A denoising method for angular acceleration sensor based on dilated convolutional neural network

Abstract

The present invention provides a denoising method for an angular acceleration sensor based on a dilated convolutional neural network. Based on the angular acceleration data measured by the angular acceleration sensor, a dilated convolutional neural network is introduced to correct the angular acceleration data, thereby designing a denoising method that combines the angular acceleration sensor with deep learning. This method takes into account the data processing capacity and training effect of deep learning, adopts operations applied in parallel to multiple instances, significantly improving the training speed. Simultaneously, a cosine annealing learning rate adjustment algorithm is adopted to improve learning efficiency and greatly reduce training time. Therefore, accurate attitude estimation can be obtained using a low-cost angular acceleration sensor. Compared with general neural network training methods, the method of the present invention is robust to overfitting, has high training efficiency, and fast training speed. The angular acceleration sensor introduced by this product is low-cost, making it easy to implement on a large scale.

Description

Technical Field

[0001] The present invention belongs to the field of automatic control technology, and in particular relates to an angular acceleration sensor denoising method based on an expanded convolutional neural network. Background Art

[0002] Attitude estimation technology is crucial and fundamental for path planning and motion control in vehicles like drones and handheld devices. Consequently, researchers have conducted extensive research on obtaining accurate attitude information for these vehicles. Within this field, angular acceleration sensors, as inertial devices, can directly acquire angular acceleration data from the vehicle's motion. This data can be integrated twice to obtain the vehicle's attitude information, improving the usability of low-cost inertial navigation systems.

[0003] As a sensor that directly measures angular acceleration, the angular acceleration sensor itself is subject to noise interference when collecting data. The attitude information obtained by integrating the angular acceleration twice is significantly affected by noise. Traditional methods try to eliminate this noise through selective modeling, but the solution process is complex. Dilated convolutional neural networks can extract features from large amounts of data and train specific parameters by establishing an appropriate loss function to achieve the desired results. Therefore, dilated convolutional neural networks can be used to denoise angular acceleration sensors and improve the accuracy of attitude estimation. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide an angular acceleration sensor denoising method based on a dilated convolutional neural network, which can use the dilated convolutional neural network to denoise the angular acceleration sensor and improve the accuracy of posture estimation.

[0005] A method for denoising an angular acceleration sensor based on a dilated convolutional neural network, comprising:

[0006] The angular acceleration sensor is used to collect the three-axis angular acceleration β at the current moment n. n , and the three-axis angular accelerations at multiple moments before are input into the dilated convolutional neural network to obtain the output value of the network, that is, the angular acceleration correction

[0007] The corrected angular acceleration value is obtained according to the defined angular acceleration sensor error compensation model

[0008]

[0009] Among them, β n is the three-axis angular acceleration measured by the n sensor at the current moment, is the angular acceleration correction, is the corrected angular acceleration value, is the learnable error coefficient matrix;

[0010] Using the corrected angular acceleration value Perform posture calculation to obtain the posture change R between the two frames of data n ;

[0011] The true value of the attitude change is obtained and the attitude change R is calculated by the corrected angular acceleration value. n Compare the results, pass the difference into the loss function, and calculate the loss value. The specific expression is as follows:

[0012]

[0013] Where y represents the reference value, Δ is the difference between the true value of the posture change and the estimated value of the posture change, and δ is the set threshold hyperparameter;

[0014] According to the loss value calculated by the loss function, the gradient is calculated and back-propagated to update the parameters of the dilated convolutional neural network and the learnable parameters of the error compensation model;

[0015] When the training stop condition is reached, the matrix based on the training completion The dilated convolutional neural network uses the input angular acceleration data to obtain the angular acceleration correction The angular acceleration sensor is denoised according to formula (1).

[0016] Preferably, the structure of the dilated convolutional neural network includes the following structures in sequence: a padding layer that performs padding operations on input data; a convolution layer that convolves the data with a set convolution kernel; a normalization layer; a nonlinear layer; a random inactivation layer; and a fully connected layer that outputs data of a predetermined dimension.

[0017] Better, learnable error coefficient matrix Initialize at the beginning of training:

[0018]

[0019] Among them, R is the matrix of corresponding dimension of random assignment, and I is the identity matrix.

[0020] Preferably, the Adam optimizer is used to update the parameters of the dilated convolutional neural network.

[0021] The present invention has the following beneficial effects:

[0022] Based on the angular acceleration data measured by an angular acceleration sensor, the present invention introduces a dilated convolutional neural network to correct the angular acceleration data, designing a denoising method that combines angular acceleration sensors with deep learning. This method takes into account the data processing capacity and training effect of deep learning, and uses operations applied in parallel to multiple instances to significantly increase the training speed. It also uses a cosine annealing learning rate adjustment algorithm to improve learning efficiency and greatly reduce training time, thereby enabling accurate posture estimation using a low-cost angular acceleration sensor. Compared to conventional neural network training methods, the present invention's method is robust to overfitting, has high training efficiency, and fast training speed. The angular acceleration sensor introduced in this product is low-cost, making it easy to implement on a large scale. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Flow chart of the method of the present invention.

[0024] Figure 2 This is a network layer relationship diagram of the expanded convolutional neural network in the present invention. DETAILED DESCRIPTION

[0025] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0026] To address the noise problem in existing angular acceleration sensor measurements, the powerful learning and processing capabilities of neural networks are utilized to correct the original angular acceleration data. This invention provides an angular acceleration sensor denoising method based on a dilated convolutional neural network. The overall steps are as follows:

[0027] Step 1: Parameter initialization. Initialize the network construction parameters such as the input signal channel, convolution kernel size, padding layer parameters, etc.; initialize the training parameters such as learning rate, regularization coefficient, etc.; initialize the optimization parameters such as T_0, T_mult, minimum learning rate, etc.; initialize the compensation model parameters such as error coefficient matrix Initialize and set them as optimizable parameters.

[0028] Step 2: Data collection. Collect the three-axis angular acceleration β through the angular acceleration sensor n .

[0029] Step 3: Data preprocessing: De-mean and normalize the angular acceleration data.

[0030] Step 4: Network extracts information. The angular acceleration data γ nThe input to the expanded convolutional neural network passes through multiple layers of the network, including: 1. A padding layer that performs padding operations on the input data to ensure that key information at the data boundary is retained and to maintain the dimensional consistency of the data when it is transmitted between neural network layers; 2. A convolution layer that convolves the incoming data with the set convolution kernel and expands the data by setting parameters to increase the receptive field of the model; 3. A normalization layer that improves the convergence speed and prevents overfitting; 4. A nonlinear layer that gives the model nonlinear characteristics and enhances its ability to approximate arbitrary nonlinear functions, thereby improving the expressiveness of the neural network; 5. A random deactivation layer that randomly assigns "0" to the convolution result according to the set parameters, and prevents the model from overfitting by deactivating the corresponding elements; 6. A fully connected layer that outputs data of predetermined dimensions. The final output is the angular acceleration correction.

[0031] Step 5: Correct the angular acceleration data. First, design the angular acceleration sensor error model. The formula is as follows:

[0032] β n =Cβ′ n +b n +η n (1)

[0033] Among them, β n is the measurement value of the angular acceleration sensor, β′ n is the true value of angular acceleration, C is the error coefficient matrix, b n is the random walk error, η n is Gaussian white noise with zero mean.

[0034] The error coefficient matrix C is calculated by formula (2):

[0035] C=S n M n (2) Among them, S n is the non-orthogonality error, M n is the proportional error.

[0036] Since C and b in formula (1) n The error compensation model for angular acceleration sensors is designed based on deep learning and uses the learning ability of neural networks to improve output accuracy. The error compensation model for angular acceleration sensors is defined as follows:

[0037]

[0038] Among them, γ nis the three-axis angular acceleration measured by the n sensor at the current moment, is the angular acceleration correction, is the corrected angular acceleration value, is the learnable error coefficient matrix.

[0039] From formula (3), we can know that the angular acceleration correction value of the neural network is Perform training, and the training model is shown in formula (4):

[0040]

[0041] in, is the angular acceleration correction output by the neural network, f(·) is the function defined by the neural network structure, β′ n-N ,…,β′ c is the input of the neural network, representing the angular acceleration data at N time points before time point n. N is determined by the neural network structure. The network parameters will be continuously optimized in step 7 until the optimal value is reached.

[0042] Learnable error coefficient matrix Initialization is performed at the beginning of training. The initialization method is shown in formula (5):

[0043]

[0044] Where R is the matrix of the corresponding dimension of the random assignment, and I is the identity matrix. This parameter will be continuously optimized in step 7 until the optimal value is reached.

[0045] The angular acceleration correction and angular acceleration γ n Input the error compensation model, that is, formula (3), to obtain the corrected angular acceleration value

[0046] Step 6: Loss value and gradient calculation. Using the corrected angular acceleration value Perform posture calculation to obtain the posture change between two frames of data. The calculation method is as follows:

[0047]

[0048] R n =R n-1 exp(ω n dt) (7)

[0049] Among them, ω0 represents the angular velocity at the initial moment, ω n represents the angular velocity of n at the current moment, R n Represents the rotation matrix of the current moment n, R n-1It represents the rotation matrix at the n-1 moment, that is, the previous moment, and dt is the time difference between the two moments.

[0050] Use formula (6) to get the angular velocity ω at the corresponding moment n , calculate formula (7), and we can get the attitude change from time n-1 to time n. The attitude true value is collected by motion capture or high-precision inertial navigation and processed into the attitude change between two frames, that is, the attitude change true value, and the attitude change R calculated by the corrected angular acceleration value. n Compare the difference and pass it into the loss function to calculate the loss value. The loss function selected in this invention is the Huber loss function, which is expressed as follows:

[0051]

[0052] Where y represents a reference value, which is usually close to 0. In this embodiment, it is set to 0. Δ is the difference between the true value of the posture change and the estimated value of the posture change. δ is a threshold hyperparameter that determines the specific expression of the formula.

[0053] According to the loss value calculated by the loss function, the gradient is calculated and backpropagated to train the neural network.

[0054] Step 7: Parameter Optimization. The parameters of the dilated convolutional neural network and the learnable parameters of the error compensation model are updated using the selected optimizer. The Adam optimizer is selected in this paper to significantly reduce training time. During training, the cosine annealing learning rate adjustment algorithm is used to adjust the optimizer learning rate and improve learning efficiency.

[0055] Step 8: Determine whether the number of training times has been reached. If so, proceed to Step 9; if not, skip to Step 4.

[0056] Step 9: Output the trained model. The trained model can then be used to denoise the input angular acceleration data to obtain more accurate angular acceleration data.

[0057] The present invention conducted experiments to verify the algorithm. Considering the following: 1. The algorithm uses attitude information from motion capture or high-precision inertial navigation as the true value, and the attitude value is used as the benchmark for denoising the angular acceleration sensor; 2. Without high-precision angular acceleration data as a reference, it is difficult to select experiments using angular acceleration as the comparison result. Therefore, this experiment selected the attitude value as the comparison result, comparing the attitude angle calculated from the original angular acceleration, the attitude angle calculated from the corrected angular acceleration, and the attitude angle calculated from the angular velocity collected by the gyroscope with the high-precision attitude angle.

[0058] During the experiment, a handheld data acquisition device containing an angular acceleration sensor was first built to collect angular velocity data. A motion capture system was also used to obtain motion information from the data acquisition device, providing the true value of the experimental attitude angle. The initial learning rate was 1e-5, the minimum learning rate was 1e-6, T_0 = 600, and T_mult = 2. The learning rate was adjusted based on the aforementioned parameters and the cosine annealing learning rate adjustment algorithm, and the total number of training cycles was set to 4200. Within the network layer, the padding layer parameter was 520, which is related to the characteristics of convolution. The convolutional layer had input dimensions of [3, 16, 32, 64] and output dimensions of [16, 32, 64, 128]. The convolution kernels were set to [7, 7, 7, 7], and the dilation parameters were [1, 4, 16, 64]. The normalization layer parameter was set to 0.1, and the dropout layer parameter was set to 0.3. The fully connected layer had input dimensions of 128, output dimensions of 3, convolution kernels of 1, and dilation parameters of 1. The hyperparameter δ in the Huber function is selected as 1e-3. The posture solution method adopted in the experiment is as follows:

[0059]

[0060] Among them, q0, q1, q2, q3 are carrier quaternions, Δt is the time difference between two states, [ω x ω y ω z ] is the three-axis angular velocity of the carrier.

[0061] Using formula (9) to obtain the quaternion, the quaternion is converted into the attitude angle parameter, and the attitude solution is completed. The specific error indicators of this experiment are as follows:

[0062]

[0063]

[0064] In the table above, IMU represents attitude calculation using data obtained from the gyroscope, CNN-AC represents attitude calculation using the method proposed in this invention after correcting angular acceleration, and traditional attitude calculation uses raw angular acceleration data. Experimental results show that compared to attitude calculation using uncorrected angular acceleration data, attitude calculation using the corrected angular acceleration data obtained by this method is more accurate, with average errors in the three axes reduced by 62.5%, 26.07%, and 22.98%, respectively.

[0065] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for denoising an angular acceleration sensor based on a dilated convolutional neural network, characterized in that: include: The angular acceleration sensor is used to collect the three-axis angular acceleration β at the current moment n. n , and the three-axis angular accelerations at multiple moments before are input into the dilated convolutional neural network to obtain the output value of the network, that is, the angular acceleration correction value The corrected angular acceleration value is obtained according to the defined angular acceleration sensor error compensation model Among them, β n is the three-axis angular acceleration measured by the n sensor at the current moment, is the angular acceleration correction, is the corrected angular acceleration value, is the learnable error coefficient matrix; Using the corrected angular acceleration value Perform posture calculation to obtain the posture change R between the two frames of data n ; The true value of the attitude change is obtained and the attitude change R is calculated by the corrected angular acceleration value. n Compare the results, pass the difference into the loss function, and calculate the loss value. The specific expression is as follows: Where y represents the reference value, Δ is the difference between the true value of the posture change and the estimated value of the posture change, and δ is the set threshold hyperparameter; According to the loss value calculated by the loss function, the gradient is calculated and back-propagated to update the parameters of the dilated convolutional neural network and the learnable parameters of the error compensation model; When the training stop condition is reached, the matrix based on the training completion The dilated convolutional neural network uses the input angular acceleration data to obtain the angular acceleration correction The angular acceleration sensor is denoised according to formula (1).

2. The angular acceleration sensor denoising method based on a dilated convolutional neural network according to claim 1, characterized in that: The structure of the dilated convolutional neural network includes the following structures in sequence: a padding layer that performs padding operations on input data; a convolution layer that convolves the data with a set convolution kernel; a normalization layer; a nonlinear layer; a random inactivation layer; and a fully connected layer that outputs data of a predetermined dimension.

3. The angular acceleration sensor denoising method based on a dilated convolutional neural network according to claim 1 or 2, characterized in that: Learnable error coefficient matrix Initialize at the beginning of training: Among them, R is the matrix of corresponding dimension of random assignment, and I is the identity matrix.

4. The angular acceleration sensor denoising method based on a dilated convolutional neural network according to claim 1 or 2, characterized in that: The Adam optimizer is used to update the parameters of the dilated convolutional neural network.

Images