Asphalt pavement paver control method based on multi-sensor array redundancy

Abstract

The application discloses an asphalt pavement paver control method based on a multi-sensor array redundancy. Balance beams are arranged on both sides of a paver, at least a GNSS sensor, a double-axis inclination sensor, an IMU measurement module and an attitude sensor are installed on each group of balance beams, at least a central, left and right cross slope sensor and a temperature sensor are installed on a screed beam, a 360-degree surround view camera and an infrared thermal imaging camera are installed on the top of the paver, a variable paving system is built, a multi-sensor redundancy is built by using various sensors, a multi-dimensional parameter coupling mathematical model is built, data fusion is carried out to obtain fused data, the fused data is converted into a paver variable, a loose paving elevation is calculated in real time, when a loose paving thickness error is greater than 5 mm, a lifting cylinder is adjusted, when an inclination angle is greater than 0.3 degrees, a differential cylinder is started to correct, and when a thickness deviation is greater than 3 mm, a spiral distributor rotating speed is adjusted. The application avoids defects of traditional single-point monitoring, automatically switches a mechanism and guarantees monitoring continuity.

Description

Technical Field

[0001] This invention belongs to the field of intelligent road paving technology, specifically relating to a control method for asphalt pavers based on multi-sensor array redundancy. Background Technology

[0002] With the continuous improvement of national transportation infrastructure construction experience, high-precision and high-quality construction has gradually been valued by engineering builders. The thickness, elevation accuracy, and smoothness of asphalt pavement paving construction directly determine the pavement's performance and service life, and are one of the core indicators for pavement engineering quality control.

[0003] Traditional construction methods typically employ single-balance beam control, slipper method, and wire method as core control means for paving thickness and elevation. These methods involve suspending slippers from the paver's tow bar, using distance sensors mounted on the slippers or pre-set wires to measure the vertical distance between the slipper and the underlying surface, indirectly providing feedback on the paving thickness, and subsequently triggering the leveling cylinders to adjust the screed's posture. However, this technology has inherent accuracy limitations: First, slipper measurement is a mechanical contact method, easily affected by local unevenness of the underlying surface, material adhesion, vehicle vibration, and errors in the wire elevation reference, resulting in elevation measurement errors exceeding ±5mm and poor paving thickness uniformity. Second, slippers rely on a pre-set reference surface; during long-distance construction, reference surface deviations are difficult to correct in real time, easily causing longitudinal wavy flatness defects, failing to meet the ±2mm thickness accuracy requirement for high-grade highways. Summary of the Invention

[0004] Purpose of the invention: To provide a control method for asphalt pavers based on multi-sensor array redundancy, which solves the above-mentioned problems existing in the prior art.

[0005] Technical solution: A control method for asphalt pavement pavers based on multi-sensor array redundancy, comprising the following steps:

[0006] Balance beams are installed on both sides of the paver. Each balance beam is equipped with at least a GNSS sensor, a dual-axis tilt sensor, an IMU measurement module, and an attitude sensor. The paver is equipped with a screed crossbeam, on which at least three cross slope sensors and temperature sensors are installed (center, left, and right). At least a 360° surround-view camera and an infrared thermal imaging camera are installed on the top of the paver. A variable paving system is established, and when sensors fail, the system automatically switches to a new mechanism to maintain continuous parameter monitoring. The variable paving system is used for spatiotemporal alignment of sensor data and dynamic parameter calculation. A multi-sensor redundancy system is constructed using sensors, a dual-axis tilt sensor, an IMU measurement module, a 360° surround-view camera, attitude sensors, and cross-slope sensors. This system acquires redundant data from multiple sensors and builds a multi-dimensional parameter-coupled mathematical model. Elevation data, cross-slope data, and attitude data are then fused redundantly to obtain elevation, cross-slope, and attitude data. These data are then converted into paver variables. During paver variable control, the multi-dimensional parameter-coupled mathematical model is used to calculate the loose paving elevation in real time and correct errors. When the loose paving thickness error exceeds 5mm, the variable paving system outputs an adjustment to the screed lifting cylinder. The multi-dimensional parameter-coupled mathematical model is also used to calculate the cross-slope in real time and adjust the tilt. When the tilt angle exceeds 0.3°, the variable paving system activates a differential cylinder for correction. A laser balance beam and an ultrasonic balance beam are used to detect the loose paving thickness in real time. When the thickness deviation exceeds 3mm, the auger distributor speed should be adjusted.

[0007] Preferably, the steps for calculating elevation data are as follows: Obtain Raw data from the sensor Filtering and smoothing are performed to eliminate jumps and noise caused by paver vibration and satellite signal interference, resulting in a smoothed output. Elevation The calculation formula is as follows: Where: preset elevation jump threshold ,when Then remove Then pre-processed The elevation data is weighted and redundantly fused with the elevation data of the balance beam and IMU, and the weights of each sensor are dynamically allocated to eliminate single-sensor system errors and output weighted redundant elevation data.

[0008] The preferred formula for calculating weighted redundant elevation data is as follows: In the formula: Indicates weighted redundant elevation; This indicates the dynamic allocation of weights for GNSS sensors; This indicates the dynamic allocation of the weights assigned to the balance beam; This indicates the dynamic allocation of weights to the IMU measurement module; This represents GNSS sensor elevation data; This indicates the elevation data of the balance beam; This represents the elevation data from the IMU measurement module.

[0009] Preferably, the calculation steps for cross slope data are as follows: Obtain the cross slope angles of three sets of cross slope sensors and perform averaging to generate a reference cross slope value, the calculation formula of which is as follows: In the formula: This represents the reference cross slope value, the average value of the cross slopes on the left and right sides, used for subsequent filtering and anomaly detection; The values ​​represent the cross slope values ​​at the left end, which are the cross slopes measured by the sensors on the left side. This represents the right-side cross slope value, which is the preset cross slope angle threshold measured by the right-side sensor. Abnormal data with excessive deviations are removed. If so, then the data in that cell will be removed. This represents the cross slope deviation threshold, which is... =0.3°; After removing the cross slope values, the resulting cross slope data is filtered using a moving average to suppress high-frequency vibration noise, yielding a smoothed single-source cross slope value. The calculation formula is as follows: In the formula: This represents the smoothed single-source cross slope value; Indicates the length of the sliding window, a typical value. =3; Indicates the first time; Indicates the first At any given time, the cross slope sensor in the middle is used as the primary cross slope sensor, while the cross slope sensors on the left and right sides serve as redundant cross slope sensors. This allows for the calculation of the primary single-source cross slope value. and redundant single-source cross slope values The main cross slope sensor and redundant cross slope sensors are dynamically weighted and fused. The weight coefficients are then determined by dynamic or static weight allocation or fault switching to eliminate the system error of a single cross slope sensor and output the fused cross slope angle.

[0010] Preferably, the static weight allocation process is as follows: obtain the measurement standard deviation of the main cross slope sensor and the redundant cross slope sensor. , Based on the measurement standard deviation of the main cross slope sensor and redundant sensors , Assigning weights and calculating primary static adjustment weights and redundant static adjustment weights The smaller the standard deviation, the greater the weight; the primary static adjustment weight. The calculation formula is as follows: Redundant static adjustment weights The calculation formula is as follows: .

[0011] Preferably, the dynamic weight allocation process is as follows: obtain the health scores of the main cross slope sensor and the redundant cross slope sensor. , The weights for the active dynamic adjustment are calculated based on the health scores of the main cross slope sensor and redundant sensors. and redundant dynamic adjustment weight Among them, the higher the health score, the greater the weight, and the weight is dynamically adjusted. The calculation formula is as follows: Redundancy dynamically adjusted weights The calculation formula is as follows: Preset the switching threshold for the cross slope sensor. When the main cross slope sensor malfunctions, when If the main sensor fault flag is True, then a weight switch will be forced. , The cross slope value is provided separately by a redundant cross slope sensor.

[0012] Preferably, the attitude data calculation steps are as follows: First, low-pass filtering is applied to the raw attitude angles independently output by each sensor for data preprocessing, outputting single-source attitude data. Then, a Kalman filter algorithm is used to weight and calculate errors in the smoothed attitude data from the IMU measurement module, the dual-axis tilt sensor, and the attitude sensor, outputting a high-precision, high-fault-tolerant fused attitude angle. The single-source attitude data processing steps are as follows: Obtain the roll angle output by the attitude sensor. Pitch angle and heading angle The preprocessing process eliminates vibration noise and corrects for deviations caused by sensor installation tilt. Then, low-pass vibration filtering, installation tilt correction, and heading angle unwrapping calculations are performed. The calculation formula for low-pass vibration filtering is as follows: In the formula: Represents the filter coefficients, with weight coefficients 0 < <1 determines the degree to which new data contributes to the output; This represents the smoothed attitude angle, the filtered output at the current moment, which serves as the input for the next moment. This represents the original attitude angle measured by the sensor at step t at the current time, and the attitude angle output by the IMU measurement module. Indicates the previous moment The filtering results of the step; This indicates the nth sampling time. Representing the previous sampling time, forming a time series. =0,1,2,...,n, a preset attitude threshold is used to remove outliers. Then remove .

[0013] in, This represents the elevation jump threshold. =0.02m; The calculation formula for installation tilt correction is as follows: In the formula: express The actual vehicle body posture angle after constant correction for installation deviations; express The original attitude angle data of the vehicle body after constant vibration and noise filtering; This indicates the fixed system deviation caused by improper installation of the vehicle body attitude sensor; the formula for calculating the unwrapping of the heading angle is as follows: In the formula: Indicates the first The paver's heading correction value at any given moment; Indicates the first The filtered value of the paver's heading at any given moment; This represents the modulo operation, which calculates the remainder after dividing θYaw(t) by 360°, and normalizes the angle value to the interval [0°, 360°].

[0014] The preferred high-precision, high-fault-tolerant fused attitude angle calculation process is as follows: Based on the fuselage attitude provided by the IMU module, the dual-axis tilt sensor and the body attitude sensor provide independent attitude data, which are then fused through Kalman filtering. The calculation formula is as follows: In the formula, This represents the fused vehicle attitude state vector, including the roll angle. Pitch angle and heading angle The mathematical dimension is n×1, where n represents the dimension of the state vector and the number of attitude parameters the system needs to estimate. A typical value for n is given if roll angle is included. and pitch angle When n=2, then n=2; if the heading angle is included. Therefore, n=3; The vector represents the vehicle body attitude state of the IMU measurement module, including noise and drift, with a mathematical dimension of n×1; This represents the vehicle attitude observations measured by the tilt sensor, with a mathematical dimension of m×1, where m represents the dimension of the observation vector and represents the number of attitude parameters measured by the tilt sensor. A dual-axis tilt sensor measures the roll angle. and pitch angle When m=2, the dual-axis tilt sensor measures the roll angle. If m=1, then m=1; The Kalman gain matrix determines the weights of the IMU and tilt sensor data in the fusion process, and its mathematical dimension is n×m. The observation matrix represents the mapping of the IMU measurement module state to the observation space of the tilt sensor, with a mathematical dimension of m×n; the attitude redundancy fusion calculation formula is as follows: In the formula: , Represents the fusion coefficient. It is dynamically adjusted based on the noise from the attitude sensor. Indicates the attitude redundancy angle; Indicates the angle of the IMU measurement module; Indicates the angle of the dual-axis tilt sensor; The angles of the vehicle body attitude sensors are represented. The fused roll angle, fused pitch angle, and fused yaw angle are calculated using the attitude redundancy fusion calculation formula. Substituting the fused roll angle and fused pitch angle into the following formula, the actual elevation of the screed is calculated. The calculation formula is as follows: In the formula, This indicates the horizontal distance between the balance beam and the ironing board; This indicates the vertical height difference between the balance beam and the screed board; Indicates the longitudinal slope angle, The cross slope angles are represented by the front-to-back tilt angle and the left-to-right tilt angle of the paver; and the cross slope correction for the screed is calculated using the following formula: In the formula: Indicates the correction of the cross slope of the ironing board; express The baseline cross slope angle after constant sensor filtering and noise reduction data preprocessing; This indicates the design thickness of the paving layer or the effective working height of the screed. This represents the effective horizontal length of one side of the screed. Through sensor extrinsic parameter calibration, the relative positions of the GNSS and the laser balance beam are measured using a total station, generating a transformation matrix. The GNSS elevation is then converted to the screed reference point. The calculation formula is as follows: In the formula: express The elevation of the screed is constantly derived and calculated from GNSS data; express Elevation data obtained from GNSS sensors at any given time; This indicates the horizontal distance between the GNSS installation location and the ironing plate; express The sine value of the pitch angle of the paver at all times.

[0015] Preferably, the data preprocessing procedure for the dual-axis tilt sensor is as follows: Obtain the lateral tilt angle of the dual-axis tilt sensor. longitudinal tilt angle Preprocessing eliminates temperature drift errors and corrects deviations caused by the non-parallelism between the mounting shaft of the dual-axis tilt sensor and the vehicle body shaft. Specifically, tilt temperature compensation is performed first, calculated using the following formula: In the formula, This represents the attitude angle after temperature compensation, or the attitude sensor output after eliminating the effects of temperature. This represents the original attitude angle, the attitude angle directly output by the attitude sensor; This represents the temperature coefficient, indicating the sensor's output sensitivity as temperature changes. =0.005° / ℃; The current temperature is indicated by either measurement from the built-in thermometer of the attitude sensor or by an external temperature sensor. This represents the calibration temperature at 25℃; the calculation formula for axial deviation correction is as follows: ;in, This represents the attitude angle after axial correction, and the attitude sensor output after eliminating installation axis deviation. This represents the attitude angle after temperature compensation, and is output using the same temperature compensation formula. This represents the vertical axis attitude angle, output by the attitude sensor in the direction perpendicular to the current axis. This indicates the angle between the attitude sensor axis and the vehicle body axis, determined through calibration, with a typical value. Then, median filtering is performed to eliminate transient noise. The calculation formula is as follows: In the formula, This represents the attitude angle after median filtering, and the sensor output after eliminating instantaneous noise; This represents the attitude angle after axial correction, and is the same as the axial correction formula.

[0016] Preferably, the specific steps for spatiotemporal alignment and dynamic parameter calculation of data from each sensor using a variable averaging system are as follows: In the embedded system, a hardware timer is used to accurately timestamp the data from each sensor. Through resampling, the high-frequency data from the IMU measurement module is downsampled to be synchronized with GNSS. The linear interpolation is calculated using the following formula. In the formula, Indicates the GNSS sampling time; This indicates the most recent IMU sampling time. Through the calibration of the extrinsic parameters of each sensor, the relative position of the GNSS sensor and the balance beam is measured using a total station. A transformation matrix is ​​generated, and the elevation of the GNSS sensor is converted to the screed reference point. The calculation formula is as follows: In the formula: This indicates the elevation conversion of the GNSS sensor to the ironing plate reference point; express The elevation data obtained from GNSS sensors at all times; when calculating the real-time elevation of the Songpu surface, the pre-processed GNSS elevation data should be input first. Design Elevation And the thickness of the pine pine The theoretical and actual elevations of the loose pavement are calculated using the following formula: ; In the formula, Indicates the compaction coefficient; Indicates the design loose layer thickness; express The screed elevation data is preprocessed by multi-sensor fusion at all times; when calculating the cross slope control, the fused cross slope is input first. Design cross slope The cross slope control quantity is calculated using the following formula: In the formula, Indicates the width of the ironing board; This indicates the cross slope control gain.

[0017] Beneficial Effects: This invention relates to a control method for asphalt pavers based on multi-sensor array redundancy. Multiple sets of sensors of various types are installed on the balance beam, screed crossbeam, and the top of the paver to construct a variable paving system. A multi-sensor redundancy system is built using sensors, dual-axis tilt sensors, IMU measurement modules, 360° surround-view cameras, attitude sensors, and cross-slope sensors. This system provides backup protection for parameter monitoring at the hardware level. When a single sensor fails, the redundant unit immediately takes over, avoiding the drawback of traditional single-point monitoring where a single failure will cause all systems to stop operating. The automatic switching mechanism requires no manual intervention, ensuring continuous monitoring. Furthermore, in conjunction with a multi-dimensional parameter coupling mathematical model, the system can quickly match the optimal parameters for different construction scenarios, adapting to complex construction conditions. It performs redundant fusion of elevation data, cross-slope data, and attitude data, eliminating abnormal data and preventing misjudgments caused by environmental interference. Attached Figure Description

[0018] Figure 1 This is a schematic diagram showing the installation of each sensor in this invention; Figure 2 This is a table showing the measurement accuracy of the attitude sensor of the present invention; Figure 3 This describes the method for inspecting the installation quality of each sensor in this invention.

[0019] Figure 1 The attached diagram is labeled as follows: 1. IMU measurement module; 2. 360° surround view camera; 3. Infrared thermal imaging camera; 4. Temperature sensor; 5. Dual-axis tilt sensor; 6. Cross slope sensor; 7. Attitude sensor. Detailed Implementation

[0020] like Figures 1 to 3As shown, this invention provides a technical solution: a control method for an asphalt pavement paver based on multi-sensor array redundancy, comprising the following steps: setting up balance beams on both sides of the paver, with at least a GNSS sensor, a dual-axis tilt sensor, an IMU measurement module, and an attitude sensor installed on each balance beam. In this embodiment, the horizontal positioning deviation of the GNSS sensor is no greater than 10mm; the vertical positioning deviation is no greater than 2mm, its protection level is IP67 or higher, it is dustproof and waterproof, and its operating temperature range is -30℃ to +60℃; the measurement range of the dual-axis tilt sensor should cover the transverse ± The static accuracy should be maintained within -0.05° to +0.05°, and within -0.02° to +0.02° in high-precision scenarios, corresponding to a paving thickness error within -1mm to +1mm. Under paver vibration, the dynamic accuracy should decrease to within -0.1° to +0.1° to ensure real-time control stability. The resolution should reach 0.001° to capture minute tilt changes. The sampling rate should be no less than 20Hz, and greater than 50Hz in high-precision scenarios. The operating temperature range should meet the following requirements. The temperature range is 40℃~+85℃, and the protection level should be IP67 or higher. At least three cross slope sensors and temperature sensors should be installed on the crossbeam of the screed: one on the middle, one on the left, and one on the right. The cross slope sensor measurement range should cover -10%~+10% to accommodate different design cross slopes and paver posture changes. The static measurement accuracy should meet the requirements of ±0.02% for slope error and ±0.03° for angle error. Under paver vibration, the dynamic accuracy attenuation should be controlled between -0.1% and +0.01% to ensure real-time control stability. The resolution should reach 0.001% or 0.001° to capture minute cross slopes. Slope variation; sampling rate should be no less than 20Hz, and no less than 50Hz for high-precision scenarios, matching the GNSS sampling rate, tilt sensor, and paver hydraulic system response frequency; temperature range should be -40℃ to +85℃; protection level should be IP67 or higher, with vibration resistance, able to withstand vibration acceleration greater than 10g, frequency 10-2000Hz, to avoid data fluctuations caused by paver vibration; at least a 360° surround view camera and an infrared thermal imaging camera should be installed on the top of the paver to build a variable paving system, wherein the installation quality detection method of each sensor is as follows: Figure 3 As shown, when a sensor fails, the variable paving system automatically switches to a switching mechanism to maintain continuous parameter monitoring, utilizing the variable paving system for spatiotemporal alignment of data from each sensor and dynamic parameter calculation; A multi-sensor redundancy system is constructed using sensors, a dual-axis tilt sensor, an IMU measurement module, a 360° surround-view camera, an attitude sensor, and a cross-slope sensor. This system acquires redundant data from multiple sensors, builds a multi-dimensional parameter-coupled mathematical model, and performs redundant fusion of elevation, cross-slope, and attitude data to obtain elevation, cross-slope, and attitude data. This allows for rapid matching of optimal parameters for different construction scenarios, adapting to complex construction conditions. The system also performs redundant fusion of elevation, cross-slope, and attitude data, eliminating abnormal data and misjudgments caused by environmental interference. The elevation, cross-slope, and attitude data are then converted into paver variables. The elevation data calculation steps are as follows: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] Raw data from the sensor Filtering and smoothing are performed to eliminate jumps and noise caused by paver vibration and satellite signal interference, resulting in a smoothed output. Elevation The calculation formula is as follows: Where: preset elevation jump threshold ,when Then remove Then pre-processed The elevation data is weighted and redundantly fused with the elevation data from the balance beam and IMU, dynamically allocating the weights of each sensor to eliminate single-sensor system errors, and outputting weighted redundant elevation data. The calculation formula for the weighted redundant elevation data is as follows: In the formula: Indicates weighted redundant elevation; This indicates the dynamic allocation of weights for GNSS sensors; This indicates the dynamic allocation of the weights assigned to the balance beam; This indicates the dynamic allocation of weights to the IMU measurement module; This represents GNSS sensor elevation data; This indicates the elevation data of the balance beam; This represents the elevation data from the IMU measurement module.

[0021] The calculation steps for cross slope data are as follows: Obtain the cross slope angles from three sets of cross slope sensors and perform averaging to generate a reference cross slope value. The calculation formula is as follows:

[0022] In the formula: This represents the reference cross slope value, the average value of the cross slopes on the left and right sides, used for subsequent filtering and anomaly detection; The values ​​represent the cross slope values ​​at the left end, which are the cross slopes measured by the sensors on the left side. This represents the right-side cross slope value, which is the preset cross slope angle threshold measured by the right-side sensor. Abnormal data with excessive deviations are removed. If so, then the data in that cell will be removed. This represents the cross slope deviation threshold, which is... =0.3°;

[0023] After removing the cross slope values, the resulting cross slope data is filtered using a moving average to suppress high-frequency vibration noise, yielding a smoothed single-source cross slope value. The calculation formula is as follows: In the formula: This represents the smoothed single-source cross slope value; Indicates the length of the sliding window, a typical value. =3; Indicates the first time; Indicates the first At any given time, the cross slope sensor in the middle is used as the primary cross slope sensor, while the cross slope sensors on the left and right sides serve as redundant cross slope sensors. This allows for the calculation of the primary single-source cross slope value. and redundant single-source cross slope values The main cross slope sensor and redundant cross slope sensors are dynamically weighted and fused. The weight coefficients are then determined by dynamic or static weight allocation or fault switching to eliminate the system error of a single cross slope sensor and output the fused cross slope angle.

[0024] In a further embodiment, the static weight allocation process is as follows: The measurement standard deviations of the main cross slope sensor and the redundant cross slope sensor are obtained. , Based on the measurement standard deviation of the main cross slope sensor and redundant sensors , Assigning weights and calculating primary static adjustment weights and redundant static adjustment weights The smaller the standard deviation, the greater the weight; the primary static adjustment weight. The calculation formula is as follows: Redundant static adjustment weights The calculation formula is as follows: .

[0025] The dynamic weight allocation process is as follows: Obtain the health scores of the main cross slope sensor and the redundant cross slope sensor. , The weights for the active dynamic adjustment are calculated based on the health scores of the main cross slope sensor and redundant sensors. and redundant dynamic adjustment weight Among them, the higher the health score, the greater the weight, and the weight is dynamically adjusted. The calculation formula is as follows: Redundancy dynamically adjusted weights The calculation formula is as follows: Preset the switching threshold for the cross slope sensor. When the main cross slope sensor malfunctions, when If the main sensor fault flag is True, then a weight switch will be forced. , The cross slope value is provided separately by a redundant cross slope sensor.

[0026] The attitude data calculation steps are as follows: First, low-pass filtering is applied to the raw attitude angles independently output by each sensor for data preprocessing, outputting single-source attitude data. Then, a Kalman filter algorithm is used to weight and calculate errors in the smoothed attitude data from the IMU measurement module, dual-axis tilt sensor, and attitude sensor, outputting a high-precision, high-fault-tolerant fused attitude angle. The single-source attitude data processing steps are as follows: Obtain the roll angle output by the attitude sensor. Pitch angle and heading angle The preprocessing process eliminates vibration noise and corrects for deviations caused by sensor installation tilt. Then, low-pass vibration filtering, installation tilt correction, and heading angle unwrapping calculations are performed. The calculation formula for low-pass vibration filtering is as follows: In the formula: Represents the filter coefficients, with weight coefficients 0 < <1 determines the degree to which new data contributes to the output; This represents the smoothed attitude angle, the filtered output at the current moment, which serves as the input for the next moment. This represents the original attitude angle measured by the sensor at step t at the current time, and the attitude angle output by the IMU measurement module. Indicates the previous moment The filtering results of the step; This indicates the nth sampling time. Representing the previous sampling time, forming a time series. =0,1,2,...,n, a preset attitude threshold is used to remove outliers. Then remove .

[0027] in, This represents the elevation jump threshold. =0.02m; The calculation formula for installation tilt correction is as follows: In the formula: express The actual vehicle body posture angle after constant correction for installation deviations; express The original angle data of the vehicle body attitude after constant vibration and noise filtering, which includes the angle data of roll angle and pitch angle; This indicates the fixed system deviation caused by improper installation of the vehicle body attitude sensor; the formula for calculating the unwrapping of the heading angle is as follows: In the formula: Indicates representative The corrected heading angle at any time; Indicates representative The heading angle obtained by filtering the time sensor; This represents the modulo operation, which calculates the remainder after dividing θYaw(t) by 360°, and normalizes the angle value within the interval [0°, 360°]. The attitude sensor measurement accuracy is shown in the table below. Figure 2 As shown, multiple sets of sensors of various types are installed on the top of the leveling beam, screed beam, and paver to build a variable paving system. The system employs multiple sensor redundancies, including sensors, dual-axis tilt sensors, IMU measurement modules, 360° surround-view cameras, attitude sensors, and cross-slope sensors. This hardware-level backup ensures parameter monitoring. When a single sensor fails, the redundant unit immediately takes over, avoiding the drawback of traditional single-point monitoring where a single failure can cause the entire system to stop operating. The automatic switching mechanism requires no manual intervention, ensuring continuous monitoring.

[0028] When implementing variable control for the paver, a multi-dimensional parameter-coupled mathematical model is used to calculate the loose paving elevation in real time and correct errors. When the loose paving thickness error exceeds 5mm, the variable paving system outputs an adjustment to the screed lifting cylinder. The multi-dimensional parameter-coupled mathematical model is also used to calculate the cross slope in real time and adjust the tilt. When the tilt angle exceeds 0.3°, the variable paving system activates a differential cylinder for correction. A laser balance beam and an ultrasonic balance beam are used to detect the loose paving thickness in real time. When the thickness deviation exceeds 3mm, the auger spreader speed should be adjusted. The calculation process for the high-precision, high-fault-tolerant fused attitude angle is as follows: Based on the machine attitude provided by the IMU module, independent attitude data is provided by the dual-axis tilt sensor and the vehicle attitude sensor. This data is then fused using a Kalman filter, and the calculation formula is as follows: In the formula, This represents the fused vehicle attitude state vector, including the roll angle. Pitch angle and heading angle The mathematical dimension is n×1, where n represents the dimension of the state vector and the number of attitude parameters the system needs to estimate. A typical value for n is given if roll angle is included. and pitch angle When n=2, then n=2; if the heading angle is included. Therefore, n=3; The vector represents the vehicle body attitude state of the IMU measurement module, including noise and drift, with a mathematical dimension of n×1; This represents the vehicle attitude observations measured by the tilt sensor, with a mathematical dimension of m×1, where m represents the dimension of the observation vector and represents the number of attitude parameters measured by the tilt sensor. A dual-axis tilt sensor measures the roll angle. and pitch angle When m=2, the dual-axis tilt sensor measures the roll angle. If m=1, then m=1; The Kalman gain matrix determines the weights of the IMU and tilt sensor data in the fusion process, and its mathematical dimension is n×m. The observation matrix represents the mapping of the IMU measurement module state to the observation space of the tilt sensor, with a mathematical dimension of m×n; the attitude redundancy fusion calculation formula is as follows: In the formula: , Represents the fusion coefficient. It is dynamically adjusted based on the noise from the attitude sensor.

[0029] Indicates the attitude redundancy angle; Indicates the angle of the IMU measurement module; Indicates the angle of the dual-axis tilt sensor; The angles of the vehicle body attitude sensors are represented. The fused roll angle, fused pitch angle, and fused yaw angle are calculated using the attitude redundancy fusion calculation formula. Substituting the fused roll angle and fused pitch angle into the following formula, the actual elevation of the screed is calculated. The calculation formula is as follows: In the formula, This indicates the horizontal distance between the balance beam and the ironing board; This indicates the vertical height difference between the balance beam and the screed board; Indicates the longitudinal slope angle, The cross slope angles are represented by the front-to-back tilt angle and the left-to-right tilt angle of the paver; and the cross slope correction for the screed is calculated using the following formula: In the formula: Indicates the correction of the cross slope of the ironing board; express The baseline cross slope angle after constant sensor filtering and noise reduction data preprocessing; This indicates the design thickness of the paving layer or the effective working height of the screed. This represents the effective horizontal length of one side of the ironing board. In this embodiment, the effective horizontal length of one side of the ironing board is half the total width of the ironing board. Through sensor extrinsic parameter calibration, the relative position of GNSS and laser balance beam is measured by a total station to generate a transformation matrix, and the GNSS elevation is converted to the ironing board reference point. The calculation formula is as follows: In the formula: express The elevation of the screed is constantly derived and calculated from GNSS data; express Elevation data obtained from GNSS sensors at any given time; This indicates the horizontal distance between the GNSS installation location and the ironing plate; express The sine value of the pitch angle of the paver at all times.

[0030] The data preprocessing process for the dual-axis tilt sensor is as follows: Obtain the lateral tilt angle of the dual-axis tilt sensor. longitudinal tilt angle Preprocessing eliminates temperature drift errors and corrects deviations caused by the non-parallelism between the mounting shaft of the dual-axis tilt sensor and the vehicle body shaft. Specifically, tilt temperature compensation is performed first, calculated using the following formula: In the formula, This represents the attitude angle after temperature compensation, or the attitude sensor output after eliminating the effects of temperature. This represents the original attitude angle, the attitude angle directly output by the attitude sensor; This represents the temperature coefficient, indicating the sensor's output sensitivity as temperature changes. =0.005° / ℃; The current temperature is indicated by either measurement from the built-in thermometer of the attitude sensor or by an external temperature sensor. This represents the calibration temperature at 25℃; the calculation formula for axial deviation correction is as follows: ;in, This represents the attitude angle after axial correction, and the attitude sensor output after eliminating installation axis deviation. This represents the attitude angle after temperature compensation, and is output using the same temperature compensation formula. This represents the vertical axis attitude angle, output by the attitude sensor in the direction perpendicular to the current axis. This indicates the angle between the attitude sensor axis and the vehicle body axis, determined through calibration, with a typical value. Then, median filtering is performed to eliminate transient noise. The calculation formula is as follows: In the formula, This represents the attitude angle after median filtering, and the sensor output after eliminating instantaneous noise; This represents the attitude angle after axial correction, and is the same as the axial correction formula.

[0031] The specific steps for spatiotemporal alignment and dynamic parameter calculation of sensor data using the variable averaging system are as follows: In the embedded system, a hardware timer is used to accurately timestamp the data of each sensor. Through resampling, the high-frequency data of the IMU measurement module is downsampled to be synchronized with GNSS. The following formula is used to calculate linear interpolation. In the formula, Indicates the GNSS sampling time; This indicates the most recent IMU sampling time. Through the calibration of the extrinsic parameters of each sensor, the relative position of the GNSS sensor and the balance beam is measured using a total station. A transformation matrix is ​​generated, and the elevation of the GNSS sensor is converted to the screed reference point. The calculation formula is as follows: In the formula: express The elevation of the screed is constantly derived and calculated from GNSS data; express Elevation data obtained from GNSS sensors at any given time; This indicates the horizontal distance between the GNSS installation location and the ironing plate; express The sine of the paver's pitch angle should be entered at all times; when calculating the real-time elevation of the loose pavement, the pre-processed GNSS elevation should be entered first. Design Elevation And the thickness of the pine pine The theoretical and actual elevations of the loose pavement are calculated using the following formula: ; In the formula, Indicates the compaction coefficient; Indicates the design loose layer thickness; express The screed elevation data is preprocessed by multi-sensor fusion at all times; when calculating the cross slope control, the fused cross slope is input first. Design cross slope The cross slope control quantity is calculated using the following formula: In the formula, Indicates the width of the ironing board; This indicates the cross slope control gain.

[0032] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A control method for asphalt pavement pavers based on multi-sensor array redundancy, characterized in that, Includes the following steps: Balance beams are installed on both sides of the paver. Each balance beam is equipped with at least a GNSS sensor, a dual-axis tilt sensor, an IMU measurement module, and an attitude sensor. The paver is equipped with a screed beam, which is equipped with at least a cross slope sensor and a temperature sensor in the middle, left, and right sides. At least a 360° surround view camera and an infrared thermal imaging camera are installed on the top of the paver. A variable paving system is built, and when the sensors fail, the variable paving system automatically switches to maintain continuous parameter monitoring. The variable paving system is used to align the data from each sensor in time and space and to dynamically calculate the parameters. by The system employs sensors, dual-axis tilt sensors, IMU measurement modules, 360° surround view cameras, attitude sensors, and cross slope sensors to build a multi-sensor redundancy, acquire multi-sensor redundancy data, build a multi-dimensional parameter coupling mathematical model, and perform redundancy fusion of elevation data, cross slope data, and attitude data to obtain elevation data, cross slope data, and attitude data. The elevation data, cross slope data, and attitude data are then converted into paver variables. When implementing variable control for the paver, a multi-dimensional parameter coupled mathematical model is used to calculate the loose paving elevation in real time and correct errors. When the loose paving thickness error is greater than 5mm, the variable paving system outputs an adjustment to the screed lifting cylinder. The multi-dimensional parameter coupled mathematical model is used to calculate the cross slope in real time and adjust the tilt. When the tilt angle is greater than 0.3°, the variable paving system activates the differential cylinder correction. The laser balance beam and ultrasonic balance beam are used to detect the loose paving thickness in real time. When the thickness deviation is greater than 3mm, the speed of the auger spreader should be adjusted.

2. The asphalt paver control system based on multi-sensor array redundancy according to claim 1, characterized in that, The steps for calculating elevation data are as follows: Get Raw data from the sensor Filtering and smoothing are performed to eliminate jumps and noise caused by paver vibration and satellite signal interference, resulting in a smoothed output. Elevation The calculation formula is as follows: ； Where: Preset elevation jump threshold ,when Then remove ; Then pre-processed The elevation data is weighted and redundantly fused with the elevation data of the balance beam and IMU, and the weights of each sensor are dynamically allocated to eliminate single-sensor system errors and output weighted redundant elevation data.

3. The asphalt paver control system based on multi-sensor array redundancy according to claim 2, characterized in that, The formula for calculating weighted redundant elevation data is as follows: ； In the formula: Indicates weighted redundant elevation; This indicates the dynamic allocation of weights for GNSS sensors; This indicates the dynamic allocation of the weights assigned to the balance beam; This indicates the dynamic allocation of weights to the IMU measurement module; This represents GNSS sensor elevation data; This indicates the elevation data of the balance beam; This represents the elevation data from the IMU measurement module.

4. The asphalt paver control system based on multi-sensor array redundancy according to claim 1, characterized in that, The calculation steps for cross slope data are as follows: The cross slope angles of three sets of cross slope sensors are acquired and averaged to generate a reference cross slope value, calculated using the following formula: ； In the formula: This represents the reference cross slope value, the average value of the cross slopes on the left and right sides, used for subsequent filtering and anomaly detection; The values ​​represent the cross slope values ​​at the left end, which are the cross slopes measured by the sensors on the left side. This represents the right-side cross slope value, which is the preset cross slope angle threshold measured by the right-side sensor. Abnormal data with excessive deviations are removed. If so, then the data in that cell will be removed. This represents the cross slope deviation threshold, which is... =0.3°; After removing the cross slope values, the resulting cross slope data is filtered using a moving average to suppress high-frequency vibration noise, yielding a smoothed single-source cross slope value. The calculation formula is as follows: ； In the formula: This represents the smoothed single-source cross slope value; Indicates the length of the sliding window, a typical value. =3; Indicates the first At that moment; Indicates the first At that moment; The cross slope sensor in the middle serves as the primary cross slope sensor, while the cross slope sensors on the left and right sides serve as redundant cross slope sensors. This allows the calculation of the primary single-source cross slope value. and redundant single-source cross slope values The main cross slope sensor and redundant cross slope sensors are dynamically weighted and fused. The weight coefficients are then determined by dynamic or static weight allocation or fault switching to eliminate the system error of a single cross slope sensor and output the fused cross slope angle.

5. The asphalt paver control system based on multi-sensor array redundancy according to claim 4, characterized in that, The process of static weight allocation is as follows: Obtain the measurement standard deviation of the primary cross slope sensor and the redundant cross slope sensor. , Based on the measurement standard deviation of the main cross slope sensor and redundant sensors , Assigning weights and calculating primary static adjustment weights and redundant static adjustment weights The smaller the standard deviation, the greater the weight; the primary static adjustment weight. The calculation formula is as follows: ； Redundant static adjustment weights The calculation formula is as follows: 。 6. The asphalt paver control system based on multi-sensor array redundancy according to claim 4, characterized in that, The process of dynamically assigning weights is as follows: Obtain health scores from the primary cross slope sensor and redundant cross slope sensors. , The weights for the active dynamic adjustment are calculated based on the health scores of the main cross slope sensor and redundant sensors. and redundant dynamic adjustment weight Among them, the higher the health score, the greater the weight, and the weight is dynamically adjusted. The calculation formula is as follows: ； Redundancy dynamic adjustment weight The calculation formula is as follows: ； Preset cross slope sensor switching threshold When the main cross slope sensor malfunctions, when If the main sensor fault flag is True, then a weight switch will be forced. , The cross slope value is provided separately by a redundant cross slope sensor.

7. The asphalt paver control system based on multi-sensor array redundancy according to claim 1, characterized in that, The steps for calculating attitude data are as follows: By performing low-pass filtering on the raw attitude angles output independently by each sensor for data preprocessing, single-source attitude data is output. Then, the smoothed attitude data from the IMU measurement module, dual-axis tilt sensor, and attitude sensor are weighted and error calculated using the Kalman filter algorithm to output high-precision, high-fault-tolerant fused attitude angles. The single-source attitude data processing steps are as follows: Obtain the roll angle output by the attitude sensor Pitch angle and heading angle Preprocessing eliminates vibration noise and corrects deviations caused by sensor installation tilt. Then, low-pass vibration filtering, installation tilt correction, and heading angle unwrapping calculations are performed. The calculation formula for low-pass vibration filtering is as follows: ； In the formula: Represents the filter coefficients, with weight coefficients 0 < <1 determines the degree to which new data contributes to the output; This represents the smoothed attitude angle, the filtered output at the current moment, which serves as the input for the next moment. This represents the original attitude angle measured by the sensor at step t at the current time, and the attitude angle output by the IMU measurement module. Indicates the previous moment The filtering results of the step; This indicates the nth sampling time. Representing the previous sampling time, forming a time series. =0,1,2,...,n, a preset attitude threshold is used to remove outliers. Then remove ; in, This represents the elevation jump threshold. =0.02m; The formula for calculating the installation tilt correction is as follows: ； In the formula: express The actual vehicle body posture angle after constant correction for installation deviations; express The original attitude angle data of the vehicle body after constant vibration and noise filtering; This indicates the fixed system deviation caused by the incorrect installation position of the vehicle body attitude sensor; The formula for calculating the unwrapping of the heading angle is as follows: ； In the formula: Indicates the first The paver's heading correction value at any given moment; Indicates the first The filtered value of the paver's heading at any given moment; This represents the modulo operation, which calculates the remainder after dividing θYaw(t) by 360°, and normalizes the angle value to the interval [0°, 360°].

8. The asphalt paver control system based on multi-sensor array redundancy according to claim 7, characterized in that, The calculation process for high-precision, high-fault-tolerant fused attitude angles is as follows: Based on the fuselage attitude provided by the IMU module, the dual-axis tilt sensor and the vehicle body attitude sensor provide independent attitude data, which are then fused using Kalman filtering. The calculation formula is as follows: ； In the formula, This represents the fused vehicle attitude state vector, including the roll angle. Pitch angle and heading angle The mathematical dimension is n×1, where n represents the dimension of the state vector and the number of attitude parameters the system needs to estimate. A typical value for n is given if roll angle is included. and pitch angle When n=2, then n=2; if the heading angle is included. Therefore, n=3; The vector represents the vehicle body attitude state of the IMU measurement module, including noise and drift, with a mathematical dimension of n×1; This represents the vehicle attitude observations measured by the tilt sensor, with a mathematical dimension of m×1, where m represents the dimension of the observation vector and represents the number of attitude parameters measured by the tilt sensor. A dual-axis tilt sensor measures the roll angle. and pitch angle When m=2, the dual-axis tilt sensor measures the roll angle. Then m=1 The Kalman gain matrix determines the weights of the IMU and tilt sensor data in the fusion process, and its mathematical dimension is n×m. The observation matrix represents the state of the IMU measurement module and maps it to the observation space of the tilt sensor. Its mathematical dimension is m×n. The formula for calculating attitude redundancy fusion is as follows: ； In the formula: , Represents the fusion coefficient. It is dynamically adjusted based on the noise from the attitude sensor. Indicates the attitude redundancy angle; Indicates the angle of the IMU measurement module; Indicates the angle of the dual-axis tilt sensor; Indicates the angle of the vehicle body attitude sensor; The fused roll angle, fused pitch angle, and fused yaw angle are calculated using the attitude redundancy fusion calculation formula. Substituting the fused roll angle and fused pitch angle into the following formula, the actual elevation of the screed is calculated. The calculation formula is as follows: ； In the formula, This indicates the horizontal distance between the balance beam and the ironing board; This indicates the vertical height difference between the balance beam and the screed board; Indicates the longitudinal slope angle, The cross slope angles represent the front-to-back tilt angle and the left-to-right tilt angle of the paver. The cross slope of the screed is corrected, and the calculation formula is as follows: ； In the formula: Indicates the correction of the cross slope of the ironing board; express The baseline cross slope angle after constant sensor filtering and noise reduction data preprocessing; This indicates the design thickness of the paving layer or the effective working height of the screed. Indicates the effective horizontal length of one side of the ironing board; Through sensor extrinsic parameter calibration, the relative positions of GNSS and laser balance beam are measured using a total station, a transformation matrix is ​​generated, and the GNSS elevation is converted to the screed reference point. The calculation formula is as follows: ； In the formula: express The elevation of the screed is constantly derived and calculated from GNSS data; express Elevation data obtained from GNSS sensors at any given time; This indicates the horizontal distance between the GNSS installation location and the ironing plate; express The sine value of the pitch angle of the paver at all times.

9. The asphalt paver control system based on multi-sensor array redundancy according to claim 7, characterized in that, The data preprocessing process for the dual-axis tilt sensor is as follows: Obtain the lateral tilt angle from the dual-axis tilt sensor longitudinal tilt angle Preprocessing eliminates temperature drift errors and corrects deviations caused by the non-parallelism between the mounting shaft of the dual-axis tilt sensor and the vehicle body shaft. Specifically, tilt temperature compensation is performed first, calculated using the following formula: ； In the formula, This represents the attitude angle after temperature compensation, or the attitude sensor output after eliminating the effects of temperature. This represents the original attitude angle, the attitude angle directly output by the attitude sensor; This represents the temperature coefficient, indicating the sensor's output sensitivity as temperature changes. =0.005° / ℃; The current temperature is indicated by either measurement from the built-in thermometer of the attitude sensor or by an external temperature sensor. This indicates the calibration temperature at 25℃; The formula for calculating axial deviation correction is as follows: ； in, This represents the attitude angle after axial correction, and the attitude sensor output after eliminating installation axis deviation. This represents the attitude angle after temperature compensation, and is output using the same temperature compensation formula. This represents the vertical axis attitude angle, output by the attitude sensor in the direction perpendicular to the current axis. This indicates the angle between the attitude sensor axis and the vehicle body axis, determined through calibration, with a typical value. ; Then, median filtering is performed to eliminate transient noise. The calculation formula is as follows: ； In the formula, This represents the attitude angle after median filtering, and the sensor output after eliminating instantaneous noise; This represents the attitude angle after axial correction, and is the same as the axial correction formula.

10. The asphalt paver control system based on multi-sensor array redundancy according to claim 1, characterized in that, The specific steps for spatiotemporal alignment and dynamic parameter calculation of data from various sensors using a variable averaging system are as follows: In embedded systems, hardware timers are used to accurately timestamp the data from each sensor. By resampling, the high-frequency data from the IMU measurement module is downsampled to be synchronized with GNSS. The following formula is used to calculate linear interpolation. ； In the formula, Indicates the GNSS sampling time; This indicates the most recent IMU sampling time; through the calibration of the extrinsic parameters of each sensor, the relative position of the GNSS sensor and the balance beam is measured using a total station, a transformation matrix is ​​generated, and the elevation of the GNSS sensor is converted to the reference point of the screed. The calculation formula is as follows: ； In the formula: This indicates the elevation conversion of the GNSS sensor to the ironing plate reference point; express Elevation data obtained from GNSS sensors at any given time; When calculating the elevation of Songpu surface in real time, the pre-processed GNSS elevation should be entered first. Design Elevation And the thickness of the pine paving The theoretical and actual elevations of the loose pavement are calculated using the following formula: ； ； In the formula, Indicates the compaction coefficient; Indicates the design loose layer thickness; express The screed elevation data is preprocessed by multi-sensor fusion at all times; When calculating the cross slope control value, first input the merged cross slope. Design cross slope The cross slope control quantity is calculated using the following formula: ； In the formula, Indicates the width of the ironing board; This indicates the cross slope control gain.

Images