A multi-tonnage unmanned vehicle steer-by-wire redundancy control method and system

Abstract

The application relates to the field of data processing, in particular to a multi-tonnage unmanned vehicle steer-by-wire redundancy control method and system, wherein the method comprises the following steps: acquiring a torque sequence of each preset monitoring point in a steering process, calculating a consistency score of each monitoring point according to the torque sequence; acquiring a vibration sequence of each wheel, calculating a correlation coefficient according to the torque sequence and the vibration sequence to obtain an environmental influence score of each monitoring point; obtaining a confidence of each monitoring point based on the ratio of the consistency score and the environmental influence score, and weighting and fusing the torque sequence according to the confidence to obtain a fused torque sequence; predicting a predicted steering angle sequence of each wheel according to the fused torque sequence and a motion state sequence; calculating a correction coefficient of each wheel according to the deviation between the predicted steering angle sequence and an expected steering angle sequence; combining a set torque effective value, the correction coefficient and a wheel deviation normalization weight to calculate a set torque value of the steering process, and driving an actuator to realize target steering.

Description

Technical Field

[0001] This application relates to the field of autonomous driving. In particular, it relates to a redundancy control method and system for steer-by-wire for multi-tonnage unmanned vehicles. Background Technology

[0002] With the rapid development of smart mines, port logistics, and other fields, multi-tonnage unmanned vehicles, such as 20-300 tonnage mining dump trucks, have become important equipment in unmanned transportation systems. These vehicles typically operate in harsh, unstructured road environments, characterized by high loads, intense vibrations, and long continuous operating times, placing extremely high demands on the reliability, safety, and fault tolerance of the steering system. Traditional mechanical steering systems, due to their fixed transmission ratio, complex mechanical structure, and susceptibility to road impacts directly transmitted to the actuators, are no longer sufficient to meet the needs of unmanned driving. Steer-by-wire systems, by eliminating the mechanical steering column and achieving direct electrical signal control, offer significant advantages such as variable transmission ratio, rapid response, easy integration with chassis domain control, and flexible cab layout. Existing unmanned vehicles often retain a simplified steering column combined with a steer-by-wire system for steering control. Examples include unmanned mining trucks, automated guided vehicles (AGVs), and unmanned heavy trucks from Sany Heavy Industry, Westwell Technology, and EasyDrive, representing the mainstream technology direction for multi-tonnage unmanned vehicle steering systems.

[0003] In existing technologies, redundant designs for steer-by-wire systems often employ simple hardware backup schemes, such as dual motors, dual power supplies, or dual sensors. However, in extreme environments like mining areas with high vibration, heavy impact, and dust corrosion, these systems are prone to issues such as momentary sensor drift, motor overheating, momentary actuator power insufficiency, or common-mode failure. Furthermore, current multi-sensor fusion strategies are relatively simplistic, typically employing simple averaging or fixed-weight fusion methods. These strategies fail to adequately consider the heterogeneous characteristics of torque sensors and their reliability under varying environmental factors, such as uneven vibration transmission. This results in significant deviations between the actual steering torque and the required steering torque, leading to decreased tracking accuracy and technical problems that can cause vehicle instability. Summary of the Invention

[0004] The main objective of this invention is to provide a redundant control method and system for steer-by-wire for multi-tonnage unmanned vehicles. This aims to solve the technical problems in existing redundant control methods for steer-by-wire, such as the inability to identify the reliability of data from multiple heterogeneous torque sensors under load conditions, and the large deviation between the target torque obtained and the steering torque required by traditional multi-sensor fusion strategies, which leads to drastic vehicle instability due to the actual steering amplitude.

[0005] To achieve the above objectives, the present invention provides a redundancy control method and system for steer-by-wire of multi-tonnage unmanned vehicles, the method and system comprising: In the first aspect, a redundant control method for steer-by-wire for multi-tonnage unmanned vehicles includes: In response to the autonomous vehicle entering the steering process, the torque sequence of each preset monitoring point is acquired, and the consistency score of the corresponding preset monitoring point is calculated based on each torque sequence. The vibration sequence of each wheel during the steering process is obtained, and the correlation coefficient is calculated based on the torque sequence and the vibration sequence to obtain the environmental impact score of each preset monitoring point; The confidence weight of each preset monitoring point is obtained based on the consistency score and the environmental impact score, and the torque sequence is weighted and fused based on the confidence weight to obtain the fused torque sequence of the steering process. The steering state sequence of each wheel is generated based on the fused torque sequence, and the environmental state sequence of each wheel is obtained. The environmental state sequence and the steering state sequence are input into a pre-trained steering angle prediction model to obtain the predicted steering angle sequence of each wheel. Obtain the expected steering angle sequence of the steering process, generate a predicted angle deviation sequence based on the predicted steering angle sequence and the expected steering angle sequence, and assign a correction coefficient for each wheel based on the predicted angle deviation sequence; Obtain the effective value of the set torque during the steering process, and obtain the target torque that enables each wheel to complete the steering based on the effective value of the set torque and the correction coefficient for each wheel; The correction weight for each wheel is calculated based on the predicted angle deviation sequence, and the target torque is corrected using the correction weight to obtain the set torque value for the steering process.

[0006] Preferably, in response to the autonomous vehicle entering the steering process, the torque sequence of each preset monitoring point is acquired, and a consistency score is calculated for each preset monitoring point based on the torque sequence, including: When the autonomous vehicle enters the turning process, the torque sequence of each preset monitoring point is acquired through multiple heterogeneous sensors. The absolute difference between the torque values ​​of each preset monitoring point at the corresponding time is calculated. The mean of the absolute difference is calculated to obtain the relative error value of each preset monitoring point at each time. The relative error values ​​are fused to obtain the consistency score of each preset monitoring point.

[0007] Preferably, the step of acquiring the vibration sequence of each wheel during the steering process, and calculating the correlation coefficient based on the torque sequence and the vibration sequence to obtain the environmental impact score for each of the preset monitoring points includes: Vibration sequences of each wheel are collected by an accelerometer. The torque sequence of each preset monitoring point and the vibration sequence of each wheel are processed by first-order difference to obtain torque difference sequences and vibration difference sequences. The Pearson correlation coefficient and distance correlation coefficient between the torque difference sequences and the vibration difference sequences are calculated. The Pearson correlation coefficient and the distance correlation coefficient are fused to obtain the environmental impact score of each preset monitoring point.

[0008] Preferably, obtaining the confidence weight of each preset monitoring point based on the consistency score and the environmental impact score includes: Calculate the ratio between the consistency score and the environmental impact score of each preset monitoring point to obtain the confidence level of each preset monitoring point. Then, sum and normalize the confidence levels to obtain the confidence weight of each preset monitoring point.

[0009] Preferably, the step of weighting and fusing the torque sequence according to the confidence weight to obtain the fused torque sequence for the steering process includes: The sum of the products of the torque value at each preset monitoring point and the confidence weight at each preset monitoring point at each moment is calculated to obtain the fused torque value at each moment. The fused torque sequence of the steering process is constructed based on the fused torque value.

[0010] Preferably, the step of generating a steering state sequence for each wheel based on the fused torque sequence, obtaining an environmental state sequence for each wheel, and inputting the environmental state sequence and the steering state sequence into a pre-trained steering angle prediction model to obtain a predicted steering angle sequence for each wheel includes: The unmanned vehicle's weight and speed sequences are acquired through the onboard control system. The actual steering angle sequences of each wheel during steering are collected using magnetic angle sensors, and the set steering angle sequences of each wheel during steering are also collected through the onboard system. The weight, speed, and vibration sequences of each wheel are then concatenated to obtain the environmental state sequence for each wheel. Finally, the fused torque sequence, the actual steering angle sequences, and the set steering angle sequences of each wheel are concatenated to obtain the steering state sequence for each wheel.

[0011] The environmental state sequence and steering state sequence of each wheel are concatenated and then input into the trained steering angle prediction model, which outputs the predicted steering angle sequence of each wheel.

[0012] Preferably, the step of obtaining the expected steering angle sequence of the steering process, generating a predicted angle deviation sequence based on the predicted steering angle sequence and the expected steering angle sequence, and applying a correction coefficient for each wheel based on the predicted angle deviation sequence includes: The unmanned vehicle control system acquires the expected steering angle sequence within the time interval corresponding to the predicted steering angle sequence. The difference between the expected steering angle sequence and the corresponding steering angle in the predicted steering angle sequence for each wheel is calculated to obtain the predicted angle deviation sequence for each wheel. The differences in the predicted angle deviation sequences are then fused to obtain the correction coefficient for each wheel.

[0013] Preferably, the step of obtaining the effective value of the set torque during the steering process, and obtaining the target torque that enables each wheel to complete the steering based on the effective value of the set torque, the correction coefficient for each wheel, and the wheel position, includes: The unmanned vehicle control system acquires a set torque sequence within the time interval corresponding to the predicted steering angle sequence during the steering process. Based on the set torque sequence, the root mean square effective value is calculated to obtain the set torque effective value. The product of the set torque effective value and the correction coefficient for each wheel is calculated separately. The product and the set torque effective value are then fused to obtain the target torque that enables each wheel to complete steering.

[0014] Preferably, the step of calculating the correction weight for each wheel based on the predicted angle deviation sequence, and using the correction weight to correct the target torque to obtain the set torque value for the steering process, includes: The absolute value of all differences in the predicted angle deviation sequence for each wheel is taken and the average value is calculated to obtain the deviation degree of each wheel. The deviation degree of each wheel is summed and normalized to obtain the correction weight of each wheel. The sum of the products between the target torque of each wheel and the correction weight of each wheel is calculated to obtain the set torque value for the steering process.

[0015] Secondly, a multi-tonnage unmanned vehicle steer-by-wire redundancy control system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the multi-tonnage unmanned vehicle steer-by-wire redundancy control method described in any one of the claims is implemented.

[0016] This application has the following beneficial effects: 1. This application achieves dynamic confidence assessment of multi-sensor signals by calculating the consistency score of four heterogeneous torque sensors and obtaining the environmental impact score based on the mean values ​​of the linear and nonlinear correlations of the vibration differential sequences of the left and right front wheels. This method effectively solves the problem of unreliable fusion data caused by sensor common-mode failure and vibration interference in the prior art. Even under extreme conditions such as complex road conditions with high vibration and heavy load impact, it can still output a stable and high-precision fused torque sequence, providing a reliable data foundation for subsequent prediction and control, and significantly improving the fault detection coverage and overall safety of the system.

[0017] 2. This application introduces the root mean square effective value (RMS) calculation to set the torque benchmark. Combined with differential correlation analysis of multi-wheel vibration sequences and torque sequences, and a relative correction mechanism for predicted angle deviation and historical deviation, it achieves dynamic adaptive adjustment of vibration and load. It comprehensively quantifies the effects of linear and nonlinear vibrations, intelligently allocates correction weights according to the degree of deviation of each wheel, making the final set torque value more consistent with the actual stress state of the vehicle, improving steering accuracy and lateral stability, and extending the service life of the actuator.

[0018] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments of this application in conjunction with the accompanying drawings. Attached Figure Description

[0019] The following sections will describe some specific embodiments of this application in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a schematic flowchart of a redundancy control method for steer-by-wire of a multi-tonnage unmanned vehicle according to an embodiment of this application; Figure 2 This is a schematic diagram of a redundant control system for steer-by-wire of a multi-tonnage unmanned vehicle according to an embodiment of this application; Detailed Implementation It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] The following reference Figures 1 to 2This application describes a redundant control method and system for steer-by-wire for a multi-tonnage unmanned vehicle, according to embodiments of the present application. In the description of this embodiment, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature, that is, include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. When a feature "includes or contains" one or more of the features it encompasses, unless otherwise specifically described, this indicates that other features are not excluded and may be further included.

[0021] In the description of this embodiment, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0023] Please see Figure 1 A schematic flowchart of a redundant control method for steer-by-wire for multi-tonnage unmanned vehicles includes steps S1-S6, as detailed below: S1: In response to the autonomous vehicle entering the steering process, obtain the torque sequence of each preset monitoring point, and calculate the consistency score of the corresponding preset monitoring point based on each torque sequence.

[0024] In redundant control of steer-by-wire for multi-tonnage unmanned vehicles, sensors of the same model are prone to unidirectional deviations due to similar temperature drift, aging characteristics, and electromagnetic interference. This makes consistency verification unable to effectively identify single-path sensor faults, reducing fault detection coverage. Therefore, torque measurement and redundant control using multiple heterogeneous sensors are crucial for the automatic steering of unmanned vehicles on complex road surfaces.

[0025] Specifically, the unmanned vehicle is an unmanned vehicle that retains a simplified steering column and uses a steer-by-wire system for steering control. The torque sensors of different models have four independent torque measurement channels that use different physical measurement principles, in order to avoid common-mode failure.

[0026] Four preset monitoring points are set at equal intervals in a ring on the same horizontal plane as the steering column of the autonomous vehicle. A different type of torque sensor is set at each preset monitoring point. When the autonomous vehicle enters the steering process, the torque data of each preset monitoring point is collected by the torque sensor during the steering process. The collected torque data is Z-score standardized to obtain the torque sequence of each preset monitoring point.

[0027] Specifically, the Z-score standardization is implemented by calculating the standard deviation of the torque data as the denominator, calculating the difference between the value to be standardized and the mean of the torque data as the numerator, and using the ratio of the two as the standardization value to achieve standardization.

[0028] Taking one of the preset monitoring points as an example, let this monitoring point be the target monitoring point. Taking one moment in the torque sequence as an example, let this moment be the target moment. Calculate the absolute difference between the torque sequence of the target monitoring point and the torque sequence of each preset monitoring point at the target moment. Calculate the average of the absolute differences to obtain the relative error value of the target monitoring point at the target moment. Iterate through each moment to obtain the relative error value of the target monitoring point at each moment. Calculate the negative exponent of the sum of the relative error values ​​to obtain the consistency score of the target monitoring point. Since the steering torque sensed by each sensor should ideally be highly consistent, the torque signals of different monitoring points will differ when road vibration, load changes, or local interference occur. By summing the above relative error values, the overall deviation between the target monitoring point and other monitoring points can be quantified: the higher the consistency score, the more consistent the torque signal of the monitoring point is with other channels, and the higher the reliability; the lower the consistency score, the more likely the channel is subject to local interference or anomalies, and its weight should be reduced in the confidence calculation.

[0029] Based on the above steps for obtaining consistency scores, the consistency score for each preset monitoring point is obtained.

[0030] S2: Obtain the vibration sequence of each wheel during the steering process, and calculate the correlation coefficient based on the torque sequence and the vibration sequence to obtain the environmental impact score of each preset monitoring point.

[0031] A three-axis accelerometer is installed on each of the left and right front wheel hubs of the autonomous vehicle. The sensors collect vibration data of each wheel during the steering process, and the collected vibration data is standardized to obtain the vibration sequence of each wheel during the steering process.

[0032] The torque sequence of the target monitoring point is subjected to first-order difference processing to obtain the torque difference sequence of the target monitoring point. The vibration sequence of each wheel is subjected to first-order difference processing to obtain the vibration difference sequence of each wheel.

[0033] Specifically, the Pearson correlation coefficient is implemented as follows: For two sequences of the same length, each sequence contains several sampling points; secondly, the arithmetic mean of the two sequences is calculated respectively; thirdly, for each sampling point, the signal value of that point is subtracted from the arithmetic mean of its corresponding sequence to obtain the respective deviation value; then, the two deviation values ​​are multiplied together and the product of all sampling points is accumulated to obtain the numerator; at the same time, the squares of the deviation values ​​of the two sequences are calculated and accumulated, and the square roots of the two accumulated values ​​are multiplied together to obtain the denominator; finally, the numerator is divided by the denominator to obtain the Pearson correlation coefficient.

[0034] Taking one of the wheels as an example, and assuming that the wheel is the control wheel, calculate the absolute value of the Pearson correlation coefficient between the torque difference sequence of the target monitoring point and the vibration difference sequence of each wheel, and calculate the mean of the absolute value of the Pearson correlation coefficient to obtain the linear correlation of the target monitoring point.

[0035] Specifically, the distance correlation coefficient is implemented as follows: For two sequences of the same length, each sequence contains several sampling points; secondly, the Euclidean distance between all sampling points within each sequence is calculated to form two independent distance matrices; thirdly, the two distance matrices are subjected to bicentering, that is, each element in the matrix is ​​subtracted from the mean of its row and the mean of its column, and then the global mean of the entire matrix is ​​added to obtain two bicentered distance matrices; then, the elements at corresponding positions in the two bicentered distance matrices are multiplied point by point and summed, and the square root is taken to obtain the numerator; at the same time, the sum of the squares of all elements in each of the two bicentered distance matrices is calculated and the square root is taken, and the two are multiplied to obtain the denominator; finally, the numerator is divided by the denominator to obtain the distance correlation coefficient.

[0036] Calculate the absolute value of the distance correlation coefficient between the torque difference sequence of the target monitoring point and the vibration difference sequence of each wheel, and calculate the mean of the absolute value of the distance correlation coefficient to obtain the nonlinear correlation of the target monitoring point.

[0037] The mean values ​​of the linear and nonlinear correlations of the target monitoring points are calculated to obtain the environmental impact score of each point. Since road vibration in the mining area is one of the main environmental factors affecting the reliability of torque sensor signals, vibration is transmitted to different monitoring points through the chassis and steering mechanism, causing local noise or deviation in the torque signal. Relying solely on a single correlation index is insufficient to fully capture the complexity of vibration impact. However, by using the mean values ​​of linear and nonlinear correlations, both the linear synchronization relationship and the nonlinear dependency between vibration and torque signals can be reflected simultaneously, thus more comprehensively and accurately quantifying the degree of vibration interference at the target monitoring point. A lower environmental impact score indicates a greater impact of vibration on the torque signal at that monitoring point, and it should be given a lower weight in confidence calculations to achieve more accurate fault isolation and data fusion. The method of averaging both correlations balances the contributions of the two types of correlations, avoiding the limitations of a single index, and preventing excessive suppression of the environmental impact score (i.e., an overly strong "veto" effect) when a correlation close to zero in the product operation. This ensures that the environmental impact score maintains reasonable sensitivity and continuity, better reflecting the gradual changes in the vibration environment of the mining area, and contributing to the stability and interpretability of subsequent confidence calculations.

[0038] Based on the above steps for obtaining environmental impact scores, the environmental impact score for each preset monitoring point is obtained.

[0039] S3: Obtain the confidence weight of each preset monitoring point based on the consistency score and the environmental impact score, and perform weighted fusion of the torque sequence based on the confidence weight to obtain the fused torque sequence of the steering process.

[0040] The confidence level of a target monitoring point is obtained by calculating the ratio between its consistency score and environmental impact score. The consistency score reflects the overall similarity between the torque signal of the target monitoring point and other monitoring points, serving as a core positive indicator of signal reliability. The environmental impact score, on the other hand, reflects the degree of interference from road vibration on the monitoring point, representing a negative indicator of signal susceptibility to environmental disturbances. Using a ratio directly reflects the relative strength of signal consistency compared to environmental interference: a larger ratio indicates that the signal consistency of the target monitoring point is far stronger than its vibration interference, resulting in higher confidence; a smaller ratio indicates that environmental interference is relatively more significant, requiring a reduction in the weight of that channel. Compared to product operations, this ratio method more clearly quantifies the relative contributions of the two factors, avoiding the difficulty in intuitively interpreting weight allocation when multiple scores work together in a product, and better meeting the needs of this invention for evaluating the dynamic reliability of torque signals in non-stationary road environments. When the environmental impact score equals 0, the confidence level is set to 1 to ensure numerical stability.

[0041] Based on the confidence level acquisition steps described above, the confidence level of each preset monitoring point is obtained. The confidence level of each preset monitoring point is summed and normalized to obtain the confidence weight of each preset monitoring point. The sum of the products between the torque value and the confidence weight of the torque sequence of each preset monitoring point at the target time is calculated to obtain the fused torque value at the target time.

[0042] Specifically, the summation and normalization is implemented by calculating the sum of the confidence levels of each preset monitoring point as the denominator, and using the confidence level of each preset monitoring point as the numerator, and calculating the ratio to complete the summation and normalization.

[0043] Based on the steps described above for obtaining the fusion torque value, the fusion torque value at each moment is obtained, and a fusion torque sequence for the steering process is constructed based on the fusion torque value.

[0044] S4: Generate a steering state sequence for each wheel based on the fused torque sequence, and obtain an environmental state sequence for each wheel. Input the environmental state sequence and the steering state sequence into a pre-trained steering angle prediction model to obtain a predicted steering angle sequence for each wheel.

[0045] The unmanned vehicle's weight and speed sequence are obtained through the unmanned vehicle control system. A magnetic angle sensor is installed on each of the left and right front wheel hubs of the unmanned vehicle. The magnetic angle sensor collects the actual steering angle sequence of each wheel during the steering process, and the unmanned vehicle system collects the set steering angle sequence of each wheel during the steering process.

[0046] Taking one wheel as an example, and assuming that wheel is the target wheel, the environmental state sequence of the target wheel is obtained by splicing the weight, speed, and vibration sequence of the unmanned vehicle. The steering state sequence of the target wheel is obtained by splicing the fused torque sequence, the actual steering angle sequence, and the set steering angle sequence of the target wheel.

[0047] Based on the above steps for obtaining the environmental state sequence and steering state sequence, the environmental state sequence and steering state sequence for each wheel are obtained.

[0048] Specifically, the turning prediction model includes the LSTM prediction model, which sequentially comprises an input layer, a first LSTM layer, a second LSTM layer, a fully connected layer, and an output layer. The input layer receives a temporal input tensor of dimension L×D, where L is the sliding time window length and D is the number of feature dimensions for the environmental state and turning state. The first LSTM layer contains 64 hidden units, used to extract long-short-term temporal dependency features from the input sequence and output a hidden state sequence. The second LSTM layer contains 32 hidden units, receives the output of the first LSTM layer, further abstracts a higher-order temporal representation, and outputs the hidden state vector of the last time step. The fully connected layer maps the hidden state vector output by the second LSTM layer to the predicted values ​​for the next N control cycles, with N neurons. The output layer outputs the predicted turning angle sequence for the next N time steps. The training of the corner prediction model includes obtaining the environmental state sequence and steering state sequence of each historical steering process as samples according to the above steps. The training set and validation set are divided according to the general proportion corresponding to the data scale. The mean squared error is used as the loss function, and the backpropagation algorithm combined with the Adam optimizer is used to iteratively update the network weights to minimize the error between the predicted steering angle sequence output by the model and the actual steering angle sequence. An early stopping mechanism is adopted during the training process. The training is terminated when the validation set loss does not decrease for a preset number of consecutive rounds to prevent overfitting. After the training is completed, the historical samples and network weight parameters are stored in a fixed manner for online access by the vehicle controller to reduce the time consumption of real-time redundant control.

[0049] The environmental state sequence and steering state sequence of the target wheel are concatenated and input into the trained steering angle prediction model, which outputs the predicted steering angle sequence of the target wheel.

[0050] Based on the above method of obtaining the predicted steering angle sequence, the predicted steering angle sequence of each wheel during the steering process is obtained.

[0051] Specifically, during the steering process, the acquisition lengths of the vehicle speed sequence, fused torque sequence, actual steering angle sequence, and set steering angle sequence for each wheel are strictly aligned with the sequence lengths before splicing the input sequences of the steering angle prediction model. When the acquisition lengths of individual sequences do not meet the input format, nearest neighbor interpolation is used for padding.

[0052] S5: Obtain the expected steering angle sequence of the steering process, generate a predicted angle deviation sequence based on the predicted steering angle sequence and the expected steering angle sequence, and assign a correction coefficient to each wheel based on the predicted angle deviation sequence.

[0053] The unmanned vehicle control system acquires the expected steering angle sequence within the time interval corresponding to the predicted steering angle sequence of the target wheel. It calculates the difference between the corresponding steering angles in the expected and predicted steering angle sequences to obtain the predicted angle deviation sequence of the target wheel. The average of all differences in the predicted angle deviation sequence is used as the numerator, and the average of the absolute values ​​of all differences is used as the denominator. The ratio between the numerator and denominator is calculated to obtain the correction coefficient for the target wheel. This denominator reflects the angle deviation level of the target wheel and serves as a normalization benchmark for the current prediction deviation. When the numerator is greater than 0, indicating a relatively small overall trend in the predicted steering angle, the correction coefficient is greater than 0, and the system appropriately increases torque output to compensate for understeering. When the numerator is less than 0, indicating a relatively large overall trend in the predicted steering angle, the correction coefficient is less than 0, and the system appropriately reduces torque output to avoid oversteering. This relative correction method can adaptively adjust the current control force based on historical performance, improving steering smoothness and accuracy. When the denominator is 0, it means that there is no turning angle deviation within the prediction time interval and the default correction coefficient is 0. When the numerator is 0, it means that there is a natural time-series self-compensation phenomenon of turning angle deviation within the prediction time interval and the default correction coefficient is 0, so no over-correction is performed.

[0054] Based on the above steps for obtaining the correction coefficients, the correction coefficients for each wheel during the steering process are obtained.

[0055] S6: Obtain the effective value of the set torque for the steering process; obtain the target torque that enables each wheel to complete the steering based on the effective value of the set torque and the correction coefficient of each wheel; calculate the correction weight of each wheel based on the predicted angle deviation sequence, and use the correction weight to correct the target torque to obtain the set torque value for the steering process.

[0056] Specifically, the effective root mean square (RMS) value is obtained by squaring the value of each sampling point in the sequence to obtain a set of squared values; then, the arithmetic mean of the set of squared values ​​is calculated, that is, the sum of all squared values ​​is calculated and then divided by the total number of sampling points to obtain the mean square value; finally, the square root of the above mean square value is performed, and the result is the effective root mean square value of the sequence.

[0057] The system acquires the set torque sequence within the time interval corresponding to the predicted steering angle sequence during the steering process through the unmanned vehicle's onboard control system. The root mean square (RMS) effective value of the set torque is then calculated based on this set torque sequence. While the set torque sequence may fluctuate within the predicted time interval, the RMS value comprehensively reflects the overall energy level of the torque during that time interval, avoiding the influence of noise or local disturbances on a single instantaneous value. The RMS value of the set torque calculated using the RMS value better represents the steady-state torque requirement during the steering process, providing a stable and reliable benchmark value for subsequent adjustments.

[0058] Specifically, since the steering process is a gradual process to smoothly achieve the steering target rather than an instantaneous one, the set torque sequence is a torque sequence generated by the unmanned vehicle control system based on a certain smoothing step size for the torque required for the steering angle in the steering command. The smoothing step size is determined based on the feedback delay between the torque and the wheel hub angle.

[0059] The absolute values ​​of all differences in the predicted angle deviation sequence of the target wheel are taken and averaged to obtain the deviation degree of the target wheel. Similarly, the deviation degree of each wheel is obtained. The deviation degree of each wheel is summed and normalized to obtain the correction weight of each wheel. Since the left and right wheels of a multi-tonnage unmanned vehicle may have different angle deviation degrees during turning due to differences in road adhesion coefficients, uneven load distribution, or vibration transmission, the correction weight obtained by summing and normalizing the deviation degrees of each wheel enables dynamic allocation of the target torque to each wheel according to the deviation ratio, making the final set torque value more consistent with the actual stress state of the vehicle. This weight allocation method embodies the adaptive logic of "the greater the deviation, the greater the correction contribution, avoiding instability," effectively improving the coordination and stability of the turning process and avoiding lateral instability caused by over-compensation or under-compensation of one side of the wheel. It is particularly suitable for complex working conditions on unstructured roads in mining areas.

[0060] Specifically, the summation and normalization is achieved by calculating the sum of the deviations of each wheel as the denominator, and using the deviation of each wheel as the numerator, and calculating the ratios to complete the summation and normalization.

[0061] Calculate the product of the set torque effective value and the correction coefficient of the target wheel, and calculate the sum of the product and the set torque effective value to obtain the target torque that the steering column needs to apply when the target wheel completes the steering.

[0062] Based on the above steps for obtaining the target torque, the target torque of each wheel is obtained, and the sum of the products between the target torque of each wheel and the correction weight of each wheel is calculated to obtain the set torque value for the steering process.

[0063] The set torque value for the steering process is input into the unmanned vehicle control system. The target torque is evenly distributed to the two windings of the dual-winding motor or the dual steering motor. The torque command of each channel is converted into the q-axis current reference of the permanent magnet synchronous motor. After current PI regulation, coordinate transformation and space vector pulse width modulation, the inverter is driven so that the steering motor outputs the corresponding electromagnetic torque. After the reducer and gear rack mechanism, it is converted into rack thrust and then drives the front wheel to achieve the target turning angle.

[0064] The flowchart provided in this embodiment is not intended to indicate that the operations of the method will be performed in any particular order, or that all operations of the method are included in every case. Furthermore, the method may include additional operations. Within the scope of the technical concept provided by the method in this embodiment, additional variations can be made to the above method.

[0065] It should be understood that in some embodiments, the components may be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented using software or firmware stored in memory and executed by a suitable instruction execution system.

[0066] This application also provides a redundant control system for steer-by-wire for multi-tonnage unmanned vehicles. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the image processing-based bolt coating quality inspection method according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface, the setup and functions of which are known in the art and will not be described further here.

[0067] The computer program instructions used to perform the operations of this application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages. The computer program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can connect to the user's computer via any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or it can connect to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, in order to perform aspects of this application, electronic circuits, including, for example, programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), can execute computer-readable program instructions to personalize the electronic circuits by utilizing state information of computer-readable program instructions.

[0068] Therefore, those skilled in the art should recognize that although many exemplary embodiments of this application have been shown and described in detail herein, many other variations or modifications conforming to the principles of this application can be directly determined or derived from the disclosure of this application without departing from the spirit and scope of this application. Thus, the scope of this application should be understood and construed as covering all such other variations or modifications.

Claims

1. A redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle, characterized in that, include: In response to the autonomous vehicle entering the steering process, the torque sequence of each preset monitoring point is acquired, and the consistency score of the corresponding preset monitoring point is calculated based on each torque sequence. The vibration sequence of each wheel during the steering process is obtained, and the correlation coefficient is calculated based on the torque sequence and the vibration sequence to obtain the environmental impact score of each preset monitoring point; The confidence weight of each preset monitoring point is obtained based on the consistency score and the environmental impact score, and the torque sequence is weighted and fused based on the confidence weight to obtain the fused torque sequence of the steering process. The steering state sequence of each wheel is generated based on the fused torque sequence, and the environmental state sequence of each wheel is obtained. The environmental state sequence and the steering state sequence are input into a pre-trained steering angle prediction model to obtain the predicted steering angle sequence of each wheel. Obtain the expected steering angle sequence of the steering process, generate a predicted angle deviation sequence based on the predicted steering angle sequence and the expected steering angle sequence, and assign a correction coefficient for each wheel based on the predicted angle deviation sequence; Obtain the effective value of the set torque during the steering process, and obtain the target torque that enables each wheel to complete the steering based on the effective value of the set torque and the correction coefficient for each wheel; The correction weight for each wheel is calculated based on the predicted angle deviation sequence, and the target torque is corrected using the correction weight to obtain the set torque value for the steering process. This includes: taking the absolute value of all differences in the predicted angle deviation sequence of each wheel and calculating the average value to obtain the deviation degree of each wheel. The deviation of each wheel is summed and normalized to obtain the correction weight of each wheel. The sum of the products between the target torque of each wheel and the correction weight of each wheel is calculated to obtain the set torque value for the steering process.

2. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, In response to the autonomous vehicle entering the steering process, the torque sequence of each preset monitoring point is acquired, and a consistency score is calculated for each preset monitoring point based on the torque sequence, including: When the autonomous vehicle enters the turning process, the torque sequence of each preset monitoring point is acquired through multiple heterogeneous sensors. The absolute difference between the torque values ​​of each preset monitoring point at the corresponding time is calculated. The mean of the absolute difference is calculated to obtain the relative error value of each preset monitoring point at each time. The relative error values ​​are fused to obtain the consistency score of each preset monitoring point.

3. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, The process of acquiring the vibration sequence of each wheel during steering, and calculating the correlation coefficient based on the torque sequence and the vibration sequence to obtain the environmental impact score for each preset monitoring point, includes: Vibration sequences of each wheel are collected by an accelerometer. The torque sequence of each preset monitoring point and the vibration sequence of each wheel are processed by first-order difference to obtain torque difference sequences and vibration difference sequences. The Pearson correlation coefficient and distance correlation coefficient between the torque difference sequences and the vibration difference sequences are calculated. The Pearson correlation coefficient and the distance correlation coefficient are fused to obtain the environmental impact score of each preset monitoring point.

4. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, The step of obtaining the confidence weight of each preset monitoring point based on the consistency score and the environmental impact score includes: Calculate the ratio between the consistency score and the environmental impact score of each preset monitoring point to obtain the confidence level of each preset monitoring point. Then, sum and normalize the confidence levels to obtain the confidence weight of each preset monitoring point.

5. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, The step of weighting and fusing the torque sequence according to the confidence weight to obtain the fused torque sequence for the steering process includes: The sum of the products between the torque value at each moment and the confidence weight of each preset monitoring point in the torque sequence of each preset monitoring point is calculated to obtain the fused torque value at each moment. The fused torque sequence of the steering process is constructed based on the fused torque value.

6. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, The step of generating a steering state sequence for each wheel based on the fused torque sequence, obtaining an environmental state sequence for each wheel, and inputting the environmental state sequence and the steering state sequence into a pre-trained steering angle prediction model to obtain a predicted steering angle sequence for each wheel includes: The unmanned vehicle's weight and speed sequences are acquired through an unmanned vehicle control system. The actual steering angle sequence of each wheel during steering is collected using a magnetic angle sensor, and the set steering angle sequence of each wheel during steering is also collected through the unmanned vehicle system. The unmanned vehicle's weight, speed, and vibration sequences of each wheel are then stitched together to obtain the environmental state sequence of each wheel. Finally, the fused torque sequence, the actual steering angle sequence of each wheel, and the set steering angle sequence are stitched together to obtain the steering state sequence of each wheel. The environmental state sequence and steering state sequence of each wheel are concatenated and then input into the trained steering angle prediction model, which outputs the predicted steering angle sequence of each wheel.

7. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, The step of obtaining the expected steering angle sequence of the steering process, generating a predicted angle deviation sequence based on the predicted steering angle sequence and the expected steering angle sequence, and applying a correction coefficient for each wheel based on the predicted angle deviation sequence includes: The system obtains the expected steering angle sequence within the time interval corresponding to the predicted steering angle sequence through the unmanned vehicle control system. It calculates the difference between the expected steering angle sequence and the corresponding position steering angle in the predicted steering angle sequence of each wheel to obtain the predicted angle deviation sequence of each wheel. The differences in the predicted angle deviation sequence are then fused to obtain the correction coefficient of each wheel.

8. The redundant control method for steer-by-wire of a multi-tonnage unmanned vehicle according to claim 1, characterized in that, The step of obtaining the effective value of the set torque during the steering process, and obtaining the target torque that enables each wheel to complete steering based on the effective value of the set torque, the correction coefficient for each wheel, and the wheel position, includes: The system obtains the set torque sequence within the time interval corresponding to the predicted steering angle sequence during the steering process through the unmanned vehicle control system. Based on the set torque sequence, the root mean square effective value is calculated to obtain the set torque effective value. The product of the set torque effective value and the correction coefficient of each wheel is calculated respectively. The product and the set torque effective value are fused to obtain the target torque that enables each wheel to complete the steering.

9. A redundant control system for steer-by-wire of a multi-tonnage unmanned vehicle, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the steps of a redundancy control method for steer-by-wire for a multi-tonnage unmanned vehicle according to any one of claims 1-8.

Images