Method and device for calibrating lateral intervention torque threshold, and vehicle

Abstract

The application provides a lateral intervention torque threshold calibration method and device and a vehicle, and relates to the technical field of intelligent driving control. When the vehicle is automatically controlled in the lateral direction and no driver lateral intervention is detected, vehicle state information and driving scene information are collected, a reference hand torque corresponding to the vehicle state information is determined based on a reference hand torque model, and the reference hand torque is weighted and calculated according to the driving scene information, so that a dynamic lateral intervention torque threshold is finally obtained. The threshold for judging the driver's lateral intervention can be adjusted in real time according to the vehicle state and the driving scene, thereby solving the problem that the fixed threshold cannot adapt to dynamic changes, and improving the accuracy and smoothness of the human-computer interaction of the lateral control of the driving assistance system.

Description

Technical Field

[0001] This application relates to the field of intelligent driving control technology, and in particular to a method, device and vehicle for calibrating the lateral intervention torque threshold. Background Technology

[0002] In the field of intelligent driving control technology, for driver assistance systems with automatic lateral control capabilities, accurately determining whether the driver intends to intervene laterally by turning the steering wheel is a key requirement for achieving smooth human-machine collaborative driving. Related technologies typically require setting a lateral intervention torque threshold. When the system detects that the torque applied by the driver to the steering wheel exceeds this threshold, it is determined as active driver intervention, and control is allocated or transferred accordingly. The appropriateness of this threshold setting directly affects the interactive experience and functional safety of the system in complex and ever-changing real-world driving scenarios.

[0003] To achieve the aforementioned judgment objectives, a common approach in related technologies is to preset a fixed lateral intervention torque threshold or a fixed threshold characteristic curve during system development or vehicle manufacturing, based on limited calibration scenarios and test data. However, because the lateral intervention torque threshold upon which such solutions rely is static during actual vehicle operation or can only change based on a pre-set curve, its inherent principle makes it difficult to adapt to complex and ever-changing real-world road environments. Specifically, a fixed lateral intervention torque threshold may not accurately reflect the driver's true intention to take over in dynamically changing driving scenarios (such as curves with different curvatures or different deceleration conditions), leading to intervention judgments that are either overly sensitive or overly sluggish.

[0004] The aforementioned technical solutions based on fixed or static calibration curves, because they cannot adaptively adjust to real-time changes in driving scenarios, ultimately lead to insufficient accuracy in lateral intervention judgments. This often manifests as unnecessary and frequent exits or difficulty in responding promptly to driver takeover operations, affecting the practicality of the driver assistance system and user trust. Therefore, how to make the lateral intervention torque threshold adaptable to dynamically changing complex driving scenarios has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application provides a method, apparatus, and vehicle for calibrating the lateral intervention torque threshold.

[0006] Firstly, a calibration method for a lateral intervention torque threshold is provided, applicable to vehicles with automatic lateral control capabilities. The method includes: when controlling the vehicle based on automatic lateral control capabilities and no driver lateral intervention is detected, collecting vehicle state information and driving scenario information; determining a reference hand torque corresponding to the vehicle state information based on a reference hand torque model, wherein the reference hand torque model defines the reference hand torque corresponding to different vehicle states; and performing a weighted calculation on the reference hand torque according to the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0007] In one embodiment, the reference hand torque model is dynamically maintained based on vehicle state information.

[0008] In one embodiment, the vehicle status information includes steering wheel angle, vehicle speed, and steering wheel torque.

[0009] In one embodiment, the reference hand torque model is a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes and steering wheel hand torque as the values. Determining the reference hand torque corresponding to the vehicle state information based on the reference hand torque model includes: determining the steering wheel hand torque that matches the steering wheel angle and vehicle speed in the vehicle state information based on the two-dimensional table, and using the determined steering wheel hand torque as the reference hand torque corresponding to the vehicle state information.

[0010] In one embodiment, the method further includes: acquiring multiple consecutive sets of vehicle state information, each set including at least two sets of vehicle state information; calculating the dispersion of steering wheel angle and vehicle speed in each set of data; when the dispersion of multiple consecutive sets is lower than a preset dispersion calibration value, and the dispersion of the middle set is lower than the dispersion of the first and last sets, updating the reference hand torque model according to the vehicle state information corresponding to the middle set.

[0011] In one embodiment, the reference hand torque is weighted and calculated based on driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle, including: determining a weighting coefficient based on driving scenario information; and weighting and calculating the reference hand torque based on the weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0012] In one embodiment, the driving scenario information includes at least one of vehicle deceleration, environmental perception information, system status information, and driver behavior data.

[0013] In one embodiment, determining the weighting coefficient based on driving scenario information includes: determining the corresponding weighting coefficient according to preset rules based on each piece of information in the driving scenario information.

[0014] In one embodiment, the reference hand torque is weighted according to a weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle, including: weighting the reference hand torque according to the maximum value among the weighting coefficients to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0015] Secondly, a calibration device for a lateral intervention torque threshold is provided, applicable to vehicles with automatic lateral control capabilities. The device includes: a data acquisition module for acquiring vehicle state information and driving scenario information when the vehicle is controlled based on automatic lateral control capabilities and no driver lateral intervention is detected; a logic judgment module for determining a reference hand torque corresponding to the vehicle state information based on a reference hand torque model, the reference hand torque model defining the reference hand torque corresponding to different vehicle states; and a weighted calculation of the reference hand torque based on the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0016] In one embodiment, the device further includes an online learning module for dynamically maintaining the reference hand torque model based on vehicle status information.

[0017] In one embodiment, the vehicle status information includes steering wheel angle, vehicle speed, and steering wheel torque.

[0018] In one embodiment, the reference hand torque model is a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes and steering wheel hand torque as the value. The logic judgment module is specifically used to determine the steering wheel hand torque that matches the steering wheel angle and vehicle speed in the vehicle status information based on the two-dimensional table, and the determined steering wheel hand torque is used as the reference hand torque corresponding to the vehicle status information.

[0019] In one embodiment, the online learning module is specifically used to acquire multiple consecutive sets of vehicle state information, each set including at least two sets of vehicle state information; calculate the dispersion of steering wheel angle and vehicle speed in each set of data; when the dispersion of multiple consecutive sets is lower than a preset dispersion calibration value, and the dispersion of the middle set is lower than the dispersion of the first and last sets, update the reference hand torque model according to the vehicle state information corresponding to the middle set.

[0020] In one embodiment, the logic judgment module is specifically used to determine the weighting coefficient based on driving scenario information; and to perform weighted calculation on the reference hand torque according to the weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0021] In one embodiment, the driving scenario information includes at least one of vehicle deceleration, environmental perception information, system status information, and driver behavior data.

[0022] In one embodiment, the logic judgment module is specifically used to determine the corresponding weighting coefficients based on each piece of information in the driving scenario information according to preset rules.

[0023] In one embodiment, the logic judgment module is specifically used to perform weighted calculation on the reference hand torque based on the maximum value among the weighting coefficients to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0024] Thirdly, a vehicle is provided that has automatic lateral control capability, including the means as described in any one of the second aspects above.

[0025] In this embodiment, when the vehicle is automatically laterally controlled and no lateral intervention by the driver is detected, vehicle status information and driving scenario information can be collected. A reference hand torque corresponding to the vehicle status information is determined based on a reference hand torque model. Then, the reference hand torque is weighted and calculated according to the driving scenario information to obtain a dynamic lateral intervention torque threshold. This allows the threshold for judging lateral intervention by the driver to be adjusted in real time according to the vehicle status and driving scenario, thereby solving the problem that a fixed threshold cannot adapt to dynamic changes and improving the accuracy and smoothness of the human-machine interaction for lateral control of the driver assistance system. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a method for calibrating a lateral intervention torque threshold provided in an embodiment of this application.

[0027] Figure 2 This is a schematic diagram of the structure of a calibration device for a lateral intervention torque threshold provided in an embodiment of this application.

[0028] Figure 3 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application. Detailed Implementation

[0029] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that “ / ” means “or,” for example, A / B can mean A or B; “and / or” in the text is merely a description of the relationship between related objects, indicating that three relationships can exist, for example, A and / or B can mean: A alone, A and B simultaneously, and B alone.

[0030] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0031] The terms "first" and "second" in the following embodiments of this application are for descriptive purposes only and should not be construed as implying relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0032] In the field of intelligent driving control technology, for driver assistance systems with automatic lateral control capabilities, accurately determining whether the driver intends to intervene laterally by turning the steering wheel is a key requirement for achieving smooth human-machine collaborative driving. Related technologies typically require setting a lateral intervention torque threshold. When the system detects that the torque applied by the driver to the steering wheel exceeds this threshold, it is determined as active driver intervention, and control is allocated or transferred accordingly. The appropriateness of this threshold setting directly affects the interactive experience and functional safety of the system in complex and ever-changing real-world driving scenarios.

[0033] To achieve the aforementioned judgment objectives, a common approach in related technologies is to preset a fixed lateral intervention torque threshold or a fixed threshold characteristic curve during system development or vehicle manufacturing, based on limited calibration scenarios and test data. However, because the lateral intervention torque threshold upon which such solutions rely is static during actual vehicle operation or can only change based on a pre-set curve, its inherent principle makes it difficult to adapt to complex and ever-changing real-world road environments. Specifically, a fixed lateral intervention torque threshold may not accurately reflect the driver's true intention to take over in dynamically changing driving scenarios (such as curves with different curvatures or different deceleration conditions), leading to intervention judgments that are either overly sensitive or overly sluggish.

[0034] The aforementioned technical solutions based on fixed or static calibration curves, because they cannot adaptively adjust to real-time changes in driving scenarios, ultimately lead to insufficient accuracy in lateral intervention judgments. This often manifests as unnecessary and frequent exits or difficulty in responding promptly to driver takeover operations, affecting the practicality of the driver assistance system and user trust. Therefore, how to make the lateral intervention torque threshold adaptable to dynamically changing complex driving scenarios has become an urgent technical problem to be solved.

[0035] To address this issue, this application provides a calibration method for the lateral intervention torque threshold. By constructing a reference hand torque model and introducing a weighted calculation mechanism based on driving scenario information, the calibration generation process of the lateral intervention torque threshold is improved. This allows the lateral intervention torque threshold to adaptively adjust in real time according to vehicle status and driving scenario without sacrificing the simplicity of the system's judgment logic or its engineering feasibility. Furthermore, this solves the problem of abnormal human-machine interaction caused by a fixed lateral intervention torque threshold failing to adapt to dynamically changing driving environments and vehicle states, thus improving the accuracy and smoothness of lateral control interaction in driver assistance systems.

[0036] The calibration method for the lateral intervention torque threshold provided in this application can be applied to vehicles with automatic lateral control capabilities. These vehicles are typically equipped with driver assistance systems such as Lane Centering Control (LCC) and Lane Keeping Assist (LKA), which can automatically apply lateral control to the vehicle under specific conditions. This method aims to dynamically determine the torque threshold value (i.e., the lateral intervention torque threshold) used to determine whether the driver intends to take over lateral control by turning the steering wheel.

[0037] The calibration method for this lateral intervention torque threshold can be performed by one or more controllers or processors within the vehicle. It should be noted that the "automatic lateral control capability" mentioned in this application refers broadly to the ability of a vehicle system to automatically adjust the steering angle to maintain the vehicle within its lane or along a predetermined path without relying on the driver's application of steering wheel torque. This capability can be achieved through an electric power steering (EPS) system, a steer-by-wire system, or other actuators capable of performing automatic steering.

[0038] The calibration method for the lateral intervention torque threshold provided in this application will be described in detail below with reference to the accompanying drawings.

[0039] Reference Figure 1 This is a flowchart illustrating a method for calibrating a lateral intervention torque threshold provided in an embodiment of this application. Figure 1 As shown, the calibration method for the lateral intervention torque threshold can specifically include the following S101-S103.

[0040] S101. When controlling the vehicle based on automatic lateral control capability and no lateral intervention by the driver is detected, collect vehicle status information and driving scenario information.

[0041] While the driver assistance system is activated and performing lateral control tasks normally, the system can continuously monitor whether the driver applies torque to the steering wheel in an attempt to take over control. When no such driver lateral intervention is detected, the system enters a data acquisition window. At this time, the vehicle's lateral movement is entirely dominated by the system, and the acquired data best reflects the steady-state characteristics of the system's automatic control of the vehicle in the current state.

[0042] As a specific implementation method, various information can be synchronously acquired from the vehicle network and various sensors. For example, vehicle status information may include, but is not limited to: steering wheel angle, steering wheel torque, vehicle speed, and vehicle deceleration. Driving scenario information may include, but is not limited to: information representing the vehicle's own dynamics, environmental perception information from sensors such as cameras and radar or high-precision maps, system status information of the driver assistance system itself, and driver behavior data acquired through driver behavior monitoring such as a Driver Monitoring System (DMS). This approach can comprehensively capture multi-dimensional factors affecting driver intervention decisions, encompassing the "human-vehicle-road-system" framework. Of course, vehicle status information and driving scenario information may also include only a portion of the information listed above; the specific configuration can be tailored to actual design requirements and is not limited here.

[0043] More generally, there are various ways to collect driving scenario information. For example, environmental perception information can be obtained through millimeter-wave radar, lidar, or vehicle-to-everything (V2X) communication; driver behavior data can also be inferred through capacitive steering wheel sensors, torque sensors, or visual tracking. These solutions can all acquire multi-dimensional scenario information, and there are no restrictions here. In practical applications, the corresponding information collection method can be configured according to the specific information required.

[0044] In this application, driver lateral intervention refers to the driver's intention to change the lateral motion state of the vehicle by operating the steering wheel. Specifically, it can refer to the driver's action of actively overcoming the steering wheel's counter-torque and turning the steering wheel, or it can encompass the driver's action of gripping the steering wheel tightly and applying a stable counter-torque, intending to intervene in or take over lateral control during system control, etc.

[0045] In some possible implementations, specific means of achieving the judgment that "no lateral intervention by the driver was detected" may include: using a steering wheel torque sensor to monitor the torque value in real time. Of course, visual recognition or other methods can also be used to detect whether the driver has made lateral intervention, and this is not a limitation here.

[0046] S102. Based on the reference hand torque model, determine the reference hand torque corresponding to the vehicle state information. The reference hand torque model defines the reference hand torque corresponding to different vehicle states.

[0047] For example, the reference hand torque model is essentially a mapping library that establishes a correspondence between the vehicle's kinematic or dynamic states (such as steering wheel angle, vehicle speed, etc.) and a baseline hand torque value. This baseline hand torque value, i.e., the reference hand torque (y_r), represents the steady-state hand torque value on the steering wheel when the system maintains automatic lateral control under the current vehicle state.

[0048] As a specific implementation, the reference hand torque model can be a two-dimensional lookup table with steering wheel angle and vehicle speed as the horizontal and vertical axes. The values ​​in the table represent the steady-state steering wheel hand torque applied by the system's lateral control at the corresponding steering wheel angle and vehicle speed. The vehicle can find the best-matching value in this two-dimensional table based on the real-time collected steering wheel angle and vehicle speed, and determine it as the reference hand torque. For example, if the current steering wheel angle is 5 degrees and the vehicle speed is 60 km / h, the system will look up the cell value corresponding to the steering wheel angle of 5 degrees and the vehicle speed of 60 km / h in the table. This approach provides a structured lookup method with minimal computational overhead, ensuring the real-time response of the system.

[0049] More generally, the reference hand torque model can be implemented in various ways. For example, it can include, but is not limited to: a regression function fitted to experimental data, a neural network model trained using machine learning algorithms, or an inference system containing a set of fuzzy logic rules. All these solutions can achieve the function of determining a reference hand torque based on vehicle state information. For instance, a regression function can be obtained using data fitting based on collected vehicle state information (such as steering wheel angle, vehicle speed, etc.) and the corresponding steering wheel hand torque applied by lateral control, serving as the reference hand torque model. The corresponding reference hand torque can then be calculated based on the vehicle state information and this regression function. Alternatively, a neural network model can be trained using collected vehicle state information (such as steering wheel angle, vehicle speed, etc.) as sample data and the corresponding steering wheel hand torque applied by lateral control as the output target. The reference hand torque can then be obtained by inputting vehicle state information into this reference hand torque model.

[0050] Of course, the above is merely an example of a reference hand torque model. In this application, the reference hand torque model can be any mathematical or data structured expression used to characterize the variation of steering wheel hand torque with vehicle state when the system is automatically laterally controlled and the driver does not intervene. It can be the aforementioned two-dimensional table, regression function, neural network model, etc., or it can be a parameterized model that covers higher dimensions, such as a model that also considers vehicle state factors such as vehicle load and tire wear rate. There are no restrictions here.

[0051] Through S102, a reference hand torque that matches the current vehicle state can be obtained, which enables the lateral intervention torque threshold obtained further based on the reference hand torque to adapt to the current vehicle state, thereby solving the local problem in related technologies that fixed thresholds cannot adapt to changes in vehicle state.

[0052] S103. The reference hand torque is weighted and calculated based on the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0053] After obtaining a basic reference hand torque in S102, this step incorporates driving scenario information to perform a weighted correction on the baseline reference hand torque value. The core logic is that the more complex the driving scenario, the higher the threshold for lateral intervention should be to avoid the system mistakenly identifying driver intervention due to road disturbances or brief torque fluctuations; conversely, in simple and stable scenarios, a lower threshold can be maintained to ensure that the driver's takeover intentions can be responded to in a timely manner.

[0054] The driving scenario information may include, but is not limited to, at least one of vehicle deceleration, environmental perception information, system status information, and driver behavior data. Therefore, the reference hand torque can be weighted and calculated based on at least one type of driving scenario information. In practical applications, the required driving scenario information can be configured based on the specific impact of the driving scenario; no restrictions are placed here.

[0055] As a specific implementation method, this weighted calculation can be decomposed into a two-stage process: First, based on the collected driving scenario information, one or more scenario weighting coefficients (a) are determined; then, multiplication or addition operations are performed on the reference hand torque (y_r) according to these scenario weighting coefficients. For example, a basic weighting coefficient (k) can be multiplied by the scenario weighting coefficient (a) to obtain a total adjustment factor, and then the total adjustment factor can be multiplied by the reference hand torque to obtain the final lateral intervention torque threshold (kay_r). The scenario weighting coefficient (a) can be determined according to preset rules based on at least one driving scenario information, such as vehicle deceleration, lane curvature, system target steering wheel angle, system target deceleration, and driver behavior data. Using this method, the system can integrate and quantify complex, multi-dimensional scenario information into a single, calculable adjustment value, which is clear in structure and easy to implement.

[0056] More generally, the weighted calculation method can be diverse. For example, it may include, but is not limited to: summing the scene weighting coefficients corresponding to each driving scenario information and multiplying by the reference hand torque; selecting the weighting coefficient of the most stringent driving scenario information as the sole weighting term; or constructing a more complex model that uses the driving scenario information and the reference hand torque as inputs to directly output the target threshold (i.e., the lateral intervention torque threshold). All these solutions can achieve the function of adjusting the final judgment threshold according to the driving scenario information.

[0057] In this application, the final obtained lateral intervention torque threshold can be used by the vehicle to determine whether the driver is making lateral interventions. For example, this lateral intervention torque threshold can be compared with the actual hand torque (y_d) applied by the driver on the steering wheel at the current moment. When the absolute value of the hand torque applied by the driver is detected to continuously exceed the lateral intervention torque threshold for a preset duration, it is determined that the driver is making effective lateral interventions, thereby adjusting or disabling the lateral control of the driver assistance system. Of course, the above is only an example of judging the driver's intervention intention based on the lateral intervention torque threshold and the hand torque applied by the driver. In practical applications, any relevant strategy can be used to judge the driver's intervention intention based on the lateral intervention torque threshold and the hand torque applied by the driver; no limitation is made here.

[0058] The calibration method for the lateral intervention torque threshold in this application, based on S102, provides a "baseline" that precisely matches the vehicle state through a reference hand torque model, providing an accurate calculation starting point for the scenario-adaptive adjustment in S103. S103 then superimposes a "correction amount" based on the driving scenario onto this baseline, ultimately generating a dynamic lateral intervention torque threshold that adapts to changes in vehicle and driving environment. This ensures that the threshold for judging driver lateral intervention can both sensitively reflect changes in vehicle state and accurately adapt to the complexity of driving scenarios, thereby improving the overall accuracy and smoothness of the human-machine interaction in lateral control of the driver assistance system.

[0059] To further optimize the long-term accuracy and adaptability of the reference hand torque model in the above embodiments, enabling it to automatically adapt to changes in vehicle steering characteristics over time, this application also provides the following preferred solutions.

[0060] In a preferred embodiment, the vehicle status information continuously collected in S101 can not only be used to determine the reference hand torque in real time, but also to dynamically maintain the reference hand torque model itself.

[0061] In this preferred embodiment, the reference hand torque model in the calibration method for the lateral intervention torque threshold is dynamically maintained based on vehicle state information. This allows the reference hand torque model to be an entity that can learn online and evolve on its own, rather than a static model that remains unchanged. The principle behind this is that the friction, damping, and other characteristics of the vehicle steering system change slowly with usage time, wear, temperature, and other factors. By continuously updating the reference hand torque model using a large amount of real data generated during the vehicle's daily operation, the reference hand torque deviation caused by vehicle characteristic drift can be corrected.

[0062] The advantage of this design is that it avoids inaccurate reference hand torque due to changes in vehicle characteristics, ensuring that the reference hand torque defined by the reference hand torque model for different vehicle states is always accurate and up-to-date. Even after the vehicle has traveled hundreds of thousands of kilometers, the system's judgment benchmark remains valid, thus guaranteeing the long-term accuracy of the lateral intervention torque threshold calculation and its adaptability to the vehicle.

[0063] Those skilled in the art will understand that methods for dynamically maintaining the reference hand torque model may include, but are not limited to, using recursive least squares to update model parameters online, or periodically batch-processing collected data to retrain the model. For example, the system can set a mileage or time period to periodically upload data to a cloud server for more complex model training, and then distribute the updated model to the vehicle via Over-The-Air (OTA) technology, or directly perform model training or online model updates locally on the vehicle.

[0064] By employing the aforementioned dynamic maintenance features, this preferred solution can continuously correct and optimize the reference hand torque model, ensuring it remains highly synchronized with the actual physical characteristics of the vehicle. This further resolves the calibration reference drift problem caused by vehicle aging, component replacement, and other factors.

[0065] In a preferred embodiment, the reference hand torque model can be a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes, and steering wheel hand torque as the value. Based on this, in this preferred embodiment, the vehicle state information used to dynamically maintain the reference hand torque model can include steering wheel angle, vehicle speed, and steering wheel hand torque. Steering wheel angle and vehicle speed together define the vehicle state, while steering wheel hand torque is the output value that needs to be recorded and learned in that state.

[0066] For example, in this preferred embodiment, the reference hand torque model can be updated accordingly based on the steering wheel angle, vehicle speed, and steering wheel hand torque, depending on the specific form of the reference hand torque model. For instance, if the reference hand torque model is a regression function, it can be dynamically fitted based on the collected steering wheel angle, vehicle speed, and steering wheel hand torque to update the corresponding regression function. Or, if the reference hand torque model is a neural network model, it can be continuously optimized based on the collected steering wheel angle, vehicle speed, and steering wheel hand torque, and so on.

[0067] To address the challenge of efficiently and conveniently determining the reference hand torque corresponding to the current vehicle state, this application provides the following preferred solution.

[0068] In a preferred embodiment, the reference hand torque model is constructed as a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes, and steering wheel hand torque as the value. Accordingly, when determining the reference hand torque corresponding to the vehicle state information based on the reference hand torque model, specifically, based on the two-dimensional table, the steering wheel hand torque matching the steering wheel angle and vehicle speed in the collected vehicle state information can be determined, and the found steering wheel hand torque can be used as the reference hand torque corresponding to the current vehicle state information.

[0069] This two-dimensional table structure discretizes the continuous vehicle state space into a finite number of grids, providing a deterministic query mechanism with extremely low computational complexity. Its advantages lie in the fact that the lookup process does not involve complex iterative calculations, has a very short turnaround time, and is completely predictable, making it ideal for automotive embedded real-time systems. Furthermore, this two-dimensional table structure is intuitive and easy to understand, facilitating initial calibration and subsequent online updates.

[0070] Those skilled in the art will understand that alternative structures for achieving the same query purpose include: using hash tables to store torque values ​​under specific state combinations, and constructing KD trees to support more efficient nearest neighbor searches. For example, when vehicle state information also includes a third dimension (such as vehicle load), the two-dimensional table can be expanded into a three-dimensional table.

[0071] It should be noted that when a vehicle leaves the factory for the first time, the initial value of this two-dimensional table can be set to a fixed value that has been calibrated based on the basic parameters; there are no restrictions on this.

[0072] By employing the aforementioned two-dimensional table structure and matching search method, this preferred solution ensures that step S102 can be completed in a very short time from both data organization and algorithm implementation perspectives. This provides speed and certainty guarantees for subsequent weighted calculations and final real-time intervention decisions. This further contributes to achieving a highly smooth driving assistance experience.

[0073] Based on the aforementioned implementation methods, when the reference hand torque model is a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes and steering wheel hand torque as the value, in order to solve the problem that when directly using the collected vehicle state data for dynamic maintenance of the reference hand torque model, non-representative steering wheel hand torque data generated during vehicle steering or transient speed changes may be mistakenly used for updating the reference hand torque model, thus causing model distortion, this application provides the following preferred solution.

[0074] In a preferred embodiment, the process of performing dynamic maintenance of the reference hand torque model is designed to include a strict data filtering mechanism to "clean" the data and retain only high-quality data that represents the vehicle in a steady-state, balanced lateral motion state.

[0075] In this preferred embodiment, the process of dynamically maintaining the reference hand torque model based on vehicle state information can be specifically as follows: Obtain multiple consecutive sets of vehicle state information, each set including at least two vehicle state information items (each vehicle state information item includes steering wheel angle, vehicle speed, and corresponding steering wheel hand torque); calculate the dispersion of steering wheel angle and vehicle speed in each set of data; only when the dispersion of multiple consecutive sets is lower than a preset dispersion calibration value, and the dispersion of the middle set is lower than the dispersion of the first and last sets, is the reference hand torque model updated based on the vehicle state information corresponding to the middle set. The middle set can be at least one set of data other than the first and last sets. The dispersion can be represented by any index characterizing data dispersion, such as range, mean deviation, variance, or standard deviation; no restriction is placed here.

[0076] For example, within a sliding time window T, M consecutive sets of N data points on steering wheel angle, vehicle speed, and steering wheel torque can be recorded. For each set of data, the standard deviation of the steering wheel angle and vehicle speed can be calculated. As its dispersion, the formula for standard deviation is: ,in This represents the i-th sample data in the m-th group. This represents the mean of the data set. The judgment logic is: if there are m consecutive groups of standard deviations... All are less than the calibration value And the standard deviation of the second group to the (m-1)th group, which is located in the middle of these groups. Smaller than the first group and last group When the standard deviation is taken, it is assumed that the vehicle is in a highly stable state during the time period corresponding to the second to the (m-1)th data group. At this point, the values ​​of the corresponding grids in the reference hand torque model (i.e., the two-dimensional table) can be updated based on the average values ​​of the steering wheel angle, vehicle speed, and steering wheel torque corresponding to the second to the (m-1)th data groups. For example, the average value can replace the old value, or a weighted average can be taken with the old value of the grid to achieve a smooth update. Of course, the above explanation uses standard deviation as an example of dispersion; in practical applications, any other dispersion index can be used, and there are no restrictions here.

[0077] This data filtering mechanism can find "grooves"-shaped segments with continuous statistical stability and fluctuations that first decrease and then increase, accurately capturing the moment when a vehicle reaches a brief steady-state equilibrium after navigating a transient. Its advantage lies in perfectly avoiding transient processes with drastic changes in steering wheel angle and significant fluctuations in hand torque, such as when the driver releases the steering wheel, the system begins to intervene, or the vehicle has just entered a curve, thus greatly improving the quality of the learning samples.

[0078] By employing the data filtering logic based on statistical dispersion and "groove" morphology recognition, this preferred solution can automatically and accurately extract the most valuable steady-state data segments from massive amounts of driving data. This effectively filters out "noise" during the driving process, resulting in a dynamically maintained reference hand torque model with extremely high accuracy and robustness, thereby indirectly ensuring that intervention decisions based on this model are always made based on reliable information.

[0079] In one possible implementation, as described in the relevant section S103 above, the weighted calculation process for the reference hand torque can be divided into two ordered and independent stages.

[0080] In this preferred embodiment, the weighted calculation of the reference hand torque based on driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle may specifically include: determining a weighting coefficient based on driving scenario information; and performing a weighted calculation of the reference hand torque based on the weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0081] This design processes at least one type of driving scenario information and outputs one or more quantified weighted coefficients. Finally, a weighted calculation of the reference hand torque is performed based on these weighted coefficients. Its advantage lies in the fact that when adding or modifying driving scenario information, developers only need to add the corresponding driving scenario information without affecting the core threshold calculation logic, greatly improving the system's maintainability and scalability.

[0082] In some possible implementations, driving scenario information may specifically include at least one of vehicle deceleration, environmental perception information, system status information, and driver behavior data. This ensures that the driving scenario covers at least one core dimension: "vehicle dynamics - external environment - system intent - driver status".

[0083] Among these metrics, vehicle deceleration reflects the intensity of the vehicle's longitudinal movement; a greater deceleration suggests a potentially more urgent situation. Environmental perception information, such as the curvature of the lane ahead (obtained through the vehicle's cameras or high-precision maps), reflects the objective requirements of road geometry for lateral control. System status information, such as the target steering wheel angle and target deceleration planned and output by the driver assistance system itself, reveals the control actions the system is about to execute, reflecting the system's "intention." Driver behavior data, such as the driver's field of vision and hand movements (e.g., the direction in which the driver actively turns the steering wheel) obtained through a driver monitoring system, is directly related to the driver's attention level and potential takeover intentions.

[0084] This design ensures the comprehensiveness of the weighted calculation of the reference hand torque based on driving scenario information, enabling it to reflect the current driving scenario from different perspectives.

[0085] By utilizing the aforementioned four dimensions of driving scenario information, this optimized solution lays a comprehensive and reliable data foundation for generating a scenario weighting coefficient that comprehensively reflects the real-world conditions of "human-vehicle-road-system." This further enhances the adaptability and accuracy of the dynamically determined lateral intervention torque threshold in various complex scenarios.

[0086] When determining the corresponding weighting coefficients based on driving scenario information, a unified and parallel processing method is needed to transform each piece of information into a specific impact on the threshold.

[0087] In a preferred embodiment, determining the weighting coefficient based on driving scenario information may specifically include: determining a corresponding weighting coefficient based on each type of information in the driving scenario information according to their respective preset rules; and then, calculating the reference hand torque by weighting according to these weighting coefficients to obtain the final lateral intervention torque threshold.

[0088] This involves defining a unique preset rule for each type of driving scenario information (such as vehicle deceleration, lane curvature, etc.) to determine the corresponding weighting coefficient. This parallel processing architecture makes the entire system highly scalable. When engineers need to introduce a new driving scenario factor (e.g., road adhesion coefficient in rainy or foggy weather), they only need to add a new, independent parallel channel, set its own rules, and have it output a new weighting coefficient, without modifying any existing logic. This allows the driving scenario information used for weighted calculations to be continuously expanded like building blocks, greatly reducing the complexity and risk of system integration.

[0089] For example, taking driving scenario information specifically including vehicle deceleration, lane curvature (belonging to environmental perception information), target steering wheel angle and target deceleration (belonging to system status information), and the driver's line of sight and the direction in which the driver actively turns the steering wheel (belonging to driver behavior data) as an example, we can pre-set the following rules for vehicle deceleration: when the absolute value of vehicle deceleration is greater than the calibration value A (which can be configured according to specific circumstances and is not limited here), the weighting coefficient is a1 and a1 > 1 (the specific value of a1 can be pre-configured according to actual conditions, as long as it is a value greater than 1); when the absolute value of vehicle deceleration is less than or equal to the calibration value A, the weighting coefficient is a1 and a1 = 1. Similarly, for lane curvature, we can pre-set the following rules: when the absolute value of lane curvature increases, the weighting coefficient is a2 and a2 > 1 (the specific value of a2 can be pre-configured according to actual conditions, as long as it is a value greater than 1); when the absolute value of lane curvature remains unchanged or decreases, the weighting coefficient is a2 and a2 = 1. For system target steering wheel angle, the following rules can be preset: when the absolute value of the system target steering wheel angle is greater than the current absolute value of the steering wheel angle, the weighting coefficient is a3 and a3 > 1 (the specific value of a3 can be pre-configured according to the actual situation, as long as it is a value greater than 1); when the absolute value of the system target steering wheel angle is less than or equal to the current absolute value of the steering wheel angle, the weighting coefficient is a3 and a3 = 1. Similarly, for system target deceleration, the following rules can be preset: when the absolute value of the system target deceleration is greater than the calibration value B (which can be configured according to the specific situation, and is not limited here), the weighting coefficient is a4 and a4 > 1 (the specific value of a4 can be pre-configured according to the actual situation, as long as it is a value greater than 1); when the absolute value of the system target deceleration is less than or equal to the calibration value B, the weighting coefficient is a4 and a4 = 1. The system can pre-determine the weighting coefficient based on the driver's field of vision and the direction in which the driver actively turns the steering wheel. When the driver's gaze deviates from the direction of the steering wheel and the direction of the driver's active steering wheel turn (indicating a driver intention to take over), a weighting coefficient of a5 and a5=1 is preset. When the driver's gaze deviates from the direction of the steering wheel and the direction of the driver's active steering wheel turn are not in the same direction (driver intention cannot be confirmed), a weighting coefficient of a5 and a5>1 (the specific value of a5 can be pre-configured according to the actual situation, as long as it is a value greater than 1). Therefore, the vehicle can determine the corresponding weighting coefficient based on the corresponding driving scenario information and its preset rules.

[0090] Based on the aforementioned implementation methods, when multiple weighting coefficients are ultimately determined, a reasonable weighting fusion strategy is needed to avoid the final threshold being over-amplified.

[0091] In a preferred embodiment, the reference hand torque is weighted and calculated according to each weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle. Specifically, the reference hand torque is weighted and calculated according to the maximum value among the calculated weighting coefficients to obtain the final lateral intervention torque threshold.

[0092] For example, if five weighting coefficients a1, a2, a3, a4, and a5 are determined based on driving scenario information, then the final scenario weighting coefficient 'a' is determined to be max(a1, a2, a3, a4, a5). Therefore, the reference hand torque can be calculated using the final determined scenario weighting coefficient 'a'.

[0093] For example, the method for weighting the reference hand torque based on the finally determined scene weighting coefficient 'a' can be to directly multiply the scene weighting coefficient 'a' and the reference hand torque 'y_r' to obtain the lateral intervention torque threshold. Optionally, the aforementioned product can also be multiplied by a preset base weighting coefficient 'k' (k is a value greater than 1 configured according to the actual situation) to obtain the lateral intervention torque threshold.

[0094] By adopting the "maximum value" strategy, the system effectively prevents excessively large weighting coefficients caused by multiplication or addition when multiple scenarios occur simultaneously, which could lead to an excessively high lateral intervention torque threshold. An excessively high threshold would make the system slow to respond to the driver's true intention to take over, requiring the driver to exert a great deal of force to disengage the system, which would pose safety risks and result in a poor user experience.

[0095] Those skilled in the art will understand that the method for integrating the weighting coefficients may also include taking the average value of the weighting coefficients, taking the weighted combination value of the weighting coefficients, or selecting the weighting coefficient corresponding to the driving scenario information with the highest priority according to the priority of different driving scenario information, etc., without limitation.

[0096] Based on the above implementation methods provided in this application, the threshold for judging lateral intervention by the driver can be adjusted in real time according to the vehicle status and driving scenario, thereby solving the problem that a fixed threshold cannot adapt to dynamic changes and improving the accuracy and smoothness of the human-machine interaction for lateral control of the driving assistance system.

[0097] Based on the examples given above, those skilled in the art will obviously be able to make various equivalent modifications or changes, which also fall within the scope of the embodiments of this application.

[0098] It is understood that, in order to achieve the above-mentioned functions, the device may include hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in a hardware-driven or software-driven manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this application.

[0099] This application embodiment can divide functional modules according to the above method example, thereby obtaining an apparatus capable of implementing the above method example. For example, Figure 2 A schematic diagram of a calibration device for a lateral intervention torque threshold provided in an embodiment of this application is shown. Figure 2 As shown, the device may include: a data acquisition module 201, used to acquire vehicle status information and driving scenario information when controlling the vehicle based on automatic lateral control capability and no driver lateral intervention is detected; a logic judgment module 202, used to determine the reference hand torque corresponding to the vehicle status information based on the reference hand torque model, the reference hand torque model defining the reference hand torque corresponding to different vehicle states; and to perform weighted calculation of the reference hand torque according to the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0100] In one embodiment, the device further includes an online learning module 203 for dynamically maintaining the reference hand torque model based on vehicle status information.

[0101] In one embodiment, the vehicle status information includes steering wheel angle, vehicle speed, and steering wheel torque.

[0102] In one embodiment, the reference hand torque model is a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes and steering wheel hand torque as the value. The logic judgment module 202 is specifically used to determine the steering wheel hand torque that matches the steering wheel angle and vehicle speed in the vehicle status information based on the two-dimensional table, and the determined steering wheel hand torque is used as the reference hand torque corresponding to the vehicle status information.

[0103] In one embodiment, the online learning module 203 is specifically used to acquire multiple consecutive sets of vehicle state information, each set including at least two sets of vehicle state information; calculate the dispersion of steering wheel angle and vehicle speed in each set of data; when the dispersion of multiple consecutive sets is lower than a preset dispersion calibration value, and the dispersion of the middle set is lower than the dispersion of the first and last sets, update the reference hand torque model according to the vehicle state information corresponding to the middle set.

[0104] In one embodiment, the logic judgment module 202 is specifically used to determine the weighting coefficient based on driving scenario information; and to perform weighted calculation on the reference hand torque according to the weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0105] In one embodiment, the driving scenario information includes at least one of vehicle deceleration, environmental perception information, system status information, and driver behavior data.

[0106] In one embodiment, the logic judgment module 202 is specifically used to determine the corresponding weighting coefficients according to preset rules based on each piece of information in the driving scenario information.

[0107] In one embodiment, the logic judgment module 202 is specifically used to perform weighted calculation on the reference hand torque based on the maximum value among the weighting coefficients to obtain the lateral intervention torque threshold corresponding to the current vehicle.

[0108] This application also provides a vehicle with automatic lateral control capability, which includes the aforementioned lateral intervention torque threshold calibration device, capable of executing the method described in the aforementioned embodiments to dynamically determine a lateral intervention torque threshold that matches the vehicle state and driving scenario.

[0109] As an example, Figure 3 A schematic diagram of the structure of a vehicle provided in an embodiment of this application is shown. Figure 3 As shown, the vehicle may include at least: a bus 301, a processor 302, a memory 303, and a communication interface 304. The processor 302, memory 303, and communication interface 304 communicate via the bus 301. The memory 303 stores computer program code, which includes computer instructions. When these computer instructions are executed by the processor 302, the vehicle performs the methods provided in the foregoing embodiments. It should be understood that this application does not limit the number of processors 302 and memories 303 in the vehicle.

[0110] Bus 301 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 301 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus 301 may be represented by a single line, but this does not mean that there is only one bus or one type of bus. The bus 301 may include a path for transmitting information between various components of an electronic device (e.g., memory 303, processor 302, communication interface 304).

[0111] The processor 302 may include any one or more of the following: central processing unit, microprocessor (MP), digital signal processor (DSP), graphics processor, microcontroller, neural network processing unit, system-on-a-chip, etc.

[0112] Memory 303 may include volatile memory, such as random access memory (RAM). Memory 303 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0113] Communication interface 304 uses a command distribution module, such as, but not limited to, a network interface card or transceiver, to enable communication between the device and other devices or communication networks.

[0114] Furthermore, in the embodiments of this application, the functional units or modules of the apparatus used to implement the aforementioned method example can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially or in other words, the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] Therefore, embodiments of this application also provide a computer program product, which includes computer program instructions that, when executed by a processor, cause the processor to perform the methods provided in the foregoing embodiments.

[0117] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for calibrating the threshold of lateral intervention torque, characterized in that, Applied to vehicles with automatic lateral control capabilities, the method includes: When the vehicle is controlled based on the automatic lateral control capability and no lateral intervention by the driver is detected, the vehicle status information and driving scenario information of the vehicle are collected. The reference hand torque corresponding to the vehicle state information is determined based on the reference hand torque model, which defines the reference hand torque corresponding to different vehicle states. The reference hand torque is weighted and calculated based on the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle.

2. The method according to claim 1, characterized in that, The reference hand torque model is dynamically maintained based on the vehicle state information.

3. The method according to claim 1 or 2, characterized in that, The vehicle status information includes steering wheel angle, vehicle speed, and steering wheel torque.

4. The method according to claim 3, characterized in that, The reference hand torque model is a two-dimensional table with steering wheel angle and vehicle speed as the horizontal and vertical axes, and steering wheel hand torque as the value. Determining the reference hand torque corresponding to the vehicle state information based on the reference hand torque model includes: Based on the two-dimensional table, the steering wheel torque that matches the steering wheel angle and vehicle speed in the vehicle status information is determined, and the determined steering wheel torque is used as the reference torque corresponding to the vehicle status information.

5. The method according to claim 4, characterized in that, The method further includes: Acquire multiple consecutive sets of vehicle status information, each set including at least two sets of vehicle status information; Calculate the dispersion of steering wheel angle and vehicle speed in each set of data; When the dispersion of multiple consecutive groups is lower than the preset dispersion calibration value, and the dispersion of the middle group is lower than the dispersion of the first and last groups, the reference hand torque model is updated according to the vehicle state information corresponding to the middle group.

6. The method according to claim 1 or 2, characterized in that, The step of weighting the reference hand torque based on the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle includes: The weighting coefficients are determined based on the driving scenario information; The reference hand torque is weighted and calculated based on the weighting coefficients to obtain the lateral intervention torque threshold corresponding to the current vehicle.

7. The method according to claim 6, characterized in that, The driving scenario information includes at least one of the following: vehicle deceleration, environmental perception information, system status information, and driver behavior data.

8. The method according to claim 7, characterized in that, The determination of weighting coefficients based on the driving scenario information includes: Based on each piece of information in the driving scenario information, the corresponding weighting coefficients are determined according to preset rules.

9. The method according to claim 8, characterized in that, The step of weighting the reference hand torque according to the weighting coefficient to obtain the lateral intervention torque threshold corresponding to the current vehicle includes: The reference hand torque is weighted and calculated based on the maximum value among the weighting coefficients to obtain the lateral intervention torque threshold corresponding to the current vehicle.

10. A calibration device for a lateral intervention torque threshold, characterized in that, For use in vehicles with automatic lateral control capabilities, the device includes: The data acquisition module is used to collect vehicle status information and driving scenario information of the vehicle when the vehicle is controlled based on the automatic lateral control capability and no lateral intervention by the driver is detected. The logic judgment module is used to determine the reference hand torque corresponding to the vehicle state information based on the reference hand torque model, wherein the reference hand torque model defines the reference hand torque corresponding to different vehicle states; and to perform weighted calculation on the reference hand torque according to the driving scenario information to obtain the lateral intervention torque threshold corresponding to the current vehicle.

11. A vehicle, characterized in that, It has automatic lateral control capability, including a calibration device for the lateral intervention torque threshold as described in claim 10.

Images