Steering signal activation method, controller and vehicle for a vehicle

Abstract

The application provides a steering signal activation method, controller and vehicle of a vehicle. It belongs to the technical field of intelligent driving. The method comprises the following steps: obtaining the behavior fusion features of a driver in real time; inputting the behavior fusion features into a preset rule engine to obtain the first steering intention of the driver and the confidence thereof; if the confidence of the first steering intention is within a first confidence range, inputting the behavior fusion features into a preset neural network model to obtain the second steering intention of the driver and the confidence thereof; determining the final steering intention and the comprehensive confidence based on the confidence of the first steering intention and the confidence of the second steering intention; if the comprehensive confidence is greater than a preset confidence threshold, controlling the steering signal activation of the corresponding direction of the vehicle based on the final steering intention. The above method can not only exert the advantage of fast reasoning speed of the rule engine, but also utilize the neural network to improve the accuracy of intention recognition for unobvious intentions, so as to balance the real-time performance and the accuracy.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to a method for activating a vehicle's steering signal, a controller, and a vehicle. Background Technology

[0002] Vehicle turn signals are crucial safety warning devices during vehicle operation. They are used to alert surrounding vehicles and pedestrians to the vehicle's driving intentions before lane changes, turns, or swerves, effectively preventing side collisions, scrapes, and other traffic accidents. They play an irreplaceable role in improving road safety and regulating driving behavior. With the rapid development of automotive intelligence and driver assistance technologies, how to automatically, accurately, and promptly activate turn signals when the driver fails to operate them in time or forgets to do so has become an important research direction in the field of intelligent driving to improve driving safety and comfort. Furthermore, automotive-grade driver assistance systems place extremely high demands on the real-time performance and reliability of turn intention recognition.

[0003] Most existing automatic turn signal activation technologies rely on neural network models to recognize the driver's steering intention and trigger the turn signal activation. This approach results in significant inference delays in intention recognition, failing to meet the real-time requirements of automotive-grade systems for turn signal activation. Summary of the Invention

[0004] This application provides a method for activating a vehicle's steering signal, a controller, and a vehicle, to solve the problem of poor real-time performance in existing automatic steering signal activation methods.

[0005] In a first aspect, this application provides a method for activating a vehicle's turn signal, comprising: Real-time acquisition of driver behavior fusion features, including driver behavior data and steering wheel angle; The behavior fusion features are input into a preset rule engine to obtain the driver's first steering intention and its confidence level. If the confidence level of the first steering intention is within the first confidence level range, the behavior fusion feature is input into a preset neural network model to obtain the driver's second steering intention and its confidence level. Based on the confidence levels of the first and second steering intentions, the final steering intention and overall confidence level are determined. If the overall confidence level is greater than a preset confidence threshold, then the steering signal for the vehicle in the corresponding direction is activated based on the final steering intention.

[0006] As can be seen from the above embodiments, this embodiment first rapidly outputs the initial steering intention and confidence level through a rule engine. Only when the accuracy of the initial steering intention is low is a neural network model activated to assist in the judgment. This method leverages the fast reasoning speed of the rule engine while utilizing the neural network to improve the accuracy of intention recognition for less obvious intentions, thus balancing real-time performance and accuracy. Simultaneously, by comprehensively determining whether to activate the steering signal based on the confidence level, the false triggering problem of a single judgment method can be avoided, ensuring timely and accurate activation of the steering signal. This effectively alerts surrounding vehicles and pedestrians, reduces the incidence of side collisions, scratches, and other traffic accidents, improves driving safety and comfort, and meets the requirements of automotive-grade intelligent driving.

[0007] In one possible implementation, the behavior fusion features include eye gaze angle, head three-dimensional pose angle, and shoulder rotation angle; The step of inputting the behavior fusion features into a preset rule engine to obtain the driver's first steering intention includes: The determination criteria for the first direction are based on the vehicle driving scenario; the first direction is either left or right; the vehicle driving scenario includes at least one of vehicle speed and road attributes. Determine whether the eye gaze angle, the head three-dimensional posture angle, the shoulder rotation angle, and the steering wheel angle meet the judgment conditions of the first direction; If the conditions are met, then the driver's first steering intention is determined to be the first direction.

[0008] As can be seen from the above embodiments, this embodiment incorporates eye gaze angle, head three-dimensional posture angle, shoulder rotation angle, and steering wheel angle into the behavior fusion features. By comprehensively capturing the driver's steering behavior signals, it avoids the one-sidedness of single feature judgment, making the intent output of the rule engine more reliable and providing a precise foundation for subsequent comprehensive confidence calculation. This further ensures the accuracy of steering signal activation and improves the practicality of the driving assistance system. Simultaneously, by combining the vehicle driving scenario to determine the judgment conditions of the first direction, the intent judgment is made more closely aligned with actual driving conditions, solving the defects of intent judgment being disconnected from the scenario and having a high misjudgment rate.

[0009] In one possible implementation, the condition for determining the first direction based on the vehicle driving scenario includes: Determine the speed range within which the vehicle speed falls; Based on the vehicle speed range, a set of judgment conditions corresponding to the first direction is determined; the set of judgment conditions includes a first feature combination, a judgment threshold corresponding to each behavior fusion feature in the first feature combination, and a duration threshold for each behavior fusion feature in the first feature combination to exceed the corresponding judgment threshold. Among them, the duration threshold corresponding to the behavior fusion feature in the judgment condition set where the vehicle speed range is in the low speed range is greater than the duration threshold corresponding to the same behavior fusion feature in the judgment condition set where the vehicle speed range is not in the low speed range, and the low speed range is the range where the vehicle speed is less than the preset speed.

[0010] As can be seen from the above embodiments, this embodiment uses a time threshold for dynamically matching behavioral fusion features within a vehicle speed range to deeply couple the judgment logic of the rule engine with the characteristics of real driving behavior. At low speeds, the driver's observation and operation rhythm is relatively slow; extending the time threshold can effectively filter interference from brief, unconscious actions and avoid false activation. At higher speeds, the driver's lane-changing decisions are faster; shortening the time threshold can capture instantaneous intentions and prevent missed or delayed judgments. This time-based matching mechanism, tailored to vehicle speed, solves the problems of high false judgment rates in low-speed scenarios and delayed response in medium- and high-speed scenarios with fixed time thresholds, improving the scenario adaptability and judgment accuracy of intent recognition from a time dimension.

[0011] In one possible implementation, the set of judgment conditions includes a first set of conditions and a second set of conditions; The set of judgment conditions for determining the first direction based on the vehicle speed range includes: If the vehicle speed is within the low speed range, then determine whether the road attribute where the vehicle is currently located is an intersection; If the road attribute where the vehicle is currently located is an intersection, then the judgment condition for determining the first direction is a first condition set; the first condition set includes a second feature combination, a judgment threshold corresponding to each behavior fusion feature in the second feature combination, and a duration threshold for each behavior fusion feature in the second feature combination to exceed the corresponding judgment threshold; the first feature combination includes the second feature combination; If the road attribute where the vehicle is currently located is not an intersection, then the judgment condition for determining the first direction is the second condition set; the second condition set includes a third feature combination, a judgment threshold corresponding to each behavior fusion feature in the third feature combination, and a duration threshold for each behavior fusion feature in the third feature combination to exceed the corresponding judgment threshold; the first feature combination includes the third feature combination; In the second feature combination, the judgment threshold corresponding to the behavior fusion feature is greater than the judgment threshold corresponding to the same behavior fusion feature in the third feature combination.

[0012] As can be seen from the above embodiments, this embodiment further subdivides the judgment conditions into two sets: intersection and non-intersection, based on road attributes within the low-speed range. In intersection scenarios, the turning action is clear; by increasing the judgment thresholds for head, eye, and steering wheel angle, the accuracy of recognizing turning intentions is improved. Simultaneously, a high judgment threshold avoids false triggers caused by frequent observation at intersections. In non-intersection scenarios, lane-changing actions are gradual; by lowering the judgment threshold, small-amplitude pre-turning actions can be sensitively captured. This differentiated strategy solves the problem of poor scenario adaptability of fixed-threshold schemes, balancing the accuracy of intersection recognition with the sensitivity of non-intersection recognition.

[0013] In one possible implementation, the first confidence range includes an upper confidence limit and a lower confidence limit; After obtaining the driver's first steering intention and its confidence level, the method further includes: If the confidence level of the first steering intention is greater than the upper limit of the confidence level, then the steering signal of the vehicle in the corresponding direction is activated according to the first steering intention. If the confidence level of the first steering intention is less than the lower confidence level, then the first steering intention is determined to be no steering.

[0014] As can be seen from the above embodiments, this embodiment clearly defines the upper and lower limits of the first confidence range. In high-confidence scenarios, the steering signal output by the preset rule engine is directly triggered to maximize response efficiency. In extremely low-confidence scenarios, no steering is directly determined, thus avoiding the risk of false triggering from the root. The neural network is activated for fine judgment only in the intermediate fuzzy interval, making the allocation of computing power more reasonable. This ensures both the response speed of clear scenarios and avoids invalid calculations, while also improving the recognition accuracy of fuzzy steering scenarios, further optimizing the stability and reliability of the system.

[0015] In one possible implementation, before activating the steering signal for the vehicle's corresponding direction based on the final steering intention if the overall confidence level is greater than a preset confidence threshold, the method further includes: If the overall confidence level is within the first confidence level range for a preset time period before the current moment, then the preset confidence threshold is increased.

[0016] As can be seen from the above embodiments, when the overall confidence level is continuously in the low confidence interval within a preset time, this embodiment actively increases the preset confidence threshold and tightens the activation conditions of the steering signal. This can effectively filter out the continuously ambiguous judgment results caused by the driver's unintentional actions and environmental interference, avoid the system from being in an uncertain state or erroneously triggering the steering signal, improve the stability and anti-interference ability of the system under complex interference scenarios, and make the steering signal activation logic more in line with actual working conditions.

[0017] In one possible implementation, determining the final steering intention and overall confidence based on the confidence levels of the first and second steering intentions includes: Take the first steering intention as the final steering intention, and if the first steering intention and the second steering intention Figure 1 If the confidence levels of the first and second turning intentions are found to be consistent, a weighted sum is taken to determine the overall confidence level. If the first steering intention and the second steering intention are inconsistent, the confidence level of the first steering intention shall be used as the overall confidence level.

[0018] As can be seen from the above embodiments, this embodiment clarifies the core judgment rules for the final turning intention and the overall confidence level, which can solve the problems of ambiguous judgment logic and insufficient security when intentions conflict in existing dual-model schemes, and always uses the first turning intention output by the rule engine as the final turning intention. Figure 1 Zhishi improves the reliability of the first turning intention by using a weighted summation method. When there is an intention conflict, it directly uses the confidence of the rule engine. This not only gives full play to the precision judgment value of the neural network, but also avoids the security risks of the black box model, thus greatly improving the security and reliability of the system.

[0019] In one possible implementation, before the weighted summation of the confidence levels of the first turning intention and the second turning intention to determine the overall confidence level, the method further includes: Based on environmental information, the weighted values ​​of the confidence levels of the first turning intention and the second turning intention are adjusted; the environmental information includes ambient light intensity and windshield wiper status.

[0020] As can be seen from the above embodiments, this embodiment accurately perceives the reliability of visual detection by considering ambient light intensity and wiper status. In scenarios where the accuracy of visual feature detection decreases, such as at night or in rainy weather, it automatically increases the weight of the rule engine's confidence and decreases the weight of the neural network to avoid misjudgments caused by visual interference. In scenarios with good visual conditions, it appropriately increases the weight of the neural network to enhance recognition accuracy. The dynamic weighting mechanism allows the system to maintain a stable recognition effect under all environmental conditions.

[0021] Secondly, this application provides a controller including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the vehicle steering signal activation method as described in the possible implementation of the first aspect above.

[0022] Thirdly, embodiments of this application provide a vehicle that includes the controller described in the second aspect above. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the implementation of the vehicle steering signal activation method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the vehicle steering signal activation device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the controller provided in an embodiment of this application. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0027] With the rapid development of automotive intelligence and driver assistance technology, driver assistance systems are being upgraded to higher levels. Automotive-grade systems have placed extremely high demands on the real-time performance, reliability, and safety of automatic activation of steering signals. How to automatically, accurately, and promptly activate steering signals when the driver does not operate in time or forgets to operate has become an important research direction in the field of intelligent driving.

[0028] In existing automatic turn signal activation technologies, related solutions typically rely on data-driven model inference for recognition, such as neural network models. These mathematical models usually have multi-layered network structures containing millions or even hundreds of millions of parameters. Performing a single forward propagation requires a large number of multiplication and accumulation operations, and their structure is relatively complex, which is particularly time-consuming on resource-constrained automotive-grade embedded chips. Secondly, the turn signal recognition task often needs to process continuous time-series data, such as changes in the driver's head and eye posture over a period of time. This means that the neural network model may need to process structures such as three-dimensional convolutional or recurrent neural networks. Their serial or quasi-serial computation mode further exacerbates the latency, and the inference process of this data model is untraceable, which does not meet automotive-grade functional safety requirements and poses safety hazards.

[0029] Based on this, this application provides a method for activating a vehicle's steering signal. This method constructs a dual-modal steering signal activation architecture that prioritizes a rule engine and supplements it with a neural network, fundamentally solving the shortcomings of existing pure neural network solutions, such as high inference latency and insufficient real-time performance.

[0030] See Figure 1 The document illustrates a flowchart of the vehicle steering signal activation method provided in this embodiment. The execution entity of this method can be an on-board controller, specifically a body domain controller or an intelligent driving domain controller, as detailed below: S101: Real-time acquisition of driver behavior fusion features, the behavior fusion features including driver behavior data and steering wheel angle.

[0031] Specifically, the behavior fusion feature is a comprehensive feature set that integrates driver's human behavior data and vehicle operation status data, used to fully characterize the driver's steering operation intention. In this embodiment, the behavior fusion feature specifically includes two core features: driver behavior data and steering wheel angle. Driver behavior data characterizes the driver's body posture and actions during driving, used to predict the driver's pre-steering operation intention. The steering wheel angle is the angle value of the steering wheel rotation around the vehicle's steering axis, which is native data directly obtainable from the vehicle's CAN bus, used to characterize the steering operation amplitude already performed by the driver.

[0032] In this embodiment, driver behavior features may include the driver's three-dimensional head posture angle, eye gaze angle, and shoulder rotation angle. The vehicle controller collects driver behavior data and steering wheel angle in real time at a preset sampling frequency. After filtering, denoising, and normalizing the collected raw data, the data is spliced ​​together to form behavior fusion features. Behavior fusion features may also include vehicle yaw rate and lane departure distance.

[0033] S102: Input the behavior fusion features into the preset rule engine to obtain the driver's first steering intention and its confidence level.

[0034] In this embodiment, the preset rule engine is a reasoning module pre-installed in the vehicle controller, which stores fixed logical judgment rules and thresholds, and has the characteristics of fast reasoning speed, interpretable logic, and traceable results.

[0035] Specifically, the first steering intention is the result of the driver's steering operation intention output by the preset rule engine after logical reasoning on the behavior fusion features. The result types include three categories: left turn, right turn, and no turn.

[0036] The vehicle controller can perform logical reasoning based on the driver's behavioral fusion characteristics to obtain the driver's first steering intention and confidence level.

[0037] S103: If the confidence level of the first steering intention is within the first confidence level range, then the behavior fusion feature is input into a preset neural network model to obtain the driver's second steering intention and its confidence level.

[0038] In this embodiment, the first confidence range is a pre-defined interval used to define the initial judgment result of the preset rule engine as fuzzy and unreliable. It includes an upper confidence limit and a lower confidence limit. The neural network is only activated for secondary judgment when the confidence level is within this interval. For example, the upper confidence limit of the first confidence range can be 55% to 70%, specifically 60%, and the lower confidence limit can be 45% to 35%, specifically 40%.

[0039] In this embodiment, the controller determines in real time whether the confidence level of the first steering intention is within a first confidence level range. If it is not within this range, the controller directly determines whether to activate the corresponding steering signal based on the first steering intention and the corresponding confidence level. If it is within the first confidence level range, the preprocessed behavior fusion features are input into a preset neural network model. The preset neural network model outputs the second steering intention and its corresponding confidence level through forward propagation inference. Based on preset priority rules and weighted calculation rules, the controller can determine the final steering intention based on the first steering intention, and calculate a comprehensive confidence level by combining the confidence levels of the first and second steering intentions. Then, the comprehensive confidence level is compared with a preset confidence threshold in real time. If the comprehensive confidence level is greater than the preset confidence threshold, a steering signal activation command for the corresponding direction is generated and sent to the vehicle body controller to control the turn signal in the corresponding direction to turn on. If the comprehensive confidence level is not greater than the preset confidence threshold, it means that the driver currently has no steering intention and no operation is performed.

[0040] Specifically, the preset neural network model is a deep learning inference model pre-trained using massive amounts of driver behavior and steering scenario data and installed within the controller. It features multi-feature nonlinear correlation inference and high accuracy in fuzzy scene recognition. The input parameters of the neural network model are driver behavior data and steering wheel angle, and the output parameter is the second steering intention. The second steering intention is the judgment result of the driver's steering operation intention output by the preset neural network model after inferring from the behavior fusion features, including three categories: left turn, right turn, and no steering.

[0041] While inputting fused behavioral features into the neural network model, vehicle driving scenario and environmental information can also be input to assist the preset neural network model in reasoning to obtain the second steering intention.

[0042] As a specific embodiment, the controller can extract a behavior fusion feature sequence according to a first preset time window, and input the behavior fusion feature sequence into a preset rule engine. The preset rule engine performs continuous logical matching on the temporal features within the first preset time window, outputs a first turning intention, and calculates the confidence level of the first turning intention based on the continuity of feature matching and the proportion of matching time within the first preset time window. That is, the higher the proportion of matching time, the greater the confidence level of the first turning intention.

[0043] S104: Determine the final steering intention and the overall confidence level based on the confidence level of the first steering intention and the confidence level of the second steering intention.

[0044] In this embodiment, the specific implementation process for determining the final turning intention and overall confidence level may include: When the first steering intention and the second steering intention are inconsistent—for example, the first steering intention is in the first direction and the second steering intention is in the second direction, and the first and second directions are inconsistent—the controller first calculates the confidence level for the first direction. In this case, the confidence level for the second steering intention being in the first direction is zero, so the overall confidence level is the confidence level of the first steering intention itself. Then, the controller calculates the confidence level for the second direction. In this case, the confidence level for the first steering intention being in the second direction is zero, so the overall confidence level for the second steering intention is the confidence level of the second steering intention itself.

[0045] Then, the controller can select the steering intention with the greater overall confidence level of the first steering intention and the overall confidence level of the first steering intention as the final steering intention.

[0046] The above method can select the recognition results with higher confidence from the preset rule engine and preset neural network model as the final steering intention, thereby improving the accuracy of vehicle steering intention recognition.

[0047] S105: If the overall confidence level is greater than the preset confidence threshold, then the steering signal of the vehicle in the corresponding direction is activated based on the final steering intention.

[0048] In this embodiment, the preset confidence threshold is a pre-defined comprehensive confidence threshold used to determine whether to activate the steering signal. The steering signal activation operation is only performed when the comprehensive confidence level is greater than this threshold. The preset confidence threshold can be 85% to 95%.

[0049] When the controller detects that the overall confidence level is greater than the preset confidence threshold, it controls the activation of the steering signal in the corresponding direction of the vehicle based on the final steering intention. After receiving the steering signal in the corresponding direction, the vehicle's lighting controller turns on the turn signal in that direction.

[0050] As can be seen from the above embodiments, this embodiment first rapidly outputs the initial steering intention and confidence level through a rule engine. Only when the accuracy of the initial steering intention is low is a neural network model activated to assist in the judgment. This method leverages the fast reasoning speed of the rule engine while utilizing the neural network to improve the accuracy of intention recognition for less obvious intentions, thus balancing real-time performance and accuracy. Simultaneously, by comprehensively determining whether to activate the steering signal based on the confidence level, the false triggering problem of a single judgment method can be avoided, ensuring timely and accurate activation of the steering signal. This effectively alerts surrounding vehicles and pedestrians, reduces the incidence of side collisions, scratches, and other traffic accidents, improves driving safety and comfort, and meets the requirements of automotive-grade intelligent driving.

[0051] In one possible implementation, the behavior fusion features include eye gaze angle, head three-dimensional pose angle, and shoulder rotation angle; the specific implementation process of S102 includes: S201: Determine the judgment condition for the first direction based on the vehicle driving scenario; the first direction is either left or right; the vehicle driving scenario includes at least one of vehicle speed and road attributes; S202: Determine whether the eye gaze angle, the head three-dimensional posture angle, the shoulder rotation angle, and the steering wheel angle meet the judgment conditions of the first direction; S203: If satisfied, then determine that the driver's first steering intention is the first direction.

[0052] In this embodiment, the eye gaze angle is the angle between the driver's line of sight and the reference direction directly in front of the vehicle, including the horizontal deflection angle and the vertical deflection angle, which is used to characterize the driver's visual gaze direction and predict the steering intention.

[0053] The three-dimensional head attitude angle is the angle of deflection of the driver's head relative to the reference attitude directly in front of the vehicle in three-dimensional space. It includes yaw angle (horizontal left and right deflection), pitch angle (vertical up and down deflection), and roll angle (twist around the head axis). In this embodiment, the yaw angle is mainly used to characterize the direction of the driver's head turning.

[0054] The shoulder rotation angle is the horizontal torsion angle of the line connecting the driver's left and right shoulders relative to the reference axis directly in front of the vehicle. It is used to characterize the lateral torsion trend of the driver's upper body and verify the driver's true steering intention.

[0055] Vehicle driving scenarios are work condition types categorized based on vehicle driving status and road environment. Specifically, they can be road scenarios and vehicle speed. Road scenarios can include intersections and non-intersections, road curvature, ambient light intensity, weather conditions, etc.

[0056] Specifically, the controller can determine whether the driver's three-dimensional head posture angle, eye gaze angle, shoulder twist angle, and steering wheel angle meet the corresponding first direction judgment conditions. If at least one behavior fusion feature meets the corresponding first direction judgment condition, the first steering intention is determined as the first direction. Furthermore, the confidence level can be determined based on the number and weight of the behavior fusion features that meet the first direction judgment conditions. The steering wheel angle has a higher weight than the driver behavior data.

[0057] Before using a preset rule engine for logical judgment, the controller can also acquire vehicle speed data and road attributes in real time to determine the current vehicle driving scenario; then, based on the current vehicle driving scenario, it retrieves a pre-stored set of judgment conditions for the first direction that matches the current vehicle driving scenario. The set of judgment conditions includes the judgment threshold and duration requirement corresponding to each behavior fusion feature.

[0058] As can be seen from the above embodiments, this embodiment incorporates eye gaze angle, head three-dimensional posture angle, shoulder rotation angle, and steering wheel angle into the behavior fusion features. By comprehensively capturing the driver's steering behavior signals, it avoids the one-sidedness of single feature judgment, making the intent output of the rule engine more reliable and providing a precise foundation for subsequent comprehensive confidence calculation. This further ensures the accuracy of steering signal activation and improves the practicality of the driving assistance system. Simultaneously, by combining the vehicle driving scenario to determine the judgment conditions of the first direction, the intent judgment is made more closely aligned with actual driving conditions, solving the defects of intent judgment being disconnected from the scenario and having a high misjudgment rate.

[0059] In one possible implementation, the specific implementation process of S201 includes: Determine the speed range within which the vehicle speed falls; Based on the vehicle speed range, a set of judgment conditions corresponding to the first direction is determined; the set of judgment conditions includes a first feature combination, a judgment threshold corresponding to each behavior fusion feature in the first feature combination, and a duration threshold for each behavior fusion feature in the first feature combination to exceed the corresponding judgment threshold. Among them, the duration threshold corresponding to the behavior fusion feature in the judgment condition set where the vehicle speed range is in the low speed range is greater than the duration threshold corresponding to the same behavior fusion feature in the judgment condition set where the vehicle speed range is not in the low speed range, and the low speed range is the range where the vehicle speed is less than the preset speed.

[0060] In this embodiment, the full range of vehicle speed can be divided into multiple speed ranges, such as a low-speed range, a medium-high-speed range, and a high-speed range. When the vehicle speed is less than a preset speed, it is determined to be in the low-speed range. When the vehicle speed is greater than or equal to the preset speed but less than a high-speed threshold, it is determined to be in the medium-high-speed range. When the vehicle speed is greater than or equal to the high-speed threshold, it is determined to be in the high-speed range. The high-speed threshold is greater than the preset speed.

[0061] Specifically, the set of judgment conditions for the first direction differs for different vehicle speed ranges. The set of judgment conditions includes a first feature combination, which includes at least one of the following: steering wheel angle, head three-dimensional posture angle, eye gaze angle, and shoulder rotation angle. Furthermore, the judgment thresholds for the same behavior fusion features in the first feature combinations corresponding to different vehicle speed ranges are the same. That is, within different vehicle speed ranges, the judgment threshold corresponding to the steering wheel angle is a direction threshold; the judgment threshold corresponding to the head three-dimensional posture angle is a head threshold; the judgment threshold corresponding to the eye gaze angle can be an eye threshold; and the judgment threshold corresponding to the shoulder rotation angle can be a shoulder threshold.

[0062] Specifically, the behavioral fusion features in the first feature combination corresponding to the low-speed range include steering wheel angle, head three-dimensional posture angle, eye gaze angle, and shoulder rotation angle. The duration threshold for the head three-dimensional posture angle in the first feature combination corresponding to the low-speed range is a first preset duration, the duration threshold for the steering wheel angle is a second preset duration, the duration threshold for the eye gaze angle is a third preset duration, and the duration threshold for the shoulder rotation angle is a fourth preset duration.

[0063] The behavioral fusion features in the first feature combination corresponding to the medium-to-high speed range include steering wheel angle, head three-dimensional posture angle, eye gaze angle, and shoulder rotation angle. The duration threshold for the head posture angle in the first feature combination corresponding to the medium-to-high speed range is the fifth preset duration, the duration threshold for the steering wheel angle is the sixth preset duration, the duration threshold for the eye gaze angle is the seventh preset duration, and the duration threshold for the shoulder rotation angle is the eighth preset duration.

[0064] The behavioral fusion features in the first feature combination corresponding to the high-speed range include steering wheel angle and eye gaze angle. The duration threshold for steering wheel angle in the first feature combination corresponding to the high-speed range is the ninth preset duration, and the duration threshold for eye gaze angle is the tenth preset duration.

[0065] Among them, the first preset duration is longer than the fifth preset duration, the second preset duration is longer than the sixth preset duration, the sixth preset duration is longer than the ninth preset duration, the third preset duration is longer than the seventh preset duration, the seventh preset duration is longer than the tenth preset duration, and the fourth preset duration is longer than the eighth preset duration.

[0066] Therefore, the specific set of judgment conditions corresponding to different vehicle speed ranges can include: Within the low-speed range, the judgment condition for the steering wheel angle is that the steering wheel angle deviates in the first direction and is greater than the direction threshold for a second preset duration; the judgment condition for the head three-dimensional posture angle is that the head deviates in the first direction and the head three-dimensional posture angle is greater than the head threshold for a first preset duration; the judgment condition for the eye gaze angle is that the eye gaze angle deviates in the first direction and is greater than the eye threshold for a third preset duration; the judgment condition for the shoulder rotation angle is that the shoulder twists in the first direction and the shoulder rotation angle is greater than the shoulder threshold for a fourth preset duration.

[0067] Within the medium-to-high speed range, the judgment condition for the steering wheel angle is that the steering wheel angle deviates in the first direction and lasts for a sixth preset duration greater than the direction threshold; the judgment condition for the head three-dimensional posture angle is that the head deviates in the first direction and lasts for a fifth preset duration greater than the head threshold; the judgment condition for the eye gaze angle is that the eye gaze angle deviates in the first direction and lasts for a seventh preset duration greater than the eye threshold; the judgment condition for the shoulder rotation angle is that the shoulder twists in the first direction and lasts for an eighth preset duration greater than the shoulder threshold.

[0068] Within the high-speed range, the judgment condition for the steering wheel angle is that the steering wheel angle deviates in the first direction and lasts for a ninth preset duration greater than the direction threshold; the judgment condition for the eye gaze angle is that the eye gaze angle deviates in the first direction and lasts for a tenth preset duration greater than the eye threshold.

[0069] Specifically, it sets complete multi-feature judgment conditions for low-speed and medium-to-high-speed ranges; for high-speed scenarios, it simplifies the judgment dimensions, retaining only two core features: steering wheel angle and eye gaze angle, adapting to the simple operation characteristics of drivers during high-speed cruising, solving the problems of lag in response in high-speed scenarios and high misjudgment rate in low-speed scenarios in existing solutions, and achieving accurate adaptation across the entire vehicle speed range.

[0070] The high-speed threshold is usually set to 60km / h or 80km / h, which can be adjusted according to the vehicle model and application scenario, and the high-speed threshold is greater than the preset speed.

[0071] The first to tenth preset durations are all pre-set thresholds used to determine the duration of driver behavior, and are used to adapt to the response speed requirements of different vehicle speed scenarios.

[0072] The fifth preset duration can be 0.2s, the seventh preset duration can be 0.2s, and the tenth preset duration can be 0.05~0.1s. The first preset duration corresponding to the steering wheel angle can be 0.05s, and the sixth and ninth preset durations can be 0s.

[0073] In one possible implementation, the set of judgment conditions includes a first set of conditions and a second set of conditions; the specific implementation process for determining the set of judgment conditions for the first direction based on the vehicle speed within the vehicle speed range includes: If the vehicle speed is within the low speed range, then determine whether the road attribute where the vehicle is currently located is an intersection; If the road attribute where the vehicle is currently located is an intersection, then the judgment condition for determining the first direction is a first condition set; the first condition set includes a second feature combination, a judgment threshold corresponding to each behavior fusion feature in the second feature combination, and a duration threshold for each behavior fusion feature in the second feature combination to exceed the corresponding judgment threshold; the first feature combination includes the second feature combination; If the road attribute where the vehicle is currently located is not an intersection, then the judgment condition for determining the first direction is the second condition set; the second condition set includes a third feature combination, a judgment threshold corresponding to each behavior fusion feature in the third feature combination, and a duration threshold for each behavior fusion feature in the third feature combination to exceed the corresponding judgment threshold; the first feature combination includes the third feature combination; In the second feature combination, the judgment threshold corresponding to the behavior fusion feature is greater than the judgment threshold corresponding to the same behavior fusion feature in the third feature combination.

[0074] In this embodiment, the second feature combination includes the three-dimensional head posture angle, the steering wheel angle, and the eye gaze angle. The judgment threshold for the three-dimensional head posture angle in the second feature combination is the first head threshold, the judgment threshold for the steering wheel angle is the first direction threshold, and the judgment threshold for the eye gaze angle is the first eye threshold.

[0075] Accordingly, the first set of conditions includes: the judgment condition for the head three-dimensional posture angle is that the head deflects in the first direction and the head three-dimensional posture angle lasts for a first preset duration greater than the first head threshold; the judgment condition for the steering wheel angle is that the steering wheel angle deflects in the first direction and lasts for a second preset duration greater than the first direction threshold; the judgment condition for the eye gaze angle is that the eye gaze angle deflects in the first direction and lasts for a third preset duration greater than the first eye threshold.

[0076] In this embodiment, the third feature combination includes the three-dimensional head posture angle, steering wheel angle, eye gaze angle, and shoulder rotation angle. The judgment threshold for the three-dimensional head posture angle in the third feature combination is the second head threshold, the judgment threshold for the steering wheel angle is the second direction threshold, the judgment threshold for the eye gaze angle is the second eye threshold, and the judgment threshold for the shoulder rotation angle is the shoulder threshold.

[0077] Accordingly, the second set of conditions includes: The judgment condition for the steering wheel angle is that the steering wheel angle deviates in the first direction and is greater than the second direction threshold for a second preset duration; the judgment condition for the head three-dimensional posture angle is that the head deviates in the first direction and the head three-dimensional posture angle is greater than the second head threshold for a first preset duration; the judgment condition for the eye gaze angle is that the eye gaze angle deviates in the first direction and is greater than the second eye threshold for a third preset duration; the judgment condition for the shoulder rotation angle is that the shoulder twists in the first direction and the shoulder rotation angle is greater than the shoulder threshold for a fourth preset duration.

[0078] Among them, the first direction threshold is greater than the second direction threshold, the first eye threshold is greater than the second eye threshold, and the first head threshold is greater than the second head threshold.

[0079] This embodiment, based on the aforementioned technical solution, further refines the specific implementation logic of setting a set of judgment conditions based on road attribute differences in low-speed driving scenarios. Specifically, low-speed driving scenarios are divided into two categories: intersections and non-intersections. For intersection scenarios, since the turning needs of vehicles are usually turning or making a U-turn, the driver needs to check the information in front and behind the oncoming lane. Therefore, the head three-dimensional posture angle, eye gaze angle, and shoulder rotation angle are all larger, requiring a higher feature threshold. Because the feature threshold is set higher, the detection is more accurate, so no excessive features are needed to complete the judgment, and it can also avoid misjudgments caused by unintentional actions of the driver at intersections. For non-intersection scenarios, since turning is usually used for lane changing, only a small movement is needed to check the rearview mirror area. Therefore, a lower feature threshold is set, and the judgment condition of shoulder rotation angle is added to adapt to the behavior characteristics of smooth turning in low-speed non-intersection scenarios. This solves the problem that fixed thresholds cannot adapt to different low-speed road scenarios and that the misjudgment rate and false negative rate cannot be balanced.

[0080] In this embodiment, the controller can determine whether the road the vehicle is currently on is an intersection or not based on the vehicle navigation system.

[0081] The controller determines the confidence level of each judgment condition based on the set of judgment conditions corresponding to different vehicle driving scenarios. For example, when the behavior fusion feature meets the corresponding judgment condition, the confidence level of the behavior fusion feature is set to 1, otherwise it is set to zero. The confidence levels of all behavior fusion features in the set of judgment conditions corresponding to the vehicle driving scenario are weighted and summed to obtain the confidence level of the driver's first steering intention.

[0082] For example, when the vehicle is at a low speed and located at an intersection, if the steering wheel angle is greater than the first direction threshold and lasts for a second preset duration, the first steering intention can be considered as the first direction, and the basic confidence score is the first value. If the head three-dimensional posture angle meets the first head threshold and lasts for a first preset duration, and the eye gaze angle meets the first eye threshold and lasts for a third preset duration, the additional confidence scores are the second and third values, respectively. The first, second, and third values ​​are weighted and summed to obtain the confidence score of the first steering intention. The weighted value of the first value is greater than the weighted value of the second and third values, and the sum of the weights of the first, second, and third values ​​is 1.

[0083] Furthermore, the weighting value can be determined based on environmental parameters. For example, in situations with weak ambient light or rainy weather, the reliability of driver behavior characteristics determined from images is reduced due to obstructed visibility. Therefore, the weight of the first value can be increased, while the weights of the second and third values ​​can be decreased. If the steering wheel angle is not greater than the first directional threshold, the first steering intention is determined to be no steering. At the same time, the confidence level of this first steering intention is determined using the above method based on other driver behavior characteristics.

[0084] When the vehicle is in a low-speed range and in a non-intersection road scenario, if the steering wheel angle is continuously greater than the second direction threshold for a second preset duration, the first steering intention is determined to be the first direction. At the same time, an additional value of the confidence level of the first steering intention is determined based on the driver's behavior characteristics. If the steering wheel angle does not meet the condition of continuously greater than the second direction threshold for a second preset duration, the first steering intention is determined to be no steering, and an additional value of the confidence level of the first steering intention is determined based on the driver's behavior characteristics.

[0085] Specifically, the preset speed can be 25km / h~35km / h. The first head threshold can be 10°~20°, the first preset duration can be 0.5s, the first direction threshold can be 10°~20°, the first eye threshold can be 10°~20°, the second head threshold can be 5°~10°, the second eye threshold can be 5°~10°, the second direction threshold can be 5°~10°, the shoulder threshold can be 5°~10°, and the second preset duration can be 0.3s.

[0086] In one specific implementation, when the road scenario is an intersection, it can also be determined whether the current situation is a regular low-speed intersection scenario or a low-speed creeping intersection scenario based on the vehicle speed. If the current scenario is a low-speed creeping intersection scenario, the determination conditions include: The system determines whether the steering wheel angle exceeds a third-direction threshold and whether the eye gaze angle exceeds a first eye threshold for an eleventh preset duration. If the steering wheel angle exceeds the third-direction threshold and the eye gaze angle exceeds the first eye threshold for a duration exceeding a seventh preset duration, then the first steering intention is determined to be in the first direction. The third-direction threshold is less than the first-direction threshold, and the eleventh preset duration is less than the third preset duration. The vehicle speed threshold for both low-speed crawling intersection scenarios and regular low-speed intersection scenarios can be 15 km / h.

[0087] In one possible implementation, the steering wheel angle is used as the primary judgment feature, and driver behavior features are used as auxiliary judgment features. When the steering wheel angle meets the judgment condition for the corresponding first direction, the first steering intention is determined to be the first direction; otherwise, it is determined to be no steering. The confidence level of the first steering intention is determined based on the confidence levels of all judgment condition outputs corresponding to the vehicle driving scenario. Alternatively, at least one of the steering wheel angle and driver behavior features can be used as the primary judgment feature. If all primary judgment features meet their respective judgment conditions for the first direction, the first steering intention is determined to be the first direction; otherwise, it is determined to be no steering. The confidence level of the first steering intention is determined based on the driver behavior features and the steering wheel angle.

[0088] As a specific embodiment, the specific implementation process of determining whether the eye gaze angle, the head three-dimensional posture angle, the shoulder rotation angle, and the steering wheel angle satisfy the determination condition of the first direction in S102 includes: Within the medium-to-high speed range, if the steering wheel angle deflects in the first direction and remains for a sixth preset duration greater than the direction threshold; and simultaneously the head deflects in the first direction and the head three-dimensional attitude angle remains for a fifth preset duration greater than the head threshold, then it is determined that the first medium-speed sub-rule is satisfied. If the steering wheel angle deflects in the first direction and continues for a sixth preset duration greater than the direction threshold, and the eye gaze angle deflects in the first direction and continues for a seventh preset duration greater than the eye threshold, then the second medium-speed sub-rule is determined to be satisfied. If the shoulder twists in the first direction and the shoulder rotation angle remains greater than the shoulder threshold for an eighth preset duration; and the eye gaze angle deflects in the first direction and remains greater than the eye threshold for a seventh preset duration, then the third medium-speed sub-rule is determined to be satisfied. If either the first or second medium-speed sub-rule is satisfied, the first turning intention is determined to be the first direction. The third medium-speed sub-rule is used to assist in determining the confidence level. For each medium-speed sub-rule satisfied, a corresponding confidence score is obtained. The confidence scores of each satisfied medium-speed sub-rule are weighted and summed to obtain the confidence level of the first turning intention.

[0089] If the current vehicle is at a low speed and at an intersection, the specific implementation process for determining the first steering intention and confidence level based on the first condition set may include: If the steering wheel angle is deflected in the first direction and is greater than the second direction threshold for a second preset duration; and the head deflects in the first direction and the head three-dimensional attitude angle is greater than the second head threshold for a first preset duration, then it is determined that the first intersection turning sub-rule is satisfied. If the eye gaze angle deviates in the first direction and continues for a third preset duration greater than the second eye threshold, and at the same time the steering wheel angle deviates in the first direction and continues for a second preset duration greater than the second direction threshold, then it is determined that the second intersection sub-rule is satisfied. If either the first intersection turning sub-rule or the second intersection turning sub-rule is satisfied, then the first turning intention is determined to be the first direction. Furthermore, for each intersection turning sub-rule satisfied, a corresponding confidence score is obtained. The confidence scores of each satisfied intersection turning sub-rule are weighted and summed to obtain the confidence score of the first turning intention.

[0090] In one possible implementation, the specific implementation process of S102 further includes: If the steering wheel angle is less than the first direction threshold and greater than the second direction threshold, and the driver's driving characteristics do not meet the judgment conditions of their respective first directions, then the first steering intention is determined to be no steering.

[0091] If the driver's driving characteristics all meet the judgment conditions of their respective first directions, but the steering wheel angle is less than the second direction threshold, then the first steering intention is determined to be no steering.

[0092] If the steering wheel turning rate is less than zero and the eye gaze angle is less than the second eye threshold, then the first steering intention is determined to be no steering.

[0093] The above method discloses the false trigger exclusion rule in the preset rule engine. That is, if any of the above judgment conditions are met, the first steering intention is determined to be no steering, thereby improving the real-time performance of steering intention recognition.

[0094] In one possible implementation, the first confidence level range includes an upper confidence level and a lower confidence level; after S102, the method provided in this embodiment further includes: If the confidence level of the first steering intention is greater than the upper limit of the confidence level, then the steering signal of the vehicle in the corresponding direction is activated according to the first steering intention. If the confidence level of the first steering intention is less than the lower confidence level, then the first steering intention is determined to be no steering.

[0095] In this embodiment, 100 is used as the full-scale value, the upper confidence level can be 80-90, and the lower confidence level can be 40. When the confidence level of the first steering intention is greater than the upper confidence level, it indicates that the preset rule engine can output a steering intention with high confidence, and this steering intention can be directly used to activate the steering signal, thereby improving the timeliness of steering signal recognition. When the first steering intention is less than the lower confidence level, it can be clearly determined that the driver has no steering intention, so no operation is required. When the confidence level of the first steering intention is within the first confidence level range, it indicates that the preset rule engine outputs a relatively vague steering intention. At this time, using the first steering intention may result in false triggering or missed triggering of the steering signal. Therefore, a preset neural network model can be further used for accurate judgment, thereby improving the accuracy of steering signal recognition.

[0096] Specifically, the controller first acquires the upper and lower confidence limits of a first confidence range. Then, at preset intervals, it acquires the driver's historical behavior fusion features for that period to determine the confidence level of the first steering intention after inputting into the preset rule engine, and the driver's manual steering operations within a preset time period following the acquisition of these behavior fusion features. For example, if the driver manually turns off the corresponding turn signal within a preset time period after the controller automatically activates the turn signal, this is recorded as the first event. Or, if the driver manually turns on the corresponding turn signal within a preset time period after the controller does not activate the turn signal, this is recorded as the second event. The controller counts the number of first and second events within the current period. If the difference between the number of first and second events is greater than a first preset difference threshold, the upper confidence limit is increased; if the difference is less than a second preset difference threshold, the upper confidence limit is decreased. The first preset difference threshold is greater than zero, and the second preset difference threshold is less than zero.

[0097] As can be seen from the above embodiments, this embodiment clearly defines the upper and lower limits of the first confidence range. In high-confidence scenarios, the steering signal output by the preset rule engine is directly triggered to maximize response efficiency. In extremely low-confidence scenarios, no steering is directly determined, thus avoiding the risk of false triggering from the root. The neural network is activated for fine judgment only in the intermediate fuzzy interval, making the allocation of computing power more reasonable. This ensures both the response speed of clear scenarios and avoids invalid calculations, while also improving the recognition accuracy of fuzzy steering scenarios, further optimizing the stability and reliability of the system.

[0098] In one possible implementation, prior to S105, the method provided in this embodiment further includes: If the overall confidence level is within the first confidence level range for a preset time period before the current moment, then the preset confidence threshold is increased.

[0099] In this embodiment, before comparing the overall confidence level with the preset confidence threshold, the controller can first retrieve all overall confidence level data within a preset time period before the current moment to form an overall confidence level time series. Then, it can determine whether all overall confidence levels in the overall confidence level time series are within the first confidence level range. If they are all within the first confidence level range, the preset confidence threshold is increased by a preset fixed step size to generate an adjusted preset confidence threshold. Finally, the current overall confidence level is compared with the adjusted preset confidence threshold. Only when the overall confidence level is greater than the adjusted preset confidence threshold is the steering signal activated. If the overall confidence level is subsequently detected to exceed the first confidence level range, the preset confidence threshold is restored to its initial value.

[0100] Specifically, before performing the final confidence level comparison, the controller can extract the comprehensive confidence level time series within a preset time period before the current moment and calculate the mean of all comprehensive confidence levels in the time series. Then, it can determine whether all comprehensive confidence levels in the comprehensive confidence level time series are within the first confidence level range. If they are, the controller determines the increase of the preset confidence threshold based on the relative position of the mean confidence level and the upper and lower limits of the first confidence level range. The closer the mean confidence level is to the upper limit of the confidence level, the greater the increase. Finally, based on the determined increase, the preset confidence threshold is adjusted upward.

[0101] As can be seen from the above embodiments, when the overall confidence level is continuously in the low confidence interval within a preset time, this embodiment actively increases the preset confidence threshold and tightens the activation conditions of the steering signal. This can effectively filter out the continuously ambiguous judgment results caused by the driver's unintentional actions and environmental interference, avoid the system from being in an uncertain state or erroneously triggering the steering signal, improve the stability and anti-interference ability of the system under complex interference scenarios, and make the steering signal activation logic more in line with actual working conditions.

[0102] In one embodiment, before activating the steering signal corresponding to the vehicle's direction based on the final steering intention, the method provided in this embodiment further includes: Upon detecting a manual operation signal from the turn signal stalk, the controller immediately disables the automatic activation command to ensure the driver's absolute control.

[0103] When the controller detects that the steering wheel speed is greater than the preset speed threshold and the driver's gaze angle is inconsistent with the direction of the steering wheel angle, it determines that it is an emergency avoidance and immediately disables the automatic activation command.

[0104] The preset speed threshold can be 100° / s to 120° / s.

[0105] In one embodiment, after the steering signal corresponding to the vehicle's direction based on the final steering intention is activated, the method provided in this embodiment further includes: When the turn signal is activated, the corresponding indicator light on the vehicle's instrument panel flashes at a preset frequency, accompanied by a short beep, to alert the driver to the system operation.

[0106] When the controller detects that the steering wheel has returned to the near-center position and has remained there for eight preset durations, it automatically turns off the turn signal for the corresponding direction and alerts the driver with a beep or text message on the instrument panel.

[0107] The above method can adopt a non-intrusive human-computer interaction approach, achieving the effect of enhancing driver trust without interfering with the driver's normal operation.

[0108] In one possible implementation, the specific implementation process of S104 includes: Take the first steering intention as the final steering intention, and if the first steering intention and the second steering intention Figure 1 If the confidence levels of the first and second turning intentions are found to be consistent, a weighted sum is taken to determine the overall confidence level. If the first steering intention and the second steering intention are inconsistent, the confidence level of the first steering intention shall be used as the overall confidence level.

[0109] In this embodiment, the controller first obtains the first steering intention and its confidence level output by the preset rule engine, and the second steering intention and its confidence level output by the preset neural network model; then, regardless of the result of the second steering intention, the first steering intention is used as the final steering intention; at the same time, the controller compares whether the first steering intention and the second steering intention are completely consistent. If both are left turns, or both are right turns, or both are no turns, then it is determined that the intention is correct. Figure 1 To; if you wish Figure 1 If the intentions are consistent, the pre-stored weighting coefficients are retrieved to perform a weighted summation of the confidence levels of the first and second turning intentions to calculate the overall confidence level. If the intentions are inconsistent, the confidence level of the first turning intention is directly used as the overall confidence level, and the confidence level of the second turning intention is not included in the calculation.

[0110] As can be seen from the above embodiments, this embodiment clarifies the core judgment rules for the final turning intention and the overall confidence level, which can solve the problems of ambiguous judgment logic and insufficient security when intentions conflict in existing dual-model schemes, and always uses the first turning intention output by the rule engine as the final turning intention. Figure 1 Zhishi improves the reliability of the first turning intention by using a weighted summation method. When there is an intention conflict, it directly uses the confidence of the rule engine. This not only gives full play to the precision judgment value of the neural network, but also avoids the security risks of the black box model, thus greatly improving the security and reliability of the system.

[0111] In one possible implementation, before determining the overall confidence level by weighted summation of the confidence levels of the first steering intention and the second steering intention, the method provided in this embodiment further includes: Based on environmental information, the weighted values ​​of the confidence levels of the first turning intention and the second turning intention are adjusted; the environmental information includes ambient light intensity and windshield wiper status.

[0112] In this embodiment, ambient light intensity is the light brightness value of the environment around the vehicle, which can be obtained by the vehicle's light sensor and on-board camera, and is used to characterize the light conditions for visual acquisition.

[0113] The wiper status refers to the working status of the vehicle's windshield wipers, including whether they are on or off, as well as different wiping speeds. This information can be directly obtained through the vehicle's CAN bus and is used to indirectly indicate whether the vehicle is in rainy or low-visibility conditions.

[0114] Specifically, before the controller performs a weighted summation of the confidence levels of the first and second steering intentions, it acquires the vehicle's current ambient light intensity data and windshield wiper status data in real time. Then, based on the ambient light intensity and windshield wiper status, it classifies the environment into three levels: the first level is a good environment, the second level is a normal environment, and the third level is a severe environment. For different environment levels, the controller retrieves the corresponding pre-stored weighted value combination. The sum of the weighted values ​​of the confidence levels of the first and second steering intentions is 1. The higher the environment level, the larger the weighted value of the confidence level of the first steering intention, and correspondingly, the smaller the weighted value of the confidence level of the second steering intention.

[0115] Specifically, under favorable conditions, the weighted confidence level of the first turning intention can be 0.7~0.8, and the weighted confidence level of the second turning intention can be 0.2~0.3; under normal conditions, the weighted confidence level of the first turning intention can be 0.5~0.6, and the weighted confidence level of the second turning intention can be 0.4~0.5; under adverse conditions, the weighted confidence level of the first turning intention can be 0.2~0.4, and the weighted confidence level of the second turning intention can be 0.6~0.8.

[0116] Finally, based on the weighted values ​​corresponding to the current environmental level, the confidence levels of the first turning intention and the second turning intention are weighted and summed to obtain the overall confidence level.

[0117] As can be seen from the above embodiments, this embodiment accurately perceives the reliability of visual detection by considering ambient light intensity and wiper status. In scenarios where the accuracy of visual feature detection decreases, such as at night or in rainy weather, it automatically increases the weight of the rule engine's confidence and decreases the weight of the neural network to avoid misjudgments caused by visual interference. In scenarios with good visual conditions, it appropriately increases the weight of the neural network to enhance recognition accuracy. The dynamic weighting mechanism allows the system to maintain a stable recognition effect under all environmental conditions.

[0118] In one possible implementation, the specific implementation process of S101 includes: The system uses an in-vehicle infrared camera to capture facial images of the driver, and extracts the eye gaze angle and three-dimensional head posture angle based on the facial images.

[0119] The driver's shoulder torsional angle is calculated by collecting triaxial acceleration and angular velocity data from the miniature inertial measurement unit of the seat belt. The driver's behavior data is marked with hardware timestamps at the time of collection, and the original values ​​are completely preserved without filtering or feature compression. The steering wheel angle is collected using a steering wheel angle sensor; the sensor's native data collection timestamp is marked. The controller performs independent preprocessing and validity verification on driver behavior data and sensor rotation angles, without changing the core values ​​of the original data or extracting high-order features. It only performs invalid data filtering, distortion correction, and compliance verification, providing effective input for subsequent spatiotemporal synchronization and data-level fusion.

[0120] After preprocessing the driver behavior data and steering wheel angle, the controller synchronizes the driver behavior data and steering wheel angle in time and space.

[0121] Specifically, using the system clock synchronized by the positioning system as a reference, the original timestamps of all data are aligned to a unified time axis, and the timestamp gaps of data with different sampling frequencies are filled by linear interpolation to ensure that the timestamp deviation of all data within the same time window is small enough. Next, based on the driving physics prior that the driver's head turns slightly earlier than the steering wheel turns, the controller sets a time association window. This window establishes a time correspondence between driver behavior data and steering wheel angles within the same time association window, thus matching driver behavior data and steering wheel angles for the same driving behavior. The length of the time association window can be 50ms to 70ms.

[0122] Then, based on the factory-calibrated transformation matrix, the controller maps the driver's gaze angle, head three-dimensional posture angle, shoulder rotation angle, and steering wheel angle to the vehicle body coordinate system to ensure that the spatial reference of all data is consistent; and outputs a multi-source raw data matrix under the same spatial reference within the same time window.

[0123] Finally, the spatiotemporally synchronized multi-source raw data matrix is ​​directly input into the Kalman filter; and the state equation constraint of the Kalman filter is set with driving behavior data preceding the steering wheel angle, to jointly suppress and smooth image jitter, steering wheel vibration, and seat belt micro inertial measurement unit noise; and outputs multi-dimensional synchronized pure behavior fusion features.

[0124] Furthermore, after obtaining the behavior fusion features, the controller can normalize the different dimensional data in the behavior fusion features into a 0-1 range according to a preset range, eliminating the influence of dimensional differences on rule matching; then, the normalized behavior fusion features are split into corresponding feature dimensions according to the rule base dimension of the rule engine, including the driver behavior data dimension and the steering wheel angle dimension.

[0125] In one possible implementation, the controller can also pre-divide the driver's gaze angle into five core gaze regions: left rearview mirror, right rearview mirror, left lane, right lane, and road surface ahead. After obtaining the eye gaze angle, the controller first determines the area the driver is gazing at based on the eye gaze angle, and then uses the specific area the driver is gazing at as a sub-feature of the behavior fusion feature, participating in the subsequent calculations of the preset rule engine and preset neural network, thereby improving the calculation effect.

[0126] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0127] The following are device embodiments of this application. For details not described in detail, please refer to the corresponding method embodiments described above.

[0128] Figure 2 A schematic diagram of the vehicle steering signal activation device provided in an embodiment of this application is shown. For ease of explanation, only the parts related to the embodiment of this application are shown, and are described in detail below: like Figure 2 As shown, the vehicle's turn signal activation device 100 includes: Feature acquisition module 110 is used to acquire driver behavior fusion features in real time, the behavior fusion features including driver behavior data and steering wheel angle; The rule engine processing module 120 is used to input the behavior fusion features into a preset rule engine to obtain the driver's first steering intention and its confidence level; The neural network processing module 130 is used to input the behavior fusion features into a preset neural network model if the confidence level of the first steering intention is within the first confidence level range, so as to obtain the driver's second steering intention and its confidence level. The final intention determination module 140 is used to determine the final turning intention and the overall confidence level based on the confidence level of the first turning intention and the confidence level of the second turning intention; The steering activation module 150 is used to activate the steering signal of the vehicle in the corresponding direction based on the final steering intention if the overall confidence level is greater than a preset confidence threshold.

[0129] In one possible implementation, the behavior fusion features include eye gaze angle, head three-dimensional pose angle, and shoulder rotation angle; the rule engine processing module 120 includes: The judgment condition determination unit is used to determine the judgment condition for a first direction based on the vehicle driving scenario; the first direction is either left or right; the vehicle driving scenario includes at least one of vehicle speed and road attributes. The rule engine unit is used to determine whether the eye gaze angle, the head three-dimensional posture angle, the shoulder rotation angle, and the steering wheel angle meet the judgment conditions of the first direction; The first steering intention determination unit is configured to determine, if satisfied, the driver's first steering intention as the first direction.

[0130] In one possible implementation, the judgment condition determination unit specifically includes: The vehicle speed range determination subunit is used to determine the vehicle speed range in which the vehicle speed falls; The judgment condition set determination subunit is used to determine the judgment condition set corresponding to the first direction based on the vehicle speed range in which the vehicle speed is located; the judgment condition set includes a first feature combination, a judgment threshold corresponding to each behavior fusion feature in the first feature combination, and a duration threshold for each behavior fusion feature in the first feature combination to exceed the corresponding judgment threshold. Among them, the duration threshold corresponding to the behavior fusion feature in the judgment condition set where the vehicle speed range is in the low speed range is greater than the duration threshold corresponding to the same behavior fusion feature in the judgment condition set where the vehicle speed range is not in the low speed range, and the low speed range is the range where the vehicle speed is less than the preset speed.

[0131] In one possible implementation, the set of judgment conditions includes a first set of conditions and a second set of conditions; the judgment condition set determining subunit is used for: If the vehicle speed is within the low speed range, then determine whether the road attribute where the vehicle is currently located is an intersection; If the road attribute where the vehicle is currently located is an intersection, then the judgment condition for determining the first direction is a first condition set; the first condition set includes a second feature combination, a judgment threshold corresponding to each behavior fusion feature in the second feature combination, and a duration threshold for each behavior fusion feature in the second feature combination to exceed the corresponding judgment threshold; the first feature combination includes the second feature combination; If the road attribute where the vehicle is currently located is not an intersection, then the judgment condition for determining the first direction is the second condition set; the second condition set includes a third feature combination, a judgment threshold corresponding to each behavior fusion feature in the third feature combination, and a duration threshold for each behavior fusion feature in the third feature combination to exceed the corresponding judgment threshold; the first feature combination includes the third feature combination; In the second feature combination, the judgment threshold corresponding to the behavior fusion feature is greater than the judgment threshold corresponding to the same behavior fusion feature in the third feature combination.

[0132] In one possible implementation, the first confidence range includes an upper confidence limit and a lower confidence limit; After obtaining the driver's first steering intention and its confidence level, the vehicle steering signal activation device 100 provided in this embodiment further includes a method and a steering activation determination module, used for: If the confidence level of the first steering intention is greater than the upper limit of the confidence level, then the steering signal of the vehicle in the corresponding direction is activated according to the first steering intention. If the confidence level of the first steering intention is less than the lower confidence level, then the first steering intention is determined to be no steering.

[0133] In one possible implementation, the vehicle steering signal activation device 100 provided in this embodiment further includes: a confidence threshold adjustment module, used for: Before activating the steering signal corresponding to the vehicle's direction based on the final steering intention if the overall confidence level is greater than a preset confidence threshold, if the overall confidence level is within the first confidence level range for a preset duration before the current moment, then the preset confidence threshold is increased.

[0134] In one possible implementation, the final intent determination module 140 is specifically used for: Take the first steering intention as the final steering intention, and if the first steering intention and the second steering intention Figure 1 If the confidence levels of the first and second turning intentions are found to be consistent, a weighted sum is taken to determine the overall confidence level. If the first steering intention and the second steering intention are inconsistent, the confidence level of the first steering intention shall be used as the overall confidence level.

[0135] In one possible implementation, the final intent determination module 140 is also used for: Before determining the overall confidence level by weighted summation of the confidence levels of the first and second steering intentions, the weighting values ​​of the confidence levels of the first and second steering intentions are adjusted based on environmental information, including ambient light intensity and windshield wiper status.

[0136] This application also provides a computer program product having program code that, when run in a corresponding processor, controller, computing device, or control unit, executes the steps in any of the above-described vehicle steering signal activation method embodiments, for example... Figure 1 Steps S101 to S105 are shown. Those skilled in the art will understand that the methods and apparatus proposed in the embodiments of this application can be implemented in various forms, including hardware, software, firmware, dedicated processors, or combinations thereof. Dedicated processors may include application-specific integrated circuits (ASICs), reduced instruction set computers (RISCs), and / or field-programmable gate arrays (FPGAs). The proposed methods and apparatus are preferably implemented as a combination of hardware and software. The software is preferably installed as an application program on a program storage device. This is typically based on a machine with a computer platform, such as one or more central processing units (CPUs), random access memory (RAM), and one or more input / output (I / O) interfaces. An operating system is also typically installed on the computer platform. The various processes and functions described herein may be part of an application program, or a portion thereof may be executed by an operating system.

[0137] Figure 3 This is a schematic diagram of the controller provided in an embodiment of this application. Figure 3 As shown, the controller 3 in this embodiment includes a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30. When the processor 30 executes the computer program 32, it implements the steps in the aforementioned vehicle steering signal activation method embodiments, for example... Figure 1 Steps S101 to S105 are shown. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 2 The functions of modules 110 to 150 are shown.

[0138] For example, the computer program 32 may be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 30 to complete / implement the solution provided in this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 32 in the controller 3.

[0139] The controller 3 may include, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of controller 3 and does not constitute a limitation on controller 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, the controller may also include input / output devices, network access devices, buses, etc.

[0140] The processor 30 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0141] The memory 31 can be an internal storage unit of the controller 3, such as a hard disk or memory of the controller 3. The memory 31 can also be an external storage device of the controller 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the controller 3. Furthermore, the memory 31 can include both internal storage units and external storage devices of the controller 3. The memory 31 is used to store the computer program and other programs and data required by the controller. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0142] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0143] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0144] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software 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 this application.

[0145] In the embodiments provided in this application, it should be understood that the disclosed devices / controllers and methods can be implemented in other ways. For example, the device / controller embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0146] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] Furthermore, the functional units in the various embodiments of this application 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.

[0148] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-described embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-described vehicle steering signal activation method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0149] Furthermore, the features of the embodiments shown in the accompanying drawings or the various embodiments mentioned in this specification should not be construed as independent embodiments. Rather, each feature described in one example of an embodiment can be combined with one or more other desired features from other embodiments to produce other embodiments not described in words or with reference to the accompanying drawings.

[0150] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for activating a vehicle's turn signal, characterized in that, include: Real-time acquisition of driver behavior fusion features, including driver behavior data and steering wheel angle; The behavior fusion features are input into a preset rule engine to obtain the driver's first steering intention and its confidence level. If the confidence level of the first steering intention is within the first confidence level range, the behavior fusion feature is input into a preset neural network model to obtain the driver's second steering intention and its confidence level. Based on the confidence levels of the first and second steering intentions, the final steering intention and overall confidence level are determined. If the overall confidence level is greater than a preset confidence threshold, then the steering signal for the vehicle in the corresponding direction is activated based on the final steering intention.

2. The method for activating a vehicle's steering signal according to claim 1, characterized in that, The behavioral fusion features include eye gaze angle, head three-dimensional pose angle, and shoulder rotation angle; The step of inputting the behavior fusion features into a preset rule engine to obtain the driver's first steering intention includes: The determination criteria for the first direction are based on the vehicle driving scenario; the first direction is either left or right; the vehicle driving scenario includes at least one of vehicle speed and road attributes. Determine whether the eye gaze angle, the head three-dimensional posture angle, the shoulder rotation angle, and the steering wheel angle meet the judgment conditions of the first direction; If the conditions are met, then the driver's first steering intention is determined to be the first direction.

3. The method for activating a vehicle's steering signal according to claim 2, characterized in that, The conditions for determining the first direction based on the vehicle driving scenario include: Determine the speed range within which the vehicle speed falls; Based on the vehicle speed range, a set of judgment conditions corresponding to the first direction is determined; the set of judgment conditions includes a first feature combination, a judgment threshold corresponding to each behavior fusion feature in the first feature combination, and a duration threshold for each behavior fusion feature in the first feature combination to exceed the corresponding judgment threshold. Among them, the duration threshold corresponding to the behavior fusion feature in the judgment condition set where the vehicle speed range is in the low speed range is greater than the duration threshold corresponding to the same behavior fusion feature in the judgment condition set where the vehicle speed range is not in the low speed range, and the low speed range is the range where the vehicle speed is less than the preset speed.

4. The method for activating a vehicle's steering signal according to claim 3, characterized in that, The set of judgment conditions includes a first set of conditions and a second set of conditions; The set of judgment conditions for determining the first direction based on the vehicle speed range includes: If the vehicle speed is within the low speed range, then determine whether the road attribute where the vehicle is currently located is an intersection; If the road attribute where the vehicle is currently located is an intersection, then the judgment condition for determining the first direction is a first condition set; the first condition set includes a second feature combination, a judgment threshold corresponding to each behavior fusion feature in the second feature combination, and a duration threshold for each behavior fusion feature in the second feature combination to exceed the corresponding judgment threshold; the first feature combination includes the second feature combination; If the road attribute where the vehicle is currently located is not an intersection, then the judgment condition for determining the first direction is the second condition set; the second condition set includes a third feature combination, a judgment threshold corresponding to each behavior fusion feature in the third feature combination, and a duration threshold for each behavior fusion feature in the third feature combination to exceed the corresponding judgment threshold; the first feature combination includes the third feature combination; In the second feature combination, the judgment threshold corresponding to the behavior fusion feature is greater than the judgment threshold corresponding to the same behavior fusion feature in the third feature combination.

5. The method for activating a vehicle's steering signal according to claim 1, characterized in that, The first confidence level range includes an upper confidence level and a lower confidence level; After obtaining the driver's first steering intention and its confidence level, the method further includes: If the confidence level of the first steering intention is greater than the upper limit of the confidence level, then the steering signal of the vehicle in the corresponding direction is activated according to the first steering intention. If the confidence level of the first steering intention is less than the lower confidence level, then the first steering intention is determined to be no steering.

6. The method for activating a vehicle's steering signal according to claim 1, characterized in that, Before activating the steering signal for the vehicle's corresponding direction based on the final steering intention if the overall confidence level is greater than a preset confidence threshold, the method further includes: If the overall confidence level is within the first confidence level range for a preset time period before the current moment, then the preset confidence threshold is increased.

7. The method for activating a vehicle's steering signal according to claim 1, characterized in that, The determination of the final steering intention and overall confidence level based on the confidence levels of the first steering intention and the second steering intention includes: The first steering intention is taken as the final steering intention. If the first steering intention and the second steering intention are consistent, the confidence levels of the first steering intention and the second steering intention are weighted and summed to determine the overall confidence level. If the first steering intention and the second steering intention are inconsistent, the confidence level of the first steering intention shall be used as the overall confidence level.

8. The method for activating a vehicle's steering signal according to claim 7, characterized in that, Before determining the overall confidence level by weighted summation of the confidence levels of the first turning intention and the second turning intention, the method further includes: Based on environmental information, the weighted values ​​of the confidence levels of the first turning intention and the second turning intention are adjusted; the environmental information includes ambient light intensity and windshield wiper status.

9. A controller comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the vehicle steering signal activation method as described in any one of claims 1 to 8.

10. A vehicle, characterized in that, Includes the controller as described in claim 9.

Images