IMU data and navigation signal fusion processing method, device, navigation equipment and medium

Abstract

The present invention discloses a method, device, navigation equipment and medium for fusion processing of IMU data and navigation signal, the method comprising: collecting data, constructing an original sampling sequence corresponding to a timestamp; performing data verification, calculating the data difference within the current sampling period, and obtaining a sampling supplement value; performing comprehensive fitting processing on the sampling supplement value; establishing a corresponding relationship between the data, and obtaining an initial fusion parameter; performing incremental calculation, establishing a corresponding relationship between the data increment and the fusion accuracy, and obtaining an incremental fusion sequence; calculating the actual difference between the IMU and navigation data, and obtaining a data deviation value; performing data calibration on the fitting sequence, and obtaining a calibration fusion sequence; performing a secondary data fitting operation, and obtaining a final fitting parameter; applying the final fitting parameter to the calibration fusion sequence, and obtaining a final fusion signal by matching the data point by point. The technical solution of the present invention improves the fusion accuracy, can adapt to complex scenarios, and has dynamic evaluation and exception handling capabilities.

Description

Technical Field

[0001] The present invention relates to the field of vehicle navigation technology, and in particular to an IMU data and navigation signal fusion processing method, device, navigation equipment and medium. Background Art

[0002] With the rapid development of high-precision navigation and positioning requirements, a single navigation system can no longer meet the requirements of practical applications. Although IMU (Inertial Measurement Unit) can provide continuous attitude and motion information, there is a problem of cumulative error; and although satellite navigation signals have high positioning accuracy, they are easily affected by the external environment and signal interruption occurs.

[0003] The existing fusion processing methods mainly use basic algorithms such as data superposition or weighted average, which fail to fully explore and utilize the complementary characteristics of IMU and navigation data; the existing fusion processing methods have the following shortcomings: 1. The simple fusion method fails to consider the reliability differences of data in different scenarios, resulting in difficulty in ensuring fusion accuracy; 2. In the case of strong nonlinear motion such as sharp turns of vehicles, traditional fusion algorithms are prone to divergence, and their high computational complexity makes it difficult to meet real-time processing requirements; 3. In an environment where the navigation signal is interfered, there is a lack of a dynamic evaluation mechanism for the quality of IMU data, and abnormal data cannot be identified and processed in a timely manner; 4. Over-reliance on measurement data ignores the auxiliary role of prior knowledge such as road information and vehicle dynamics characteristics, which limits the further improvement of fusion accuracy. Summary of the invention

[0004] The present invention provides an IMU data and navigation signal fusion processing method, which can achieve deep fusion of IMU data and navigation signals, improve fusion accuracy, adapt to complex scenarios, and have dynamic evaluation and exception processing capabilities.

[0005] To achieve the above purpose, the IMU data and navigation signal fusion processing method proposed in the present invention is applied to navigation equipment, including:

[0006] Collect IMU data and navigation data, and use the multi-frequency sampling mechanism to construct the original sampling sequence corresponding to the timestamp;

[0007] Performing data verification on the original sampling sequence, calculating the data difference within the current sampling period, and obtaining a sampling supplement value;

[0008] The sampling supplementary values ​​are subjected to comprehensive fitting processing by adopting a three-level fitting calculation to obtain a fitting sequence;

[0009] Establishing a corresponding relationship between the IMU data and the navigation data according to the fitting sequence to obtain initial fusion parameters;

[0010] Performing incremental calculation on the initial fusion parameters, establishing a corresponding relationship between data increment and fusion accuracy, and obtaining an incremental fusion sequence;

[0011] Calculating the actual difference between the IMU and the navigation data using the incremental fusion sequence to obtain a data deviation value;

[0012] Using the data deviation value to perform data calibration on the fitting sequence to obtain a calibrated fusion sequence;

[0013] Perform a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters;

[0014] The final fitting parameters are applied to the calibration fusion sequence, and the final fusion signal is obtained by point-by-point matching of the data.

[0015] The present invention also proposes an IMU data and navigation signal fusion processing device, which is applied to navigation equipment, comprising:

[0016] The acquisition module is used to collect IMU data and navigation data, and use the multi-frequency sampling mechanism to construct the original sampling sequence corresponding to the timestamp;

[0017] A verification module, used to perform data verification on the original sampling sequence, calculate the data difference within the current sampling period, and obtain a sampling supplement value;

[0018] A first fitting module is used to perform a comprehensive fitting process on the sampled supplementary values ​​by using a three-level fitting calculation to obtain a fitting sequence;

[0019] A first calculation module, used for establishing a corresponding relationship between the IMU data and the navigation data according to the fitting sequence, and obtaining an initial fusion parameter;

[0020] A second calculation module is used to perform incremental calculation on the initial fusion parameters, establish a corresponding relationship between data increment and fusion accuracy, and obtain an incremental fusion sequence;

[0021] A deviation determination module, used for calculating the actual difference between the IMU and the navigation data using the incremental fusion sequence to obtain a data deviation value;

[0022] A calibration module, used to perform data calibration on the fitting sequence using the data deviation value to obtain a calibration fusion sequence;

[0023] A second fitting module, used for performing a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters;

[0024] The matching module is used to apply the final fitting parameters to the calibration fusion sequence, and obtain the final fusion signal by matching the data point by point.

[0025] The present invention also proposes a navigation device, which includes an IMU sensor, a navigation module, and a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the above-mentioned IMU data and navigation signal fusion processing method are implemented.

[0026] The present invention also proposes a medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned IMU data and navigation signal fusion processing method are implemented.

[0027] The technical solution of the IMU data and navigation signal fusion processing method of the present invention systematically establishes a complete set of dynamic data quality evaluation system through multi-frequency sampling mechanism, multi-level data verification and three-level fitting calculation strategy; this progressive processing framework effectively overcomes the defect of insufficient data reliability evaluation of traditional methods, and significantly improves the fusion accuracy in complex environments; by establishing multi-dimensional data correspondence, combining incremental calculation and deviation analysis, and finally through the construction of calibration fusion sequence, the fusion divergence problem under strong nonlinear motion is effectively solved; and this multi-step progressive processing method effectively improves the data fusion performance under complex environments. In addition, in the quadratic fitting process, an accurate description of different motion characteristics is achieved, and at the same time, through the hierarchical matching strategy, the smooth characteristics of the ideal trajectory are retained without losing the effective information in the actual data, effectively improving the accuracy of navigation positioning; this data matching method based on quadratic fitting successfully solves the fusion problem under strong nonlinear motion. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of an embodiment of a method for fusion processing of IMU data and navigation signals of the present invention;

[0029] Figure 2 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S20 in the embodiment;

[0030] Figure 3 for Figure 2 A schematic diagram of a flow chart of an embodiment of step S23 in the embodiment;

[0031] Figure 4 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S40 in the embodiment;

[0032] Figure 5 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S50 in the embodiment;

[0033] Figure 6 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S80 in the embodiment;

[0034] Figure 7 A schematic diagram of the module structure of an embodiment of an IMU data and navigation signal fusion processing device of the present invention;

[0035] Figure 8 It is a schematic diagram of the structure of an electronic device in a hardware operating environment involved in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of 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.

[0037] The present invention proposes a method for fusing IMU data and navigation signals, which is applied to navigation equipment. The implementation terminal of the method for fusing IMU data and navigation signals of the present invention can be a navigation equipment or an electronic device, which can be a controller of the navigation equipment, or a computing device such as a single-chip microcomputer, a desktop computer, a notebook, a palm computer, and a server. In the following, the implementation terminal of the method for fusing IMU data and navigation signals takes an electronic device as an example.

[0038] like Figure 1 As shown, Figure 1 It is a flow chart of the first embodiment of the IMU data and navigation signal fusion processing method of the present invention.

[0039] In this embodiment, the IMU data and navigation signal fusion processing method includes:

[0040] Step S10, collecting IMU data and navigation data, and constructing an original sampling sequence corresponding to the timestamp using a multi-frequency sampling mechanism.

[0041] Collecting dual-source data (i.e., IMU data and navigation data) from IMU (Inertial Measurement Unit) and navigation (such as satellite navigation module), and using a multi-frequency sampling mechanism to construct an original sampling sequence corresponding to the timestamp is the primary link in the entire fusion electronic device.

[0042] Among them, the IMU data can be collected through the built-in three-axis accelerometer and three-axis gyroscope. The accelerometer range is set to ±16g to capture linear acceleration changes, with an accuracy better than ±0.1mg, zero bias stability better than ±0.05mg, and noise density better than 50μg / √Hz. The gyroscope range is set to ±2000° / s to obtain angular velocity information, zero bias stability better than ±0.1° / h, and random walk error better than ±0.02° / √h. The analog signal output by the sensor is converted by the internal 16-bit ADC and transmitted to the data processing unit using the SPI interface at a clock frequency of 20MHz, ensuring the real-time and integrity of data transmission.

[0043] The collection of satellite data can be achieved by relying on a multi-mode and multi-frequency GNSS (Global Navigation Satellite System) receiver. The receiver adopts a broadband RF front-end design and supports parallel processing of multiple frequency band signals such as GPS L1 / L2 / L5, Beidou B1 / B2 / B3, Galileo E1 / E5 / E6. The noise coefficient of the RF front-end is better than 2.5dB, and the gain dynamic range reaches 100dB, which can effectively process weak navigation signals. The signal capture adopts a parallel FFT algorithm, and the 512-point FFT transform is combined with non-coherent integration to achieve a signal capture sensitivity of -160dBm. The navigation signal is processed by a carrier tracking loop and a code tracking loop. The carrier tracking adopts a third-order PLL design, and the loop bandwidth can be dynamically adjusted between 10-18Hz; the code tracking adopts a second-order DLL, and the correlator interval is set to 0.5 code chips to ensure measurement accuracy.

[0044] In view of the different characteristics of dual-source data, the electronic device adopts a multi-frequency sampling mechanism. For example, the IMU data sampling frequency is set to 200Hz, that is, a set of data is collected every 5ms. This frequency selection is based on the analysis of the carrier motion characteristics: through the spectrum analysis of typical application scenarios, it is found that the main frequency components of the carrier motion are concentrated in the range of 0-40Hz. According to the Nyquist sampling theorem, the minimum sampling frequency needs to be greater than 80Hz, so the sampling frequency of 200Hz not only meets the requirements, but also provides 2.5 times the signal reconstruction margin. The sampling adopts an equal interval trigger method, and the sampling clock is generated by a precise timer circuit, and the clock jitter is controlled within 100ns. For example, the navigation data is sampled at a frequency of 10Hz, and a complete set of navigation solution results is obtained every 100ms. The electronic device uses a loose combination strategy based on Kalman filtering to process the navigation data, and the state vector contains 15 dimensions including position, velocity and attitude. The measurement update of the filter is synchronized with the navigation data sampling, and the prediction update is synchronized with the IMU data sampling. Considering the continuity characteristics of the carrier motion, the navigation data sampling rate of 10Hz can accurately capture the trajectory changes of the carrier.

[0045] The time synchronization of the sampling process uses a local high-precision clock source, and the PPS signal output by the GNSS receiver is used for period calibration. The rising edge accuracy of the PPS signal is better than 50ns, and the short-term stability reaches the order of 10^(-11). The electronic equipment uses a temperature-compensated crystal oscillator as the local clock source, and the phase noise is better than -130dBc / Hz@1kHz. After the PPS signal period calibration, a 200Hz IMU sampling clock and a 10Hz navigation data sampling clock are generated. The phase-locked loop filter bandwidth of the clock source is set to 2Hz, which ensures fast locking and suppresses clock jitter.

[0046] The electronic device establishes a unified sequence organization structure for the collected dual-source data. In the IMU sampling sequence, each data frame contains 64 bytes of information: 8 bytes of timestamp, 24 bytes of measurement data (acceleration and angular velocity), 4 bytes of status flag, and 4 bytes of checksum. The data is stored in little-endian format and arranged continuously at 5ms intervals. Each data frame of the navigation data sequence contains 128 bytes: 8 bytes of timestamp, 24 bytes of position data, 24 bytes of velocity data, 32 bytes of status information, and 4 bytes of checksum. All data are stored in the floating-point format of the IEEE754 standard.

[0047] The electronic device synchronously samples the IMU and navigation data through a high-precision clock reference, establishes a one-to-one correspondence between data with different sampling rates through timestamps, and forms a unified original sampling sequence. In this sequence, each timestamp node contains complete dual-source data information: the navigation data records macro states such as position and speed, and the IMU data provides the posture and motion details at the corresponding moment. Through this timestamp-based sequence construction method, the electronic device realizes the time sequence correspondence of the dual-source data and completes the construction of the original sampling sequence.

[0048] Step S20, performing data verification on the original sampling sequence, calculating the data difference within the current sampling period, and obtaining the sampling supplement value.

[0049] When checking the data of the original sampling sequence, if the electronic device finds that the GNSS signal quality has decreased due to the obstruction of tall buildings near a certain road section (such as a ring road section), and / or, near a certain road section (such as an S-shaped road section), some GNSS data has jumped due to the obstruction of dense trees; the electronic device will supplement these abnormal data by calculating the data difference within the sampling period. For example, when the GNSS signal is poor, it relies more on IMU data for supplementation; when the IMU data has abnormal fluctuations, it is corrected in combination with GNSS data.

[0050] Combination Figure 2 As shown, in some embodiments, step S20 may include: step S21, step S22 and step S23; wherein:

[0051] Step S21, performing multi-level data verification on the original sampling sequence, wherein the multi-level data verification includes performing data integrity verification, time stamp verification and data validity verification;

[0052] The data verification of the original sampling sequence first needs to ensure the integrity and validity of the data. The electronic device has designed a multi-level data verification mechanism to conduct comprehensive verification from the bottom-level data integrity to the top-level data consistency. At the hardware level, the raw data output by the IMU and GNSS receivers is CRC-checked. The check polynomial uses CRC-32, which can detect 99.99999997% of sudden errors. The verification process is implemented by the table lookup method. 256 32-bit CRC values ​​are pre-calculated and stored in the lookup table. Each time, only table lookup and XOR operation are needed to quickly complete the CRC check, which greatly improves the real-time performance. The electronic device also implements a data packet header verification mechanism, which identifies the starting position of the data packet through specific synchronization words and packet length fields, effectively preventing parsing errors caused by data misalignment. For IMU data, the electronic device introduces a range verification mechanism: the acceleration measurement value a_meas must be in the range of [-16g, +16g], and the angular velocity measurement value ω_meas must be in the range of [-2000° / s, +2000° / s]. Data outside this range will be marked as an outlier and require further processing.

[0053] In order to ensure the time consistency of the data, the electronic device strictly checks the timestamp. Since the IMU sampling frequency is 200Hz, the time interval δt between adjacent sampling points should be constant at 5 ms in theory. The electronic device sets a time tolerance of ±100μs. Once the time interval is found to exceed the range of 4.9 ms to 5.1 ms, the data compensation mechanism will be triggered. The timestamp check uses a sliding window technique to calculate the statistical characteristics of the time interval in a 320 ms window (corresponding to 64 IMU sampling points), including the mean μ_t, variance σ²_t and skewness S_t. This statistical analysis can effectively identify electronic device problems such as clock drift and timer jitter. For 10Hz navigation data, the electronic device also sets a time tolerance of ±1 ms to ensure the timing accuracy of the navigation solution results. The electronic device maintains a timestamp index structure based on a red-black tree, which can complete the search and insertion of timestamps within the time complexity of O(log n).

[0054] In terms of data validity verification, the electronic device also implements a sensor self-test mechanism. For IMU, the zero bias stability and noise level of the sensor are evaluated by analyzing the output characteristics in a stationary state. When the electronic device is started, it will automatically perform a 30-second static sampling process to collect enough samples to calculate the noise characteristics of the sensor. Through Allan variance analysis, the electronic device can identify error characteristics such as angle random walk (σ_θ), rate random walk (σ_ω), bias instability (σ_b), and dynamically adjust subsequent data processing parameters based on the analysis results. For GNSS receivers, the electronic device evaluates the quality of the navigation solution by monitoring parameters such as carrier-to-noise ratio (C / N0), multipath index (MP), and geometric dilution of precision (GDOP).

[0055] Step S22, on the basis of ensuring data validity, adopting a hierarchical difference calculation strategy, calculates the data difference within the current sampling period to obtain data difference features;

[0056] On the basis of ensuring the validity of the data, the electronic device begins to calculate the data difference within the current sampling period. Taking into account the sampling frequency difference between IMU and navigation data (200Hz vs 10Hz), the electronic device adopts a hierarchical difference calculation strategy. For IMU data, there are 20 IMU sampling points in each 100 ms navigation sampling period. The electronic device first calculates the statistical characteristics of these 20 sampling points, including the mean μ_x, standard deviation σ_x, maximum value x_max and minimum value x_min. In particular, for acceleration data, the electronic device calculates the root mean square value RMS_a = √(a_x² + a_y² + a_z²) of the three-axis composite acceleration as an indicator of the carrier's motion intensity. Double-precision floating-point operations are used in the calculation process, and the Kahan summation algorithm is used to reduce the cumulative error. For angular velocity data, the electronic device calculates the modulus change rate of the angular velocity vector Δω= |ω(t) -ω(t-Δt)| / Δt, which is used to characterize the steering characteristics of the carrier.

[0057] The calculation of data differences uses adaptive window technology. In a state of stillness or uniform linear motion, the electronic device uses a 500 ms time window (corresponding to 100 IMU sampling points) for difference analysis to improve noise suppression capabilities. When violent motion is detected (such as sharp turns or emergency braking), the electronic device automatically narrows the time window to 100 ms (corresponding to 20 IMU sampling points) to improve the ability to capture dynamic characteristics. The adaptive adjustment of the window size is based on the variance analysis of the IMU data: when the variance σ²_a exceeds the preset threshold σ²_threshold, it is considered that the carrier is in a violent motion state and the observation window needs to be narrowed.

[0058] Step S23, obtaining a sampling supplement value using the data difference feature, and supplementing the navigation data.

[0059] After obtaining the detailed data difference characteristics, the electronic device supplements the navigation data based on the data difference characteristics analysis of the sampling supplement value.

[0060] Combination Figure 3 As shown, in some embodiments, the step of supplementing the navigation data may include:

[0061] Step S231, using a prediction mechanism based on Kalman filtering, between two adjacent navigation sampling points, using the acceleration and angular velocity information provided by the IMU data, and predicting the position and attitude changes of the carrier through the state equation;

[0062] Step S232: Using the prediction result of the state equation as a supplementary value of the navigation data to fill in the data gaps between the navigation sampling points of the navigation data.

[0063] The electronic device can use a prediction mechanism based on Kalman filtering to predict the position and attitude changes of the carrier through the state equation using the acceleration and angular velocity information provided by the IMU between two adjacent navigation sampling points. The state vector x contains 15 components: 3D position (p_x, p_y, p_z), 3D velocity (v_x, v_y, v_z), 3D attitude angle (θ_x, θ_y, θ_z) and 6D sensor errors (accelerometer bias b_a_x, b_a_y, b_a_z and gyroscope bias b_g_x, b_g_y, b_g_z). The electronic device equation takes into account the effects of the earth's rotation and gravity changes, and the integration step is set to 1 ms to ensure calculation accuracy. The prediction results are used as supplementary values ​​for navigation data to fill in the data gaps between 10Hz sampling. At the same time, the electronic device will dynamically adjust the noise parameters in the prediction process according to the quality of the IMU data to optimize the prediction effect.

[0064] Through the above data verification and difference calculation process, the electronic device fully processes the original sampling sequence, calculates the data difference within the sampling period, and finally obtains the required sampling supplementary values. These supplementary values ​​not only ensure the integrity of the data, but also maintain the time consistency of the data.

[0065] Step S30, using a three-level fitting calculation to perform a comprehensive fitting process on the sampled supplementary values ​​to obtain a fitting sequence;

[0066] The electronic device adopts three levels of fitting calculation, mainly based on the following considerations: IMU and GNSS data have different characteristics and advantages, and optimal fusion can only be achieved through a progressive fitting process. Step S30 may include: firstly, single data fitting is performed on the IMU data and navigation data in the sampled supplementary values, wherein the single data fitting of the IMU data adopts piecewise polynomial fitting, and the single data fitting of the navigation data adopts cubic B-spline interpolation; then, double data fitting is performed on the sampled supplementary values, and the preliminary association between the two types of data in different sections is established through the extended Kalman filter framework; finally, the single data fitting results and the double data fitting results are unified under the Bayesian estimation framework for combined data fitting to obtain a fitting sequence.

[0067] When entering the fitting calculation stage, the electronic equipment first performs single data fitting on the IMU and GNSS data respectively to retain their respective characteristic advantages; then, through dual data fitting, a preliminary association between the two types of data in different road sections such as loop sections and straight sections is established; finally, in the combined data fitting stage, the results of the first two steps are optimized as a whole to obtain a complete fitting sequence.

[0068] First, single data fitting is performed to retain the advantages of each type of characteristics, such as the high dynamic response of IMU data and the long-term stability of GNSS data; secondly, dual data fitting can establish the correlation between the two types of data and preliminarily solve the data fusion problem; the final combined data fitting is based on the first two steps for overall optimization to obtain the final fitting sequence. This progressive fitting strategy can give full play to the complementary advantages of the two types of data. For example, when a vehicle enters an S-shaped road from a loop, the single data fitting retains the IMU's accurate description of the steering process and the GNSS's absolute position reference, while the dual data fitting establishes the corresponding relationship between the steering angle and the position change. Finally, the accurate motion trajectory is obtained through combined data fitting.

[0069] In the single data fitting stage, the electronic device first performs piecewise polynomial fitting on the IMU supplementary value provided in step 2. According to the motion state characteristics identified in the supplementary value, the data sequence is divided into multiple continuous segments. The selection of segmentation points is based on three judgment criteria: the acceleration change rate |da / dt| exceeds the threshold of 2 m / s³, the angular velocity mutation |dω / dt| exceeds the threshold of 30° / s², and the statistical characteristic difference exceeds 2σ. For each data segment, the electronic device uses a polynomial fitting: f(t) =∑(αi·φi(t)), i = 0, 1, ..., m, where φi(t) is the i-order basis function for state expression, and αi is the fitting coefficient to be determined, indicating the weight of each order state. The electronic device adaptively selects the order of the basis function according to the motion characteristics: m = 3 is selected for the uniform motion segment, and m = 5 is selected for the variable acceleration motion segment. In order to ensure the stability of numerical calculations, orthogonal polynomial electronic devices (such as Legendre polynomials) are used as basis functions.

[0070] For the GNSS supplementary value provided in step 2, considering its low sampling rate characteristics, cubic B-spline interpolation is used for fitting: B(t) =Σ(i=0 to n) c_i·N_i,3(t), where N_i,3(t) is the third-order B-spline basis function, c_i is the control point coefficient, and n is the number of control points. The electronic device dynamically adjusts the node distribution according to the accuracy index of GNSS data (such as GDOP value), appropriately reduces the number of nodes in areas with good data quality to smooth the noise, and increases the node density in areas with poor accuracy to maintain local characteristics.

[0071] The dual data fitting stage mainly solves the coordination problem of IMU and GNSS data. The electronic device constructs a 15-dimensional state vector x = [p^T, v^T, θ^T, b_a^T, b_g^T]^T, where p = [p_x, p_y, p_z]^T is a 3-dimensional position vector, v = [v_x, v_y, v_z]^T is a 3-dimensional velocity vector, θ = [θ_x, θ_y, θ_z]^T is a 3-dimensional attitude angle vector, b_a = [b_ax, b_ay, b_az]^T is a 3-dimensional accelerometer bias vector, and b_g = [b_gx, b_gy, b_gz]^T is a 3-dimensional gyroscope bias vector. State estimation is performed based on the extended Kalman filter framework: x(k+1) = f(x(k), u(k)) + w(k), z(k) = h(x(k)) + v(k), where f(·) represents the system state equation, h(·) is the observation equation, u(k) is the system input, w(k) is the process noise that obeys the N(0, Q) distribution, and v(k) is the observation noise that obeys the N(0, R) distribution. The electronic device dynamically adjusts the noise covariance matrix according to the data quality assessment result of step S20, thereby realizing adaptive optimization of the filter performance.

[0072] In the combined data fitting stage, the electronic device unifies the results of single data fitting and double data fitting into the Bayesian estimation framework: x* = argmax(P(x|z)) = argmax(P(z|x)P(x)), where P(z|x) represents the observation likelihood function and P(x) represents the state prior distribution. The electronic device adopts a block processing strategy to improve the computational efficiency, divides the data sequence into several overlapping time blocks, performs Bayesian estimation independently in each data block, and then achieves smooth transition between blocks through the weighted average method.

[0073] Through this multi-level fitting calculation, the electronic device performs a comprehensive fitting process on the sample supplementary value provided in step S20, and finally obtains a fitting sequence. This sequence not only retains the effective characteristics of various types of data, but also realizes the organic fusion of data through multi-level fitting, and prepares for the subsequent establishment of the corresponding relationship between IMU data and navigation data.

[0074] Step S40, establishing a corresponding relationship between the IMU data and the navigation data according to the fitting sequence to obtain initial fusion parameters.

[0075] Based on the fitting sequence, the electronic device establishes the relationship between the time domain correspondence and the state quantity correspondence. At the time domain level, data alignment is ensured through precise time synchronization; at the state quantity level, multi-dimensional mapping is established in combination with different motion characteristics. For example, in the circular road section, the mapping relationship between angular velocity and position change is focused on, and in the straight road section, the correspondence between acceleration and speed change is focused on. In the S-shaped road section, the coupling relationship between multiple state quantities such as position, speed, and posture is considered at the same time, and finally a complete initial fusion parameter is obtained.

[0076] From the fitting sequence, the electronic device makes full use of the respective characteristics retained by single data fitting, the preliminary association established by double data fitting, and the state information after the combined data fitting optimization, and constructs a complete corresponding relationship through the two levels of time domain correspondence and state quantity correspondence.

[0077] Combination Figure 4 As shown, in some embodiments, step S40 may include step S41, step S42 and step S43; wherein:

[0078] Step S41, using the timestamp information in the fitting sequence to establish a time synchronization mapping between the IMU data and the navigation data;

[0079] In the time domain, the electronic device first uses the timestamp information in the fitting sequence to establish a time synchronization mapping between the IMU and navigation data. Considering the difference in sampling frequency between the two data, the electronic device uses a piecewise linear interpolation method to align the time reference. The time synchronization process can be expressed as:

[0080] T_sync =α·T_imu + (1-α)·T_gnss

[0081] Where α is the time weight coefficient, which is determined by minimizing the variance of the time deviation. The electronic device can use the sliding window technique to calculate the local time deviation, and the window size is set to 200 IMU data points, corresponding to 10 navigation data points. When the time deviation is detected to exceed the threshold (typical value is 1ms), the electronic device will trigger the time reference recalibration mechanism.

[0082] On the basis of establishing the time correspondence, the electronic device needs to build a differentiated correspondence model for different motion characteristics. In the turning motion segment, due to the rapid accumulation of inertial navigation attitude errors, the electronic device focuses on establishing the mapping relationship between the angular velocity integral and the GNSS position change; in the uniform motion segment, due to the stable motion state, the electronic device mainly focuses on the consistency of the acceleration integral and the GNSS velocity; in the variable acceleration motion segment, it is necessary to consider the coupling between multiple state quantities at the same time and establish a more complex correspondence. The electronic device establishes differentiated state mapping relationships for different road sceneries. For example, when the vehicle makes a large angle turn, it focuses on building the correspondence between the angular velocity output by the gyroscope and the GNSS displacement change, and sets a higher gyroscope data weight; in the straight acceleration and deceleration section, it focuses on establishing the mapping relationship between the accelerometer output and the GNSS velocity, and increases the weight coefficient of the acceleration data; in the complex continuous change of direction section, the electronic device considers the coupling effects of multiple dimensions of position, velocity and attitude through the complete mapping of the state vector. This scene-adaptive mapping mechanism ensures that the electronic device can flexibly adjust the fusion strategy according to the actual motion characteristics.

[0083] Step S42, constructing a three-layer correspondence model based on the established data association in the fitting sequence, wherein the three layers of correspondence are position correspondence, velocity correspondence and posture correspondence;

[0084] At the state level, based on the established data association in the fitting sequence, the electronic device builds a three-layer correspondence model. This process is divided into three levels: position correspondence, speed correspondence, and attitude correspondence. For the position correspondence, the electronic device uses a spatial transformation model:

[0085] P_fusion = R·P_imu + T +ε

[0086] Among them, R is a 3×3 orthogonal rotation matrix that satisfies R·R^T = I (I is the unit matrix), T is the translation vector, and ε is the residual term. The initial values ​​of R and T are determined by least squares estimation, and then the Huber estimator is used for robust optimization to effectively suppress the influence of outliers. The optimization process adopts an iterative weighted strategy, and the weight function is designed as:

[0087] w(x) = { 1, |x| ≤ kk / |x|, |x|>k}

[0088] Among them, k is the threshold parameter, which is dynamically determined according to the interquartile range of the data distribution.

[0089] The establishment of the velocity correspondence needs to consider the cumulative effect of the IMU integral error. The electronic device has designed an error compensation mechanism based on zero velocity detection. When the carrier is detected to be in a stationary state (the standard deviation of the acceleration and angular velocity are both less than the preset threshold), this moment is used as the velocity error calibration point. The velocity correspondence model is expressed as:

[0090] V_fusion = V_imu +∫ε_a dt +δv

[0091] Among them, ε_a is the acceleration error term, and δv is the velocity compensation. The electronic device establishes the error state equation by analyzing the time series characteristics of the velocity error:

[0092] δv(k+1) = Φ·δv(k) + w(k)

[0093] Where Φ is the 3×3 state transfer matrix and w(k) is the process noise. This error modeling method effectively compensates for the electronic device error in the IMU integration process.

[0094] The establishment of attitude correspondence is expressed by quaternion. The electronic device first uses IMU data to calculate the attitude change rate:

[0095]

[0096] Among them, q is the attitude quaternion, ω is the angular velocity vector, Represents quaternion multiplication. Then compare the calculated result with the attitude calculated by navigation to establish the attitude error model:

[0097]

[0098] Among them, q_gnss^(-1) represents the conjugate of the quaternion of the GNSS solution attitude. represents quaternion multiplication, and R(·) represents the conversion of quaternion into equivalent rotation matrix. This representation method directly gives the deviation angle of the IMU-solved attitude relative to the GNSS-solved attitude.

[0099] Step S43, after constructing the model of the three-layer correspondence relationship, perform initial fusion parameter estimation to obtain initial fusion parameters;

[0100] After the three-layer correspondence is fully established, the electronic device begins to estimate the initial fusion parameters. First, the sensor error characteristic parameters are obtained: when the vehicle is stationary (such as waiting for a traffic light or parking in a parking lot), the electronic device collects the IMU output data and obtains the zero bias vector b of the accelerometer and gyroscope through statistical analysis. For the scale factor matrix S, the electronic device uses the data of the vehicle's uniform straight-line driving section for calibration. At this time, the accelerometer output should be close to zero, and the gyroscope output should also be close to zero. By comparing the deviation between the actual output and the theoretical value, the proportional relationship between the sensor output and the real physical quantity is established.

[0101] Then the electronic device estimates other fusion parameters, including the state transfer matrix F, the process noise matrix Q and the measurement noise matrix R. The estimation of the state transfer matrix is ​​based on the linearization result of the corresponding relationship model:

[0102]

[0103] Where f(·) is the state equation, is the state estimate. The electronic device uses numerical difference method to calculate partial derivatives to reduce the complexity of analytical derivation. The estimation of process noise matrix Q uses innovative sequence method:

[0104] Q = E[e_k·e_k^T] - H·P·H^T

[0105] Here e_k is the innovation sequence, H is the measurement matrix, and P is the state covariance matrix. The estimation of the measurement noise matrix R is based on the accuracy index of the navigation data:

[0106] R = diag(σ_p², σ_v², σ_θ²)

[0107] Among them, σ_p, σ_v, and σ_θ represent the measurement standard deviations of position, velocity, and attitude, respectively.

[0108] Through the above correspondence modeling and parameter estimation, the electronic device finally obtains a complete set of initial fusion parameters θ, including:

[0109] 1. State transfer matrix F: describes the law of the evolution of the state of the electronic device over time. It has a dimension of 15×15 and contains the mapping relationship between state quantities such as position, speed, and posture;

[0110] 2. Process noise matrix Q: It characterizes the uncertainty of the dynamic characteristics of the electronic device. It is also a 15×15 dimensional matrix, and its diagonal elements correspond to the noise intensity of each state quantity.

[0111] 3. Measurement noise matrix R: describes the uncertainty of the observation data, which is a 9×9 dimensional matrix corresponding to the observation errors of position, velocity and attitude;

[0112] 4. State covariance matrix P: reflects the accuracy of state estimation. The initial value is obtained by statistical analysis of the fitting sequence in step 3.

[0113] 5. Electronic device bias parameter vector b: includes accelerometer bias ba and gyroscope bias bg, which are estimated by analyzing the static data segment;

[0114] 6. Scale factor matrix S: describes the proportional relationship between the sensor output and the real physical quantity, in the form of a diagonal matrix.

[0115] This set of initial fusion parameters θ = {F, Q, R, P, b, S} fully describes the necessary information required for the fusion of IMU and navigation data. Compared with traditional methods, this parameterized representation has the following advantages: First, it retains the characteristics of the two types of data. The high dynamic characteristics of IMU data are reflected by the F and Q matrices, and the long-term stability of navigation data is characterized by the R matrix; second, the parameter set contains a complete uncertainty description, so that the subsequent fusion processing can fully consider the data reliability; finally, this parameterized structure is convenient for subsequent online updates and optimizations.

[0116] Step S50, performing incremental calculation on the initial fusion parameters, establishing a corresponding relationship between data increment and fusion accuracy, and obtaining an incremental fusion sequence.

[0117] By calculating the increments of the initial fusion parameters, the electronic device processes the full data of the preset road section (for example, from the roundabout to the parking lot). For example, when waiting to enter the roundabout, the electronic device calculates the state increment in the static state and determines the fusion accuracy based on the stability of the acceleration and angular velocity; in the straight section of the main road, the incremental change of the state parameters is calculated based on the uniform motion characteristics, and the accuracy correspondence between the speed increment and the position change is established; and in the roundabout and S-shaped road sections, the electronic device calculates the incremental change of the posture parameters for large-scale steering and dynamically adjusts the fusion weights to adapt to violent movements. Through this piecewise incremental calculation, the electronic device finally generates an incremental fusion sequence containing state estimates and corrections.

[0118] Combination Figure 5 As shown, in some embodiments, step S50 may include step S51 and step S52.

[0119] Step S51, constructing an incremental state expression form using a state transfer matrix;

[0120] The electronic device first constructs the incremental state expression using the state transfer matrix F. Define the incremental state vector , where δx(k) is a 15-dimensional state increment vector [δp^T,δv^T,δθ^T,δb_a^T,δb_g^T]^T, including position increment δp=[δp_x,δp_y,δp_z]^T, velocity increment δv=[δv_x,δv_y,δv_z]^T, attitude increment δθ=[δθ_x,δθ_y,δθ_z]^T, accelerometer bias increment δb_a=[δb_ax,δb_ay,δb_az]^T and gyroscope bias increment δb_g=[δb_gx,δb_gy,δb_gz]^T; x(k) is the real state, This incremental state expression describes the evolution law of each state component through the state transfer matrix F, and realizes the complete characterization of the actual deviation of state quantities such as position, velocity and posture.

[0121] Step S52, performing incremental calculation based on the state incremental expression form, and adopting an adaptive fusion strategy based on motion features to obtain an incremental fusion sequence;

[0122] The electronic device uses the constructed state increment vector to design the increment calculation method. Since the increment vector contains the actual deviations of 15-dimensional state quantities such as position, velocity and attitude, in order to optimize these deviations, the electronic device designs a calculation function based on the maximum likelihood estimation principle: L(δx) = -0.5·(δx^T·P^(-1)·δx + ln|P|). Where P is the 15×15 state covariance matrix and P^(-1) is its inverse matrix. The calculation function contains two terms: the first term δx^T·P^(-1)·δx is the weighted quadratic form of the state increment, and each state component is weighted by the inverse matrix of the covariance matrix, reflecting the size of the state deviation; the second term ln|P| is the logarithmic term of the determinant of the covariance matrix, which characterizes the degree of uncertainty in the state estimation. This design realizes a comprehensive evaluation of state deviation and estimation uncertainty. The electronic device uses an improved gradient descent method to optimize the likelihood function and introduces a dynamic learning rate adjustment mechanism to avoid parameter oscillation while ensuring the convergence speed. Specifically, when the gradient directions of multiple consecutive iterations are similar, the learning rate is appropriately increased to accelerate convergence; when the gradient direction changes drastically, the learning rate is reduced to improve stability.

[0123] Based on this calculation function framework, the electronic device optimizes the calculation of each state quantity respectively. When processing the position increment, the electronic device combines the measurement noise matrix R for adaptive optimization. The horizontal position increment uses 0.5m as the reference threshold, and the vertical position increment uses 1.0m as the reference threshold. When the position increment exceeds these thresholds, the electronic device adjusts the corresponding elements in the R matrix according to the current GDOP value. The adjustment formula is σ_p = α·GDOP, where the proportional coefficient α changes dynamically with the number of satellites. For example, when the number of visible satellites is greater than 12, α is 0.1; when the number of satellites is between 8-12, α is 0.15; when the number of satellites is less than 8, α is 0.2. This segmented adjustment strategy effectively balances the positioning accuracy under different satellite observation conditions.

[0124] The velocity increment processing optimizes both the process noise matrix Q and the accelerometer parameters. The velocity increment evaluation model uses: σv = k·|δv| + c·|a|, where k and c are coefficients related to the noise intensity of the velocity component in the Q matrix. The electronics analyze the acceleration data in segments, focusing on calibrating the zero bias b_a in the stationary segment, optimizing the scale factor S_a in the uniform speed segment, and focusing on the noise characteristics in the variable speed segment. A dedicated parameter optimization strategy is used for each motion state, such as using the recursive averaging method to estimate the zero bias in the stationary segment, optimizing the scale factor through the least squares method in the uniform speed segment, and using an adaptive Kalman filter to update the noise parameters in the variable speed segment.

[0125] During the attitude increment analysis and calculation process, the electronic equipment focuses on optimizing the gyroscope parameters S_g and b_g. The electronic equipment collects angular velocity data for 10 consecutive seconds, obtains the attitude change by integration, and then compares it with the observed attitude to evaluate the gyroscope performance. During this analysis process, the electronic equipment found that during continuous steering movement, due to the cumulative effect of the gyroscope zero bias, the heading angle solved by the IMU will drift significantly; and in areas blocked by tall buildings or dense trees, the GNSS signal quality decreases, resulting in reduced positioning accuracy. In response to these characteristic deviations, the electronic equipment takes compensation measures respectively: for gyroscope drift, when the temperature changes significantly (the temperature change rate exceeds 2°C / minute), the temperature compensation mechanism is activated, and a third-order polynomial is used to describe the relationship between temperature and gyroscope parameters; for GNSS signal quality issues, the observation weight is adjusted to reduce its impact.

[0126] Step S53: Based on the optimization calculation result and the motion characteristics, an adaptive state fusion strategy is used to perform state fusion to obtain an incremental fusion sequence.

[0127] Based on the above optimization calculation results, the electronic device adopts an adaptive state fusion strategy according to the differences in motion characteristics. Specifically: when continuous uniform linear motion is detected, the electronic device adopts a linear fusion model to directly perform weighted fusion of position increments, velocity increments and attitude increments due to the linear characteristics of position and velocity changes; when continuous steering motion is identified, due to the complex attitude changes and velocity coupling, the electronic device automatically switches to a nonlinear fusion model, which captures more state change details by increasing the sampling frequency, and quickly responds to state changes by shortening the calculation window, thereby achieving accurate fusion of nonlinear motion characteristics.

[0128] At the same time, the electronic device realizes the adaptive update of the initial fusion parameter set θ: the state transfer matrix F is mainly updated at the motion state transition point; the noise matrices Q and R are adjusted based on the statistical characteristics of the data residual; the sensor parameters b_a, b_g, S_a, S_g are dynamically updated according to their respective optimization strategies. These updated parameters are directly used in the generation process of the incremental fusion sequence to ensure that each data point in the sequence is based on the optimal parameter estimation.

[0129] Through the above incremental calculation and state fusion process, the electronic device finally obtains the incremental fusion sequence {X_k, k=1, 2, ..., N}. Each element X_k of the incremental fusion sequence contains a fusion state vector obtained based on the linear / nonlinear fusion model. The vector integrates the optimization results of various state quantities such as position, speed, and attitude, and reflects the complete state information of the electronic device at that moment.

[0130] Step S60, using the incremental fusion sequence to calculate the actual difference between the IMU and navigation data to obtain a data deviation value.

[0131] The electronic device analyzes the incremental fusion sequence and calculates the actual difference in the data. For example, in the loop section, it is found that the IMU data has accumulated errors due to continuous steering, and the attitude and angular velocity deviations are calculated; in the area blocked by buildings, the position and velocity deviations caused by the GNSS signal jump are analyzed to obtain a complete set of data deviation values.

[0132] The electronic device first builds a difference calculation framework based on the incremental fusion sequence. The specific approach is to perform state decomposition on each element X_k of the incremental fusion sequence and extract the state quantities such as position, velocity, and attitude. The state decomposition adopts the quaternion representation method, obtains the attitude information through the conversion of quaternion and direction cosine matrix, and extracts the position and velocity components at the same time. For the extraction of IMU data, the electronic device first preprocesses the original data, including noise reduction and outlier removal. At the same time, the original navigation solution results at the corresponding time are obtained. The electronic device uses the sliding window method for data alignment, with the window size set to 100ms and the step size of 10ms to ensure the time accuracy of data alignment. In each window, the electronic device uses the linear interpolation method to synchronize data with different sampling rates. For acceleration data, the deviation model is expressed as: δa =δa_d +δa_r, where δa_d is the deterministic deviation, including electronic device errors such as zero bias and scale factor, and δa_r is the random deviation. By comparing the acceleration component in X_k with the original IMU measurement value, the electronic device extracts these two types of error characteristics. When performing error separation, the electronic device first pre-processes the data using a 20th-order FIR filter, with the filter cutoff frequency set to 1 / 10 of the original sampling rate, effectively reducing the impact of high-frequency noise. For the filtered data, the electronic device uses the least squares method to fit the deterministic deviation component.

[0133] In order to improve the reliability of deviation analysis, the electronic device performs segmented processing on the incremental fusion sequence. Based on the motion state information in the incremental fusion sequence, the data sequence is divided into static segment, uniform speed segment and dynamic segment. When identifying the static segment, the electronic device sets strict judgment conditions: the data segment with acceleration standard deviation less than 0.02m / s² and duration longer than 3 seconds is marked as a static segment. During the static segment judgment process, the electronic device uses the sliding variance calculation method with a window length of 1 second and a step length of 0.1 second. In order to avoid misjudgment, the electronic device also introduces angular velocity auxiliary judgment, requiring the angular velocity standard deviation to be less than 0.1° / s at the same time. In each static segment, the electronic device calculates the mean and standard deviation of the three-axis acceleration and angular velocity. The calculation adopts a block processing method, with 500 sampling points per block and 100 points overlapping adjacent blocks. For each data block, the electronic device first performs outlier detection, uses the 3σ criterion to eliminate outliers, and then calculates the mean and standard deviation. Finally, the reliability of the feature quantity is improved by the consistency test of multiple block results, and the consistency test adopts the χ² test method.

[0134] In the analysis of the uniform motion segment, the electronic device focuses on the scale factor error, which is expressed as: δS = (a_meas - a_ref) / a_ref where a_meas is the IMU measurement value and a_ref is the reference value in the incremental fusion sequence. The identification of the uniform segment adopts a triple criterion: the rate of change of speed is less than 0.1m / s², the rate of change of heading angle is less than 1° / s, and this state must last for at least 5 seconds. The electronic device pre-processes both the speed and heading data with a smoothing window of 0.5 seconds to reduce the impact of instantaneous fluctuations. Once the uniform segment is identified, the electronic device divides it into several 2-second analysis units, with adjacent units overlapping by 0.5 seconds. Within each analysis unit, the electronic device uses a piecewise linear fitting method to evaluate the stability of the scale factor. The weighted least squares algorithm is used for fitting, and the weight is proportional to the signal-to-noise ratio of the measurement value. For the fitting results, the electronic device not only calculates the root mean square value of the residual, but also analyzes the probability distribution characteristics of the residual to evaluate the degree of certainty of the scale factor.

[0135] The analysis of the dynamic segment is more complicated and needs to consider the influence of multiple factors at the same time. The electronic device extracts the state increment information from the incremental fusion sequence and focuses on analyzing the dynamic response characteristics of the acceleration and angular velocity data. For the gyroscope data, the deviation model is: δω=δω_d +δω_r. At this time, the electronic device divides the data into three categories according to the intensity of the movement: low dynamic (angular velocity <15° / s), medium dynamic (15° / s-30° / s) and high dynamic (>30° / s). When classifying, the electronic device uses a 0.2 second sliding window to calculate the mean of the angular velocity to avoid misclassification caused by instantaneous fluctuations. For each type of data, the electronic device adopts a different processing strategy: the low dynamic segment focuses on zero bias stability and uses the recursive mean method to estimate the zero bias; the medium dynamic segment analyzes the linearity characteristics and uses piecewise polynomial fitting to evaluate the nonlinear error; the high dynamic segment focuses on the bandwidth characteristics and evaluates the dynamic performance by calculating the amplitude-frequency characteristics of the input and output signals. The processing of each type of data uses a basic analysis unit of 100ms, and the dynamic response performance is evaluated by comparing the state changes of adjacent units.

[0136] For the deviation analysis of navigation data, the electronic equipment makes full use of the state estimation information in the incremental fusion sequence. The calculation of the position deviation is expressed as: δp = p_gnss - p_fusion, where p_gnss is the original navigation solution result and p_fusion is the position value in the incremental fusion sequence. The electronic equipment adopts a two-layer analysis strategy: at the macro level, a 1-second data window is used to calculate the statistical characteristics of the position deviation, including the mean, standard deviation and autocorrelation coefficient; at the micro level, the navigation update cycle (usually 100ms) is used as the basic unit to analyze the time-varying characteristics of the deviation, focusing on the variation law of the deviation under different satellite configurations. The electronic equipment processes the deviations in the horizontal and vertical directions separately and establishes their respective error characteristic models.

[0137] The calculation of velocity deviation requires the combination of IMU data integration and navigation velocity solution results: δv = v_gnss -v_imu -∫δa dt. The electronic device decomposes the velocity deviation into three main components: the constant term caused by the initial alignment error, the drift term caused by the integration of the accelerometer error, and the random error term in the navigation solution. For the integration process, the electronic device uses a basic integration step of 5ms, and performs acceleration zero bias compensation before each integration step. Error compensation is performed once per solution cycle (100ms), and the compensation amount is based on the acceleration deviation characteristics obtained from the previous analysis. The electronic device implements a zero-speed detection mechanism based on variance analysis, with a detection window set to 0.5 seconds. When a stationary state is detected, the speed correction is performed immediately, effectively suppressing the speed drift caused by integration.

[0138] By analyzing the electronic equipment of the incremental fusion sequence, the data deviation value set is finally obtained: D = {δa, δω, δp, δv}, where δa is the accelerometer deviation vector, δω is the gyroscope deviation vector, δp is the position deviation vector, and δv is the velocity deviation vector.

[0139] Step S70: Use the data deviation value to perform data calibration on the fitting sequence to obtain a calibrated fusion sequence.

[0140] Based on these deviation values, the electronic equipment performs segmented calibration processing on the fitting sequence. For example, it corrects the IMU data segment by segment when turning in a loop, performs segmented corrections in areas where the GNSS signal is unstable, and uses a weighted method to achieve a smooth transition between intervals, ultimately generating a calibration fusion sequence.

[0141] The electronic device first builds a calibration calculation framework. For each data point in the fitting sequence, the corresponding deviation calibration is applied. For the IMU data part, the acceleration calibration adopts: a_cal = a_fit -δa, where a_fit is the acceleration value in the fitting sequence and δa is the corresponding acceleration deviation. The calibration process adopts a segmented processing strategy. The electronic device divides the data into basic intervals of 200ms, and adjacent intervals overlap by 50ms. In each interval, the electronic device uses a linear interpolation method to calculate the calibration parameters to ensure the continuity of parameter changes. For overlapping areas, a cosine-weighted smooth transition strategy is used to avoid jumps in the calibration results. The calibration of the acceleration data by the electronic device pays special attention to the dynamic changes of the zero bias and scale factor. By accumulating the statistical characteristics of multiple intervals, a dynamic update mechanism for the calibration parameters is established.

[0142] The calibration of angular velocity data adopts a similar method: ω_cal =ω_fit -δω, and the electronic equipment adopts differentiated calibration strategies for different motion stages. In the static and low-speed motion stages (angular velocity less than 5° / s), the zero bias error is mainly compensated, and the calibration parameter update cycle is set to 1 second; in the medium dynamic stage (5° / s to 15° / s), the influence of zero bias and scale factor is considered at the same time, and the update cycle is shortened to 0.5 seconds; in the high-dynamic turning process (angular velocity greater than 15° / s), more attention is paid to the calibration of scale factor and nonlinear error, and the update cycle is further reduced to 0.2 seconds. For each dynamic condition, the electronic equipment has established a special parameter update strategy to ensure the real-time and accuracy of calibration.

[0143] For the calibration of navigation data, the electronic device first processes the position data: p_cal = p_fit -δp. The position calibration takes into account the spatial correlation of the navigation error. The electronic device divides the calibration interval according to the navigation update period (for example, 1 second). In each interval, the calibration parameters are adjusted according to the change of the GDOP value. When the GDOP is less than 3, the complete deviation calibration is directly applied; when the GDOP is between 3 and 5, the calibration amount is reduced to 80% of the deviation value; when the GDOP is greater than 5, the calibration amount is further reduced to 50%. This GDOP-based adaptive calibration strategy ensures the reliability of position calibration. At the same time, the electronic device establishes a distance-based error attenuation model, and the calibration amount decays exponentially with the increase of the distance from the reference point.

[0144] The calibration of velocity data needs to consider the influence of IMU and navigation at the same time: v_cal = v_fit -δv. During the velocity calibration process, the electronic equipment pays special attention to the drift problem caused by the acceleration integral. The calibration calculation uses a sliding window technology, with the window length set to 2 seconds and the step length set to 0.5 seconds. In each window, the electronic equipment first analyzes the changing trend of the velocity deviation, and then selects the appropriate calibration strategy based on the trend characteristics. For linear drift, a first-order correction model is used; for nonlinear drift, a second-order correction model is used. The calibration parameters are updated using a recursive averaging method, which not only ensures the stability of the parameters, but also reflects the dynamic changes of the drift characteristics.

[0145] In order to ensure the continuity and smoothness of the calibration process, the electronic device adopts a multi-level buffer mechanism. At the boundary of adjacent calibration intervals, a weighted average method is used to achieve smooth transition: x_trans = w1·x_prev + w2·x_next, where the weight factors w1 and w2 vary cosine with the time interval from the boundary point. Specifically, when the time interval from the boundary point is Δt, the calculation formula of the weight factor is: w1 = 0.5*(1 + cos(πΔt / T)) w2 = 1 -w1, where T is the length of the transition interval, which is usually set to 5 times the original sampling period.

[0146] In terms of time synchronization, the electronic equipment has established an accurate time correspondence. The calibration results of IMU data (200Hz) and navigation data (10Hz) need to be unified to the same time base. The electronic equipment chooses the IMU data timestamp as the base and aligns all data through the cubic spline interpolation method. The interpolation process uses the Not-a-knot boundary condition to avoid interpolation oscillation at the endpoints. This time synchronization mechanism ensures the time consistency of the calibration results of different sensor data.

[0147] Finally, the electronic device organizes the calibration results into a calibration fusion sequence C = {c_k, k=1, 2, ..., N}. Each element c_k of the calibration fusion sequence contains complete state information: c_k = [p_cal, v_cal, q_cal], where p_cal is the calibrated position, v_cal is the calibrated velocity, and q_cal is the calibrated attitude quaternion. Each state quantity in this calibration fusion sequence has been calibrated by the electronic device, effectively eliminating various electronic device errors in the original data.

[0148] Step S80, performing a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters.

[0149] When performing secondary fitting, the electronic device builds a segmented fitting model based on the vehicle's motion characteristics. For example, when turning on a roundabout, the parameters are optimized by considering the steering angle and speed constraints; when entering an S-shaped road, the continuity of the fitting is ensured based on the vehicle's dynamic characteristics.

[0150] Through preliminary processing, the electronic device has obtained a fusion sequence that has been calibrated for deviations. However, due to the many complex factors affecting the vehicle's motion, such as road geometry constraints, vehicle dynamics, etc., data calibration alone may not fully reflect the actual motion laws. Therefore, the electronic device needs to perform secondary fitting to take these physical constraints and actual motion characteristics into account in order to obtain more realistic fusion parameters.

[0151] Combination Figure 6 In some embodiments, step S80 may include: step S81 and step S82.

[0152] Step S81, constructing a mathematical model of quadratic fitting, and establishing corresponding fitting models for the position component, the velocity component and the attitude component respectively;

[0153] The electronic device first constructs a mathematical model of quadratic fitting, and establishes corresponding fitting models for the position component p_cal, the velocity component v_cal, and the attitude component q_cal. For each state component, a fitting basis function based on a polynomial is established: f(t) =∑(αi·φi(t)), i = 0, 1, ..., m, where φi(t) is the i-order basis function and αi is the fitting coefficient to be determined. The electronic device adaptively selects the order of the basis function according to the motion characteristics, selects the third order for the uniform motion segment, and increases to the fifth order for the variable acceleration motion segment. The selection of the basis function adopts an orthogonal polynomial electronic device, such as Legendre polynomials or Chebyshev polynomials. This orthogonal basis function electronic device can significantly improve the condition number of the fitting problem, improve the calculation stability, and avoid the generation of ill-conditioned equations. Based on the above basis function characteristics, the electronic device applies f(t) to the fitting of the position, velocity, and attitude components, respectively, and the corresponding fitting functions are recorded as f_p(t), f_v(t), and f_q(t). At the same time, the electronic device also standardized the basis functions, uniformly mapping the time interval to [-1, 1] to further improve numerical stability.

[0154] Step S82, segmenting the data based on the mutation points of the motion state, the significant change points of acceleration and the singular points of curvature, using the fitting function to solve the fitting parameters by piecewise fitting, and performing iterative optimization in the process of solving the fitting parameters to obtain the final fitting parameters.

[0155] In order to improve the fitting accuracy, the electronic device adopts a segmented fitting strategy. Based on the motion characteristics of the calibration fusion sequence, the data is divided into multiple continuous segments. The selection of segmentation points is based on three criteria: the mutation point of the motion state, the significant change point of acceleration, and the singular point of curvature. The electronic device calculates the gradient of these characteristic quantities and sets the corresponding threshold for automatic segmentation. Specifically, when the acceleration change rate exceeds 2m / s³, or the heading angle change rate exceeds 30° / s, or the curvature change exceeds the preset threshold, the electronic device will segment at this point. This adaptive segmentation strategy ensures that the fitting model can accurately describe the characteristics of different motion stages.

[0156] For the fitting of position data, the electronic device establishes the error equation: E_p = ||p_cal - f_p(t)||², where p_cal is the position data in the calibration sequence and f_p(t) is the position fitting function. The fitting process uses the weighted least squares method, and the design of the weights takes into account the reliability of the data. The weight values ​​are normalized to ensure the balance of data of different dimensions. The electronic device solves the least squares problem through an analytical solution method and obtains the position fitting parameters: α_p = (Φ^T·W·Φ)^(-1)·Φ^T·W·p_cal, where Φ is the basis function matrix and W is the weight matrix. In this calculation process, the electronic device uses the QR decomposition method to solve the normal equation, which is more stable than direct inversion.

[0157] Velocity fitting also uses a similar approach, but it is necessary to consider the consistency constraint with position fitting: E_v = ||v_cal - f_v(t)||², f_v(t) = d(f_p(t)) / dt. This constraint ensures the physical rationality of the position and velocity fitting results. By deriving the position fitting function, the electronic device automatically obtains the basis function form of velocity fitting. During the optimization process, the electronic device considers both the velocity fitting error and the position derivative constraint, and balances the two objectives by weighting. This joint optimization strategy ensures kinematic consistency.

[0158] The fitting of the attitude data is represented by quaternions. Since quaternions satisfy the unit norm constraint, the electronic device uses the exponential mapping method to transform the problem into the optimization of three-dimensional parameters:

[0159]

[0160] Among them, θ is the three-dimensional rotation vector and q_ref is the reference attitude. This parameterization method avoids the singularity problem that may be caused by directly interpolating the quaternion. The electronic device establishes a fitting model based on the rotation vector:

[0161]

[0162] This error metric directly corresponds to the geodesic distance in the rotational space and is more consistent with the geometric nature of rotational motion.

[0163] In the process of solving the fitting parameters, the electronic device adopts an iterative optimization strategy. First, the linearization method is used to obtain the initial solution, and then the parameters are optimized by the Newton iteration method. In each iteration, the gradient and Hessian matrix of the objective function with respect to the parameters are first calculated, and then the linear equations are solved to obtain the parameter increment. During the iterative solution process, the electronic device monitors the convergence performance in real time and dynamically adjusts the step size according to the convergence speed. The step size adjustment adopts the Armijo criterion to ensure the monotonic decrease of the objective function.

[0164] In order to ensure the continuity of the fitting results, the electronic device imposes derivative continuity constraints at the segmentation points. For position fitting, the zero-order to second-order derivatives are required to be continuous; for attitude fitting, the angular velocity and angular acceleration are required to be continuous. These constraints are introduced into the optimization problem through the Lagrange multiplier method: L = E +λ^T·g(α), where g(α) is the constraint equation and λ is the Lagrange multiplier. The electronic device uses the augmented Lagrangian method to solve this constrained optimization problem, and finally obtains the optimal solution that satisfies all constraints by iteratively updating the original variables and Lagrange multipliers.

[0165] Through the above optimization calculations, the electronic device finally obtains the fitting parameter set: θ_fit = {α_p, α_v, α_q}, where α_p is the position fitting parameter, α_v is the velocity fitting parameter, and α_q is the attitude fitting parameter. This set of parameters fully describes the motion characteristics of the carrier, ensuring both the accuracy of the fitting results and the physical rationality of the results.

[0166] Step S90: Apply the final fitting parameters to the calibration fusion sequence, and obtain the final fusion signal by matching the data point by point.

[0167] Based on the final fitting parameters, the electronic device matches the overall driving trajectory at the macro level, fine-tunes the speed and steering posture at the micro level, and dynamically adjusts the fusion weights of IMU and GNSS data according to the characteristics of the road section, ultimately generating a complete fusion signal.

[0168] The basic principle of this step is to first use the fitting parameters to reconstruct an ideal motion trajectory with good continuity and smoothness, and then match it with the actual calibration sequence. Through this "template matching" method, the excellent characteristics of the ideal trajectory are retained without losing the effective information in the actual data.

[0169] The electronic device first constructs a signal reconstruction model based on the fitting parameters. For the position component, the signal is reconstructed according to the position fitting parameter α_p: p_rec(t) =∑(α_p, i·φi(t)), i = 0, 1, ..., m, where p_rec(t) represents the reconstructed ideal position trajectory, φi(t) is the Legendre polynomial basis function electronic device used for orthogonal decomposition of the signal, and α_p, i is the i-th order coefficient in the position fitting parameter. The reconstruction process adopts a recursive calculation method. The electronic device first calculates the values ​​of each order of basis functions, and then obtains the reconstruction result by accumulation. In order to improve the calculation efficiency, the electronic device pre-establishes a basis function value table, and directly obtains the calculation at a fixed time point by looking up the table. In practical applications, for example, when a vehicle is driving on an "S"-shaped turning section, the electronic device can generate a smooth ideal trajectory through this reconstruction method, effectively avoiding jumps and noise in the original data.

[0170] The reconstruction of the velocity signal requires consideration of both the position fitting parameter α_p and the velocity fitting parameter α_v: v_rec(t) = w1·d(p_rec(t)) / dt + w2·∑(α_v, i·ψi(t)), where v_rec(t) represents the reconstructed ideal velocity trajectory, w1 and w2 are complementary fusion weights (w1 + w2 = 1), and ψi(t) is the velocity reconstruction basis function that matches the derivative of φi(t). For example, when a vehicle is traveling at a constant speed on a highway, the electronic device will increase the weight of w1 (typically 0.8) because the position derivative can better reflect the actual speed; while in the case of stop-and-go driving on urban roads, the electronic device will increase the weight of w2 (to 0.6) to better utilize the directly measured velocity information.

[0171] The reconstruction of the attitude signal uses a piecewise quaternion spline based on the attitude fitting parameter α_q:

[0172]

[0173] Where q_rec(t) represents the reconstructed ideal attitude quaternion, q_k is the attitude value at the node, θk is the local rotation parameter, u(t) is the time normalization factor in the interval [0, 1], and α_q is the fitting parameter that controls the attitude interpolation characteristics. For example, when the vehicle makes an "S" turn, the electronic device sets the attitude nodes at the key positions of each turn: the starting point, the maximum turning point and the end point of the first turn, and the corresponding position of the second turn. For a typical "S" turn (each turn is about 45 degrees), the electronic device usually sets an interpolation node every 15 degrees to ensure a smooth transition of the attitude change.

[0174] The data matching process adopts a hierarchical processing strategy. First, trajectory matching is performed at the macro level: p_match = argmin ||p_rec(t) - p_cal(t)||², where p_cal(t) is the actual position data in the calibration sequence, and ||·||² represents the Euclidean distance norm. In specific implementation, the electronic device first divides the entire trajectory into different feature segments. Taking the "S" turn as an example, the electronic device identifies five feature segments of "straight-left turn-straight-right turn-straight". A 5-second matching window is used in the straight segment to ensure the overall smoothness of the trajectory; it is reduced to 2 seconds in the turning segment to better capture the turning details.

[0175] At the micro level, the electronic device performs fine matching of speed and attitude. Speed ​​matching needs to satisfy: v_match = arg min (||v_rec(t) - v_cal(t)||² +λ·||d(p_match) / dt - v_rec(t)||²) , where v_cal(t) is the actual speed data in the calibration sequence and λ is the weight coefficient for balancing the measured data with the kinematic constraints. When performing speed matching, the electronic device uses a basic time step of 20ms to ensure that high-frequency motion features are captured.

[0176] Attitude matching considers the geometric properties of quaternions:

[0177]

[0178] Where q_cal(t) is the actual posture data in the calibration sequence, log(·) represents the quaternion logarithmic mapping, Represents quaternion multiplication. The electronic device adopts a two-stage matching strategy: first, a rough matching is performed using a step size of 100ms to quickly determine the overall posture change trend; then a fine matching is performed using a step size of 10ms to accurately reconstruct the posture changes during the steering process.

[0179] The final fusion process is expressed as: x_fusion(t) =β(t)·x_match(t) + (1-β(t))·c_k(t), where x_match(t) is the state after matching, c_k(t) is the corresponding data in the calibration sequence, and β(t) is the fusion coefficient adaptively adjusted based on the motion state. The electronic device dynamically adjusts β(t) according to the motion state and reasonably allocates the weights of the matching results and calibration data at different motion stages.

[0180] Through the above processing, the electronic device finally obtains the fusion signal sequence F = {f_k, k=1, 2, ..., N}, each element of which contains complete state information: f_k = [p_fusion, v_fusion, q_fusion, t_k], where p_fusion, v_fusion, q_fusion represent the fused position, velocity and attitude data respectively, and t_k is the corresponding timestamp. This fusion signal sequence optimizes the calibration fusion sequence through point-by-point matching, and obtains the final fusion result with good continuity and smoothness.

[0181] The technical solution of the IMU data and navigation signal fusion processing method of the embodiment of the present invention systematically establishes a complete set of dynamic data quality evaluation system through multi-frequency sampling mechanism, multi-level data verification and three-level fitting calculation strategy; this progressive processing framework effectively overcomes the defects of insufficient data reliability evaluation of traditional methods and significantly improves the fusion accuracy in complex environments; by establishing multi-dimensional data correspondence, combining incremental calculation and deviation analysis, and finally through the construction of calibration fusion sequence, the fusion divergence problem under strong nonlinear motion is effectively solved; and this multi-step progressive processing method effectively improves the data fusion performance under complex environments. In addition, in the quadratic fitting process, the adaptive order polynomial basis function and piecewise quaternion spline are innovatively used to achieve accurate description of different motion characteristics; at the same time, through the hierarchical matching strategy, the smooth characteristics of the ideal trajectory are retained without losing the effective information in the actual data, effectively improving the accuracy of navigation positioning. This data matching method based on quadratic fitting successfully solves the fusion problem under strong nonlinear motion.

[0182] It should be noted that all the above-mentioned embodiments of the IMU data and navigation signal fusion processing method of the present invention can be arbitrarily combined or combined to form a new embodiment when there is no contradiction or conflict between them.

[0183] Reference Figure 7 The present invention also proposes an IMU data and navigation signal fusion processing device 100, which is applied to navigation equipment.

[0184] In this embodiment, the IMU data and navigation signal fusion processing device 100 includes:

[0185] The acquisition module 101 is used to acquire IMU data and navigation data, and construct an original sampling sequence corresponding to the timestamp using a multi-frequency sampling mechanism;

[0186] The verification module 102 is used to perform data verification on the original sampling sequence, calculate the data difference in the current sampling period, and obtain the sampling supplement value;

[0187] The first fitting module 103 is used to perform a comprehensive fitting process on the sampled supplementary values ​​by using a three-level fitting calculation to obtain a fitting sequence;

[0188] A first calculation module 104 is used to establish a corresponding relationship between the IMU data and the navigation data according to the fitting sequence to obtain an initial fusion parameter;

[0189] The second calculation module 105 is used to perform incremental calculation on the initial fusion parameters, establish a corresponding relationship between data increment and fusion accuracy, and obtain an incremental fusion sequence;

[0190] A deviation determination module 106 is used to calculate the actual difference between the IMU and navigation data using the incremental fusion sequence to obtain a data deviation value;

[0191] A calibration module 107, configured to perform data calibration on the fitting sequence using the data deviation value to obtain a calibration fusion sequence;

[0192] A second fitting module 108 is used to perform a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters;

[0193] The matching module 109 is used to apply the final fitting parameters to the calibration fusion sequence, and obtain the final fusion signal by matching the data point by point.

[0194] The IMU data and navigation signal fusion processing device 100 of the present invention can implement the steps of the above-mentioned IMU data and navigation signal fusion processing method, and therefore has at least all the beneficial effects brought about by the technical solution of the above-mentioned IMU data and navigation signal fusion processing method embodiment, which will not be described one by one here.

[0195] The present invention also provides a navigation device, which includes an IMU sensor, a navigation module and an electronic device, see Figure 8 , Figure 8 It is a schematic diagram of the structure of an electronic device in a hardware operating environment involved in an embodiment of the present invention.

[0196] The electronic device of the embodiment of the present invention may be a controller of a navigation device, or a computing device such as a single chip microcomputer, a notebook, a palm computer, and a server. Figure 8As shown, the electronic device may include: a processor 1001 (such as a CPU), a network interface 1004, a user interface 1003, a memory 1005 and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit, such as a keyboard (Keyboard), and the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0197] Those skilled in the art will understand that Figure 8 The electronic device structure shown in the figure does not constitute a limitation of the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0198] like Figure 8 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module and a computer program.

[0199] exist Figure 8 In the electronic device shown, the network interface 1004 is mainly used to connect to the background server and communicate data with the background server; the user interface 1003 is mainly used to connect to the client (user end) and communicate data with the client; and the processor 1001 can be used to call the computer program stored in the memory 1005. When the computer program is called and executed by the processor 1001, the above-mentioned steps of IMU data and navigation signal fusion processing are implemented.

[0200] Based on the computer program proposed in the aforementioned embodiment, the present invention further proposes a medium storing the computer program. When the computer program is executed by a controller, the IMU data and navigation signal fusion processing method recorded in the aforementioned embodiment is implemented.

[0201] Since the navigation device and medium of the present invention can implement the above-mentioned steps of IMU data and navigation signal fusion processing, they at least have all the beneficial effects brought by the technical solutions of the above-mentioned IMU data and navigation signal fusion processing method embodiments, which will not be described one by one here.

[0202] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0203] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0204] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0205] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0206] The above are only partial or preferred embodiments of the present invention. Neither the text nor the drawings can limit the scope of protection of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the overall concept of the present invention, or direct / indirect application in other related technical fields are included in the scope of protection of the present invention.

Claims

1. A method for fusion processing of IMU data and navigation signals, applied to navigation equipment, characterized in that: include: Collect IMU data and navigation data, and use the multi-frequency sampling mechanism to construct the original sampling sequence corresponding to the timestamp; Performing data verification on the original sampling sequence, calculating the data difference within the current sampling period, and obtaining a sampling supplement value; The sampling supplementary values ​​are subjected to comprehensive fitting processing by using a three-level fitting calculation to obtain a fitting sequence; Establishing a corresponding relationship between the IMU data and the navigation data according to the fitting sequence to obtain initial fusion parameters; Performing incremental calculation on the initial fusion parameters, establishing a corresponding relationship between data increment and fusion accuracy, and obtaining an incremental fusion sequence; Calculating the actual difference between the IMU and the navigation data using the incremental fusion sequence to obtain a data deviation value; Using the data deviation value to perform data calibration on the fitting sequence to obtain a calibrated fusion sequence; Perform a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters; The final fitting parameters are applied to the calibration fusion sequence, and the final fusion signal is obtained by point-by-point matching of the data.

2. The IMU data and navigation signal fusion processing method according to claim 1, characterized in that: The performing of data verification on the original sampling sequence, calculating the data difference within the current sampling period, and obtaining the sampling supplement value includes: Performing multi-level data verification on the original sampling sequence, wherein the multi-level data verification includes performing data integrity verification, time stamp verification and data validity verification; On the basis of ensuring the validity of data, a hierarchical difference calculation strategy is adopted to calculate the data difference in the current sampling period and obtain the data difference characteristics; The sampling supplement value is obtained by utilizing the data difference feature, and the navigation data is supplemented.

3. The IMU data and navigation signal fusion processing method according to claim 2, characterized in that: The obtaining the sampling supplement value by utilizing the data difference feature and supplementing the navigation data includes: A prediction mechanism based on Kalman filtering is adopted to predict the position and attitude changes of the carrier through a state equation between two adjacent navigation sampling points using the acceleration and angular velocity information provided by the IMU data; The prediction result of the state equation is used as a supplementary value of the navigation data to fill the data gaps between the navigation sampling points of the navigation data.

4. The IMU data and navigation signal fusion processing method according to claim 1, characterized in that: The three-level fitting calculation is used to perform a comprehensive fitting process on the sample supplementary values ​​to obtain a fitting sequence, including: Performing single data fitting on the IMU data and the navigation data in the sampled supplementary values ​​respectively, wherein the single data fitting of the IMU data adopts piecewise polynomial fitting, and the single data fitting of the navigation data adopts cubic B-spline interpolation; Performing double data fitting on the sampled supplementary values, and establishing preliminary associations between two types of data in different road sections through an extended Kalman filter framework; The single data fitting result and the double data fitting result are unified into a Bayesian estimation framework to perform combined data fitting to obtain the fitting sequence.

5. The IMU data and navigation signal fusion processing method according to any one of claims 1 to 4, characterized in that: The step of establishing a corresponding relationship between the IMU data and the navigation data according to the fitting sequence to obtain initial fusion parameters includes: Using the timestamp information in the fitting sequence, a time synchronization mapping between the IMU data and the navigation data is established; Based on the established data association in the fitting sequence, a three-layer correspondence model is constructed, wherein the three-layer correspondence is respectively a position correspondence, a speed correspondence and a posture correspondence; After constructing the model of the three-layer correspondence relationship, an initial fusion parameter estimation is performed to obtain an initial fusion parameter; The initial fusion parameters include: a state transfer matrix, a process noise matrix, a measurement noise matrix, a state covariance matrix, a system deviation parameter vector and a scale factor matrix.

6. The IMU data and navigation signal fusion processing method according to claim 5, characterized in that: The incremental calculation of the initial fusion parameters is performed to establish a corresponding relationship between data increment and fusion accuracy to obtain an incremental fusion sequence, including: Use the state transfer matrix to construct the incremental state representation; Designing a likelihood calculation function based on the incremental state expression form, and optimizing the position increment, velocity increment and attitude increment to obtain an optimized calculation result; Based on the optimization calculation result and in accordance with the motion characteristics, an adaptive state fusion strategy is adopted to perform state fusion to obtain the incremental fusion sequence.

7. The IMU data and navigation signal fusion processing method according to any one of claims 1 to 4, characterized in that: The performing a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters includes: Constructing a mathematical model of quadratic fitting, establishing corresponding fitting models for position component, velocity component and attitude component respectively, wherein a polynomial-based fitting basis function is established for each state component, and the fitting basis function is applied to each component to obtain a corresponding fitting function, and the order of the basis function is adaptively selected according to the motion characteristics; The data is segmented based on the mutation points of the motion state, the significant change points of acceleration and the singular points of curvature, the fitting function is used to solve the fitting parameters by a segmented fitting method, and iterative optimization is performed in the process of solving the fitting parameters to obtain the final fitting parameters.

8. An IMU data and navigation signal fusion processing device, applied to navigation equipment, characterized in that: include: The acquisition module is used to collect IMU data and navigation data, and use the multi-frequency sampling mechanism to construct the original sampling sequence corresponding to the timestamp; A verification module, used to perform data verification on the original sampling sequence, calculate the data difference within the current sampling period, and obtain a sampling supplement value; A first fitting module is used to perform a comprehensive fitting process on the sampled supplementary values ​​by using a three-level fitting calculation to obtain a fitting sequence; A first calculation module, used for establishing a corresponding relationship between the IMU data and the navigation data according to the fitting sequence, and obtaining an initial fusion parameter; A second calculation module is used to perform incremental calculation on the initial fusion parameters, establish a corresponding relationship between data increment and fusion accuracy, and obtain an incremental fusion sequence; A deviation determination module, used for calculating the actual difference between the IMU and the navigation data using the incremental fusion sequence to obtain a data deviation value; A calibration module, used to perform data calibration on the fitting sequence using the data deviation value to obtain a calibration fusion sequence; A second fitting module, used for performing a secondary data fitting operation based on the calibration fusion sequence to obtain final fitting parameters; The matching module is used to apply the final fitting parameters to the calibration fusion sequence, and obtain the final fusion signal by matching the data point by point.

9. A navigation device, comprising an IMU sensor and a navigation module, characterized in that: The navigation device also includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the IMU data and navigation signal fusion processing method as described in any one of claims 1 to 7 are implemented.

10. A medium, characterized in that The medium stores a computer program, which, when executed by a processor, implements the steps of the IMU data and navigation signal fusion processing method according to any one of claims 1 to 7.

Images