114105056A (en) * | 2021-11-02 | 2022-03-01 | 杭州远视智能科技有限公司 | High-cargo-level forklift safety operation control system and method |
| CN114526756A (en) * | 2022-01-04 | 2022-05-24 | 华南理工大学 | Unmanned aerial vehicle airborne multi-sensor correction method and device and storage medium |
| CN114526756B (en) * | 2022-01-04 | 2023-06-16 | 华南理工大学 | Unmanned aerial vehicle-mounted multi-sensor correction method, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109459586A (en) | A kind of unmanned plane accelerometer scaling method based on LM algorithm | |
| CN106767917B (en) | A kind of oblique redundant inertial navigation system calibrated error model modelling approach | |
| CN109489853B (en) | High-precision multichannel platinum resistor temperature measurement module and method based on constant current source | |
| US5166882A (en) | System for calibrating a gyro navigator | |
| CN110006460A (en) | Star sensor and magnetometer relative Calibration method and system | |
| CN109188026A (en) | The deep learning method of automatic Calibration suitable for mems accelerometer | |
| EP3542131A1 (en) | Self-calibrating inertial measurement system and method | |
| CN113916256A (en) | Calibration method for triaxial MEMS gyroscope combined inertial measurement unit | |
| CN107588785B (en) | Star sensor internal and external parameter simplified calibration method considering image point error | |
| WO2020164206A1 (en) | Calibration method for gravity gradiometer of rotating accelerometer | |
| CN110361031A (en) | A kind of IMU population parameter error quick calibrating method theoretical based on backtracking | |
| CN108562305A (en) | A kind of quick thick scaling method in inertia/five position of astronomy deep integrated navigation system installation error | |
| CN113916257B (en) | Calibration method for triaxial MEMS (micro-electromechanical systems) addition-calculation combined inertial measurement unit | |
| US6594559B2 (en) | Iterative method of aircraft sideslip compensation for multi-function probe air data systems | |
| Gu et al. | Design and implementation of attitude and heading reference system with extended Kalman filter based on MEMS multi-sensor fusion | |
| CN113433600B (en) | Method for calibrating installation error angle of gravimeter | |
| CN105571666B (en) | Flow-compensated method and compensation device, flow sensor | |
| Zhong et al. | A new drone accelerometer calibration method | |
| CN115727874A (en) | Temperature error integrated calibration and compensation method for electronic compass sensor | |
| CN111780787B (en) | MEMS inertial measurement unit calibration method based on optical fiber inertia assistance | |
| CN114705184A (en) | Nine-axis attitude sensor integrated intelligent error compensation method based on neural network | |
| CN108415244B (en) | Multi-degree-of-freedom electrostatic suspension system geometric symmetry approximation method | |
| CN112325902B (en) | Method for establishing system-level online calibration reference coordinate system of gravimeter inertial component | |
| CN111928844B (en) | Model system of MEMS gyroscope on AGV application | |
| CN112698416B (en) | Gravitational acceleration measuring method and system suitable for strapdown inertial navigation system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190312 |