Polarization trailing error estimation method based on meta-learning in high dynamic environment

Abstract

The invention provides a polarization trailing error estimation method based on meta-learning in a high dynamic environment, which belongs to the field of bionic polarization navigation, and comprises the following steps: establishing a polarization angle error database caused by trailing interference at different rotating speeds, training a rotating speed correlation and rotating speed irrelevant network based on a meta-learning strategy in an offline stage, and estimating the polarization angle error. The polarization angle error is decoupled into rotation speed-related and rotation speed-independent characteristics, and in the online stage, during linear motion, polarization / inertial navigation is used for correcting the three-dimensional attitude error of the dynamic model; in the turning motion, the rotation speed related characteristics are used as states, the polarization angle inverted by the carrier kinetic model is used as a measurement quantity, the rotation speed independent characteristics are used as a measurement matrix, the rotation speed related characteristics are estimated, the rotation speed related characteristics and the rotation speed independent characteristics are fused, and the polarization trailing error is estimated. According to the method, the self-adaptive estimation and compensation of the polarization trailing error under different rotating speeds in a high-dynamic environment are completed by combining offline and online learning, and the navigation performance of the bionic polarization integrated navigation system is improved.

Description

Technical Field

[0001] The present invention belongs to the field of bionic polarization navigation, and specifically relates to a polarization trailing error estimation method based on meta-learning in a high-dynamic environment. The method realizes adaptive estimation and compensation of polarization angle error in an environment with different speed trailing interference, thereby improving the navigation performance of a bionic polarization integrated navigation system in a turning maneuvering environment. Background Art

[0002] Many creatures in nature, such as bees, dragonflies, and monarch butterflies, can sense polarized light from the sky for autonomous navigation and positioning. Inspired by this biological navigation mechanism, bionic polarization navigation has garnered widespread attention in recent years. It offers advantages such as immunity to electromagnetic interference and high dynamics, enabling passive, autonomous navigation. The polarization angle is calculated by integrating the polarization sensor within a single sampling period. However, during large maneuvers involving a vehicle, the polarization sensor's sampling frequency is lower than the vehicle's angular velocity, resulting in an untimely response and a polarization angle trailing error, which reduces the performance of bionic polarization navigation. Therefore, compensating for this polarization angle trailing error during turning maneuvers is crucial for improving the performance and environmental adaptability of bionic polarization navigation.

[0003] The invention, "A Multi-Source Error Calibration Method for a Bionic Polarization Sensor Based on Adaptive UKF" (Application No. CN201811414147.X), establishes a system state model based on the installation error, scale factor, polarization coefficient, polarization angle, and polarization degree as state variables of the polarization navigation system. A polarization navigation system measurement model is established using light intensity measurements containing multi-source errors as output. An adaptive unscented Kalman filter is designed to estimate the installation error, scale factor, polarization angle, and polarization degree, thereby estimating and compensating for the polarization angle error under multi-source errors. The invention, "A Combined Inertial Navigation / Polarization Navigation Method Based on Non-Rayleigh Scattering Model Error" (Application No. CN202111417764.7), introduces the non-Rayleigh scattering model error and combines it with the strapdown inertial navigation error state to construct an extended-dimensional system state vector. Furthermore, the polarization angle error caused by the non-Rayleigh scattering model is modeled in the measurement equation, enabling online estimation and compensation of the polarization angle error under the non-Rayleigh scattering model, improving the system's adaptability to adverse weather conditions. There is also existing technology that targets outliers caused by large maneuvers. Based on the time-varying non-Gaussian characteristics of polarization navigation measurement noise, an adaptive random filtering method that takes into account measurement correlation entropy is designed to improve the system's anti-interference ability under large maneuvers.

[0004] The aforementioned prior art polarization sensor models account for the impact of various environmental factors on the polarization sensor. However, they fail to consider the impact of the trailing error generated by the polarization sensor during high-dynamic turning maneuvers. This results in a significant increase in polarization angle error during turning maneuvers, significantly reducing the accuracy of heading and attitude estimation. To address this issue, research is urgently needed to estimate and compensate for polarization trailing error at different rotational speeds and improve the heading and attitude accuracy of bionic polarization navigation. Summary of the Invention

[0005] To overcome the shortcomings of the existing technology, the present invention proposes a polarization trailing error estimation method based on meta-learning in a high-dynamic environment. It establishes a database of polarization angle errors caused by trailing interference at different speeds, including angular velocity, acceleration, polarization intensity, and polarization angle error. In the offline stage, the speed-dependent network and the speed-independent network are trained separately based on the meta-learning strategy to decouple the polarization angle error into speed-dependent features and speed-independent features. In the online stage, during the linear motion of the carrier, the polarization sensor and inertial navigation are first used for combined navigation. Then, the estimated heading and horizontal attitude are used to correct the three-dimensional attitude derived by dynamic recursion. During the turning motion of the carrier, the speed-dependent features are used as the state, the polarization angle derived by the carrier's dynamic inversion is used as the quantity measurement, and the speed-independent features are used as the measurement to complete the online correction of the speed-dependent features. Finally, the speed-dependent features and the speed-independent features are fused to complete the adaptive estimation and compensation of the polarization angle error in the polarization trailing interference environment.

[0006] To achieve the above objectives, the present invention adopts a technical solution: a polarization smearing error estimation method based on meta-learning in a high-dynamic environment, comprising the following steps:

[0007] Step 1: Establish a database of polarization angle errors at different rotation speeds, including angular velocity , acceleration , polarized light intensity , and polarization angle error ,in are the components of angular velocity on the three axes, are the components of acceleration on the three axes, is the light intensity of a channel, is the number of channels;

[0008] Step 2: Build a speed-related network based on offline meta-learning strategy Speed-independent network , with angular velocity , acceleration and polarized light intensity As input to the network, using contrastive learning loss , Generate Adversarial Loss and reconstruction losses Constrain and train and network, the polarization angle error Decoupling into speed-dependent characteristics Speed-independent characteristics , where m is the dimension of speed-related and irrelevant features;

[0009] Step 3: In the online stage, when the carrier moves in a straight line, the polarization / inertial navigation combined navigation is used to estimate the three-dimensional attitude of the carrier and correct the three-dimensional attitude of the carrier dynamic model. and angular velocity error As a state , establish the system state equation , with the heading angle of polarization / inertial navigation fusion , roll angle and pitch angle As a measurement, establish the measurement equation , correct the three-dimensional posture of the carrier dynamics model , and Represents Related functions, Represents When the carrier turns, the speed-related features are corrected online based on the dynamic model and the speed-independent features. As the state vector , establish the system state equation , based on the speed-independent characteristics Polarization angle error of carrier dynamics model inversion , establish the system measurement equation , realizing speed-related features Online adaptive update to obtain the corrected speed-related characteristics , and Represents Related functions, Represents Related measurement values;

[0010] Step 4: and The coupling completes the estimation and compensation of polarization angle tailing error.

[0011] The beneficial effects of the present invention compared with the prior art are:

[0012] This paper proposes a meta-learning-based polarization smearing error estimation method in highly dynamic environments. This method achieves adaptive estimation and compensation of polarization angle error under different speed smearing interference conditions, improving the navigation performance of a bionic polarization integrated navigation system during turning maneuvers. Based on a meta-learning strategy, this paper decouples the speed-dependent and speed-independent features of the polarization angle error in the offline phase. In the online phase, a speed-dependent feature model is constructed to achieve adaptive estimation and compensation of polarization angle error under different speed smearing interference conditions, improving the navigation performance of the bionic polarization integrated navigation system during turning maneuvers. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This invention provides a polarization smearing error estimation method based on meta-learning in a high dynamic environment.

[0014] Figure 2 A simulation trajectory setting diagram for an embodiment of the present invention;

[0015] Figure 3 This is a simulation angular velocity diagram of an embodiment of the present invention;

[0016] Figure 4 A simulated heading angle diagram according to an embodiment of the present invention;

[0017] Figure 5 A simulated roll angle diagram according to an embodiment of the present invention;

[0018] Figure 6 This is the comparison result of heading angle error. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] The polarization angle is obtained by integrating and solving the polarization sensor in one sampling period. However, during the turning maneuver, due to the rapid change of the environment, the polarization angle solution error is amplified, which in turn affects the navigation accuracy of the polarization sensor. Figure 1 As shown, the present invention proposes a polarization smearing error estimation method based on meta-learning in a high dynamic environment. The specific implementation steps are as follows:

[0021] Step 1: Establish a database of polarization angle errors at different rotation speeds, including the angular velocity output by the gyroscope , the acceleration output by the accelerometer , polarized output light intensity , and the polarization angle error obtained based on the reference system ,in are the components of angular velocity on the three axes, are the components of acceleration on the three axes, is the light intensity of a channel of the polarization sensor, is the number of channels of the polarization sensor;

[0022] Step 2: Build a speed-related network based on offline meta-learning strategy Speed-independent network , with angular velocity , acceleration and polarized light intensity As input to the network, using contrastive learning loss , Generate Adversarial Loss and reconstruction losses Constrain and train and network, the polarization angle error Decoupling into speed-dependent characteristics Speed-independent characteristics , where m is the dimension of speed-related and irrelevant features;

[0023] Step 3: In the online stage, when the carrier moves in a straight line, the polarization / inertial navigation combined navigation is used to estimate the three-dimensional attitude of the carrier and correct the three-dimensional attitude of the carrier dynamic model. and angular velocity error As a state , establish the system state equation , with the heading angle of polarization / inertial navigation fusion , roll angle and pitch angle As a measurement, establish the measurement equation , correct the three-dimensional posture of the carrier dynamics model , and Represents Related functions, Represents When the carrier turns, the speed-related features are corrected online based on the dynamic model and the speed-independent features. As the state vector , establish the system state equation Based on the speed-independent feature Polarization angle error of carrier dynamics model inversion , establish the system measurement equation , realizing speed-related features Online adaptive update to obtain the corrected speed-related characteristics , and Represents Related functions, Represents Related measurement values;

[0024] Step 4: and The coupling completes the estimation and compensation of polarization angle trailing error. By compensating the polarization angle error under trailing interference, the fusion accuracy of polarization / inertial navigation can be effectively improved in maneuvering environments.

[0025] Specifically, in step 1:

[0026] Sun Vector Polarization vector under carrier system b The relationship can be expressed as:

[0027] (1)

[0028] Among them, the sun vector The astronomical almanac can be used to calculate the polarization vector under the carrier system b. It can be expressed as , is the attitude matrix, which can be calculated from the heading angle and horizontal attitude angle provided by the reference, from which the polarization angle can be solved , combined with the polarization angle measured by the polarization sensor , and then we can get the label of polarization angle error , which can be expressed as:

[0029] (2)

[0030] In this way, a database of polarization angle errors at different rotation speeds can be established.

[0031] Specifically, in step 2,

[0032] and The angular velocity output by the gyroscope , the acceleration output by the accelerometer , polarized output light intensity As input, and The outputs are speed-related features Speed-independent characteristics , thus, the polarization angle error can be decoupled as and , the above process can be expressed as:

[0033] (3)

[0034] Speed-related network Speed-independent network The loss function can be expressed as:

[0035] (4)

[0036] in, Represent the coefficients of reconstruction loss, contrastive learning loss and generative adversarial loss, respectively. It can be expressed as:

[0037] (5)

[0038] in, is 1-norm, contrastive learning loss It can be expressed as:

[0039] (6)

[0040] in, is the positive sample loss, which can be expressed as:

[0041] (7)

[0042] in, is the 2-norm, and From the same speed The data below and , is the negative sample loss, which can be expressed as:

[0043] (8)

[0044] in, and From different speeds and The contrastive learning loss can be improved by pulling in the speed-related features of the same speed and alienating the speed-related features of different speeds. The ability to extract speed-related features. Generate adversarial loss It can be expressed as:

[0045] (9)

[0046] in, is the discriminator, is the data from zero speed, Is from non-zero speed Under the data, generate adversarial loss This can be improved by making the discriminator unable to distinguish speed-independent features from zero-speed and non-zero-speed data. The ability to extract speed-independent features. Thus, it is possible to train a machine that can accurately extract speed-independent and speed-dependent features. and .

[0047] Specifically, in step 3:

[0048] When the carrier moves in a straight line, the polarization / inertial navigation combined navigation is used to estimate the three-dimensional attitude of the carrier , and is used to correct the three-dimensional posture of the carrier dynamics model. The system state equation of the dynamics model can be expressed as:

[0049] (10)

[0050] in, , is the three-axis attitude error angle, is the three-axis angular velocity error, represents the noise of the state equation, is the state transfer matrix, which can be expressed as:

[0051] (11)

[0052] in, is the inertia matrix, Represents a diagonal matrix, using the three-dimensional attitude of polarization / inertial navigation combined navigation The measurement model of the modified dynamic model can be expressed as:

[0053] (12)

[0054] Therefore, the dynamic model recursion error is corrected by polarization / inertial navigation combined navigation in the straight line segment to ensure that the dynamic model can deliver accurate three-dimensional posture. .

[0055] When the carrier turns, the state equation of the system can be expressed as:

[0056] (13)

[0057] in, , represents the noise of the state equation, is the state transfer matrix, which can be expressed as:

[0058] (14)

[0059] in, is the attenuation coefficient, the measurement equation can be expressed as:

[0060] (15)

[0061] The corrected speed-related characteristics can be obtained from this .

[0062] Specifically, the step 4 includes: and The coupling completes the estimation and compensation of polarization angle trailing error. By compensating the polarization angle error under trailing interference, the fusion accuracy of polarization / inertial navigation can be effectively improved in maneuvering environments.

[0063] The polarization angle after compensating the tailing error can be expressed as:

[0064] (16)

[0065] in, is the polarization angle measured by the polarization sensor, To compensate for the polarization angle after tailing interference, the polarization angle error compensation under tailing interference can effectively improve the fusion accuracy of polarization / inertial navigation in maneuvering environments.

[0066] Example:

[0067] The simulation trajectory consists of 8 sets of maneuvers, each of which includes: rolling to the left at 0.05° / s for 2.5 seconds; turning to the left at a specific speed for 10 seconds; rolling to the left at 0.05° / s for 2.5 seconds; uniform linear motion for 5 seconds; rolling to the left at 0.05° / s for 2.5 seconds; turning to the right at a specific speed for 10 seconds; rolling to the left at 0.05° / s for 2.5 seconds; uniform linear motion for 5 seconds. The speeds of each set of maneuvers are 1 / 7 / 2.5 / 11.5 / 16 / 20.5 / 23.5 / 25° / s respectively. The accelerometer zero bias is 10ug, and the speed is a random walk. , gyroscope zero bias 10° / h, angle random walk , the polarized light intensity is added with a uniformly distributed noise with an amplitude of 10Lux. Figure 2 Shown is the simulation trajectory setting diagram of this embodiment, Figure 3 The simulation angular velocity diagram of this embodiment is shown as follows. Figure 4 This is the simulation heading angle diagram of this embodiment, Figure 5 2 is a diagram of the simulated roll angle of this embodiment.

[0068] Using the method of the present invention, in the offline phase, a meta-learning strategy is used to decouple the polarization angle error into speed-dependent and speed-independent features. In the online phase, when the carrier is in linear motion, polarization / inertial navigation fusion is used to correct the three-dimensional attitude error of the carrier dynamics model. During cornering, the polarization angle and speed-independent features inverted from the carrier dynamics are used to estimate the speed-dependent features, completing the adaptive estimation and compensation of the polarization angle error in an environment with polarization smearing interference. The compensation method of the present invention was compared with the method without smearing error compensation and with the method using only offline compensation.

[0069] The comparison results of heading angle error are as follows: Figure 6 As shown, the straight line represents the uncompensated trailing error, the circled result represents the result using offline learning, and the cross represents the result using both offline and online learning. It is clear that offline learning effectively compensates for the trailing error, while the addition of online learning stabilizes the heading error around 0. Quantitative analysis shows that the heading angle absolute mean error (RMSE) without compensation is 0.60°, while the RMSE after offline learning compensation is 0.24°, a 60.0% improvement. The RMSE after both offline and online learning compensation is 0.10°, a 58.3% improvement compared to offline learning alone. This completes the meta-learning-based polarization angle error estimation under trailing interference conditions.

[0070] Although the above describes the illustrative specific embodiments of the present invention to facilitate understanding of the present invention by those skilled in the art, and it should be clear that the present invention is not limited to the scope of the specific embodiments, it is obvious to those skilled in the art that as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

Claims

1. A polarization smearing error estimation method based on meta-learning in a high dynamic environment, characterized by: The steps include: Step 1: Establish a database of polarization angle errors at different rotation speeds, including angular velocity , acceleration , polarized light intensity , and polarization angle error ,in are the components of angular velocity on the three axes, are the components of acceleration on the three axes, is the light intensity of a channel, is the number of channels; Step 2: Build a speed-related network based on offline meta-learning strategy Speed-independent network , with angular velocity , acceleration and polarized light intensity As input to the network, using contrastive learning loss , Generate Adversarial Loss and reconstruction losses Constrain and train and network, the polarization angle error Decoupling into speed-dependent characteristics Speed-independent characteristics , where m is the dimension of speed-related and irrelevant features; Step 3: In the online stage, when the carrier moves in a straight line, the polarization / inertial navigation combined navigation is used to estimate the three-dimensional attitude of the carrier and correct the three-dimensional attitude of the carrier dynamic model. and angular velocity error As a state , establish the system state equation , with the heading angle of polarization / inertial navigation fusion , roll angle and pitch angle As a measurement, establish the measurement equation , correct the three-dimensional posture of the carrier dynamics model , and Represents Related functions, Represents When the carrier turns, the speed-related features are corrected online based on the dynamic model and the speed-independent features. As the state vector , establish the system state equation , based on the speed-independent feature Polarization angle error of carrier dynamics model inversion , establish the system measurement equation , realizing speed-related features Online adaptive update to obtain the corrected speed-related characteristics , and Represents Related functions, Represents Related measurement values; Step 4: and The coupling completes the estimation and compensation of polarization angle tailing error.

2. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 1, characterized in that: In the step (1), a database of polarization angle errors at different rotation speeds is established, including the angular velocity output by the gyroscope. , the acceleration output by the accelerometer , polarized output light intensity , and the polarization angle error obtained based on the reference system ,in, is the light intensity of a channel of the polarization sensor, is the number of channels of the polarization sensor; Sun Vector Polarization vector under carrier system b The relationship is expressed as: (1) Among them, the sun vector Using the astronomical almanac, the polarization vector of the carrier system b is calculated. Expressed as , is the attitude matrix, which is calculated from the heading angle and horizontal attitude angle provided by the reference system, and the polarization angle is solved from it , combined with the polarization angle measured by the polarization sensor , and then get the polarization angle error , expressed as: (2) Thus, a database of polarization angle errors at different rotation speeds is established.

3. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 2, characterized in that: In the step (2), and The angular velocity output by the gyroscope , the acceleration output by the accelerometer , polarized light intensity of polarized output As input, and The outputs are speed-related features Speed-independent characteristics , thus, the polarization angle error is decoupled as and , the above process is expressed as: (3)。 4. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 1, characterized in that: Speed-related network Speed-independent network The loss function is expressed as: (4) in, Represent the coefficients of reconstruction loss, contrastive learning loss and generative adversarial loss, respectively. Expressed as: (5) in, is 1-norm, contrastive learning loss Expressed as: (6) in, is the positive sample loss, expressed as: (7) in, and From the same speed The data below and , is the negative sample loss, expressed as: (8) in, is the 2-norm, and From different speeds and The following data; Generative Adversarial Loss Expressed as: (9) in, is the discriminator, is the data from zero speed, Is from non-zero speed The data under this condition is used to train a method to extract speed-independent and speed-dependent features. and .

5. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 3, characterized in that: In step (3), when the carrier moves linearly, the three-dimensional attitude of the carrier is estimated using polarization / inertial navigation combined navigation. , and is used to correct the three-dimensional posture of the carrier dynamics model. The system state equation of the dynamics model is expressed as: (10) in, , is the three-axis attitude error angle, is the three-axis angular velocity error, represents the noise of the state equation, is the state transition matrix.

6. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 5, characterized in that: State transition matrix Expressed as: (11) in, is the inertia matrix, represents a diagonal matrix.

7. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 6, characterized in that: 3D attitude using polarization / INS integrated navigation The measurement model of the modified dynamic model is expressed as: (12) Therefore, the dynamic model recursive error is corrected by polarization / inertial navigation combined navigation in the straight line segment, so that the dynamic model can recurse the three-dimensional attitude. .

8. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 7, characterized in that: When the carrier turns, the state equation of the system is expressed as: (13) in, , represents the noise of the state equation, is the state transition matrix.

9. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 8, characterized in that: State transition matrix Expressed as: (14) in, is the attenuation coefficient; The measurement equation is expressed as: (15) Thus, the corrected speed-related characteristics are obtained .

10. The polarization smearing error estimation method based on meta-learning in a high dynamic environment according to claim 9, characterized in that: In step (4), the polarization angle for compensating the trailing error is expressed as: (16) in, is the polarization angle measured by the polarization sensor, To compensate for the polarization angle after tailing interference.

Images