CN110749337B - MIMU error compensation method based on deep neural network - Google Patents
MIMU error compensation method based on deep neural network Download PDFInfo
- Publication number
- CN110749337B CN110749337B CN201910961889.2A CN201910961889A CN110749337B CN 110749337 B CN110749337 B CN 110749337B CN 201910961889 A CN201910961889 A CN 201910961889A CN 110749337 B CN110749337 B CN 110749337B
- Authority
- CN
- China
- Prior art keywords
- mimu
- output
- neural network
- error
- mems
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 11
- 230000001133 acceleration Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 7
- 230000007787 long-term memory Effects 0.000 claims description 4
- 230000006403 short-term memory Effects 0.000 claims description 4
- 238000010438 heat treatment Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 abstract 1
- 239000011159 matrix material Substances 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- 238000001816 cooling Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000005476 size effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C25/00—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
- G01C25/005—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass initial alignment, calibration or starting-up of inertial devices
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Manufacturing & Machinery (AREA)
- Gyroscopes (AREA)
Abstract
The invention discloses a MIMU error compensation method based on a deep neural network. And then, carrying out MIMU data acquisition aiming at the MIMU error model, and making a MIMU error data set. Finally, building an LSTM neural network model, training the neural network by adopting the manufactured data set, and continuously probing and adjusting the network structure; and predicting the MIMU error through the trained neural network, and using the network predicted value for MIMU error compensation. According to the invention, the LSTM neural network is adopted to carry out error compensation on the MIMU, so that the traditional multi-position calibration method is replaced, and the accuracy and the efficiency of the MIMU error compensation are improved.
Description
Technical Field
The invention relates to a MIMU error compensation method based on a deep neural network, and belongs to the technical field of MIMU output error compensation.
Background
The MEMS inertial sensor has the advantages of small volume, low power consumption, low cost, mass production and the like, and is gradually and widely applied to the fields of aerospace, national defense, consumer electronic products and the like. The MEMS accelerometer and the gyroscope respectively measure the acceleration and the angular velocity of the sensor in the sensitive axis direction, and the micro inertial measurement unit (Micro Inertial Measurement Unit, MIMU) is formed by three-dimensional configuration of the accelerometer and the gyroscope. The MEMS inertial device has large measurement error and is easily influenced by environmental factors, and in addition, non-orthogonal error and temperature drift caused by an integration process during three-dimensional orthogonal configuration form an important error source of MIMU, and a corresponding error model is required to be established to compensate the error so as to improve the MIMU precision.
The traditional MIMU calibration method needs to carry out position test, speed test and temperature test, and independently calibrates all inertial sensors according to the procedures of normal temperature and then high and low temperature, which is extremely tedious and takes huge time. And the MEMS device has low precision and low zero offset repeatability, and the performance of the MEMS device is increased and decreased along with time, so that a larger error exists between an error model and laboratory calibration.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the MIMU error compensation method based on the deep neural network is used for mining the characteristics hidden in time sequence data, constructing the deep neural network and compensating the strong nonlinear error of MIMU output.
The invention adopts the following technical scheme for solving the technical problems:
a MIMU error compensation method based on a deep neural network comprises the following steps:
step 1, establishing an error model of MIMU, analyzing influence factors causing MIMU output errors according to the error model, and determining input variables and output variables of a deep and long-term memory neural network;
step 2, acquiring specific data of the input variable and the output variable in the MIMU working state according to the input variable and the output variable determined in the step 1, and constructing a data set;
step 3, establishing a deep long short-term memory neural network prediction model, and training the deep long-term memory neural network prediction model by utilizing the data set constructed in the step 2 to obtain a trained prediction model;
and 4, acquiring MIMU output data in real time, calculating an output error predicted value by using the trained predicted model, and performing error compensation on the MIMU output data by using the output error predicted value.
As a preferred embodiment of the present invention, the step 1 specifically includes:
respectively establishing angular rate errors epsilon of three orthogonal MEMS gyroscopes a And acceleration error v of three orthogonal MEMS accelerometers a ModelAccording to the respective error models, analyzing influence factors causing output errors of the MEMS gyroscope and the MEMS accelerometer, thereby determining input variables and output variables of the deep long-short time memory neural network;
the error models of the MEMS gyroscope and the MEMS accelerometer are as follows:
wherein ε 0a And delta 0a Constant zero offset vector epsilon of gyroscope and accelerometer respectively Ta And delta Ta The drift vectors of the gyroscope and the accelerometer with zero offset along with the temperature respectively,is a coefficient matrix->Coefficient of omega a For the angular velocity vector of the gyroscope output, +.>Is a coefficient matrix->Coefficients of f a Acceleration vector output for accelerometer, +.>Is a coefficient matrix->Coefficients of-> Is a coefficient matrix->Coefficient epsilon of (a) ra And delta ra A=x, y, z, which is a random error.
As a preferable scheme of the invention, the input variable and the output variable of the deep long short-time memory neural network are specifically as follows:
the input variables of the deep long short-time memory neural network comprise the output of the MEMS gyroscope, the output of the MEMS accelerometer, the temperature of the MEMS gyroscope and the temperature of the MEMS accelerometer;
the output variables of the deep long-short time memory neural network comprise the output error predicted value of the MEMS gyroscope and the output error predicted value of the MEMS accelerometer.
As a preferable scheme of the invention, the specific data of the input variable and the output variable in the MIMU working state is collected, and the specific method comprises the following steps:
fixing the MIMU on a rotary table with a temperature control box; starting a turntable, controlling the inner ring of the turntable to uniformly accelerate to 200 ℃ per second, and simultaneously starting a temperature control box to heat the MIMU; when the speed of the inner ring of the turntable reaches 200 DEG per second, the outer ring of the turntable is controlled to be evenly accelerated to 200 DEG per second, and when the MIMU temperature reaches 80 ℃, heating is stopped; keeping the turntable in a current motion state for a preset time; controlling the outer ring of the turntable to uniformly decelerate to 0, and controlling the temperature control box to cool the MIMU; when the speed of the outer ring of the turntable reaches 0, controlling the inner ring of the turntable to uniformly reduce to 0, and stopping reducing the temperature when the MIMU temperature reaches-45 ℃; and acquiring data of the MIMU in the whole process, and repeatedly acquiring the data until the data quantity meets the data set requirement.
As a preferable scheme of the invention, the loss function of the deep long short-time memory neural network is a cross entropy function, and the Optimizer is an Optimizer.
Compared with the prior art, the technical scheme provided by the invention has the following technical effects:
according to the method, firstly, error modeling is carried out on MIMU, possible factors causing MIMU output errors are analyzed, input output quantity of an LSTM network is determined, then data acquisition is carried out on MIMU, a data set is made, then the data set is used for training of a deep Short-Term Memory (LSTM) neural network, the error of the MIMU is predicted better through a model, and finally accuracy of MIMU error compensation is improved.
Drawings
Fig. 1 is a flowchart of a MIMU error compensation method based on a deep neural network in accordance with the present invention.
Fig. 2 is a flow chart of data acquisition in accordance with the present invention.
Fig. 3 is a block diagram of an LSTM neural network unit of the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below by referring to the drawings are exemplary only for explaining the present invention and are not to be construed as limiting the present invention.
As shown in fig. 1, a flowchart of a MIMU error compensation method based on a deep neural network according to the present invention includes the following steps:
step 1, performing error modeling on MIMU, analyzing possible factors causing MIMU output errors, and determining the input output quantity of an LSTM network;
step 2, acquiring multiple groups of time sequence historical data of MIMU, wherein each group of time sequence data comprises inertial sensor output data, MIMU output error data and MIMU working environment data, the inertial sensor output data comprises three-axis accelerometer output and three-axis gyroscope output, the MIMU output error data comprises three-axis accelerometer error and three-axis gyroscope error, and the MIMU working environment data comprises temperature and time;
step 3, constructing the time series data into a data set;
step 4, building an LSTM neural network prediction model, training the LSTM neural network by utilizing the data set, and adjusting, optimizing and determining the number of input layers, hidden layers and output layers;
step 5, acquiring real-time MIMU output data, and calculating a predicted MIMU output error;
and 6, compensating the MIMU output error in real time by utilizing the predicted data.
Step 1, carrying out error modeling on MIMU, exciting different error items through an experimental method, and establishing three orthogonal MEMS gyroscope angular rate errors epsilon and three orthogonal MEMS accelerometer acceleration errors V models as follows:
wherein the first term ε 0 And 0 constant zero offset vector of gyroscope and accelerometer respectively, second term epsilon T And delta T Drift vectors with zero bias temperature respectively, and third term is coefficient matrix respectivelyProduct of angular velocity vector omega output by gyro andthe product of the output f of the accelerometer and the output f of the accelerometer respectively represents nonlinear errors related to angular motion and linear acceleration, and the fourth term is a coefficient matrix respectively>Multiplication product and coefficient matrix with accelerometer output f>The product of carrier angular velocity omega, the first formula represents the mechanical sensitivity error term of the gyro angular velocity and the linear motion acceleration (vibration/impact acceleration), and the second formula represents the size effect error of the MEMS accelerometerThe first four terms of the two equations are functionally regular, and belong to deterministic errors, ε r And delta r Is a random error.
Step 2, acquiring historical time series data, which specifically comprises the following steps:
a set of data acquisition system is designed by taking a singlechip STM32F407 as a processor, and a plurality of serial ports of the singlechip are utilized to simultaneously acquire MIMU and turntable output data, so that the problem of time synchronization between the MIMU and the turntable is solved, the data synchronization of the MIMU and the turntable is realized, and the data acquisition flow is shown in figure 2. A set of turntable motion trail is designed, so that the acquired time series data is closer to MIMU data in a working state:
(1) starting a turntable, controlling the inner ring of the turntable to uniformly accelerate to 200 DEG/s, and simultaneously starting a temperature control box to heat the MIMU;
(2) the outer ring of the rotary table is controlled to uniformly accelerate to 200 DEG/s, and heating is stopped when the temperature reaches 80 ℃;
(3) maintaining the turntable in motion for a certain period of time;
(4) controlling the outer ring of the turntable to uniformly decelerate to 0, controlling the incubator, and cooling the MIMU;
(5) controlling the inner ring of the turntable to uniformly slow down to 0, and stopping cooling when the temperature reaches-45 ℃;
and repeatedly collecting data until the data quantity meets the data set requirement.
Step 4, constructing an LSTM neural network prediction model, wherein the LSTM neural network unit is shown in fig. 3, and the LSTM neural network unit controls the updating and losing of the cell state through an input gate, a forgetting gate and an output gate, and the specific formula is as follows:
i t ＝σ(W xi x t +W hi h t-1 +d)
O t ＝σ(W xo x t +W ho h t-1 +d)
f t ＝σ(W xf x t +W hf h t-1 +d)
wherein i is t 、O t 、f t Input gate output, output gate output and forget gate output at the moment t respectively; h is a t-1 、c t-1 The hidden layer output and the memory unit state value at the time t-1 are respectively; sigma is an activation function; d is a bias value; the rest is corresponding weight coefficient matrix.
The input gate and the forget gate determine the memory state c at the current moment t ＝f t c t-1 +i t tanh(W xc x t +W hc h t-1 +d) and obtaining a final output value h under the action of an output gate t ＝o t tanh(c t )。
Searching a hidden layer structure with the best performance by adopting a heuristic method, building a plurality of different models for training, and taking the mean square error (mse) of an output layer error vector as a measurement index of the precision of a network training result:
from the above table, it can be seen that when the number of hidden layers is 4 and the number of nodes in each layer is 50, the LSTM network model training effect is optimal, and the complexity problem of the network structure is comprehensively considered, so that the network structure is determined to be hidden layer 3, and the number of nodes in each layer is 50. The LSTM network loss function selects the cross entropy function, and the Optimizer selects the Optimizer, and the learning rate is designed to decrease exponentially with training times. And finally, acquiring real-time MIMU output data, calculating a predicted MIMU output error through the prediction model, and using the predicted data to compensate the MIMU output error in real time.
The above embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereto, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the present invention.
Claims (4)
1. The MIMU error compensation method based on the deep neural network is characterized by comprising the following steps of:
step 1, establishing an error model of MIMU, analyzing influence factors causing MIMU output errors according to the error model, and determining input variables and output variables of a deep and long-term memory neural network; the method comprises the following steps:
respectively establishing angular rate errors epsilon of three orthogonal MEMS gyroscopes a And acceleration error v of three orthogonal MEMS accelerometers a The model is used for analyzing influence factors causing output errors of the MEMS gyroscope and the MEMS accelerometer according to the respective error models, so that input variables and output variables of the deep long-short time memory neural network are determined;
the error models of the MEMS gyroscope and the MEMS accelerometer are as follows:
wherein ε 0a And delta 0a Constant zero offset vector epsilon of gyroscope and accelerometer respectively Ta And delta Ta The drift vectors of the gyroscope and the accelerometer with zero offset along with the temperature respectively,is a coefficient matrix->Coefficient of omega a For the angular velocity vector of the gyroscope output, +.>Is a coefficient matrix->Is of (1)Number f a For the acceleration vector output by the accelerometer,is a coefficient matrix->Coefficients of-> Is a coefficient matrix->Coefficient epsilon of (a) ra And delta ra A=x, y, z, which is a random error;
step 2, acquiring specific data of the input variable and the output variable in the MIMU working state according to the input variable and the output variable determined in the step 1, and constructing a data set;
step 3, establishing a deep long short-term memory neural network prediction model, and training the deep long-term memory neural network prediction model by utilizing the data set constructed in the step 2 to obtain a trained prediction model;
and 4, acquiring MIMU output data in real time, calculating an output error predicted value by using the trained predicted model, and performing error compensation on the MIMU output data by using the output error predicted value.
2. The MIMU error compensation method based on a deep neural network according to claim 1, wherein the input variable and the output variable of the deep long short-time memory neural network are specifically:
the input variables of the deep long short-time memory neural network comprise the output of the MEMS gyroscope, the output of the MEMS accelerometer, the temperature of the MEMS gyroscope and the temperature of the MEMS accelerometer;
the output variables of the deep long-short time memory neural network comprise the output error predicted value of the MEMS gyroscope and the output error predicted value of the MEMS accelerometer.
3. The MIMU error compensation method based on a deep neural network as claimed in claim 1, wherein said collecting specific data of an input variable and an output variable in a MIMU operating state comprises:
fixing the MIMU on a rotary table with a temperature control box; starting a turntable, controlling the inner ring of the turntable to uniformly accelerate to 200 ℃ per second, and simultaneously starting a temperature control box to heat the MIMU; when the speed of the inner ring of the turntable reaches 200 DEG per second, the outer ring of the turntable is controlled to be evenly accelerated to 200 DEG per second, and when the MIMU temperature reaches 80 ℃, heating is stopped; keeping the turntable in a current motion state for a preset time; controlling the outer ring of the turntable to uniformly decelerate to 0, and controlling the temperature control box to cool the MIMU; when the speed of the outer ring of the turntable reaches 0, controlling the inner ring of the turntable to uniformly reduce to 0, and stopping reducing the temperature when the MIMU temperature reaches-45 ℃; and acquiring data of the MIMU in the whole process, and repeatedly acquiring the data until the data quantity meets the data set requirement.
4. The MIMU error compensation method based on a deep neural network of claim 1 wherein the loss function of the deep long short-term memory neural network is a cross entropy function and the Optimizer is an Optimizer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910961889.2A CN110749337B (en) | 2019-10-11 | 2019-10-11 | MIMU error compensation method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910961889.2A CN110749337B (en) | 2019-10-11 | 2019-10-11 | MIMU error compensation method based on deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110749337A CN110749337A (en) | 2020-02-04 |
| CN110749337B true CN110749337B (en) | 2023-10-31 |
Family
ID=69277975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910961889.2A Active CN110749337B (en) | 2019-10-11 | 2019-10-11 | MIMU error compensation method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110749337B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11205112B2 (en) | 2019-04-01 | 2021-12-21 | Honeywell International Inc. | Deep neural network-based inertial measurement unit (IMU) sensor compensation method |
| CN111505543B (en) * | 2020-03-18 | 2023-02-28 | 中国电力科学研究院有限公司 | Method and system for compensating dynamic hysteresis based on recurrent neural network |
| CN111561929B (en) * | 2020-04-26 | 2022-05-03 | 南京工业大学 | Time delay and noise reduction method for vehicle-mounted MEMS inertial sensor |
| EP3940342B1 (en) * | 2020-07-13 | 2022-08-31 | Siemens Aktiengesellschaft | Measuring device |
| CN112580257B (en) * | 2020-12-10 | 2022-03-29 | 东北大学 | Method, device and equipment for constructing prediction model of complex industrial system based on deep learning and storage medium |
| CN113091768B (en) * | 2021-03-12 | 2024-04-12 | 南京理工大学 | MIMU integral dynamic intelligent calibration compensation method |
| CN113008291A (en) * | 2021-03-15 | 2021-06-22 | 北京信息科技大学 | Data generation method and device, storage medium and electronic device |
| CN113503899A (en) * | 2021-07-12 | 2021-10-15 | 清华大学 | Nonlinear error suppression system and sensor based on neural network |
| CN113984043B (en) * | 2021-09-14 | 2024-05-10 | 哈尔滨工程大学 | Dynamic error correction method of mining inertial navigation system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101158588A (en) * | 2007-11-16 | 2008-04-09 | 北京航空航天大学 | MEMS gyroscopes error compensation method for micro satellite based on integration nerval net |
| CN107330149A (en) * | 2017-05-27 | 2017-11-07 | 哈尔滨工业大学 | MIMU Modelling of Random Drift of Gyroscopes Forecasting Methodologies based on ARMA and BPNN built-up patterns |
| CN109188026A (en) * | 2018-10-25 | 2019-01-11 | 北京航空航天大学 | The deep learning method of automatic Calibration suitable for mems accelerometer |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11024041B2 (en) * | 2018-12-10 | 2021-06-01 | Intel Corporation | Depth and motion estimations in machine learning environments |
-
2019
- 2019-10-11 CN CN201910961889.2A patent/CN110749337B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101158588A (en) * | 2007-11-16 | 2008-04-09 | 北京航空航天大学 | MEMS gyroscopes error compensation method for micro satellite based on integration nerval net |
| CN107330149A (en) * | 2017-05-27 | 2017-11-07 | 哈尔滨工业大学 | MIMU Modelling of Random Drift of Gyroscopes Forecasting Methodologies based on ARMA and BPNN built-up patterns |
| CN109188026A (en) * | 2018-10-25 | 2019-01-11 | 北京航空航天大学 | The deep learning method of automatic Calibration suitable for mems accelerometer |
Non-Patent Citations (2)
| Title |
|---|
| Changhui Jiang等.A MEMS IMU De-Noising Method Using Long Short Term Memory Recurrent Neural Networks (LSTM-RNN).sensors.2018,第1-12页. * |
| 许德新等.MEMS 惯性测量组件的温度误差补偿模型研究.哈尔滨工程大学学报.2010,第31卷(第5期),第607-613页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110749337A (en) | 2020-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110749337B (en) | MIMU error compensation method based on deep neural network | |
| CN109188026B (en) | Automatic calibration deep learning method suitable for MEMS accelerometer | |
| CN108120451A (en) | Based on silicon micro accerometer temperature-compensation method, the system for improving PSO optimization neural networks | |
| CN104281056B (en) | The gyroscope Robust Adaptive Control method learnt based on the neutral net upper bound | |
| CN110039373B (en) | Method and system for predicting thermal deformation of spindle of numerical control machine tool | |
| CN113091768B (en) | MIMU integral dynamic intelligent calibration compensation method | |
| CN112596533B (en) | MEMS gyroscope data driving high-precision active disturbance rejection control method | |
| CN104503246B (en) | Indirect adaptive neural network sliding-mode control method for micro-gyroscope system | |
| Xu et al. | A temperature compensation method for MEMS accelerometer based on LM_BP neural network | |
| CN112181002B (en) | Micro gyroscope dual-recursion disturbance fuzzy neural network fractional order sliding mode control method | |
| CN107831660A (en) | Gyroscope self-adaption high-order super-twisting sliding mode control method | |
| CN111308979B (en) | U control system based on multi-rate delay state space model | |
| Zhang et al. | Design and implementation of novel fractional-order controllers for stabilized platforms | |
| Wang et al. | Event-triggered adaptive terminal sliding mode tracking control for drag-free spacecraft inner-formation with full state constraints | |
| Gang-Qiang et al. | Real-time temperature drift compensation method of a MEMS accelerometer based on deep GRU and optimized monarch butterfly algorithm | |
| Khan et al. | Analysis and time-delay synchronisation of chaotic satellite systems | |
| CN110262237A (en) | Gyroscope super-twisting sliding mode control method based on double feedback fuzzy neural networks | |
| Vertzberger et al. | Adaptive attitude estimation using a hybrid model-learning approach | |
| CN110262242B (en) | Micro gyroscope self-adaptive fractional order sliding mode control method | |
| Zhong et al. | A new drone accelerometer calibration method | |
| CN103472725B (en) | The full control method regulating of a kind of neutral net based on nominal controller | |
| Xing et al. | Offline calibration for MEMS gyroscope G-sensitivity error coefficients based on the newton iteration and least square methods | |
| Liang et al. | System‐Level Temperature Compensation Method for the RLG‐IMU Based on HHO‐RVR | |
| Wang et al. | Compounded calibration based on FNN and attitude estimation method using intelligent filtering for low cost MEMS sensor application | |
| CN110703611B (en) | Micro-gyroscope sensor terminal sliding mode control system based on recursive fuzzy neural 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |