CN112113566B - Inertial navigation data correction method based on neural network - Google Patents
Inertial navigation data correction method based on neural network Download PDFInfo
- Publication number
- CN112113566B CN112113566B CN202011017174.0A CN202011017174A CN112113566B CN 112113566 B CN112113566 B CN 112113566B CN 202011017174 A CN202011017174 A CN 202011017174A CN 112113566 B CN112113566 B CN 112113566B
- Authority
- CN
- China
- Prior art keywords
- inertial navigation
- data
- neural network
- carrier
- layer
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
- G01C21/165—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation combined with non-inertial navigation instruments
-
- 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
- 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
- G01S19/45—Determining position by combining measurements of signals from the satellite radio beacon positioning system with a supplementary measurement
- G01S19/47—Determining position by combining measurements of signals from the satellite radio beacon positioning system with a supplementary measurement the supplementary measurement being an inertial measurement, e.g. tightly coupled inertial
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Manufacturing & Machinery (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Automation & Control Theory (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Navigation (AREA)
Abstract
The invention discloses an inertial navigation data correction method based on a neural network, which utilizes accurate carrier geographical position information provided by GNSS in an unobstructed environment as calibration information, and realizes the feature learning of the artificial neural network on the accumulated error of an inertial navigation system by training the designed artificial neural network through a reward model, thereby eliminating the accumulated error of the inertial navigation system.
Description
Technical Field
The invention relates to the field of inertial navigation, in particular to an inertial navigation data correction method based on a neural network.
Background
The rapid development of the related application of the internet of things drives the demand of Location Based Services (LBS) to rise, which makes the demand for high-precision real-time positioning schemes increasingly urgent. An accurate, stable and real-time positioning system is an important guarantee for realizing the application of the Internet of things such as Virtual Reality (VR), commodity retail, robot control, unmanned driving and the like.
Although positioning by a Global Navigation Satellite System (GNSS) such as a Global Positioning System (GPS), a beidou satellite navigation system (BDS), a Galileo satellite positioning system (Galileo), and a GLONASS (GLONASS) positioning system has been very popular at present, positioning by satellites may be inaccurate for some outdoor environments with severe shielding (such as tunnels, forests, etc.) or indoor scenes affected by shielding and interference of building structures to satellite signals. An Inertial Navigation System (INS) using an Inertial Measurement Unit (IMU) is an effective way to solve the above problems, and has many advantages of no limitation to the external environment, wide applicable scenes, low cost, and the like, and is a common means for implementing positioning applications.
However, due to the change of the relative pose of the carrier obtained by twice integrating the acceleration and the angular velocity, the inertial navigation system may generate an accumulated error, which causes the trajectory estimation to diverge as the movement distance increases. Therefore, the key to realize the high-precision inertial navigation system is to effectively eliminate the accumulated error of the inertial navigation system. At present, common methods for relieving the accumulated error of the inertial navigation system include: 1) the motion characteristics of the carrier are modeled. For example, a pedestrian dead reckoning system (PDR) can utilize the motion characteristics of biped organisms (namely, the biped can be decomposed into a swinging state and a static state in the motion process), and an error correction scheme based on zero-speed updating (ZUPT) is designed. 2) And the inertial navigation system is corrected by using other kinds of information through a data fusion method. Such as visual-inertial odometer (VIO), are high-precision positioning systems that eliminate inertial measurement unit data drift by incorporating image information. However, the first solution for eliminating the accumulated error of the inertial navigation system has certain limitations, and mainly shows that the modeling of the motion mode of the carrier does not have universality. For example, for vehicles, Unmanned Aerial Vehicles (UAVs), and the like, the absence of significant motion features may provide additional observations and enable error correction. Therefore, the accumulated error elimination of the inertial navigation system by the data fusion algorithm depending on a specific additional signal source is a general and effective scheme. In addition, for the realization of the error correction of the inertial navigation system based on data fusion, the key for improving the positioning accuracy of the inertial navigation system is to select a proper correction signal source and design an effective correction algorithm.
Disclosure of Invention
Aiming at the defects in the prior art, the inertial navigation data correction method based on the neural network provided by the invention designs a reinforcement learning method based on the inertial navigation correction neural network by utilizing the geographical position information of the carrier provided by the GNSS, learns an error model of the pose estimation of the inertial navigation system and realizes the accumulated error correction of the inertial navigation system.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that: an inertial navigation data correction method based on a neural network comprises the following steps:
s1, acquiring carrier position and attitude data of an inertial navigation system and carrier geographic position data of a GNSS;
s2, carrying out grouping preprocessing on the carrier pose data and the carrier geographic position data to obtain a plurality of groups of carrier data;
s3, constructing an inertial navigation correction neural network, and training the inertial navigation correction neural network by adopting a plurality of groups of carrier data until the average reward obtained by the reward model is the maximum to obtain the trained inertial navigation correction neural network;
and S4, correcting the position and posture data of the carrier by adopting the trained inertial navigation correction neural network to obtain inertial navigation correction data.
Further, the method of packet preprocessing in step S2 is: grouping the geographic position data of the carriers at two adjacent time points and a plurality of position and posture data of the carriers between the two adjacent time points into a group, and obtaining the nth group of carrier data as follows:wherein the GNSSnFor the carrier geographical position data of the nth point in time, GNSSn+1Is the carrier geographical position data of the (n + 1) th time point,toIs the L carrier pose data between the nth time point and the (n + 1) th time point.
Further, the inertial navigation correction neural network in step S3 includes: an input layer, a hidden layer and an output layer;
the input layer, the hidden layer and the output layer are all full-connection layers and are connected in sequence.
Further, the dimension of the input layer is 6, the dimension of the hidden layer is 20, and the dimension of the output layer is 3.
The beneficial effects of the above further scheme are: due to the fact that the dimensionality of input data is low, the problem of overfitting can be effectively relieved by adopting the related network, meanwhile, the network is simple, real-time operation can be achieved, and the mobile terminal is easy to deploy to a mobile terminal with limited computing power.
Further, the input and output relationship of the fully connected layer is as follows:
wherein, FmFor the output of the m-th fully-connected layer, Relu (. cndot.) is the activation function, x1×iFor input data matrix, dimension is 1 × i, Wj×iA weight matrix to be trained for a fully-connected layer with dimensions of j x i, b1×jThe method comprises the steps of taking a bias matrix of a full connection layer, wherein the dimension is 1 xj, and T is transposition operation of the matrix; is a matrix multiplication operation and + is a matrix addition operation.
Further, the excitation model in step S3 is:
wherein the content of the first and second substances,is input intoInertial navigation modified neural network output, GNSSnFor the carrier geographical position data of the nth point in time, GNSSn+1Is the carrier geographic position data of the (n + 1) th time point, epsilon1For the weight of Euclidean distance in the reward, e2For the weight of cosine distance in the reward, argmax is the neural network parameter θ for seeking inertial navigation corrections*So thatThe maximum function, M is the number of sets of incoming carrier data, i.e. the total number of training rounds,the average reward, | | | | is a two-norm, W is all weight parameters of the inertial navigation correction neural network, and b is all bias parameters of the inertial navigation correction neural network.
The beneficial effects of the above further scheme are: the designed reward function formula considers the difference between the estimated displacement vector and the calibration displacement vector corresponding to the GNSS from two angles of Euclidean distance and direction cosine distance, thereby not only ensuring the accuracy of the estimated moving distance, but also ensuring the accuracy of the estimated moving direction.
In conclusion, the beneficial effects of the invention are as follows:
(1) the method and the device utilize accurate carrier geographical position information provided by the GNSS in an unobstructed environment as calibration information, and realize the feature learning of the inertial navigation system accumulative error by the artificial neural network through the artificial neural network designed by the reward model training, thereby eliminating the accumulative error of the inertial navigation system.
(2) Compared with other inertial navigation error correction schemes based on data fusion, the method has the advantages that accurate calibration information can be provided by using the GNSS, meanwhile, the complexity of an inertial navigation error correction algorithm is greatly reduced and the robustness of a model is improved by using the inertial navigation correction neural network for deep reinforcement learning, and accordingly, the positioning correction precision of an inertial navigation system is improved.
Drawings
FIG. 1 is a flow chart of a neural network based inertial navigation data correction method;
FIG. 2 is a schematic structural diagram of an inertial navigation correction neural network.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 1, an inertial navigation data correction method based on a neural network includes the following steps:
s1, acquiring carrier position and attitude data of an inertial navigation system and carrier geographic position data of a GNSS;
s2, carrying out grouping preprocessing on the carrier pose data and the carrier geographic position data to obtain a plurality of groups of carrier data;
the method of the grouping preprocessing in the step S2 is: grouping the geographic position data of the carriers at two adjacent time points and a plurality of position and posture data of the carriers between the two adjacent time points into a group, and obtaining the nth group of carrier data as follows:wherein the GNSSnFor the carrier geographical position data of the nth point in time, GNSSn+1Is the carrier geographical position data of the (n + 1) th time point,toIs the L carrier pose data between the nth time point and the (n + 1) th time point.
GNSS with one set of data acrossnAnd GNSSn+1As a reference signal, the intermediate plurality of carrier pose data serves as training data.
S3, constructing an inertial navigation correction neural network, and training the inertial navigation correction neural network by adopting a plurality of groups of carrier data until the average reward obtained by the reward model is the maximum to obtain the trained inertial navigation correction neural network;
as shown in fig. 2, the inertial navigation correction neural network in step S3 includes: an input layer, a hidden layer and an output layer;
the input layer, the hidden layer and the output layer are all full-connection layers and are connected in sequence.
The dimension of the input layer is 6, the dimension of the hidden layer is 20, and the dimension of the output layer is 3.
The input and output relationship of the full connection layer is as follows:
wherein, FmIs fully connected with the m-th layerThe output of the layer, Relu (·) is the activation function, x1×iFor input data matrix, dimension is 1 × i, Wj×iA weight matrix to be trained for a fully-connected layer with dimensions of j x i, b1×jThe method comprises the steps of taking a bias matrix of a full connection layer, wherein the dimension is 1 multiplied by j, and the method is a transposition operation of the matrix; is a matrix multiplication operation and + is a matrix addition operation.
In step S3, the excitation model is:
wherein the content of the first and second substances,is input intoInertial navigation modified neural network output, GNSSnFor the carrier geographical position data of the nth point in time, GNSSn+1Is the carrier geographic position data of the (n + 1) th time point, epsilon1For the weight of Euclidean distance in the reward, e2For the weight of cosine distance in the reward, argmax is the neural network parameter θ for seeking inertial navigation corrections*So thatThe maximum function, M is the number of sets of incoming carrier data, i.e. the total number of training rounds,the average reward, | | | | is a two-norm, W is all weight parameters of the inertial navigation correction neural network, and b is all bias parameters of the inertial navigation correction neural network.
In this embodiment, the reward model may be solved by a gradient descent method.
And S4, correcting the position and posture data of the carrier by adopting the trained inertial navigation correction neural network to obtain inertial navigation correction data.
Claims (4)
1. An inertial navigation data correction method based on a neural network is characterized by comprising the following steps:
s1, acquiring carrier position and attitude data of an inertial navigation system and carrier geographic position data of a GNSS;
s2, carrying out grouping preprocessing on the carrier pose data and the carrier geographic position data to obtain a plurality of groups of carrier data;
the method of the grouping preprocessing in the step S2 is: grouping the geographic position data of the carriers at two adjacent time points and a plurality of position and posture data of the carriers between the two adjacent time points into a group, and obtaining the nth group of carrier data as follows:wherein the GNSSnFor the carrier geographical position data of the nth point in time, GNSSn+1Is the carrier geographical position data of the (n + 1) th time point,toThe L carrier pose data between the nth time point and the (n + 1) th time point are obtained;
s3, constructing an inertial navigation correction neural network, and training the inertial navigation correction neural network by adopting a plurality of groups of carrier data until the average reward obtained by the reward model is the maximum to obtain the trained inertial navigation correction neural network;
in step S3, the excitation model is:
wherein the content of the first and second substances,is input intoOf inertial navigation correction neural network, GNSSnFor the carrier geographical position data of the nth point in time, GNSSn+1Is the carrier geographic position data of the (n + 1) th time point, epsilon1For the weight of Euclidean distance in the reward, e2For the weight of cosine distance in the reward, argmax is the neural network parameter θ for seeking inertial navigation corrections*So thatThe maximum function, M is the number of sets of incoming carrier data, i.e. the total number of training rounds,the average reward is, | | | | is a two-norm, W is all weight parameters of the inertial navigation correction neural network, and b is all bias parameters of the inertial navigation correction neural network; and S4, correcting the position and posture data of the carrier by adopting the trained inertial navigation correction neural network to obtain inertial navigation correction data.
2. The inertial navigation data correction method based on neural network according to claim 1, wherein the inertial navigation correction neural network in step S3 includes: an input layer, a hidden layer and an output layer;
the input layer, the hidden layer and the output layer are all full-connection layers and are connected in sequence.
3. The inertial navigation data modification method based on neural network according to claim 2, characterized in that the dimension of the input layer is 6, the dimension of the hidden layer is 20, and the dimension of the output layer is 3.
4. The inertial navigation data correction method based on neural network according to claim 2, characterized in that the input and output of the fully-connected layer are related as follows:
wherein, FmFor the output of the m-th fully-connected layer, Relu (. cndot.) is the activation function, x1×iFor input data matrix, dimension is 1 × i, Wj×iA weight matrix to be trained for a fully-connected layer with dimensions of j x i, b1×jThe method comprises the steps of taking a bias matrix of a full connection layer, wherein the dimension is 1 xj, and T is transposition operation of the matrix; is a matrix multiplication operation and + is a matrix addition operation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011017174.0A CN112113566B (en) | 2020-09-24 | 2020-09-24 | Inertial navigation data correction method based on neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011017174.0A CN112113566B (en) | 2020-09-24 | 2020-09-24 | Inertial navigation data correction method based on neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112113566A CN112113566A (en) | 2020-12-22 |
CN112113566B true CN112113566B (en) | 2022-04-08 |
Family
ID=73800940
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011017174.0A Active CN112113566B (en) | 2020-09-24 | 2020-09-24 | Inertial navigation data correction method based on neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112113566B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112762932B (en) * | 2021-04-07 | 2021-07-13 | 智道网联科技(北京)有限公司 | Trajectory calculation method and device based on neural network model |
CN113984043B (en) * | 2021-09-14 | 2024-05-10 | 哈尔滨工程大学 | Dynamic error correction method of mining inertial navigation system |
CN114136317B (en) * | 2022-02-07 | 2022-04-19 | 中国人民解放军军事科学院国防科技创新研究院 | Underwater long-range navigation positioning method and device based on pure inertial measurement |
CN114577204B (en) * | 2022-02-09 | 2024-01-02 | 中科禾华(扬州)光电科技有限公司 | Anti-interference self-alignment method and device for strapdown inertial navigation system based on neural network |
CN115560757B (en) * | 2022-09-01 | 2023-08-22 | 中国人民解放军战略支援部队信息工程大学 | Unmanned aerial vehicle direct positioning correction method based on neural network under random attitude error condition |
CN116358562B (en) * | 2023-05-31 | 2023-08-01 | 氧乐互动(天津)科技有限公司 | Disinfection operation track detection method, device, equipment and storage medium |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8275193B2 (en) * | 2004-08-04 | 2012-09-25 | America Gnc Corporation | Miniaturized GPS/MEMS IMU integrated board |
CN107390246A (en) * | 2017-07-06 | 2017-11-24 | 电子科技大学 | A kind of GPS/INS Combinated navigation methods based on genetic neural network |
US10262224B1 (en) * | 2017-07-19 | 2019-04-16 | The United States Of America As Represented By Secretary Of The Navy | Optical flow estimation using a neural network and egomotion optimization |
WO2019048034A1 (en) * | 2017-09-06 | 2019-03-14 | Swiss Reinsurance Company Ltd. | Electronic logging and track identification system for mobile telematics devices, and corresponding method thereof |
US11688160B2 (en) * | 2018-01-17 | 2023-06-27 | Huawei Technologies Co., Ltd. | Method of generating training data for training a neural network, method of training a neural network and using neural network for autonomous operations |
CN108932840B (en) * | 2018-07-17 | 2021-09-07 | 北京理工大学 | Urban intersection passing method for unmanned vehicles based on reinforcement learning |
CN109059909A (en) * | 2018-07-23 | 2018-12-21 | 兰州交通大学 | Satellite based on neural network aiding/inertial navigation train locating method and system |
CN110362089A (en) * | 2019-08-02 | 2019-10-22 | 大连海事大学 | A method of the unmanned boat independent navigation based on deeply study and genetic algorithm |
CN111340190A (en) * | 2020-02-23 | 2020-06-26 | 华为技术有限公司 | Method and device for constructing network structure, and image generation method and device |
-
2020
- 2020-09-24 CN CN202011017174.0A patent/CN112113566B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN112113566A (en) | 2020-12-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112113566B (en) | Inertial navigation data correction method based on neural network | |
CN112097763B (en) | Underwater vehicle combined navigation method based on MEMS IMU/magnetometer/DVL combination | |
CN111707260B (en) | Positioning method based on frequency domain analysis and convolutional neural network | |
CN110595466B (en) | Lightweight inertial-assisted visual odometer implementation method based on deep learning | |
CN106979781B (en) | High-precision transfer alignment method based on distributed inertial network | |
CN111238467A (en) | Bionic polarized light assisted unmanned combat aircraft autonomous navigation method | |
CN113253325B (en) | Inertial satellite sequential tight combination lie group filtering method | |
Van et al. | Application of street tracking algorithm in an INS/GPS integrated navigation system | |
CN102506875B (en) | The air navigation aid of unmanned plane and device | |
Goppert et al. | Invariant Kalman filter application to optical flow based visual odometry for UAVs | |
Gao et al. | An integrated land vehicle navigation system based on context awareness | |
CN114689047A (en) | Deep learning-based integrated navigation method, device, system and storage medium | |
Tong et al. | Vehicle inertial tracking via mobile crowdsensing: Experience and enhancement | |
CN114018254A (en) | SLAM method for integrating laser radar and rotary inertial navigation integrated framework and information | |
CN112325878A (en) | Ground carrier combined navigation method based on UKF and air unmanned aerial vehicle node assistance | |
CN109387196B (en) | Remote underwater geomagnetic bionic navigation method | |
CN114993296B (en) | High dynamic integrated navigation method for guided projectile | |
CN115479605A (en) | High-altitude long-endurance unmanned aerial vehicle autonomous navigation method based on space target directional observation | |
CN112284379B (en) | Inertial pre-integration method of combined motion measurement system based on nonlinear integral compensation | |
CN114646993A (en) | Data fusion algorithm for accurate positioning based on GNSS, vision and IMU | |
CN108344410B (en) | Gyroscope-assisted method for improving output frequency of star sensor | |
Zhou et al. | IMU dead-reckoning localization with RNN-IEKF algorithm | |
Gustavsson | UAV pose estimation using sensor fusion of inertial, sonar and satellite signals | |
Liu | Vision sensor aided navigation for ground vehicle applications | |
CN117091457B (en) | Guided projectile navigation method and system based on deep learning |
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 |