Vehicle-mounted imu zero speed detection method based on exponential smoothing filter and boundary constraint

Abstract

This invention provides a zero-speed detection method for an onboard IMU based on exponential smoothing filtering and boundary constraints, comprising: calculating the real-time variance of the six-axis data of the inertial measurement unit on an axis-by-axis basis using a sliding window; continuously tracking the historical minimum variance of each axis as a noise baseline reference; adaptively determining the dynamic stable fluctuation range of each axis based on the noise baseline reference; using a first-order exponential smoothing filter to perform weighted iterative updates on the historical minimum variance reference to eliminate reference step jumps caused by random minimum values ​​of the window; applying preset upper and lower boundary constraints to the smoothed updated reference; setting adaptive coefficients for each axis according to the noise characteristics differences of the accelerometer and gyroscope; generating dynamic decision thresholds for each axis in combination with the smoothed and constrained reference; dynamically adjusting the multi-axis voting thresholds according to the current global noise level; jointly voting on the decision results of each axis; completing the determination of the vehicle's stationary state and outputting static data.

Description

Technical Field

[0001] This invention relates to the field of integrated navigation technology, and in particular to a zero-speed detection method for an onboard IMU based on exponential smoothing filtering and boundary constraints. Background Technology

[0002] Inertial Measurement Units (IMUs) based on Micro-Electro-Mechanical Systems (MEMS) have advantages such as small size, low cost, and low power consumption, and have been widely used in vehicle-mounted integrated navigation systems. When Global Navigation Satellite System (GNSS) signals are good, vehicles can obtain high-precision positioning results through GNSS / INS integrated navigation. However, when vehicles travel in environments where GNSS signals are denied, such as tunnels, underground parking garages, or urban canyons, the error of pure inertial navigation diverges rapidly with integration, resulting in position drift on the order of meters or even tens of meters within seconds. To suppress navigation error divergence, Zero Velocity Update (ZUPT) during brief periods of vehicle stillness, such as at traffic lights or in congested parking, is an effective technique. The core prerequisite for ZUPT is accurately identifying the vehicle's stationary state, i.e., zero-velocity detection. The performance of zero-velocity detection directly determines the effectiveness of ZUPT correction, thus affecting the positioning accuracy of the entire navigation system during periods of GNSS signal denial.

[0003] Zero-speed detection in automotive environments faces numerous unique challenges. When a vehicle is stationary, engine idling generates continuous low-frequency vibrations, residual vibrations due to uneven road surfaces, and sensor noise all cause non-zero fluctuations in the IMU output. Traditional fixed-threshold detection methods (such as acceleration sliding variance detectors) struggle to adapt to these dynamic noise environments, often leading to false positives or false negatives. In recent years, researchers have proposed various adaptive zero-speed detection methods, among which adaptive threshold methods based on statistical characteristics have attracted attention due to their simple structure and the elimination of the need for manually preset thresholds. These methods analyze historical information such as the variance of each axis of the IMU in real time (e.g., the minimum value within the tracking window) and multiply it by a fixed safety factor to generate a dynamic threshold, achieving automatic tracking of the sensor noise floor. However, practical engineering applications show that these methods suffer from unidirectional lock-up after a threshold step change, severely limiting their stability and reliability under complex automotive conditions.

[0004] The closest existing technology to this invention is an adaptive thresholding method based on statistical features. The core process of this solution is as follows: First, the variance time series of the six axes of the IMU (three-axis accelerometer, three-axis gyroscope) is calculated axis by axis using a fixed-length sliding window; second, the global minimum variance of each axis since system power-on is continuously tracked, and this minimum value is used as the noise basis benchmark; then, a dynamic decision threshold for each axis is generated by multiplying this benchmark by a preset safety factor (usually 1.2~1.5); finally, the current window variance is compared with the dynamic threshold to determine whether the vehicle is stationary. The key features of this type of algorithm are: the threshold is entirely adaptively generated by historical statistics (such as the minimum variance), eliminating the need for manual setting of the initial threshold; the algorithm only lowers the threshold when a smaller variance is detected, and the threshold change direction is unidirectional convergence. This solution achieves good detection results in static laboratory environments or ideal stationary conditions, and has low computational complexity and is easy to implement in embedded systems.

[0005] However, the aforementioned existing technologies suffer from a significant problem: unidirectional locking after a threshold step jump, making them difficult to apply directly to complex automotive environments. Specifically, when the sliding window accidentally encounters an extremely small variance value, the global minimum variance benchmark instantly jumps to that minimum value, causing a step-like abrupt change in the dynamic threshold. Simultaneously, due to the monotonically non-increasing nature of the global minimum variance sequence, once the threshold jumps to its minimum value, it becomes locked at that level and cannot automatically rise with increasing noise levels. When the sensor experiences extremely low noise conditions and the engine generates normal vibrations, the variance of all windows exceeds the locked minimum benchmark, causing the dynamic threshold to remain consistently below the actual noise level, permanently rendering the algorithm unable to detect stationary states. Manually setting conditions for the threshold to rise again requires manually setting multiple parameters, significantly reducing the system's robustness and further limiting the engineering practicality of such methods. These shortcomings severely restrict the engineering application of existing adaptive thresholding methods based on statistical features in automotive environments.

[0006] In summary, existing technologies generally suffer from the following drawbacks: unidirectional locking after a threshold step transition leads to zero-velocity detection failure that cannot be recovered, making it difficult to meet the needs of practical applications. Summary of the Invention

[0007] This invention provides a zero-speed detection method for vehicle-mounted IMUs based on exponential smoothing filtering and boundary constraints, which solves the defect in the prior art where unidirectional lock-up after a threshold step change leads to zero-speed detection failure and cannot be recovered.

[0008] In a first aspect, the present invention provides a zero-speed detection method for an onboard IMU based on exponential smoothing filtering and boundary constraints, comprising: The real-time variance of the six-axis data of the inertial measurement unit is calculated one axis at a time based on a sliding window. The historical minimum variance of each axis is continuously tracked as the noise baseline benchmark. Based on the noise baseline benchmark, the dynamic stable fluctuation range of each axis is adaptively determined to screen for suspected stationary periods of the vehicle. The historical minimum variance benchmark is updated by weighted iteration using a first-order exponential smoothing filter to eliminate benchmark step jumps caused by random minimum values ​​in the window. Preset upper and lower boundary constraints are applied to the smoothed benchmark to prevent the benchmark from locking down unidirectionally or drifting upwards for a long time, thereby achieving bidirectional adaptive adjustment of the benchmark. Based on the noise characteristics differences between the accelerometer and the gyroscope, adaptive coefficients are set for each axis. Combined with the benchmark after the smoothing constraint, dynamic decision thresholds for each axis are generated. The multi-axis voting thresholds are dynamically adjusted according to the current global noise level. The decision results of each axis are jointly voted to complete the determination of the vehicle's stationary state and output static data.

[0009] According to the zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints provided by the present invention, the real-time variance calculation of the six-axis data of the inertial measurement unit on an axis-by-axis basis includes: A fixed-length sliding window is used to extract continuous sampling data for each axis. The sample variance of the data for each axis within the window is calculated. The sample variance is used to characterize the degree of signal fluctuation within the window. The smaller the variance, the closer the sensor state is to ideal stillness.

[0010] According to the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints provided by the present invention, the step of continuously tracking the historical minimum variance of each axis as a noise basis benchmark includes: From the moment the system is powered on, the current window variance of each axis is compared with the historical minimum variance recorded before. If the current window variance is smaller, the historical minimum variance is replaced with the current window variance; otherwise, the historical minimum variance is kept unchanged. The historical minimum variance corresponds to the optimal static operating condition experienced by the sensor.

[0011] According to the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints provided by the present invention, the step of using first-order exponential smoothing filtering to perform weighted iterative updates on the historical minimum variance benchmark includes: The original historical minimum variance of the current window is combined with the smoothing benchmark of the previous window using a preset smoothing factor in a convex combination. The smoothing factor controls the rate of decay of historical data weights, so that the smoothing benchmark tracks the original minimum variance sequence asymptotically, rather than jumping to a new value instantaneously.

[0012] According to the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints provided by the present invention, the application of preset upper and lower boundary constraints to the smoothed updated reference includes: Set a lower threshold, which is the physical minimum value of the sensor's noise floor variance, to prevent the reference from decaying to zero or a negative value; Set an upper limit threshold, which is a preset multiple of the initial reference or a reasonable maximum variance value set according to the sensor range, to prevent the reference from growing indefinitely. After boundary constraints, the datum always remains within the closed interval.

[0013] The zero-speed detection method for an on-board IMU based on exponential smoothing filtering and boundary constraints provided by the present invention includes setting adaptive coefficients for each axis according to the noise characteristic differences between the accelerometer and the gyroscope, which includes: A first coefficient value is set for each axis of the accelerometer, and a second coefficient value is set for each axis of the gyroscope. The first and second coefficient values ​​are preset according to the dimensional differences of the accelerometer and the gyroscope and their different characteristics of motion sensitivity, respectively, without the need for online calibration.

[0014] According to the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints provided by the present invention, the step of dynamically adjusting the multi-axis voting threshold according to the current global noise level includes: The mean of the three-axis smoothing reference of the accelerometer is compared with the nominal static variance reference value. When the mean is significantly less than the reference value, the voting threshold is set to require all three axes to pass. When the mean value approaches or exceeds the reference value, the voting threshold is automatically lowered to avoid missed detections due to increased noise levels.

[0015] According to the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints provided by the present invention, the output static data includes: The original inertial measurement unit (IMU) data within the window that meets the static determination criteria is marked as a high-quality static data segment and output to the IMU mounting angle calibration module or zero-velocity correction module for subsequent navigation error correction.

[0016] Secondly, the present invention also provides a vehicle-mounted IMU zero-speed detection system based on exponential smoothing filtering and boundary constraints, comprising: The statistics and tracking module is used to calculate the real-time variance of the six-axis data of the inertial measurement unit based on a sliding window, continuously track the historical minimum variance of each axis as the noise baseline benchmark, and adaptively determine the dynamic stable fluctuation range of each axis based on the noise baseline benchmark, which is used to screen for suspected stationary periods of the vehicle. The smoothing and constraint module is used to perform weighted iterative updates on the historical minimum variance benchmark using a first-order exponential smoothing filter to eliminate benchmark step jumps caused by random minimum values ​​in the window. It applies preset upper and lower boundary constraints to the smoothed benchmark to prevent the benchmark from locking down unidirectionally or drifting upwards for a long time, thereby achieving bidirectional adaptive adjustment of the benchmark. The voting and decision module is used to set adaptive coefficients for each axis based on the noise characteristics differences between the accelerometer and the gyroscope, generate dynamic decision thresholds for each axis in combination with the smoothed constraint benchmark, dynamically adjust the multi-axis voting thresholds according to the current global noise level, and jointly vote on the decision results of each axis to complete the determination of the vehicle's stationary state and output static data.

[0017] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints as described above.

[0018] This invention provides a zero-speed detection method for vehicle-mounted IMUs based on exponential smoothing filtering and boundary constraints. Addressing the threshold step jump and unidirectional lock-up problems in vehicle-mounted zero-speed detection, it proposes an exponential smoothing filtering and boundary constraint mechanism to achieve smooth benchmark updates and bidirectional adaptation. It employs axis-specific adaptive coefficients and dynamic multi-axis voting, eliminating the need for manual calibration. The algorithm is computationally lightweight, significantly improving the stability and robustness of zero-speed detection under complex operating conditions, making it suitable for embedded real-time applications. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the zero-speed detection method for vehicle-mounted IMUs based on exponential smoothing filtering and boundary constraints provided by the present invention. Figure 2 This is a schematic diagram of the main interferences in zero-speed detection of the vehicle-mounted IMU provided by the present invention; Figure 3 This is a schematic diagram of the core module and data flow process of the adaptive zero-speed detection algorithm provided by the present invention; Figure 4 This is a schematic diagram of the minimum variance benchmark exponential smoothing filter and boundary constraint update process provided by the present invention; Figure 5This is a schematic diagram of the structure of the vehicle-mounted IMU zero-speed detection system based on exponential smoothing filtering and boundary constraints provided by the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] Figure 1 This is a flowchart illustrating the zero-speed detection method for an onboard IMU based on exponential smoothing filtering and boundary constraints provided in an embodiment of the present invention. Figure 1 As shown, it includes: Step 100: Calculate the real-time variance of the six-axis data of the inertial measurement unit based on the sliding window, continuously track the historical minimum variance of each axis as the noise baseline, and adaptively determine the dynamic stable fluctuation range of each axis based on the noise baseline to screen for suspected stationary periods of the vehicle. Step 200: The historical minimum variance benchmark is updated by weighted iteration using a first-order exponential smoothing filter to eliminate benchmark step jumps caused by random minimum values ​​in the window. Preset upper and lower boundary constraints are applied to the smoothed updated benchmark to prevent the benchmark from locking down unidirectionally or drifting upwards for a long time, thereby achieving bidirectional adaptive adjustment of the benchmark. Step 300: Based on the noise characteristics difference between the accelerometer and the gyroscope, set adaptive coefficients for each axis, generate dynamic decision thresholds for each axis in combination with the smoothed constraint benchmark, dynamically adjust the multi-axis voting thresholds according to the current global noise level, and jointly vote on the decision results of each axis to complete the vehicle stationary state determination and output static data.

[0023] Specifically, such as Figure 2 The diagram illustrating the main interferences in zero-speed detection by an onboard IMU shows engine idling vibration and IMU measurement noise at the top, both of which affect the time series curve below. When the vehicle is stationary, the theoretically constant IMU measurement value is affected by these two types of interference, exhibiting continuous small fluctuations, which can easily lead to misjudgments in zero-speed detection and is a significant contributing factor to errors in onboard inertial navigation.

[0024] like Figure 3The diagram showing the core module and data flow of the adaptive zero-velocity detection algorithm illustrates the overall execution flow of the method of this invention. After continuously acquiring IMU data, the system sequentially performs sliding window variance calculation, variance benchmark smoothing constraint, and multi-axis joint static decision. When the system is determined to be stationary, it outputs valid static data for sensor calibration and zero-velocity correction. In non-stationary states, the sliding window is updated, and the next round of calculation begins.

[0025] This invention first establishes a sliding window variance statistical model. By real-time calculation of six-axis variance time-series data and continuous tracking of the global minimum variance benchmark, a dynamically stable fluctuation range is adaptively generated to filter effective vehicle stationary periods. Then, a first-order exponential smoothing filter is used to perform weighted iterative updates on the minimum variance benchmark to eliminate step jumps. At the same time, upper and lower boundary constraints are set to prevent benchmark locking and long-term drift. Finally, based on the differences in noise characteristics between the accelerometer and gyroscope, axial adaptive coefficients are set to generate decision thresholds for each axis. Combined with real-time operating conditions, the multi-axis voting thresholds are dynamically adjusted to complete the determination of the vehicle's stationary state and output high-quality static data.

[0026] Based on the above embodiments, step 100 includes: A sliding window variance statistical model is established. By solving the six-axis variance time series data in real time and continuously tracking the global minimum variance benchmark, a dynamic stable fluctuation range is adaptively generated to filter the effective stationary period of the vehicle.

[0027] 1. Sliding window variance statistics and adaptive static screening 1.1 Sliding Window Variance Calculation Assume the IMU uses a sampling frequency Continuously output six-axis raw measurement values, defining the first... Triaxial accelerometer measurement vector at each sampling time With the three-axis gyroscope measuring vector .

[0028] For the axis( ), using a length of A sliding window, window number is The window contains sampling points. Define the sample mean within the window as:

[0029] in Indicates the first The first window The original measurement value of the sampling point. Then the first... Axis at the The variance estimates for each window are:

[0030] This variance physically characterizes the degree of signal fluctuation within the window: when the sensor is in an ideal static state, The variance is determined by the sensor's inherent noise level and is relatively small; however, it increases significantly when there is motion or vibration.

[0031] 1.2 Continuous tracking of the global minimum variance benchmark Define the time from system power-on to the current window Up to now, the The minimum historical variance experienced by the axis is:

[0032] The recursive relation satisfies:

[0033] This represents the sensor's "optimal static reference" on that axis, typically corresponding to the ideal operating condition where the vehicle is stationary for an extended period without any external vibration. It serves as the primary basis for subsequent adaptive threshold generation.

[0034] 1.3 Adaptive static screening of fluctuation range In a real-world vehicle environment, even when the vehicle is stationary, the variance will not be exactly equal to the variance due to engine idling vibrations, road incline, or environmental noise. Instead, it fluctuates within a certain range around it. Therefore, the dynamic stationary fluctuation range is defined as follows:

[0035] in This is the axis-dependent fluctuation coefficient, with a typical value of... .when At that time, it was believed that the first The axis meets the "suspected stationary" condition in the current window.

[0036] The lower bound of this interval is the historical minimum variance, and the upper bound is a linear scaling of the minimum variance. The entire interval is generated adaptively from the data, without any manually preset thresholds.

[0037] Based on the above embodiments, step 200 includes: A first-order exponential smoothing filter is used to perform weighted iterative updates on the minimum variance benchmark to eliminate step jumps. At the same time, upper and lower boundary constraints are set to prevent benchmark locking and long-term drift.

[0038] 2. Exponential Smoothing Iteration and Boundary Constraint Benchmark Update 2.1 Analysis of the shortcomings of the traditional minimum variance benchmark Use directly Using it as a benchmark would raise two types of problems: (1) Step transition: whenever At that moment, the benchmark immediately jumps to the new value. This abrupt change causes subsequent thresholds to change drastically, which can easily lead to oscillations in the detection state under scenarios with alternating noise levels.

[0039] (2) One-way lock-up: It is a monotonically non-increasing sequence, which can only decrease and not increase. Once the sensor experiences extremely low noise (such as being stationary for a long time after the engine is turned off), the reference is locked at a minimum value; thereafter, normal vibrations such as engine starting and road bumps will cause all variances to be greater than the reference, causing the algorithm to permanently lose its ability to detect stationary states.

[0040] 2.2 Iteration of First-Order Exponential Smoothing Filter To eliminate step jumps, first-order exponential smoothing is introduced to perform low-pass filtering on the original minimum variance benchmark. The smoothed benchmark is defined as... Its iterative formula is:

[0041] in As a smoothing factor, a typical value is... .

[0042] Theoretical property: Expanding the recurrence relation yields:

[0043] Right now It is the historical global minimum variance sequence The weighted average is calculated, with weights decaying exponentially. The transfer function (Z-transform) of this filter is:

[0044] For a first-order low-pass filter, its The cutoff frequency is:

[0045] in The sampling period is [value]. This filtering effectively suppresses high-frequency abrupt changes caused by random minimum values ​​in the window, resulting in a smooth and gradual change in the baseline update.

[0046] Property 1 (Unbiasedness): If It remains stable at a certain constant over a long period of time. ,but The smoothed baseline converges to the true basis value.

[0047] Property 2 (Eliminating Step): When When a step change occurs, At an exponential rate It smoothly approaches the new value without any instantaneous jumps.

[0048] 2.3 Boundary Constraint Mechanism While exponential smoothing mitigates jumps, if... Unchanged for a long time It will gradually decay towards this value, which may become too small. Therefore, a hard boundary constraint is introduced:

[0049] in The lower threshold is the physical minimum value of the sensor's noise floor variance. As the upper limit threshold, take a reasonable maximum variance (such as the initial baseline). times).

[0050] Boundary constraints and exponential smoothing together constitute a two-way adaptive mechanism: Adjust downwards: When ambient noise is reduced (vehicle is turned off and stationary). Decrease, after exponential smoothing Smooth descent until the lower limit is reached. .

[0051] Adjust upwards: When ambient noise increases (engine starts). Keeping the original value unchanged, but in the exponential smoothing recurrence relation The item is dominant. It will maintain its original high value and will not be locked due to historical minimum values. If it is necessary to actively adapt to higher noise, it can be prevented from growing indefinitely by upper limit constraints, or incremental logic can be added.

[0052] Theorem 1 (Basic Boundedness): After boundary constraints, Always in a closed interval Within, and when the noise floor is stable, It converges to the corresponding value of the basis (constrained by the boundary).

[0053] Based on the above embodiments, step 300 includes: Based on the difference in noise characteristics between the accelerometer and gyroscope, the axis-specific adaptive coefficients are set to generate decision thresholds for each axis. The multi-axis voting thresholds are dynamically adjusted in combination with real-time operating conditions to complete the determination of the vehicle's stationary state and output high-quality static data.

[0054] 3. Split-axis adaptive threshold and dynamic multi-axis static decision 3.1 Generation of Dynamic Decision Thresholds for Split Axes Based on smoothing benchmark , define the first The standard deviation threshold of the axis in the current window is:

[0055] in This is an adaptive coefficient related to the axis. This threshold has a clear physical meaning: the standard deviation of the current window. If less than or equal to If the axis is stationary, then it is considered to be in a static state.

[0056] Coefficient differentiation design: Because the dimension of acceleration measurement is The dimensions of a gyroscope are The two have different noise characteristics and motion sensitivity, so they need to be set separately: - Accelerometer coefficients:

[0057] - Gyroscope coefficients:

[0058] Once the coefficients are preset at the factory, no online calibration is required, and the entire detection algorithm is not sensitive to their values ​​(robustness is guaranteed by the subsequent voting mechanism).

[0059] 3.2 Single-axis stationary decision Define a single-axis stationary indicator variable:

[0060] This decision rule is equivalent to testing whether the current window variance is within the dynamic threshold range. Because It is continuously updated online, and the threshold can adapt to changes in the sensor noise floor in real time.

[0061] 3.3 Dynamic Multi-Axis Voting Mechanism Single-axis decision-making is susceptible to misjudgments due to abnormal noise on individual axes; therefore, multi-axis voting is introduced. The number of passes for both the accelerometer (three axes) and gyroscope (three axes) is counted separately:

[0062] Define accelerometer channel voting threshold Voting threshold of gyroscope channel Unlike traditional fixed thresholds (such as requiring full pass across all three axes), this invention employs a dynamic voting threshold, based on the current global noise level.

[0063] in: The mean value of the triaxial smoothing reference for acceleration; This is the mean value of the three-axis smoothing reference of the gyroscope; , This is the nominal static variance reference value (which can be pre-calibrated using a small amount of factory static data).

[0064] The theoretical basis for this design is: when the noise level is low ( )hour, Approaching 3, meaning all three axes must pass through to ensure confidence in the output static range; when the noise level is high ( Approaching or exceeding )hour, Automatically reduce to 1 or 2 to avoid missed detections.

[0065] Ultimately, the necessary and sufficient condition for the current window to be considered a stationary vehicle is:

[0066] 3.4 Static Data Output Once the above conditions are met, the raw IMU data in the current window will be... Mark as high-quality static data segment and output to: IMU mounting angle self-calibration module: Calculates the average gravitational acceleration and gyroscope zero bias using multiple segments of static data, and solves the IMU mounting deviation angle relative to the vehicle coordinate system through geometric relationships.

[0067] ZUPT Zero Velocity Correction Module: Updates the navigation state with the velocity observation set to zero in the Kalman filter, suppressing inertial navigation integration error.

[0068] 3.5 Algorithm Complexity Analysis The computational cost per window includes: Six-axis variance: Each addition, subtraction, multiplication, and division can be completed in linear time. Exponential smoothing update: 6 multiply-accumulate operations; Boundary constraints: 6 comparisons and assignments; Single-axis decision: 6 comparisons; Voting and output: constant-time operation.

[0069] The overall computational complexity is The memory only needs to store the historical minimum variance of the six axes. Smoothing reference And a few intermediate variables, to meet the real-time operation requirements of the vehicle-mounted embedded platform.

[0070] like Figure 4 The diagram illustrating the minimum variance benchmark exponential smoothing filter and boundary constraint update process details the update flow of the minimum variance benchmark. First-order exponential smoothing filtering is used to achieve weighted iteration, eliminating abrupt changes in the benchmark value. Upper and lower boundary constraints then limit the amplitude of the results, effectively preventing one-way benchmark locking and long-term drift, ultimately outputting a stable variance benchmark.

[0071] The following describes the vehicle-mounted IMU zero-speed detection system based on exponential smoothing filtering and boundary constraints provided by the present invention. The vehicle-mounted IMU zero-speed detection system based on exponential smoothing filtering and boundary constraints described below can be referred to in correspondence with the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints described above.

[0072] Figure 5 This is a schematic diagram of the structure of the vehicle-mounted IMU zero-speed detection system based on exponential smoothing filtering and boundary constraints provided in an embodiment of the present invention, as shown below. Figure 5 As shown, it includes: a statistics and tracking module 51, a smoothing and constraint module 52, and a voting and decision module 53, wherein: The statistics and tracking module 51 is used to calculate the real-time variance of the six-axis data of the inertial measurement unit based on a sliding window, continuously track the historical minimum variance of each axis as the noise baseline benchmark, and adaptively determine the dynamic stable fluctuation range of each axis based on the noise baseline benchmark to screen suspected stationary periods of the vehicle; the smoothing and constraint module 52 is used to perform weighted iterative updates on the historical minimum variance benchmark using a first-order exponential smoothing filter to eliminate the benchmark step jump caused by the random minimum value of the window, and apply preset upper and lower boundary constraints to the smoothed benchmark to prevent the benchmark from locking down unidirectionally or drifting upward for a long time, so as to realize the bidirectional adaptive adjustment of the benchmark; the voting and decision module 53 is used to set adaptive coefficients for each axis according to the noise characteristics difference of the accelerometer and gyroscope, generate the dynamic decision threshold of each axis in combination with the smoothed and constrained benchmark, dynamically adjust the multi-axis voting threshold according to the current global noise level, and jointly vote on the decision results of each axis to complete the determination of the vehicle's stationary state and output static data.

[0073] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logic instructions in the memory 630 to execute a vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints. This method includes: calculating the real-time variance of the six-axis data of the inertial measurement unit based on a sliding window; continuously tracking the historical minimum variance of each axis as a noise baseline; adaptively determining the dynamic stable fluctuation range of each axis based on the noise baseline to screen for suspected stationary periods of the vehicle; using a first-order exponential smoothing filter to perform weighted iterative updates on the historical minimum variance baseline to eliminate baseline step jumps caused by random minimum values ​​in the window; applying preset upper and lower boundary constraints to the smoothed updated baseline to prevent the baseline from unidirectionally locking downwards or drifting upwards for a long time, achieving bidirectional adaptive adjustment of the baseline; setting adaptive coefficients for each axis according to the noise characteristics differences between the accelerometer and gyroscope; generating dynamic decision thresholds for each axis based on the smoothed and constrained baseline; dynamically adjusting the multi-axis voting thresholds according to the current global noise level; jointly voting on the decision results of each axis to complete the vehicle stationary state determination and output static data.

[0074] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, 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 a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0075] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints provided by the above methods. The method includes: calculating the real-time variance of the six-axis data of the inertial measurement unit based on a sliding window, continuously tracking the historical minimum variance of each axis as a noise baseline, adaptively determining the dynamic stable fluctuation range of each axis based on the noise baseline to screen suspected stationary periods of the vehicle; using a first-order exponential smoothing filter to perform weighted iterative updates on the historical minimum variance baseline to eliminate the baseline step jump caused by the random minimum value of the window, applying preset upper and lower boundary constraints to the smoothed updated baseline to prevent the baseline from unidirectionally locking downwards or drifting upwards for a long time, and realizing bidirectional adaptive adjustment of the baseline; setting adaptive coefficients for each axis according to the noise characteristics difference of the accelerometer and gyroscope, generating dynamic decision thresholds for each axis in combination with the smoothed and constrained baseline, dynamically adjusting the multi-axis voting thresholds according to the current global noise level, jointly voting on the decision results of each axis, completing the vehicle stationary state determination and outputting static data.

[0076] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0077] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A zero-speed detection method for an onboard IMU based on exponential smoothing filtering and boundary constraints, characterized in that, include: The real-time variance of the six-axis data of the inertial measurement unit is calculated one axis at a time based on a sliding window. The historical minimum variance of each axis is continuously tracked as the noise baseline benchmark. Based on the noise baseline benchmark, the dynamic stable fluctuation range of each axis is adaptively determined to screen for suspected stationary periods of the vehicle. The historical minimum variance benchmark is updated by weighted iteration using a first-order exponential smoothing filter to eliminate benchmark step jumps caused by random minimum values ​​in the window. Preset upper and lower boundary constraints are applied to the smoothed benchmark to prevent the benchmark from locking down unidirectionally or drifting upwards for a long time, thereby achieving bidirectional adaptive adjustment of the benchmark. Based on the noise characteristics differences between the accelerometer and the gyroscope, adaptive coefficients are set for each axis. Combined with the benchmark after the smoothing constraint, dynamic decision thresholds for each axis are generated. The multi-axis voting thresholds are dynamically adjusted according to the current global noise level. The decision results of each axis are jointly voted to complete the determination of the vehicle's stationary state and output static data.

2. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The real-time variance of the six-axis data of the inertial measurement unit calculated axis by axis includes: A fixed-length sliding window is used to extract continuous sampling data for each axis. The sample variance of the data for each axis within the window is calculated. The sample variance is used to characterize the degree of signal fluctuation within the window. The smaller the variance, the closer the sensor state is to ideal stillness.

3. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The continuous tracking of the historical minimum variance of each axis as a noise baseline includes: From the moment the system is powered on, the current window variance of each axis is compared with the historical minimum variance recorded before. If the current window variance is smaller, the historical minimum variance is replaced with the current window variance; otherwise, the historical minimum variance is kept unchanged. The historical minimum variance corresponds to the optimal static operating condition experienced by the sensor.

4. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The step of using a first-order exponential smoothing filter to perform a weighted iterative update of the historical minimum variance benchmark includes: The original historical minimum variance of the current window is combined with the smoothing benchmark of the previous window using a preset smoothing factor in a convex combination. The smoothing factor controls the rate of decay of historical data weights, so that the smoothing benchmark tracks the original minimum variance sequence asymptotically, rather than jumping to a new value instantaneously.

5. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The application of preset upper and lower boundary constraints to the smoothly updated benchmark includes: Set a lower threshold, which is the physical minimum value of the sensor's noise floor variance, to prevent the reference from decaying to zero or a negative value; Set an upper limit threshold, which is a preset multiple of the initial reference or a reasonable maximum variance value set according to the sensor range, to prevent the reference from growing indefinitely. After boundary constraints, the datum always remains within the closed interval.

6. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The method of setting adaptive coefficients for each axis based on the noise characteristics differences between the accelerometer and the gyroscope includes: A first coefficient value is set for each axis of the accelerometer, and a second coefficient value is set for each axis of the gyroscope. The first and second coefficient values ​​are preset according to the dimensional differences of the accelerometer and the gyroscope and their different characteristics of motion sensitivity, respectively, without the need for online calibration.

7. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The dynamic adjustment of the multi-axis voting threshold based on the current global noise level includes: The mean of the three-axis smoothing reference of the accelerometer is compared with the nominal static variance reference value. When the mean is significantly less than the reference value, the voting threshold is set to require all three axes to pass. When the mean value approaches or exceeds the reference value, the voting threshold is automatically lowered to avoid missed detections due to increased noise levels.

8. The zero-speed detection method for vehicle-mounted IMU based on exponential smoothing filtering and boundary constraints according to claim 1, characterized in that, The output static data includes: The original inertial measurement unit (IMU) data within the window that meets the static determination criteria is marked as a high-quality static data segment and output to the IMU mounting angle calibration module or zero-velocity correction module for subsequent navigation error correction.

9. A vehicle-mounted IMU zero-speed detection system based on exponential smoothing filtering and boundary constraints, characterized in that, include: The statistics and tracking module is used to calculate the real-time variance of the six-axis data of the inertial measurement unit based on a sliding window, continuously track the historical minimum variance of each axis as the noise baseline benchmark, and adaptively determine the dynamic stable fluctuation range of each axis based on the noise baseline benchmark, which is used to screen for suspected stationary periods of the vehicle. The smoothing and constraint module is used to perform weighted iterative updates on the historical minimum variance benchmark using a first-order exponential smoothing filter to eliminate benchmark step jumps caused by random minimum values ​​in the window. It applies preset upper and lower boundary constraints to the smoothed benchmark to prevent the benchmark from locking down unidirectionally or drifting upwards for a long time, thereby achieving bidirectional adaptive adjustment of the benchmark. The voting and decision module is used to set adaptive coefficients for each axis based on the noise characteristics differences between the accelerometer and the gyroscope, generate dynamic decision thresholds for each axis in combination with the smoothed constraint benchmark, dynamically adjust the multi-axis voting thresholds according to the current global noise level, and jointly vote on the decision results of each axis to complete the determination of the vehicle's stationary state and output static data.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the vehicle-mounted IMU zero-speed detection method based on exponential smoothing filtering and boundary constraints as described in any one of claims 1 to 8.

Images