Intelligent pop-up control method and device of electric hidden pop-up ceiling handrail for vehicle

By combining multimodal perception and intent recognition capacitive sensing arrays and time-of-flight sensors with vehicle dynamic data, the problems of false triggering and dynamic operating condition adaptability of electric retractable roof handrails have been solved, realizing intelligent pop-up control of the handrails and improving control accuracy and passenger safety.

CN123067110APending Publication Date: 2026-11-10CHERY INTELLIGENT VEHICLE TECH (HEFEI) CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611150606.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-30
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing electric retractable roof grab bars are prone to accidentally popping out due to unintentional hand movements, clothing rubbing against them, or light objects floating by. Furthermore, they cannot adaptively adjust the pop-out strategy under dynamic vehicle conditions, resulting in inaccurate control and a high rate of false triggering.

Method used

Employing a multimodal perception and intent recognition method, the system acquires hand posture and motion trajectory through a capacitive sensing array and a time-of-flight sensor. Combined with vehicle dynamic data, it performs closed-loop control, generates a dynamically compensated grip trajectory bundle, identifies valid grip intentions, and adaptively adjusts the armrest pop-out strategy.

Benefits of technology

It significantly reduces the probability of false triggering, improves the accuracy and reliability of the armrest pop-out control, shortens the occupant's grip response time, enhances the timeliness and stability of auxiliary support under conditions of sharp turns, tilting and bumps, and improves the human-machine interaction experience and riding safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123067110A_ABST
    Figure CN123067110A_ABST
Patent Text Reader

Abstract

The application discloses an intelligent pop-up control method for an electric hidden pop-up ceiling handrail of a vehicle, and belongs to the field of vehicle interior control. The method comprises the following steps: extracting a capacitance feature atlas and a hand space posture feature in the capacitance feature atlas, comparing the features with a hand gripping state capacitance model respectively, and finally outputting a first confidence degree; obtaining a real-time motion trajectory of a close object and a natural gripping trajectory bundle; receiving a posture offset trend for correcting the natural gripping trajectory bundle, and correcting the natural gripping trajectory bundle into a dynamic compensation gripping trajectory bundle; comparing the real-time motion trajectory with the dynamic compensation gripping trajectory bundle according to a deviation degree, and outputting a second confidence degree; judging whether an effective gripping intention signal is outputted; setting a prediction time window, calculating a direction parameter and an amplitude parameter of an upper body of an occupant relative to the handrail in the prediction time window, and finally composing the posture offset trend; and obtaining a multi-stage pop-up strategy based on the posture offset trend.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application belongs to the field of vehicle interior control, and specifically relates to an intelligent pop-up control method and device for an electrically retractable pop-up headliner armrest for vehicles. Background Technology

[0002] The roof grab handle, located above the door frames and on the interior roof, provides support for passengers getting in and out of the vehicle and stabilizes their bodies during bumps and curves. Traditional fixed grab handles are exposed, disrupting the overall aesthetic of the roof interior and creating a feeling of confinement. Therefore, an electrically retractable solution has emerged: it retracts into the roof panel when not in use and pops out when needed. However, existing products rely solely on infrared or ultrasonic detection of approaching objects to trigger the pop-up, which has significant drawbacks: unintentional hand movements, clothing scratches, or light debris can all cause false pop-ups, interfering with the driving experience. Furthermore, the control logic uses fixed open-loop parameters, failing to distinguish between active gripping and unintentional approach. It also cannot adaptively adjust the pop-up strategy under dynamic conditions such as vehicle tilting, swaying, and bumps, making it difficult to match the real-time support needs of passengers. Therefore, we propose an intelligent pop-up control method and device for electrically retractable pop-up roof grab handles in vehicles. Summary of the Invention

[0003] To address the aforementioned problems, this application provides an intelligent pop-up control method for an electrically retractable pop-up roof grab handle for vehicles, comprising the following: Multimodal perception and intent recognition: Extract the capacitance feature map and the hand spatial posture features in the capacitance feature map, compare them with the capacitance model of human hand grasping state, and finally output the first confidence score; Acquire the real-time motion trajectory of the approaching object, as well as the natural grip trajectory bundle; and receive the attitude offset trend used to correct the natural grip trajectory bundle, and correct the natural grip trajectory bundle into a dynamically compensated grip trajectory bundle. The deviation is compared with the real-time motion trajectory and the dynamically compensated grip trajectory bundle to output a second confidence level. Determine whether to output a valid grasping intent signal based on the first confidence level and the second confidence level; Context awareness and strategy decision-making: Set a prediction time window, calculate the directional and amplitude parameters of the occupant's upper body relative to the handrail within the prediction time window, and finally form the attitude deviation trend; based on the attitude deviation trend, obtain a multi-stage ejection strategy.

[0004] Preferably, the multimodal perception and intent recognition are as follows: By deploying a capacitive sensing array around the hidden area of ​​the electric retractable pop-out roof armrest in the vehicle, the three-dimensional electric field disturbance caused by the approaching object is continuously sensed, and the capacitive feature spectrum reflecting the surface contour and internal dielectric properties of the approaching object is extracted. Extract hand spatial pose features from the capacitance feature map, including projected coordinates, vertical distance, and approach direction; A capacitance model of the human hand grasping state, including a spatial structure template and a signal change rate template, is obtained. Based on the capacitance feature map, the similarity between the hand spatial posture features and the capacitance model of the human hand grasping state is compared. Then, a weighted fusion is performed to output a first confidence level representing the probability that the approaching object is a human hand.

[0005] Preferably, the capacitive sensing array is composed of Composed of several capacitive sensing units arranged in a row and column rule, the sensing range of multiple capacitive sensing units together covers the hidden area of ​​the electric retractable pop-out roof armrest of the vehicle and its adjacent space to form a sensitive area for approaching objects; when a human hand enters the sensitive area, the water and electrolytes in the hand tissue change the coupling capacitance value between the sensing electrodes in the capacitive sensing unit, triggering a quantifiable three-dimensional electric field disturbance, and finally obtaining the capacitance feature spectrum. The hand spatial pose features in the capacitance feature map include projected coordinates, vertical distance, and approach direction, as detailed below: Projected coordinates in hand spatial posture characteristics: The projected coordinates of the hand in the direction parallel to the sensor array plane are obtained by weighting the changes in capacitance of each capacitive sensing unit. Vertical distance in hand spatial pose features: Vertical distance between hand and array: Calculate the overall capacitance response amplitude index of the array by using the weighted average of the capacitance changes of each capacitance sensing unit; The vertical distance between the hand and the array plane is obtained using a capacitance distance mapping function; The approach direction in the spatial pose features of the hand is calculated by the difference in hand position at two consecutive sampling times.

[0006] Preferably, the first confidence level is: The spatial structure similarity is calculated by comparing the capacitance feature map with the spatial structure template in the pre-stored human hand gripping state capacitance model template. Extract the time rate of change feature of the capacitance feature spectrum and define the time step. Signal rate of change matrix: The similarity between the real-time rate of change matrix and the signal rate of change template corresponding to the human hand grasping state capacitance model is calculated to obtain the signal rate of change similarity. By weighting and fusing spatial structure similarity and signal rate of change similarity, a comprehensive similarity in the biocapacitance dimension is obtained: Hand spatial pose feature matching degree calculation: The matching degree of hand spatial pose features is calculated by comparing them with a pre-stored reference gripping pose model. Define the hand spatial pose range as follows: desired projected coordinate region, desired vertical distance interval, and desired approach direction angle interval. Calculate the conformity of each sub-dimension separately, and then weight and fuse the conformity of each sub-dimension to obtain the comprehensive matching degree of hand spatial pose features. The biocapacitance dimension similarity and the hand spatial posture feature matching degree are weighted and fused to output the first confidence score. Preferably, a continuous coordinate sequence of the approaching object in three-dimensional space is obtained through a time-of-flight sensor, and the real-time motion trajectory and velocity vector of the approaching object are fitted based on the continuous coordinate sequence; Assume the time-of-flight sensor is at the sampling time The obtained three-dimensional original coordinates of the near object surface are ; Trajectory smoothing and velocity vector fitting: for continuous Original coordinate sequence of sampling points ( We perform weighted least squares smoothing fitting to obtain the smoothed coordinate sequence. ,in The smoothed coordinate sequence forms a near-real-time motion trajectory of the object. ; velocity vector Calculated by combining the difference between smoothed adjacent coordinates with moving average filtering; for continuous... For a preferred velocity sampling value, the dynamic compensation grip trajectory bundle is: Receive synchronously transmitted attitude offset trend, including direction participation amplitude parameter; The preset natural grip trajectory bundle is composed of Composed of reference trajectories; Dynamically compensate and correct the natural grip trajectory bundle to generate a dynamically compensated grip trajectory bundle. The corrections include translation corrections and rotation corrections, where: Translation correction: The reference points in the natural grip trajectory bundle are translated in the opposite direction to the occupant offset direction; Rotation correction: Rotate the reference point of the natural grip trajectory bundle around the vertical axis of the vehicle coordinate system by an angle proportional to the offset magnitude.

[0007] Preferably, the second confidence level: Calculate the real-time motion trajectory of an approaching object With dynamic compensation grip trajectory bundle The deviation between them; the average minimum distance is used as the trajectory deviation; trajectory deviation Convert to trajectory matching score ; Determine the magnitude of the smooth velocity vector Does it fall within the preset natural grip speed range? Obtain speed compliance : Calculate the second confidence level : ; In the formula The weight coefficients for the trajectory matching dimension, with values ​​ranging from... .

[0008] Preferably, the context awareness and strategy decision-making are as follows: Receive valid grip intention signals; acquire vehicle dynamic data; the vehicle dynamic data includes at least the lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate of the vehicle body, which together form the vehicle dynamic state vector: Based on the amplitudes of lateral acceleration and yaw rate, the first and second judgment thresholds for adjusting the vehicle dynamic condition index are defined. Based on vehicle dynamic data, within a preset prediction time window, the expected displacement direction and magnitude of the occupant's upper body relative to the position of the roof armrest are calculated according to the inertial coupling relationship between the occupant's upper body and the vehicle body, generating an attitude offset trend that includes direction and magnitude parameters; and the natural grip trajectory bundle is dynamically compensated and corrected based on the attitude offset trend.

[0009] Preferably, the first determination threshold and the second determination threshold are adjusted: The vehicle's dynamic state vector is composed of the vehicle's lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate. The vehicle dynamic state vector is low-pass filtered to obtain the filtered vehicle dynamic state vector. The vehicle dynamic condition index is defined by the filtered vehicle dynamic state vector. Adjust the first judgment threshold Second determination threshold : ; in, This serves as a reference value for the filtered lateral acceleration. This serves as a reference value for the filtered yaw rate. For a moment The lateral acceleration of the vehicle body after low-pass filtering For a moment The yaw rate of the vehicle body after low-pass filtering; When the vehicle dynamic operating condition index Exceeding the preset threshold for severe operating conditions When, the first judgment threshold is set. Second determination threshold Multiply by a factor less than Temporary modulation coefficient The first and second determination thresholds after modulation are obtained.

[0010] Preferably, the attitude offset trend is as follows: The lateral, longitudinal, and vertical offsets are calculated separately, and then combined to represent the attitude offset trend. ; The magnitude parameter of the attitude deviation trend is: The directional parameter is: ,in, This represents the lateral offset magnitude. This represents the longitudinal offset magnitude. Vertical offset magnitude ; Based on the aforementioned attitude offset trend, the natural grip trajectory bundle is dynamically compensated and corrected, including the pre-ejection drive speed parameter. Initial assist torque parameters Target lock-up holding torque parameters ,in: Receive attitude offset trend Define the strategy strength factor. Its value range is arrive ; Based on strategy strength factor At the preset minimum pre-ejection speed With maximum pre-ejection speed Interpolation calculation of pre-pop-up drive speed parameters between ; Initial assist torque parameters Based on strategy strength factor At the preset minimum initial assist torque With maximum initial assist torque Interpolation calculations are performed between them; Target lock-up holding torque parameters Based on strategy strength factor At the preset minimum lock-up holding torque With maximum lock-up holding torque Interpolation calculations are performed between them; And using direction parameters For the initial assist torque parameters Perform directional modulation: Extracting the lateral component of the directional parameter of the vehicle ,Right now In the vehicle coordinate system Projected values ​​on the axis; define the direction modulation coefficient. Initial assist torque parameters after directional modulation ; Pre-pop up drive speed parameters Initial assist torque parameters Target lock-up holding torque parameters Combined into a multi-stage pop-up strategy parameter set Output.

[0011] Compared with the prior art, this application has the following advantages: This invention provides an intelligent pop-up control method for an electrically retractable pop-up roof armrest in vehicles. It performs two-dimensional similarity matching between capacitance feature maps acquired by a capacitance sensor array and hand spatial posture features, and a pre-stored capacitance model of human hand gripping states. Furthermore, it combines real-time motion trajectories acquired by a time-of-flight sensor with dynamically compensated gripping trajectory bundles for trajectory consistency analysis. A first confidence level and a second confidence level are used to jointly determine a valid gripping intention. The armrest is only triggered to pop up when the occupant approaches the target and simultaneously meets both biocapacitance characteristics and natural gripping motion patterns. This effectively distinguishes between the occupant's genuine gripping action and non-target behaviors such as clothing swaying, objects passing by, or unconscious waving, significantly reducing the false trigger probability of traditional approach-based pop-up methods and improving the accuracy and reliability of the armrest pop-up control. Simultaneously, this invention further… The system acquires vehicle dynamic data such as lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate. Based on the vehicle's dynamic state, it predicts the occupant's posture deviation relative to the armrest within the prediction time window. It then performs translational and rotational dynamic compensation on the natural grip trajectory bundle, enabling the grip intention recognition process to adjust the reference grip model in real time according to changes in vehicle posture. Based on the posture deviation trend, it generates a corresponding multi-stage pop-up strategy, adaptively adjusting the pre-pop-up drive speed, initial assist torque, and locking torque. This ensures that the armrest pop-up action is consistent with the occupant's actual grip position and the vehicle's dynamic conditions, thereby shortening the occupant's grip response time and improving the timeliness and stability of the armrest's auxiliary support under conditions of sharp turns, tilting, and bumps. This further improves the human-machine interaction experience and passenger safety during vehicle operation.

[0012] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A schematic diagram of the overall workflow of the present invention is shown. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] A headliner grab handle specifically refers to an auxiliary handle installed on the side of the vehicle's headliner, usually located above the door frame, for passengers to grip. It provides a support point when getting in and out of the vehicle and helps stabilize the passenger's posture when driving on bumpy roads or cornering at high speeds. However, traditional fixed headliner grab handles are exposed in the headroom area when not in use, disrupting the overall smoothness and visual simplicity of the headliner interior and creating a feeling of confinement for some passengers. To address this, the industry has proposed the concept of an electrically retractable headliner grab handle, which retracts and hides inside the headliner panel when not in use, only popping out when needed. However, existing electric retractable roof handrails typically employ a single infrared or ultrasonic proximity sensor to detect an approaching object and trigger a pop-up control logic. This simple approach-based pop-up response mode exposes the following prominent problems in practical applications: First, any object that meets the reflection conditions—including an unconscious wave of an occupant's hand, a casual brush of clothing, or even a light object passing by—may cause the handrail to pop up inappropriately, resulting in a significant false triggering problem that seriously affects user experience and system reliability. Second, existing pop-up controls are all open-loop, fixed-parameter trigger executions, unable to distinguish between occupant's conscious grip and unconscious approach, nor can they proactively adjust the pop-up strategy to provide auxiliary support that matches the conditions when the vehicle is tilting, swaying, or experiencing severe bumps. The role of the handrail remains limited to a passive pop-up mechanism rather than an intelligent active safety assistance device.

[0017] Based on the shortcomings of the existing technology, and referring to Figure 1As shown in the illustration, this specific embodiment proposes an intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles. This method fundamentally solves the technical problems of inaccurate intent recognition, high false triggering rate, and inability to dynamically match vehicle operating conditions in existing solutions by constructing a closed-loop collaborative and computational interaction network between multimodal perception and intent recognition, as well as context perception and strategy decision-making. Its more detailed working principle is as follows: Multimodal perception and intent recognition: Extract the capacitance feature map and the hand spatial posture features in the capacitance feature map, compare them with the capacitance model of human hand grasping state, and finally output the first confidence score; A decision-level fusion architecture combining capacitive sensing and optical time-of-flight sensing is employed. The capacitive sensing array detects the three-dimensional electric field disturbances caused by a nearby object, extracting a capacitance feature map reflecting the object's surface contours and internal dielectric properties. This map is compared with a capacitance model of a human hand gripping the object, outputting a first confidence level. The time-of-flight sensor acquires a continuous three-dimensional coordinate sequence of the approaching object, fits its real-time motion trajectory and velocity vector, and compares this with a natural gripping trajectory bundle, outputting a second confidence level. When both the first confidence level and the second confidence level exceed a first decision threshold, a valid gripping intention signal is generated.

[0018] Furthermore, during operation, multimodal perception and intent recognition engage in deep interaction with context perception and policy decision-making, adapting to the current scenario. Specifically, after receiving the attitude offset trend synchronously transmitted from context perception and policy decision-making, multimodal perception and intent recognition, before comparing the deviation between the real-time motion trajectory of the approaching object and the preset natural grip trajectory bundle, first utilizes the direction and amplitude parameters contained in the attitude offset trend to perform translational correction on the reference position of the preset natural grip trajectory bundle in the vehicle coordinate system and rotational correction on the reference direction, generating a dynamically compensated grip trajectory bundle. Then, deviation comparison calculation is performed based on this dynamically compensated grip trajectory bundle.

[0019] The aforementioned interaction mechanism is based on a physical understanding of the changes in the motion characteristics of occupants' hands under dynamic vehicle conditions. When the vehicle is stationary or traveling smoothly in a straight line, the natural gripping motion path of the occupant's hands toward the overhead armrest is stable and pre-programmable in the vehicle coordinate system. At this time, the pre-programmed natural gripping trajectory bundle can accurately represent the ideal range of motion of the occupant's hands in space. However, when the vehicle experiences dynamic conditions such as tilting, yawing, rapid acceleration, or emergency braking, the occupant's upper body will move relative to the vehicle body due to its own inertia. Observed in the vehicle coordinate system, the starting position of the occupant's hands has deviated from the original calibration position as the body shifts, and the direction of their gripping action is also deflected due to the torso twisting. If the natural grip trajectory bundle, whose position and direction are fixed in the vehicle coordinate system, is still used as the reference benchmark for trajectory comparison, the actual gripping action made by the occupant under dynamic vehicle conditions will be incorrectly judged as deviating from the ideal trajectory in the vehicle coordinate system, resulting in missed judgment. That is, the occupant does have the intention to grip, but because the trajectory comparison benchmark fails to adjust synchronously with the occupant's body offset, the gripping intention cannot be correctly identified.

[0020] In the process of estimating occupant posture deviation trends during situational awareness and strategy decision-making, a posture deviation trend is generated, which quantitatively characterizes the expected deviation direction and magnitude of the occupant's upper body relative to the vehicle coordinate system. Since the movement path of the occupant's hands in space directly follows the spatial position of their upper body, this posture deviation trend essentially reflects the overall deflection of the starting reference position and pointing direction of the occupant's hand gripping path. Therefore, by using this to compensate for the reference position and reference pointing of the natural gripping trajectory bundle in the opposite direction, the corrected trajectory bundle can be realigned with the reachable area of ​​the occupant's hands after the deviation: translation correction moves the reference position of the trajectory bundle by a distance equal to the deviation magnitude in the opposite direction of the occupant's deviation direction, so that the starting area of ​​the trajectory bundle rematches the actual position of the occupant's hands after the deviation; rotation correction rotates the reference pointing of the trajectory bundle around the vertical axis by an angle proportional to the deviation magnitude, so that the spatial pointing of the trajectory bundle rematches the natural movement direction of the occupant's hands after the deviation. The dynamically compensated grip trajectory bundle generated after the above correction is restored to match the occupant's current actual body position and reachable area, thereby ensuring the effectiveness and accuracy of trajectory deviation comparison under vehicle dynamic conditions and avoiding the problem of missed judgment due to mismatch of comparison benchmark; The working principles of multimodal perception and intent recognition are explained in more detail below: Capacitance feature acquisition and comparison: By deploying a capacitive sensing array around the hidden area of ​​the electric retractable pop-out roof armrest in the vehicle, the three-dimensional electric field disturbance caused by the approaching object is continuously sensed, and the capacitive feature spectrum reflecting the surface contour and internal dielectric properties of the approaching object is extracted. Extract hand spatial pose features from the capacitance feature map, including projected coordinates, vertical distance, and approach direction; Obtain a pre-stored capacitance model of a human hand grasping state; perform similarity comparisons between the capacitance feature map and the hand spatial pose features in it and the capacitance model of the human hand grasping state, and then perform weighted fusion to output a first confidence level representing the possibility that the approaching object is a human hand; In more detail, the working principle of sub-step S1.1 is as follows: By utilizing the perturbation properties of the human hand, a biological body with high dielectric constant and conductivity, on the electric field in space, this technology achieves the dual functions of identifying the physical properties of approaching objects and spatial positioning. Its working principle is as follows: Capacitive sensing array is composed of Composed of multiple capacitive sensing units arranged in rows and columns, the sensing range of multiple capacitive sensing units together covers the hidden area of ​​the electric retractable pop-out roof armrest of the vehicle and its adjacent space to form a sensitive area for approaching objects; when a human hand enters the sensitive area, the water and electrolytes in the hand tissue change the coupling capacitance value between the sensing electrodes in the capacitive sensing unit, thereby triggering a quantifiable three-dimensional electric field disturbance. Furthermore, the physical position of each capacitive sensing unit in the capacitive sensing array is known and fixed. When an object approaches, the capacitive sensing units at different positions in the capacitive sensing array will output differentiated capacitance change values ​​due to the different distances from different parts of the object. Finally, the differentiated capacitance change values ​​together constitute a capacitance feature map that uniquely corresponds to the surface contour and internal dielectric properties of the approaching object. It reflects the surface contour morphology and internal dielectric properties of the approaching object and is used to analyze the physical properties of whether the approaching object is a hand. Hand spatial posture features are extracted from the capacitance feature map, reflecting the hand's position and posture relative to the capacitance sensing array in three-dimensional space. Since the capacitance change of each sensing unit is highly sensitive to the distance to the hand—the closer the distance, the stronger the coupling and the greater the capacitance change—the projection coordinates of the hand in the direction parallel to the array plane can be obtained by calculating the centroid of the spatial distribution of the capacitance change of each unit. The vertical distance between the hand and the array plane can be calculated through the attenuation mapping relationship between the overall amplitude of the capacitance change and the vertical distance. The approach direction of the hand can be obtained through position difference at consecutive time points. Finally, the collection of the above spatial coordinates and direction information constitutes the hand spatial posture features. The capacitive model of the human hand gripping state is obtained by: pre-collecting three-dimensional electric field disturbance data of multiple occupants when their hands approach the vehicle's electrically retractable pop-out roof armrest in a natural gripping motion; and establishing a benchmark template after feature extraction and statistical modeling to characterize the spatial distribution pattern and dynamic evolution law of the electric field disturbance induced by the capacitive sensing array when the human hand is in a gripping ready posture; and the capacitive model of the human hand gripping state includes at least the following two parts: Spatial structure template: When the human hand is in a typical grasping posture, the spatial distribution matrix of the capacitance change of each unit of the capacitance sensing array after normalization reflects the two-dimensional projection characteristics of the hand contour shape and internal dielectric properties in the array plane, and its corresponding capacitance feature spectrum. Signal rate of change template: a reference rate matrix of capacitance characteristic spectrum changes over time as a human hand approaches the handrail at a natural speed, reflecting the dynamic evolution of the hand approach action; Based on the capacitance feature map, and the hand spatial pose features therein, similarity comparisons are performed with the pre-stored capacitance model of human hand grasping state. Then, a weighted fusion is performed to output a first confidence score representing the probability that the approaching object is a human hand, as follows: Suppose that the capacitive sensing array includes OK Lieutenant General The first capacitive sensing unit, the first Line 1 The physical plane coordinates of the column capacitive sensing unit are ,in , At that moment The capacitance change output by the capacitance sensing unit is This value is the difference between the measured capacitance value and the reference capacitance value when no object is near.

[0021] Capacitance feature map extraction: Spatial filtering and normalization are performed on the original capacitance change matrix to obtain the capacitance feature map. : ; In the formula: For the spatial Gaussian smoothing filter kernel in the th... Line 1 The coefficients at the column are used to suppress the sensor's background noise and local abnormal disturbances; and They are time points The maximum and minimum values ​​of capacitance changes of all capacitive sensing units are used for MinMax normalization processing to map the feature values ​​to... Interval.

[0022] Spatial structure similarity calculation: Capacitance feature map Spatial structure template in the pre-stored human hand grasping state capacitance model template Spatial structure similarity is calculated. A two-dimensional correlation coefficient is used as a measure of spatial structure similarity. ; In the formula: and They are respectively and The mean; and They are respectively and Standard deviation; numerator is and The covariance reflects the degree of linear correlation between the spatial distribution patterns of the two.

[0023] The range of values ​​is The closer the value is The better the spatial structure matches.

[0024] Signal rate of change similarity calculation: Extracting the time rate of change characteristics of the capacitance feature map to reflect the dynamic evolution characteristics of electric field disturbances during the approach of an object. Defining time... Signal change rate matrix : ; In the formula is the sampling time interval of the capacitive sensing array.

[0025] Real-time rate of change matrix Signal rate of change template corresponding to the capacitance model of human hand grasping state Perform similarity calculation: ; The range of values ​​is The closer the value is The closer the dynamic evolution features match, the better.

[0026] Biocapacitance dimension comprehensive similarity: We weighted and fused spatial structure similarity and signal rate of change similarity to obtain a comprehensive similarity score based on biocapacitance dimension. : ; In the formula The weighting coefficients for the spatial structure dimension, with values ​​ranging from [value range missing]. The value is determined by the system calibration experiment.

[0027] Hand spatial pose feature extraction and matching degree calculation: Calculation of the projected coordinates of the hand in the array plane: The projected coordinates of the hand in the direction parallel to the sensor array plane are obtained by weighting the changes in capacitance of each capacitive sensing unit. ; ; Physical meaning: The larger the change in capacitance of a capacitive sensing unit, the higher its weight in coordinate calculation. The weighted center position physically approximates the projection of the position where the electric field disturbance caused by the hand is strongest onto the array plane.

[0028] Calculation of the vertical distance between the hand and the array: Calculate the overall capacitive response amplitude of the array. The weighted average of the capacitance changes of each capacitance sensing unit is used: ; In the formula The sensitivity calibration coefficients for each capacitive sensing unit are determined by factory calibration and are used to compensate for sensitivity differences between units at different locations.

[0029] Vertical distance between the hand and the array plane Through a preset capacitance distance mapping function get: ; In practical implementation, the mapping function can adopt an exponential decay model: ; In the formula: The coupling coefficient is related to the dielectric properties of the hand and the electrode area; The distance attenuation constant is determined by the electrode geometry and electric field distribution characteristics. This is used to offset background noise at a distance.

[0030] , , All parameters are predetermined and stored by the system calibration experiment.

[0031] Hand approach direction calculation: Hand approach direction vector Calculated by decomposing the hand position difference between two consecutive sampling times:

[0032] To suppress instantaneous measurement noise, continuous... The position sequence of each sampling point is smoothed by a sliding window before the difference is calculated.

[0033] Hand spatial pose feature matching degree calculation: The real-time calculated hand spatial pose features—including projected coordinates—are used to calculate the hand spatial pose features. Vertical distance and approach direction —Calculate the matching degree with the pre-stored reference grip posture model.

[0034] Define the range of hand spatial pose: desired projected coordinate region Expected vertical distance interval Expected approximation direction angle range .

[0035] Calculate the compliance of each sub-dimension separately: Projection position accuracy: ; In the formula The shortest distance between the current projected coordinates and the boundary of the desired region. This is the allowable deviation threshold.

[0036] Vertical distance compliance:

[0037] Approach direction conformity: ; In the formula The angle between the current approach direction vector and the reference center direction. The center angle of the desired approximation direction. When the directions are perfectly aligned... When the deviation exceeds hour .

[0038] Comprehensive matching degree of hand spatial pose features The above three sub-dimensions are weighted and fused together: ; In the formula , , The weight coefficients for each sub-dimension satisfy... The value is determined by the system calibration experiment.

[0039] First confidence level calculation: Combining biocapacitance dimension with similarity Comprehensive matching degree with hand spatial pose features Perform weighted fusion and output the first confidence level. : ; In the formula: These are the weighting coefficients for the biocapacitance dimension; These are the weighting coefficients for the spatial pose dimension; satisfying... The weight allocation is determined by the system calibration experiment.

[0040] The range of values ​​is Or mapped to after normalization The larger the value, the higher the overall degree of consistency between the approaching object and the known human hand model in terms of both physical properties and spatial posture, meaning that the approaching object is more likely to be a human hand.

[0041] The real-time motion trajectory of the approaching object in three-dimensional space is obtained by using a time-of-flight sensor, and the real-time motion trajectory and velocity vector of the approaching object are fitted based on the continuous coordinate sequence. The system acquires a preset natural grip trajectory bundle, specifically a set of three-dimensional spatial curves pre-established and stored in the system memory. This serves as a reference model for determining whether the real-time motion trajectory of an approaching object indicates a gripping intention. It depicts the range of motion paths that an occupant's hand should follow in three-dimensional space when gripping the roof armrest naturally and consciously from a normal sitting position under ideal conditions without vehicle dynamic interference. This trajectory bundle points to the central area of ​​the vehicle's electrically retractable pop-up roof armrest, and its spatial reference position and direction are defined with reference to the vehicle's coordinate system in a stationary state. The process of establishing the natural grip trajectory bundle is as follows: In the calibration stage, three-dimensional motion trajectory data of the hands of multiple occupants are collected when they reach for the electrically retractable pop-up roof armrest of the vehicle in a stationary vehicle with a natural grip action; statistical modeling and cluster analysis are performed on the collected multiple motion trajectories to extract several of the most representative reference trajectories; the above-mentioned reference trajectories are parameterized in the form of a spatial reference point sequence in the vehicle coordinate system to jointly constitute the preset natural grip trajectory bundle.

[0042] The system receives attitude offset trends including direction and amplitude parameters. When the vehicle tilts or yaws, the upper body of the occupant shifts relative to the vehicle coordinate system due to inertia, causing the natural gripping path of the occupant's hands toward the armrest to deflect as a whole relative to the vehicle coordinate system. Therefore, the system uses the direction and amplitude parameters included in the attitude offset trend to perform translational correction on the reference position of the preset natural gripping trajectory bundle in the vehicle coordinate system and rotational correction on the reference direction. The corrected natural gripping trajectory bundle is defined as a dynamically compensated gripping trajectory bundle. The real-time motion trajectory is compared with the dynamically compensated grip trajectory bundle to determine the deviation. At the same time, it is determined whether the velocity vector falls within the preset natural grip speed range, and a second confidence level is output to represent the possibility that the approaching object has a gripping intention.

[0043] The working principle of the second confidence level is explained in more detail below: By using a time-of-flight sensor, a continuous sequence of coordinates of an approaching object in three-dimensional space is acquired. Based on this continuous coordinate sequence, the real-time motion trajectory and velocity vector of the approaching object are fitted. Specifically, an optical time-of-flight sensor is used to acquire the three-dimensional spatial coordinate sequence of the approaching object, and intention recognition is achieved through trajectory fitting and kinematic analysis. The physical basis lies in the fact that the time-of-flight sensor emits modulated light pulses towards the approaching object and measures the phase difference or time difference between the emitted and received reflected light pulses, thereby acquiring the precise coordinates of the object's surface in three-dimensional space point by point. When the object moves continuously within the sensor's field of view, the coordinate points at each sampling moment constitute a continuous coordinate sequence, which contains kinematic information such as the object's motion path, velocity, and acceleration.

[0044] By smoothly fitting a continuous coordinate sequence, a near-real-time trajectory and velocity vector of the object are obtained. The magnitude and direction of the velocity vector reflect the object's current motion state and the predicted subsequent motion trend.

[0045] Simultaneously, the system receives attitude deviation trends, including directional and amplitude parameters. These trends quantitatively characterize the directional and amplitude parameters of the occupant's upper body relative to the vehicle coordinate system caused by vehicle dynamics. When the vehicle tilts or yaws, the occupant's upper body shifts relative to the vehicle coordinate system due to inertia, causing a general deflection of the natural grip path of their hands towards the armrest relative to the vehicle coordinate system. Therefore, the preset natural grip trajectory bundle is dynamically compensated and corrected to match the occupant's reachable hand area after the shift.

[0046] The specific compensation and correction method is as follows: using the direction and amplitude parameters contained in the attitude offset trend, the reference position of the preset natural grip trajectory bundle in the vehicle coordinate system is translated and the reference direction is rotated. The direction of translation correction is opposite to the occupant offset direction, and the magnitude of the translation is equal to the offset amplitude; the angle of rotation correction is proportional to the offset amplitude. The corrected trajectory bundle is defined as the dynamic compensation grip trajectory bundle.

[0047] Finally, the real-time motion trajectory of the approaching object is compared with the dynamic compensation grip trajectory bundle to determine the deviation. At the same time, it is determined whether the magnitude of its velocity vector falls within the preset natural grip speed range. The second confidence score is output in combination to quantify the possibility that the approaching object has a gripping intention. The specific principle is as follows: Assume the time-of-flight sensor is at the sampling time The obtained three-dimensional original coordinates of the near object surface are Each coordinate axis is defined within the vehicle coordinate system.

[0048] Trajectory smoothing and velocity vector fitting: for continuous Original coordinate sequence of sampling points ( We perform weighted least squares smoothing fitting to obtain the smoothed coordinate sequence. ,in The smoothed coordinate sequence forms a near-real-time motion trajectory of the object. .

[0049] velocity vector Calculated by combining the difference between smoothed adjacent coordinates with a moving average filter: ; To suppress measurement noise, for continuous A smoothed velocity vector is obtained by performing a moving average on each velocity sample value. : ; Dynamically compensated grip trajectory bundle generation: receiving synchronously transmitted attitude offset trends This includes directional parameters. With amplitude parameter .

[0050] The preset natural grip trajectory bundle is composed of It consists of several reference trajectories, of which the first one is the first one. The reference trajectory is composed of Definition of a spatial reference point sequence: ; Natural grasp trajectory bundle Perform dynamic compensation corrections to generate a dynamic compensation grip trajectory bundle. The correction operations include translation correction and rotation correction, where: Translation Correction: Each reference point in the natural grip trajectory bundle is translated in the opposite direction to the occupant offset direction. The translation vector is: ; Rotation correction: aligns the reference point of the natural grip trajectory bundle with the vertical axis around the vehicle coordinate system. The axis rotates by an angle proportional to the offset magnitude. : ; In the formula This is the preset reference arm length parameter.

[0051] Rotation matrix Defined as: ; Dynamic compensation for each reference point in the grasp trajectory bundle Obtained through a composite transformation of rotation followed by translation: ; The resulting dynamically compensated grip trajectory bundle is: ; Trajectory deviation calculation: Calculate the real-time motion trajectory of the approaching object. With dynamic compensation grip trajectory bundle The deviation between them. The average minimum distance is used as a measure of trajectory deviation: ; In the formula This represents the Euclidean distance. For the trajectory... Each point on In dynamic compensation grasp trajectory bundle Find the reference point with the smallest spatial distance from it, and take the average of the minimum distances of all points. The smaller the value, the closer the real-time motion trajectory is to the ideal grip trajectory bundle; Second confidence level calculation: Trajectory deviation degree Convert to trajectory matching score A linear decay function with threshold truncation is used: ; In the formula This is the preset tolerance threshold for deviation. When hour , indicating that the trajectories completely overlap; when hour This indicates that the trajectory deviation exceeds the tolerance range.

[0052] Simultaneously determine the magnitude of the smooth velocity vector. Does it fall within the preset natural grip speed range? Obtain speed compliance : ; Calculate the second confidence level : ; In the formula The weight coefficients for the trajectory matching dimension, with values ​​ranging from... The value is determined by the system calibration experiment. The range of values ​​is The higher the value, the more likely the person is to have the intention to grasp the object.

[0053] Fusion determination: The first confidence level is compared with a preset first determination threshold, and the second confidence level is compared with a preset second determination threshold. Only when the first confidence level exceeds the first determination threshold and the second confidence level exceeds the second determination threshold is the currently approaching object determined to simultaneously meet the two conditions of being a human hand and having a grasping intention, and a valid grasping intention signal is output, as follows: Let the first confidence level of the output be... The second confidence level is .in, It is a comprehensive confidence level that characterizes the overall degree of fit between the current approaching object and the pre-stored human hand model in two dimensions: biocapacitance features and spatial posture features. It is a comprehensive confidence level that characterizes the degree of matching between the real-time motion trajectory of the currently approaching object and the dynamically compensated grip trajectory bundle, as well as the degree of conformity between its motion speed and the natural grip speed range; This represents the current sampling time.

[0054] The preset first judgment threshold is The preset second judgment threshold is . It is the minimum confidence threshold for determining that the approaching object is indeed a human hand. These are the minimum confidence thresholds for determining whether an approaching object indicates a genuine intention to grasp it. Both thresholds are determined by the system during offline calibration through extensive testing with numerous positive and negative samples, and can be adaptively adjusted during use based on statistics of false triggers and missed triggers.

[0055] Effectively grasp intention signals The generation logic is defined by the following expression: ; in, Values When this signal is received, it indicates that the valid grasping intention has been confirmed, and the signal is output to the context awareness and strategy decision-making stage. Values At the same time, it continues to maintain the monitoring status and continuously senses and judges approaching objects.

[0056] Contextual Awareness and Strategy Decision-Making: Upon receiving a valid grasping intention signal, real-time dynamic data of the vehicle is acquired via the vehicle communication bus. Based on the inertial coupling relationship between the occupant and the vehicle body, the expected displacement direction and magnitude of the occupant's upper body relative to the armrest are calculated within a preset prediction time window, generating a posture offset trend. This posture offset trend is input into the strategy mapping model to determine a multi-stage ejection strategy that includes pre-ejection drive speed parameters, initial assist torque parameters, and target locking torque parameters, as detailed below: Vehicle dynamic data acquisition: Receiving valid grip intention signals, indicating that the occupant is about to or is in the process of gripping the roof armrest; at this time, the vehicle itself may be stationary or in a dynamic state such as tilting, swaying, or bumping during movement. Regardless of the vehicle's state, the occupant's gripping action is completed on the moving platform of the vehicle, and the movement trend of the occupant's upper body relative to the vehicle body is directly related to the vehicle's current motion state. Therefore, before executing the armrest ejection action, vehicle dynamic data is acquired in real time from the vehicle's inertial measurement unit or vehicle dynamic control system via the vehicle communication bus; the vehicle dynamic data includes at least the lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate of the vehicle body, so as to predict possible changes in the occupant's posture and determine a matching ejection strategy.

[0057] The aforementioned vehicle inertial measurement unit or vehicle dynamic control system is a standard electronic device already equipped on the vehicle. It continuously senses changes in the vehicle's acceleration in all directions and changes in its rotational angular velocity around each axis through built-in sensors such as accelerometers and gyroscopes. Specifically, it connects to the communication network via the vehicle communication bus interface to read the required vehicle dynamic data, including: Lateral acceleration: This reflects the intensity of the lateral inertial force acting on the vehicle body in the left-right direction when turning or changing lanes laterally. When a vehicle turns left, the body experiences lateral acceleration pointing to the right, and vice versa. Lateral acceleration is the main mechanical cause of the occupants' upper bodies leaning towards the outside of the curve; Longitudinal acceleration: reflects the intensity of the inertial force acting on the vehicle body in the longitudinal direction during acceleration or braking. When a vehicle accelerates rapidly, the upper body of the occupants leans backward due to inertia; when the vehicle brakes suddenly, the upper body of the occupants leans forward. Longitudinal acceleration is the main mechanical cause of the forward and backward leaning of the occupants' upper body.

[0058] Vertical acceleration of the vehicle body: This reflects the intensity of vibration and impact experienced by the vehicle body in the vertical direction when traversing uneven or bumpy road surfaces. Vertical acceleration is the main mechanical cause of the up-and-down movement of the occupants' bodies and their bouncing in the seats.

[0059] Yaw rate: Reflects the angular velocity of the vehicle's rotation about its vertical axis. Yaw rate is directly related to the severity of the vehicle's turn and is a precursor signal of vehicle roll, as well as a key parameter for predicting the tendency of occupants to sway laterally.

[0060] The vehicle's dynamic state vector is composed of the vehicle's lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate. : ; in: For a moment Lateral acceleration of the vehicle body. Its physical meaning is the acceleration of the vehicle body in the vehicle coordinate system. The acceleration component along the axial direction (towards the vehicle). A positive value indicates that the acceleration is directed to the left of the vehicle, and a negative value indicates that it is directed to the right of the vehicle.

[0061] For a moment The longitudinal acceleration of the vehicle body. Its physical meaning is the acceleration of the vehicle body in the vehicle coordinate system. The acceleration component along the axial direction (vehicle longitudinal direction). A positive value indicates that the acceleration direction is towards the front of the vehicle (usually corresponding to acceleration conditions), and a negative value indicates that the acceleration direction is towards the rear of the vehicle (usually corresponding to braking conditions).

[0062] For a moment The vertical acceleration of the vehicle body. Its physical meaning is the acceleration of the vehicle body in the vehicle coordinate system. The acceleration component in the axial direction (vertical to the vehicle). A positive value indicates that the acceleration direction is upward towards the vehicle (usually corresponding to bumpy conditions where the vehicle body moves upward), and a negative value indicates that the acceleration direction is downward towards the vehicle.

[0063] For a moment The yaw rate of the vehicle body. Its physical meaning is the yaw rate of the vehicle body around the vehicle coordinate system. The angular velocity of the rotation of the vertical axis. A positive value indicates that the vehicle is rotating counterclockwise around the vertical axis (usually corresponding to a left turn), and a negative value indicates that it is rotating clockwise (usually corresponding to a right turn).

[0064] To suppress high-frequency noise and transient spikes in vehicle dynamic data, the raw data is low-pass filtered. Taking lateral acceleration as an example, the filtered lateral acceleration... The calculation method is as follows: ; in, This is the smoothing coefficient of the low-pass filter, and its value range is... arrive between. The closer The faster the filtered data responds to the current sample value; The closer The smoother the filtered data, the greater the response delay. The sampling time interval for vehicle dynamic data is the time period between two data updates on the vehicle communication bus.

[0065] longitudinal acceleration Vertical acceleration and yaw rate All were subjected to low-pass filtering in the same manner to obtain the corresponding filtered values. , and .

[0066] Filtered vehicle dynamic state vector for: ; The filtered vehicle dynamic state vector As data input for estimating occupant attitude deviation trends.

[0067] Furthermore, while acquiring vehicle dynamic data, based on the amplitude of lateral acceleration and yaw rate, it can also predict whether the vehicle is currently in a severe dynamic condition, and accordingly perform feedforward modulation on the judgment threshold of multimodal perception and intent recognition.

[0068] Define vehicle dynamic operating condition index : ; in, This serves as a reference value for lateral acceleration. These are the reference values ​​for yaw rate. The two reference values ​​were determined through calibration experiments and represent the typical upper limits of lateral acceleration and yaw rate under normal, smooth driving conditions, respectively. For a moment The lateral acceleration of the vehicle body after low-pass filtering For a moment Yaw rate of the vehicle body after low-pass filtering Vehicle dynamic operating condition index It comprehensively reflects the intensity of the vehicle's lateral movement; its value is dimensionless, and the closer it is to... The higher the value, the more stable the vehicle is; the higher the value, the more turbulent the vehicle's dynamics.

[0069] When the vehicle dynamic operating condition index Exceeding the preset threshold for severe operating conditions When, the first judgment threshold is set. Second determination threshold Multiply by a factor less than Temporary modulation coefficient This temporarily lowers the judgment threshold, improving the sensitivity of recognizing grasping intentions in emergency situations. ; ; in, This is a temporary modulation coefficient, with a value range of [value range missing]. arrive between, and These are the first and second judgment thresholds after modulation, respectively. When the vehicle dynamic condition index falls back to... If the threshold is below this, the original judgment threshold is restored. This mechanism achieves a dynamic balance between high specificity for routine operating conditions and high sensitivity for emergency operating conditions.

[0070] Occupant posture deviation trend prediction: Based on vehicle dynamic data, within a preset prediction time window, according to the inertial coupling relationship between the upper body of the occupant and the vehicle body, the expected displacement direction and expected displacement amplitude of the upper body of the occupant relative to the position of the roof armrest are predicted, and an attitude deviation trend including direction parameters and amplitude parameters is generated. The natural grip trajectory bundle is dynamically compensated and corrected based on the attitude deviation trend, as follows: Received filtered vehicle dynamic state vector : ; in, This is the filtered lateral acceleration. This is the filtered longitudinal acceleration. This is the filtered vertical acceleration. This is the filtered yaw rate. This represents the current sampling time.

[0071] Let the equivalent mass of the upper body of the passenger be... This parameter represents the total mass of the occupant's head and torso. Let the equivalent stiffness at the connection between the occupant's upper body and the seat be... The equivalent damping coefficient is . , and All parameters are preset from human biomechanical measurement data, forming the core parameters of the occupant's upper body inertial coupling model.

[0072] Define the prediction time window length as This parameter represents the time span calculated from the current moment into the future. Its value is determined by the mechanical response time of the armrest pop-out mechanism from receiving the instruction to moving to the semi-expanded ready position. The typical value range is 200 milliseconds to 500 milliseconds.

[0073] Parameter settings for the equivalent inertial model of the occupant's upper body: Assume the equivalent mass of the occupant's upper body is... This parameter represents the total mass of the occupant's head and torso. Let the equivalent stiffness at the connection between the occupant's upper body and the seat be... The equivalent damping coefficient is . , and All parameters are preset from human biomechanical measurement data, forming the core parameters of the occupant's upper body inertial coupling model; Lateral offset calculation: The relative motion tendency of the occupant's upper body in the lateral direction of the vehicle is mainly caused by lateral acceleration and yaw rate. Lateral acceleration This causes the upper body of the occupant to experience a lateral inertial force, the direction of which is opposite to the direction of acceleration; yaw rate. The vehicle body rotates around a vertical axis, causing the upper body of the occupants to experience tangential displacement relative to the vehicle body due to inertial lag.

[0074] Through lateral acceleration Calculate lateral inertial force : ; The negative sign indicates that the direction of the inertial force is opposite to the direction of acceleration. When a vehicle turns left and generates lateral acceleration to the right, the upper body of the occupant experiences an inertial force to the left.

[0075] From yaw rate The resulting equivalent tangential acceleration of the upper body of the occupant for: ; in, The time rate of change of yaw rate (i.e., yaw acceleration) is obtained by differentiating the yaw rate at consecutive sampling times; The equivalent horizontal distance from the occupant's upper body center of mass to the vehicle's rotation center is preset based on the vehicle seat position and human sitting posture measurement data.

[0076] In the prediction time window Inside, the lateral displacement of the occupant's upper body in the horizontal direction. The calculation method is as follows: ; The first term of the calculation formula reflects the acceleration component of the occupant's upper body under the action of the lateral resultant force, and the second term reflects the effect of seat damping on the attenuation of motion. A positive value indicates that the upper body of the occupant is expected to shift to the left side of the vehicle, while a negative value indicates that the occupant is expected to shift to the right side of the vehicle.

[0077] Longitudinal offset magnitude calculation: The relative motion tendency of the occupant's upper body in the longitudinal direction of the vehicle is mainly caused by longitudinal acceleration. When the vehicle accelerates rapidly, the occupant's upper body leans backward due to inertia; when the vehicle brakes suddenly, the occupant's upper body leans forward.

[0078] Longitudinal inertial force The calculation method is as follows: ; In the prediction time window Inside, the longitudinal displacement of the occupant's upper body in the vertical direction The calculation method is as follows: ; A positive value indicates that the upper body of the occupant is expected to shift forward of the vehicle, while a negative value indicates that the occupant is expected to shift backward of the vehicle.

[0079] Vertical offset magnitude calculation: The tendency of the upper body of the occupant to move relative to the vertical direction of the vehicle is mainly caused by vertical acceleration. When the vehicle passes over a bumpy road, the vehicle body generates vertical acceleration, and the upper body of the occupant moves up and down relative to the seat due to inertia.

[0080] Vertical inertial force The calculation method is as follows: ; In the prediction time window Inside, the vertical displacement of the occupant's upper body in the vertical direction The calculation method is as follows: ; A positive value indicates that the upper body of the occupant is expected to move upward relative to the vehicle body, while a negative value indicates that it is expected to move downward.

[0081] Attitude offset trend generation: Combine the above-mentioned lateral offset magnitude, longitudinal offset magnitude, and vertical offset magnitude into a three-dimensional vector to generate the attitude offset trend. : The direction parameter of this attitude offset trend is determined by... , and The jointly determined three-dimensional spatial orientation represents the spatial direction of the expected offset of the occupant's upper body; the magnitude parameter of this attitude offset trend is the Euclidean length of the aforementioned three-dimensional vector: ; This indicates the total expected displacement of the occupant's upper body.

[0082] The unit direction vector of the direction parameter for: ; Pop-up Strategy Decision: The attitude offset trend is input into a preset strategy mapping model, which determines a multi-stage pop-up strategy that matches the current occupant attitude offset trend. The multi-stage pop-up strategy includes at least the following characteristic parameters: a pre-pop-up drive speed parameter to control the execution speed of the pre-pop-up stage, an initial assist torque parameter to control the initial output force of the active assist stage, and a target locking holding torque parameter to control the continuous support force after the vehicle's electrically retractable pop-up roof armrest is locked. A more detailed working principle is as follows: The attitude deviation trend is input into a preset strategy mapping model, which determines a multi-stage ejection strategy that matches the current occupant attitude deviation trend. The various characteristic parameters of the strategy are then output to control the ceiling handrail.

[0083] The attitude deviation trend is composed of both amplitude and direction parameters. The amplitude parameter quantitatively describes the total magnitude of the expected deviation of the occupant's upper body within the prediction time window. The larger the amplitude, the more severe the occupant's body imbalance and the more urgent the need for handrail support. The direction parameter quantitatively describes the spatial orientation of the expected deviation of the occupant's upper body, which determines the direction of the inertial force that should be primarily counteracted when the handrail is pushed.

[0084] The strategy mapping model is a pre-built decision model that maps occupant attitude deviation trends to ejection strategy parameters. Its core mapping logic is as follows: when the magnitude of the occupant attitude deviation trend is small, a gentler ejection strategy is decided, characterized by a slower pre-ejection drive speed, lower initial assist torque, and lower target locking torque, to avoid unnecessary disturbance to the occupant; when the magnitude of the occupant attitude deviation trend is large, a stronger ejection strategy is decided, characterized by a faster pre-ejection drive speed, higher initial assist torque, and higher target locking torque, to provide sufficient auxiliary support in time before significant occupant body displacement occurs.

[0085] The pre-expansion drive speed parameter is used to measure the motor speed during the pre-expansion phase. This parameter determines how quickly the handrail moves from the fully concealed position to the semi-expanded ready position. When the occupant's posture is stable, the pre-expansion speed is slower, and the handrail reaches the ready position quietly and smoothly. When the occupant's posture shifts drastically, the pre-expansion speed is increased to shorten the time it takes for the handrail to reach the ready position, thus providing a window of opportunity for the occupant to grab it urgently.

[0086] The initial assist torque parameter is used to determine the initial output torque of the motor during the active assist locking phase. This parameter determines the initial force required to push the armrest from the semi-deployed ready position to the fully deployed position. When the occupant's posture is stable, the initial assist torque is lower, and the armrest is gently pushed into the occupant's hands; when the occupant's posture shifts sharply, the initial assist torque is higher, and the armrest actively meets the occupant's hands with greater force, providing an assist force to counteract inertia.

[0087] The target locking holding torque parameter controls the continuous output of holding torque by the motor after the armrest reaches the fully extended position and completes mechanical locking. This parameter determines the continuous support the armrest can provide to the occupant in the locked state. When the occupant's posture is stable, the target locking holding torque is lower, maintaining only a basic mechanical locking state; when the occupant's posture shifts sharply, the target locking holding torque is higher to provide stable traction support to the occupant during continuous vehicle bumps or cornering, preventing the occupant from losing their grip due to body swaying.

[0088] The strategy mapping model is established as follows: During the calibration phase, by collecting the posture data of passengers gripping the handrail under different vehicle dynamic conditions and the corresponding ideal pop-up strategy parameters, a mapping relationship table or mapping function from posture offset trend to pop-up strategy parameters is established and stored for real-time retrieval.

[0089] The above specifically refers to: Receive attitude offset trend This includes amplitude parameters. and direction parameters .in, The total estimated displacement of the occupant's upper body; This is the unit direction vector for the expected offset of the occupant's upper body.

[0090] Mapping from offset magnitude to policy strength: Define strategy strength factor Its value range is arrive This is used to uniformly represent the overall strength level of the ejection strategy corresponding to the current crew attitude deviation trend. The gentlest pop-up strategy, The most powerful pop-up strategy.

[0091] Strategy strength factor Through a preset nonlinear mapping function From the offset magnitude Calculation yielded: ; In the specific implementation, the mapping function An S-shaped function with saturation characteristics is used to ensure smooth saturation at both ends of the offset range, avoiding abrupt changes in policy strength near the critical value. ; in, The steepness parameter of the mapping function controls how quickly the strength of the control strategy changes with the offset magnitude, and is determined by calibration experiments. The midpoint offset magnitude of the mapping function, when Time strategy strength factor Values It is determined by calibration experiments. Represented by natural constant An exponential function with base 0.

[0092] when Less than hour, Approaching This indicates that the occupants are in a stable posture, and the gentlest ejection strategy is employed; when Greater than hour, Approaching This indicates a severe shift in the occupant's attitude, prompting the use of the most aggressive ejection strategy.

[0093] Pre-pop up driver speed parameters: Pre-pop-up drive speed parameters Based on strategy strength factor At the preset minimum pre-ejection speed With maximum pre-ejection speed Interpolation calculations are performed between them: ; in, This is the lower limit of the pre-ejection drive speed, corresponding to the gentle ejection speed under stable occupant posture conditions; This is the upper limit of the pre-ejection drive speed, corresponding to the rapid ejection speed under conditions of severe occupant attitude deviation.

[0094] When the strategy strength factor hour, The handrails will pop out at the slowest possible speed; when hour, The handrails will pop out as quickly as possible.

[0095] Initial assist torque parameters: Initial assist torque parameters Based on strategy strength factor At the preset minimum initial assist torque With maximum initial assist torque Interpolation calculations are performed between them: ; in, This is the lower limit of the initial assist torque, corresponding to the gentle pushing force under stable occupant posture conditions; This represents the upper limit of the initial assist torque, corresponding to the powerful pushing force under conditions of severe occupant posture deviation.

[0096] When the strategy strength factor hour, The armrest is pushed with minimal torque; when hour, The handrail is pushed with maximum torque.

[0097] Target lock-up holding torque parameter decision: Target lock-up holding torque parameters Based on strategy strength factor At the preset minimum lock-up holding torque With maximum lock-up holding torque Interpolation calculations are performed between them: ; in, The lower limit of the target locking holding torque corresponds to the basic holding force that maintains the mechanical locking state under stable occupant posture conditions. This represents the upper limit of the target locking torque, corresponding to the maximum holding force that provides strong and continuous support to the occupants under conditions of severe occupant posture displacement.

[0098] When the strategy strength factor hour, The handrail is only subjected to basic holding torque; when hour, The handrail applies maximum holding torque.

[0099] Directional modulation of the initial assist torque by directional parameters: To further match the spatial orientation of occupant offset, orientation parameters are used. For the initial assist torque parameters Directional modulation is performed to make the direction of the assist torque correspond to the direction of the occupant's offset, so as to more effectively counteract the inertial offset of the occupant's body.

[0100] Extracting the lateral component of the directional parameter of the vehicle ,Right now In the vehicle coordinate system Projection value on the axis. The range of values ​​is arrive A positive value indicates that the occupant is deflected to the left side of the vehicle, while a negative value indicates that the occupant is deflected to the right side of the vehicle.

[0101] Define the directional modulation coefficient : ; in, This is the directional modulation gain coefficient, with a value range of [value range missing]. arrive It is determined by calibration experiments. for The absolute value of.

[0102] Initial assist torque parameters after direction modulation for: ; When the occupant's offset direction is mainly lateral. Larger, directional modulation coefficient A larger initial assist torque results in a correspondingly larger initial assist torque to provide stronger lateral assist force; when the occupant's offset direction is predominantly longitudinal or vertical, Smaller, directional modulation coefficient Approaching The initial assist torque remains basically at the original level.

[0103] Output of the multi-stage pop-up strategy: The above-determined characteristic parameters (pre-ejection drive speed parameters) Initial assist torque parameters Target lock-up holding torque parameters This is combined into a multi-stage pop-up strategy parameter set. : ; in, To pre-pop up the drive speed parameters, These are the initial assist torque parameters after direction modulation. The target lock-up holding torque parameter.

[0104] 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; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A smart pop-up control method for an electrically retractable pop-up roof armrest for vehicles, characterized in that, Including the following: Multimodal perception and intent recognition: Extract the capacitance feature map and the hand spatial posture features in the capacitance feature map, compare them with the capacitance model of human hand grasping state, and finally output the first confidence score; Acquire the real-time motion trajectory of approaching objects, as well as the natural grasping trajectory bundle; It also receives the attitude offset trend used to correct the natural grip trajectory bundle, and corrects the natural grip trajectory bundle into a dynamically compensated grip trajectory bundle. The deviation is compared with the real-time motion trajectory and the dynamically compensated grip trajectory bundle to output a second confidence level. Determine whether to output a valid grasping intent signal based on the first confidence level and the second confidence level; Contextual awareness and strategic decision-making: Set a prediction time window, calculate the directional and amplitude parameters of the occupant's upper body relative to the handrail within the prediction time window, and finally form the posture deviation trend; A multi-stage pop-up strategy is derived based on the attitude offset trend.

2. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 1, characterized in that: The multimodal perception and intent recognition are as follows: By deploying a capacitive sensing array around the hidden area of ​​the electric retractable pop-out roof armrest in the vehicle, the three-dimensional electric field disturbance caused by the approaching object is continuously sensed, and the capacitive feature spectrum reflecting the surface contour and internal dielectric properties of the approaching object is extracted. Extract hand spatial pose features from the capacitance feature map, including projected coordinates, vertical distance, and approach direction; A capacitance model of the human hand grasping state, including a spatial structure template and a signal change rate template, is obtained. Based on the capacitance feature map, the similarity between the hand spatial posture features and the capacitance model of the human hand grasping state is compared. Then, a weighted fusion is performed to output a first confidence level representing the probability that the approaching object is a human hand.

3. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 2, characterized in that: The capacitive sensing array is composed of It consists of capacitive sensing units arranged in rows and columns. The sensing range of multiple capacitive sensing units together covers the hidden area of ​​the electric retractable pop-out roof armrest of the vehicle and its adjacent space, so as to form a sensitive area for approaching objects. When a human hand enters the sensitive area, the water and electrolytes in the hand tissue change the coupling capacitance value between the sensing electrodes in the capacitive sensing unit, triggering a quantifiable three-dimensional electric field disturbance, and finally obtaining the capacitance feature spectrum. The hand spatial pose features in the capacitance feature map include projected coordinates, vertical distance, and approach direction, as detailed below: Projected coordinates in hand spatial posture characteristics: The projected coordinates of the hand in the direction parallel to the sensor array plane are obtained by weighting the changes in capacitance of each capacitive sensing unit. Vertical distance in hand spatial pose features: Vertical distance between hand and array: Calculate the overall capacitance response amplitude index of the array by using the weighted average of the capacitance changes of each capacitance sensing unit; The vertical distance between the hand and the array plane is obtained using a capacitance distance mapping function; The approach direction in the spatial pose features of the hand is calculated by the difference in hand position at two consecutive sampling times.

4. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 3, characterized in that: The first confidence level is: The spatial structure similarity is calculated by comparing the capacitance feature map with the spatial structure template in the pre-stored human hand grasping state capacitance model template. Extract the time rate of change feature of the capacitance feature spectrum and define the time step. Signal rate of change matrix: The similarity between the real-time rate of change matrix and the signal rate of change template corresponding to the human hand grasping state capacitance model is calculated to obtain the signal rate of change similarity. By weighting and fusing spatial structure similarity and signal rate of change similarity, a comprehensive similarity in the biocapacitance dimension is obtained: Hand spatial pose feature matching degree calculation: The matching degree of hand spatial pose features is calculated by comparing them with a pre-stored reference gripping pose model. Define the hand spatial pose range as follows: desired projected coordinate region, desired vertical distance interval, and desired approach direction angle interval. Calculate the conformity of each sub-dimension separately, and then weight and fuse the conformity of each sub-dimension to obtain the comprehensive matching degree of hand spatial pose features. The biocapacitance dimension similarity and the hand spatial posture feature matching degree are weighted and fused to output the first confidence score.

5. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 1, characterized in that: By using a time-of-flight sensor, a continuous sequence of coordinates of the approaching object in three-dimensional space is acquired, and the real-time motion trajectory and velocity vector of the approaching object are fitted based on the continuous coordinate sequence. Assume the time-of-flight sensor is at the sampling time The obtained three-dimensional original coordinates of the near object surface are ; Trajectory smoothing and velocity vector fitting: for continuous Original coordinate sequence of sampling points ( We perform weighted least squares smoothing fitting to obtain the smoothed coordinate sequence. ,in The smoothed coordinate sequence forms a near-real-time motion trajectory of the object. ; velocity vector Calculated by combining the difference between smoothed adjacent coordinates with moving average filtering; For continuous A smoothed velocity vector is obtained by performing a moving average on the velocity samples.

6. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 5, characterized in that: The dynamically compensated grip trajectory bundle: Receive synchronously transmitted attitude offset trend, including direction participation amplitude parameter; The preset natural grip trajectory bundle is composed of Composed of reference trajectories; Dynamically compensate and correct the natural grip trajectory bundle to generate a dynamically compensated grip trajectory bundle. The corrections include translation corrections and rotation corrections, where: Translation correction: The reference points in the natural grip trajectory bundle are translated in the opposite direction to the occupant offset direction; Rotation correction: Rotate the reference point of the natural grip trajectory bundle around the vertical axis of the vehicle coordinate system by an angle proportional to the offset magnitude.

7. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 2, characterized in that: The context awareness and strategy decision-making are as follows: Receive valid grip intention signals; acquire vehicle dynamic data; the vehicle dynamic data includes at least the lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate of the vehicle body, which together form the vehicle dynamic state vector: Based on the amplitudes of lateral acceleration and yaw rate, the first and second judgment thresholds for adjusting the vehicle dynamic condition index are defined. Based on vehicle dynamic data, within a preset prediction time window, the expected displacement direction and magnitude of the upper body of the passenger relative to the position of the roof armrest are calculated according to the inertial coupling relationship between the upper body of the passenger and the vehicle body, and the attitude offset trend including direction parameters and magnitude parameters is generated. The natural grip trajectory bundle is dynamically compensated and corrected based on the attitude deviation trend.

8. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 7, characterized in that: Adjust the first and second determination thresholds: The vehicle's dynamic state vector is composed of the vehicle's lateral acceleration, longitudinal acceleration, vertical acceleration, and yaw rate. The vehicle dynamic state vector is low-pass filtered to obtain the filtered vehicle dynamic state vector. The vehicle dynamic condition index is defined by the filtered vehicle dynamic state vector. Adjust the first judgment threshold Second determination threshold : ; in, This serves as a reference value for the filtered lateral acceleration. This serves as a reference value for the filtered yaw rate. For a moment The lateral acceleration of the vehicle body after low-pass filtering For a moment The yaw rate of the vehicle body after low-pass filtering; When the vehicle dynamic operating condition index Exceeding the preset threshold for severe operating conditions When, the first judgment threshold is set. Second determination threshold Multiply by a factor less than Temporary modulation coefficient The first and second determination thresholds after modulation are obtained.

9. The intelligent pop-up control method for an electrically retractable pop-up roof armrest for vehicles according to claim 8, characterized in that: The attitude shift trend is as follows: The lateral, longitudinal, and vertical offsets are calculated separately, and then combined to represent the attitude offset trend. ; The magnitude parameter of the attitude deviation trend is: The directional parameter is: ,in, This represents the lateral offset magnitude. This represents the longitudinal offset magnitude. Vertical offset magnitude ; Based on the aforementioned attitude offset trend, the natural grip trajectory bundle is dynamically compensated and corrected, including the pre-ejection drive speed parameter. Initial assist torque parameters Target lock-up holding torque parameters ,in: Receive attitude offset trend Define the strategy strength factor. Its value range is arrive ; Based on strategy strength factor At the preset minimum pre-ejection speed With maximum pre-ejection speed Interpolation calculation of pre-pop-up drive speed parameters between ; Initial assist torque parameters Based on strategy strength factor At the preset minimum initial assist torque With maximum initial assist torque Interpolation calculations are performed between them; Target lock-up holding torque parameters Based on strategy strength factor At the preset minimum lock-up holding torque With maximum lock-up holding torque Interpolation calculations are performed between them; And using direction parameters For the initial assist torque parameters Perform directional modulation: Extracting the lateral component of the directional parameter of the vehicle ,Right now In the vehicle coordinate system Projected values ​​on the axis; define the direction modulation coefficient. Initial assist torque parameters after directional modulation ; Pre-pop up drive speed parameters Initial assist torque parameters Target lock-up holding torque parameters Combined into a multi-stage pop-up strategy parameter set Output.

10. An intelligent pop-out control device for an electrically retractable pop-out roof armrest for vehicles, characterized in that, include: A capacitive sensing array is arranged around the hidden area of ​​the electric retractable pop-out roof armrest for vehicles. It consists of M×N capacitive sensing units arranged in a row and column rule. The sensing range of multiple capacitive sensing units together covers the hidden area and its adjacent space to form a sensitive area for approaching objects. It is used to continuously sense the three-dimensional electric field disturbance caused by the approaching object and output a capacitance feature spectrum. A time-of-flight sensor, positioned near the hidden area, is used to acquire a continuous sequence of coordinates of the approaching object in three-dimensional space; The handrail actuator includes at least a drive motor and a mechanical locking assembly for performing pre-ejection, active assistance push and locking holding actions of the handrail in response to a multi-stage ejection strategy; The controller is communicatively connected to the capacitive sensing array, the time-of-flight sensor, and the handrail actuator, respectively, and is configured to: Extract the hand spatial posture features from the capacitance feature map, compare them with the pre-stored human hand grasping state capacitance model, and output the first confidence score after weighted fusion. Based on the continuous coordinate sequence, the real-time motion trajectory and velocity vector of the approaching object are fitted, and the attitude offset trend is received to dynamically compensate and correct the preset natural grip trajectory bundle to generate a dynamically compensated grip trajectory bundle. The deviation between the real-time motion trajectory and the dynamic compensation grip trajectory bundle is compared, and a second confidence level is output by combining the speed conformity. When the first confidence level exceeds the first determination threshold and the second confidence level exceeds the second determination threshold, a valid grasping intent signal is output; In response to the effective grip intention signal, vehicle dynamic data is acquired, and within a preset prediction time window, the expected displacement direction and expected displacement amplitude of the upper body of the occupant relative to the position of the armrest are calculated based on the inertial coupling relationship between the upper body of the occupant and the vehicle body, and an attitude offset trend including direction parameters and amplitude parameters is generated. Based on the attitude deviation trend, a multi-stage pop-up strategy is determined, which includes pre-pop-up drive speed parameters, initial assist torque parameters, and target locking holding torque parameters, and the multi-stage pop-up strategy is output to the handrail actuator.